Stepwise photocatalytic dissociation of methanol and water on TiO 2(110)

Article abstract of DOI:10.1021/ja304049x

We have investigated the photocatalysis of partially deuterated methanol (CD₃OH) and H₂O on TiO₂(110) at 400 nm using a newly developed photocatalysis apparatus in combination with theoretical calculations. Photocatalyzed products, CD₂O on Ti_5c sites, and H and D atoms on bridge-bonded oxygen (BBO) sites from CD₃OH have been clearly detected, while no evidence of H₂O photocatalysis was found. The experimental results show that dissociation of CD₃OH on TiO₂(110) occurs in a stepwise manner in which the O-H dissociation proceeds first and is then followed by C-D dissociation. Theoretical calculations indicate that the high reverse barrier to C-D recombination and the facile desorption of CD₂O make photocatalytic methanol dissociation on TiO₂(110) proceed efficiently. Theoretical results also reveal that the reverse reactions, i.e, O-H recombination after H₂O photocatalytic dissociation on TiO₂(110), may occur easily, thus inhibiting efficient photocatalytic water splitting.

Full text of DOI:10.1021/ja304049x

Article
pubs.acs.org/JACS
Stepwise Photocatalytic Dissociation of Methanol and Water on
TiO2(110)
Qing Guo,† Chenbiao Xu,†,§ Zefeng Ren,† Wenshao Yang,† Zhibo Ma,† Dongxu Dai,† Hongjun Fan,*,†
Timothy K. Minton,‡,∥,* and Xueming Yang*,†
†State Key Laboratory of Molecular Reaction Dynamics, Dalian Institute of Chemical Physics, 457 Zhongshan Road, Dalian 116023,
Liaoning, P. R. China
‡Department of Chemistry and Biochemistry, Montana State University, Bozeman, Montana 59717, United States
S
* Supporting Information
ABSTRACT: We have investigated the photocatalysis of
partially deuterated methanol (CD3OH) and H2O on
TiO2(110) at 400 nm using a newly developed photocatalysis
apparatus in combination with theoretical calculations. Photo-
catalyzed products, CD2O on Ti5c sites, and H and D atoms on
bridge-bonded oxygen (BBO) sites from CD3OH have been
clearly detected, while no evidence of H2O photocatalysis was
found. The experimental results show that dissociation of
CD3OH on TiO2(110) occurs in a stepwise manner in which
the O−H dissociation proceeds first and is then followed by
C−D dissociation. Theoretical calculations indicate that the
high reverse barrier to C−D recombination and the facile
desorption of CD2O make photocatalytic methanol dissocia-
tion on TiO2(110) proceed efficiently. Theoretical results also
reveal that the reverse reactions, i.e, O−H recombination after H2O photocatalytic dissociation on TiO2(110), may occur easily,
thus inhibiting efficient photocatalytic water splitting.
catalysis of CH3OH on TiO2(110) was investigated in a two-
photon photoemission (2PPE) experiment, which inferred the
presence of an excited electronic state on the surface.20,21 Zhou
et al. attributed this surface state to a photocatalytic dissociated
state of methanol using a time-dependent 2PPE (TD-2PPE)
technique.22 They also used a combination of photoexcitation
with STM and found that 400 nm light could induce
dissociation of methanol on the surface, and they assigned
the dissociated state as methoxy (CH3O) on a Ti5c site and a
hydrogen atom on a BBO site. Shen and Henderson
subsequently conducted a TPD study of CH3OH photo-
catalysis on TiO2(110), and they detected a formaldehyde
(H2CO) product and suggested that CH3O is the active
intermediate that leads to H2CO.23 While the mechanism of
methanol photocatalysis on TiO2(110) is slowly being revealed,
the results to date do not unambiguously reveal the reactions
that lead from CH3OH to H2CO. By understanding this
reaction sequence in detail and comparing it to analogous
photocatalytic reactions of H2O on TiO2(110), we hope to
learn the important features of the processes that give rise to
the difference in the photocatalytic activity of CH3OH and
H2O on TiO2(110).
INTRODUCTION
■
Titanium dioxide has been extensively investigated as a catalyst
or photocatalyst,1−11 particularly in applications involving
photodegradation of organic molecules and water split-
ting,5,12,13 which have important implications in environmental
remediation and clean energy. Pure TiO2 is apparently not
photocatalytically active for splitting water to produce hydro-
gen,14 but the addition of methanol to water can dramatically
enhance the photocatalytic activity for hydrogen production.15
Therefore, understanding the key differences between the
photocatalytic chemistry of methanol and water on a model
TiO2 surface at the molecular level may provide valuable insight
into the dynamics of photocatalysis that would enhance efforts
for developing new and efficient photocatalysts for water
splitting.
Theoretical and experimental studies often focus on
TiO2(110) as a model surface,6,16 with the methanol/
TiO2(110) system serving as a model for photocatalysis on
TiO2.17−20 Henderson and co-workers18 conducted a temper-
ature-programmed desorption (TPD) study of CH3OH on
TiO2(110) and concluded that the majority of the CH3OH
molecules are adsorbed in molecular form. This conclusion is
consistent with a scanning tunneling microscopy study by
Dohnalek et al.10 that showed that methanol molecules are
adsorbed molecularly on the Ti5c sites and are dissociated only
at bridge-bonded oxygen (BBO) vacancy sites. The photo-
Received: April 27, 2012
Published: July 13, 2012
© 2012 American Chemical Society
13366
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373

Journal of the American Chemical Society
Article
Figure 1. Schematic diagram of the surface photocatalysis apparatus.
10−12 Torr. This low ambient pressure allows the measurement of
temperature-programmed desorption (TPD) signals with very high
sensitivity.
The TiO2(110) surfaces used in the experiments were prepared by
repeated cycles of Ar+ sputtering and UHV annealing at 850 K and
were characterized by Auger spectroscopy and LEED to make sure
that the surfaces were clean and flat. According to the temperature-
programmed desorption (TPD) of water, the TiO2(110) surface had
an oxygen defect density of about 4%.24
The photocatalytic methanol dissociation process was studied by
irradiating TiO2(110) surfaces that were dosed with deuterated
methanol (CD3OH) at a surface temperature of 100 K and then
conducting TPD measurements of various desorbed species to probe
the parent CD3OH molecules and the reaction products. CD3OH,
with a purity of >99.9%, was purchased from Sigma-Aldrich. Before
use, it was purified further by several freeze−pump−thaw cycles. The
surface was dosed with a 0.5 ML (1 ML = 5.2 × 1014 molecules cm−2)
coverage of CD3OH using a home-built, calibrated molecular beam
doser. The irradiating laser light had a nominal wavelength of 400 nm,
a pulse width of ∼50 fs and a bandwidth of ∼20 nm. The average
power of the light was 460 mW. The laser irradiation area on the
surface was the elliptical projection of a 6 mm diameter laser beam
with an angle of about 30° between the surface parallel and laser
direction. The laser flux on the surface was about 1.3 × 1018 photons
cm−2 s−1. The TiO2(110) surface was annealed in vacuum at 850 K for
20 min between TPD experiments to flatten and clean the surface.
We have conducted a combined experiment/theory study
using laser-induced photocatalysis and temperature-pro-
grammed desorption (TPD) mass spectrometry to probe
nascent products from the photocatalytic dissociation of
methanol on the TiO2 surface and density functional theory
to calculate stationary-point structures and energies along the
reaction path on the ground state. The results clarify a two-step
dissociation mechanism for methanol and water and show the
importance of the ground-state energetics in determining the
irreversibility of the dissociation.
EXPERIMENTAL DETAILS
■
The combination of laser-induced photocatalysis with a mass
spectrometer having extremely low background provides a tool to
probe photocatalytic chemistry. Figure 1 shows a schematic diagram of
the apparatus, which comprises (1) an ultrahigh vacuum system,
including a chamber for sample preparation and characterization, a
main chamber with a base pressure of 6 × 10−11 Torr where the
sample resides for the experiments, an extremely high vacuum
quadrupole mass spectrometer for TPD universal detection, and (2)
an amplified 1 kHz Ti/sapphire femtosecond laser system with a
frequency doubling setup. The sample preparation and character-
ization chamber is equipped with an Ar+ ion gun for sample cleaning
and a LEED/AES detector for LEED pattern and contamination
check. The key element of the apparatus is the extremely low
background quadrupole mass spectrometer detector with electron-
bombardment ionization. The ionization region is housed in a liquid
nitrogen cooled vacuum region that employs a 500 L s−1 turbo
molecular pump. A liquid nitrogen cooled titanium sublimation pump
is also used for this region. Under normal operating conditions, the
region with the electron-bombardment ionizer is maintained at 1.5 ×
COMPUTATIONAL METHODS
■
All calculations were performed with the Vienna ab initio simulation
package (VASP)25,26 and PAW potential.27 The wave function was
expanded by plane wave with kinetic cutoff of 400 eV and density
cutoff of 650 eV. The generalized gradient approximation (GGA) with
13367
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373

Journal of the American Chemical Society
Article
Figure 2. Models used in the calculations. Left: optimized structure for CH3OH adsorbed on a TiO2(110) surface. Right: optimized structure for
H2O adsorbed on a TiO2(110) surface.
the spin-polarized Perdew−Burke−Ernzerhof (PBE) functional28 was
used for all of the calculations. The surface was modeled with a six-
layer slab cut out of a TiO2 crystal to expose the 110 surface. All Ti5c
sites on the bottom layer were saturated by water molecules to
maintain the coordination environment in bulk. The periodically
repeated slabs were decoupled by 12 Å vacuum gaps. A Monkhorst−
Pack grid29 of 2 × 1 × 2 k points was used for the 4 × 2 surface unit
cell. Two molecules were adsorbed on the top layer, and one of them
dissociated during the simulation (Figure 2). The chemisorbed species
and the top four TiO2 layers were allowed to relax until the residual
forces were less than 0.03 eV/Å, while the remaining atoms were fixed
at their bulk truncated positions. Transition states (TSs) were located
by constrained minimization and climbing-image nudged elastic band
(CI-NEB) methods.30,31 Isolated gas-phase molecules were optimized
in a (15 × 15 × 15) unit cell with a single k-point.
The dissociation energy of methanol slightly depends on the
methanol concentration on the TiO2 surface. According to the
calculations, in the case of 1/8 ML adsorption the methanol
dissociation is slightly exoergic, while for 1/4 ML adsorption or
more the methanol dissociation is slightly endoergic. Because in the
experiment the actual absorption was significantly larger than 1/8 ML,
we used a model with two coadsorbed molecules on the surface to
better mimic the experimental condition. In the experiment, only a
small fraction of the methanol undergoes dissociation; therefore we
treat one molecule as inactive, while the other one dissociates.
As reported previously, the adsorption energies oscillate between
odd and even numbers of layers and converge slowly as the number of
layers is increased.32−34 Zhao et al.35 have studied up to 11-layer-thick
slabs, for which the energy differences with respect to the
corresponding ten-layer slab are smaller than 0.03 eV. The TiO2
slab has two sides, where one side is for modeling the surface with the
Ti5c sites. The other side is for modeling the bulk, where all Ti atoms
are six-coordinated. We use H2O to saturate the Ti atoms on one side,
and these Ti atoms thus have the same coordination number as in the
bulk. In test calculations, we found the current model to have less
dependence on slab thickness than the model which does not have
water saturation. The adsorption and dissociation energy difference
between the six-layer slab and the seven-layer slab is smaller than 0.02
eV. The layer convergences of our models were evaluated on a slab
with a 2 × 1 surface unit cell. For six- and seven-layer models, the
calculated adsorption energies are 0.71 and 0.72 eV, respectively. The
dissociation energies are +0.06 eV and +0.05 eV, respectively. The
energy differences between six- and seven-layer models are smaller
than 0.02 eV, so the six-layer model is adequate to provide reliable
results. The adsorption energy and dissociation energy of CH3OH
were calculated with Monkhorst−Pack grids of 2 × 1 × 2, 3 × 1 × 2,
and 2 × 1 × 3 k points. The results are essentially the same, which
suggest that the Monkhorst−Pack grid of 2 × 1 × 2 k points is
sufficient for the current model. The reported energies are electronic
energies without ZPE correction; therefore, they are the same for
CH3OH and CD3OH, which is used in the experiment.
EXPERIMENTAL RESULTS
■
Figure 3A shows TPD spectra collected at a mass-to-charge
ratio (m/z) of 33 (CD2OH+) after surfaces of TiO2(110) were
dosed with CD3OH and then irradiated by the laser for various
durations. The signal profile from m/z = 33 is exactly the same
as that from m/z = 35 (CD3OH+, not shown), indicating that
the m/z = 33 is simply the result of dissociative ionization of
the desorbed parent CD3OH molecule in the electron-
bombardment ionizer. The desorbed CD3OH signal, depicted
in Figure 3A, decreases monotonically as the laser irradiation
time increases, suggesting that the CD3OH molecules that are
adsorbed on the Ti5c sites of TiO2(110) are photocatalytically
dissociated. The high-temperature side of the CD3OH TPD
peak decreases significantly as the laser irradiation time
increases, resulting in a peak shift to lower temperatures. As
shown in Figure S1 of Supporting Information (SI), this peak
shift is the result of an increased concentration of H or D atoms
at BBO sites on the TiO2(110) surface.
Concomitant to the decrease of the CD3OH TPD peak, the
TPD signal for m/z = 32 (CD2O+) increases with increasing
laser irradiation times (Figure 3B). The CD2O+ signal may
come from three sources. The formaldehyde (CD2O) product
that is adsorbed molecularly on the surface desorbs with a peak
13368
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373

Journal of the American Chemical Society
Article
Figure 4. Integrated TPD signals of the desorbed parent methanol
(CD3OH) signal (from m/z = 33) and the formaldehyde (CD2O)
product as a function of laser irradiation time. The CD2O signals were
integrated from the m/z = 32 TPD spectra after the contributions
from CD3OH and CD3OD had been subtracted. The decrease of the
parent methanol signal correlates almost perfectly with the rise of the
formaldehyde signal.
Figure 3. (A) Typical TPD spectra collected at m/z = 33 (CD2OH+)
following different laser irradiation times. CD2OH+ is formed by
dissociative ionization of the desorbed parent CD3OH molecule in the
electron-bombardment ionizer. (B) Typical TPD spectra collected at
m/z = 32 (CD2O+) following different laser irradiation times. The m/z
= 32 signal has three components: the parent ion signal of
formaldehyde (CD2O), as well as the ion-fragment signals of the
parent CD3OH molecule and the photocatalyzed CD3OD product.
CD3OH binds to TiO2(110), and an H atom. Because the
adsorption of H atoms on the surface-bonded oxygen (SBO)
sites is energetically much less favorable than adsorption on the
BBO sites,37 the dissociated H atoms from CD3OH would be
expected to adsorb on the BBO sites. Secondary dissociation of
a D atom from intermediate CD3O produces CD2O on a Ti5c
site and a D atom, which also likely binds to a BBO site. This
reaction sequence thus explains the formation of CD2O as well
as CD3OD. It is also consistent with our previous conjecture
from a TD-2PPE study22 that photocatalysis of methanol
occurs mostly on the Ti5c sites and not on BBO defect sites.
The stepwise dissociation of CD3OH and the affinity of H
and D for the BBO sites are confirmed by TPD measurements
of the water product isotopologues, H2O, HOD, and D2O.
in the TPD spectrum near 270 K.36 The other two sources of
CD2O+ signal are dissociative ionization of adsorbed parent
CD3OH molecules and CD3OD reaction products. Both
CD3OH and CD3OD desorb in the same temperature range,
with a broad peak near 300 K, but the presence of both species
is confirmed by their detection at their respective parent mass-
to-charge ratios of 35 and 36 (see Figure S2, Supporting
Information). After a laser irradiation time of 30 min, the
CD3OD signal at m/z = 36 is about 15% of the CD3OH signal
at m/z = 35 (Figure S2, Supporting Information). The CD3OD
products are evidently formed through recombination reactions
following dissociation of CD3OH. In particular, the detection of
CD3OD suggests that CD3O combines with a D atom.
The large TPD signal for desorbed CD2O seen in Figure 3B
indicates that formaldehyde is an important photocatalytic
product. Figure 4 shows that the increase in the formaldehyde
signal with laser irradiation time is almost perfectly
anticorrelated with the decrease in the signal from the adsorbed
parent CD3OH, strongly suggesting that formaldehyde is the
main product of photocatalyzed dissociation of CD3OH on
TiO2(110). The decrease of the CD3OH signal with increased
irradiation time is not the result of desorption of CD3OH
during laser irradiation, as we observed no CD3OH signal
during irradiation. In fact, a tiny signal from formaldehyde was
detected during irradiation (by time-of-flight methods),
indicating that photocatalysis can lead to the production of
formaldehyde. In order to produce formaldehyde from
methanol, the CD3OH molecule must lose one hydrogen
atom and one deuterium atom from the hydroxyl group and the
fully deuterated methyl group, respectively.
TPD spectra were collected at m/z = 18 (H2O+, CD3 ), 19
+
(HOD+), and 20 (D2O+) following laser irradiation of varying
duration. Typical TPD spectra for these three m/z ratios are
shown in Figure 5. The TPD spectra of these products have
maxima near 450 K and the magnitudes of the signals clearly
depend on the laser irradiation time. The observed TPD
spectra for H2O, HOD, and D2O are very similar to the H2O
TPD peak that is known to originate from H atoms adsorbed
on the BBO sites of TiO2(110),38 suggesting that the TPD
signals of all the water isotopologues that we observe also come
from BBO sites with adsorbed H and D atoms, which are
produced by photocatalyzed dissociation of CD3OH on Ti5c
sites. The TPD signal as a function of laser irradiation time for
the different water isotopologues provides a comparison
between the rates of producing hydrogen atoms bound to
bridge-bonded oxygen atoms (BBO-H) and deuterium atoms
bound to bridge-bonded oxygen atoms (BBO-D). Using ionizer
fragmentation patterns for the water isotopologues that have
been measured in our laboratory, the OD+ contributions to the
signal at m/z = 18 that come from HOD and D2O were
subtracted. The relative TPD signals of H2O, HOD, and D2O
as a function of laser irradiation were then determined (see
Figure 6), assuming that the detection efficiencies for the three
isotopologues are the same. The results show that the TPD
signals rise monotonically as laser irradiation time increases.
The rate of increase in TPD signal intensity with laser
Thus, the observation of both CD3OD and CD2O points to a
two-step dissociation of CD3OH through a CD3O intermediate.
CD3OH first dissociates to CD3O on a Ti5c site, where the
13369
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373

Journal of the American Chemical Society
Article
Figure 7. TPD spectra of H2O on TiO2(110) with laser surface
irradiation (400 nm, 460 mW) for 0, 30, and 60 min.
30, and 60 min, suggesting that water is not photocatalytically
active on the TiO2(110) surface.
Figure 5. TPD spectra collected at m/z = 18, 19, and 20 from the
photocatalysis of CD3OH/TiO2(110) with the four different laser
irradiation times indicated.
THEORETICAL RESULTS
■
The optimized CH3OH and H2O adsorbed structures are
shown in Figure 2. The average adsorption energy for CH3OH
is calculated to be 0.75 eV. The O−H dissociation energy is
slightly endoergic by 0.03 eV, with a barrier of 0.25 eV. The
second dissociation step is C−H bond breaking to produce
CH2O and H. This step is highly endoergic (1.03 eV), and the
barrier is also much higher (1.57 eV). The barrier for the
reverse reaction is therefore 0.54 eV. The optimized transition
states and dissociated structures are shown in Figures S3 and S4
(Supporting Information), respectively. After O−H dissocia-
tion, the Ti5c site at the reaction position is elevated (by 0.57 Å)
and the Ti−O bond becomes shorter. The following C−H
dissociation step results in weakly coordinated CH2O; hence,
the elevated Ti5c returns to its original position. The optimized
Ti−OCH2 distance, 2.48 Å, is longer than the 2.26 Å for Ti−
O(H)CH3, implying that the adsorption energy of CH2O is
weaker than that for CH3OH, which is consistent with our
experimental observations.
The average adsorption energy of H2O, 0.69 eV, is slightly
smaller than that for CH3OH. The energetics of the first O−H
dissociation step are quite similar to those for CH3OH, with a
dissociation energy and a dissociation barrier of 0.03 and 0.25
eV, respectively. The second O−H dissociation step is
endoergic by 0.35 eV, with a barrier of 0.46 eV. The reverse
reaction for the second O−H dissociation has a low barrier of
only 0.11 eV. The optimized transition states and dissociated
structures for H2O dissociation on TiO2 are shown in Figures
S5 and S6 (Supporting Information), respectively. The
geometry changes for the first O−H dissociation step are
similar to those for the analogous O−H dissociation step in
CH3OH, featured by the elevation of the Ti5c at the reaction
position (by 0.63 Å) and the change of the Ti−O bond length.
However, the second dissociation step is different for H2O and
CH3OH. The second O−H dissociation step for H2O results in
a very strong TiO bond, as indicated by the computed bond
length, which is only 1.69 Å. As the Ti−O bond becomes
stronger, the elevated Ti5c atom does not go back; it elevates
even further by 0.86 Å.
Figure 6. Integrated TPD signals of the water isotopologues as a
function of laser irradiation time. The inset shows the same data where
the integrated TPD signals of HOD and D2O have been normalized to
the TPD signal of H2O with 60 min of laser irradiation time.
irradiation time for the HOD and D2O signals are similar to
each other, while the TPD signal intensity for H2O increases
initially with a significantly higher rate (see inset in Figure 6,
where the TPD signals of HOD and D2O are normalized to the
TPD signal of H2O with 60 min of laser irradiation time). The
relatively high rate of H2O formation indicates that H-atom
transfer from the hydroxyl group of CD3OH to the BBO sites
occurs faster than D-atom transfer from the D3-methyl group.
Therefore, photocatalyzed H-atom and D-atom dissociation
from CD3OH, the two steps toward the formation of
formaldehyde, do not occur simultaneously or concertedly.
We have attempted to observe evidence for photocatalysis of
H2O on TiO2(110) under the same conditions as those used
for the CD3OH experiments. Figure 7 shows the TPD spectra
of H2O on TiO2(110) with and without laser irradiation. The
H2O TPD spectra are identical for laser irradiation times of 0,
To check the reliability of PBE functional, we have calculated
the dissociation energy and barrier of CH3OH with the hybrid
13370
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373

Journal of the American Chemical Society
Article
Figure 8. Calculated energetics of the two-step dissociation of (A) CD3OH and (B) H2O on the TiO2(110) surface (see details of theoretical
calculations in the Supporting Information). The structures shown are simplified schematics. The exact calculated structures for the adsorption state,
dissociated states, and transition states of both methanol and water can be found in Figure 2 and Figures S4−S7, Supporting Information.
functional PBE0,39 using six-layer slabs made of 2 × 1 unit cells
and one CH3OH molecule adsorbed. The hybrid functional
predicts an O−H dissociation energy of 0.07 eV, with a barrier
of 0.32 eV. For the same model, the corresponding values for
GGA with the PBE functional are 0.06 and 0.20 eV,
respectively. These results suggest that GGA with the PBE
functional predict reasonable values and trends for the current
system. The error for the barrier, about 0.12 eV, does not
change the conclusions or the conceptual understanding put
forward by this paper.
the fully dissociated methanol has formaldehyde bound to a
Ti5c site and the hydroxyl and methyl hydrogen atoms having
been transferred to two different neighboring BBO rows rather
than the same row. Most relevant to photocatalytic dissociation
of methanol is the reverse barrier for the D atom to jump back
to the adsorbed formaldehyde after any excess energy has been
dissipated in the surface. This reverse barrier is quite high,
about 0.54 eV, and thus prevents the recombination from
occurring efficiently. In addition, the formaldehyde product
may desorb from the surface rather than recombine with an H
atom because of its weak adsorption on the surface (the
adsorption energy of formaldehyde on Ti5c is calculated to be
about 0.45 eV). Furthermore, as will be discussed in a future
publication, the formaldehyde product may react with methoxy
to produce methyl formate, which will desorb from the surface.
The desorption and reaction of the formaldehyde product thus
make methanol dissociation on the TiO2(110) surface
irreversible.
The experimental results suggest that water is not photo-
catalytically active on TiO2(110) at all at 400 nm, in striking
contrast to methanol. This observation is consistent with
previous observations under real photocatalysis conditions.13,14
The calculated energetics of H2O dissociation on TiO2(110)
are shown in Figure 8B. The minimum energy path for the first
O−H bond dissociation in H2O is nearly the same as that for
O−H dissociation in methanol, and also in essential agreement
with earlier calculations.40,41 Both processes are easily reversible
because of the low barriers. However, the dissociation of the
second O−H bond in H2O has a lower endoergicity and lower
Oviedo et al.40 have reported that for a stoichiometric
surface, the dissociation of methanol or water on TiO2 to form
a hydroxyl group on in-plane oxygen is much less preferred
than on bridge-bonded oxygen. Therefore we only explored the
transition states associated with the dissociation to form a
hydroxyl group on bridge-bonded oxygen.
DISCUSSION
■
Schematic energy level diagrams for stepwise dissociation of
CH3OH and H2O are shown in Figure 8. As seen in Figure 8A,
our theoretical results show that H-atom dissociation from
CD3OH to a BBO site is nearly thermoneutral with a small
barrier (0.25 eV), suggesting that H-atom transfer between
methanol and methoxy is facile. This result basically agrees with
previous theoretical calculations by Oviedo and co-workers37,40
and by Sanchez et al.41 The second step, C−D bond
dissociation from CD3O−Ti5c, is endoergic by 1.03 eV with a
barrier of 1.60 eV, likely making it a slower process. The
theoretical calculations predict that the lowest energy state for
13371
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373

Journal of the American Chemical Society
Article
barrier than those for the analogous C-D dissociation step in
methanol. Another important difference is the reverse barrier,
which is only 0.11 eV for recombination of O and H but 0.54
eV for recombination of CD2O and D. The low barrier to
recombination of O and H makes this reaction prone to occur
even after any excess energy has been dissipated, and the high
adsorption energy of O on the Ti5c site (>2 eV)42 guarantees
that the O will remain on the surface long enough to ensure
that recombination does occur. In the case of methanol, the
high barrier to recombination of CD2O and D and the weak
binding between CD2O and the Ti5c site make recombination
unlikely. The ground-state picture in Figure 8 is not meant to
imply that the observed stepwise dissociation of methanol or
water necessarily occurs on the ground state. Nevertheless, the
ground-state energetics will govern the reverse barriers after
two hydrogen atoms have been dissociated, and these reverse
barriers explain why methanol photocatalyzes easily on
TiO2(110) while photocatalytic water splitting on this surface
is inefficient.
All calculated dissociation barriers and recombination
barriers for O−H bonds are low, while for C−H bonds these
barriers are much higher. The dissociation reactions involve
electrophilic attack of the bridge-bonded oxygen atom by the
hydrogen in the O−H/C−H bond. The hydrogen in O−H
bond is more electrophilic than in C−H bond (because the
electronegativity difference for O/H is larger than that for C/
H); therefore, the barrier for O−H dissociation is lower. There
is a similar trend for the recombination of H and O or C, which
involve electrophilic attack of O/C by hydrogen. Since oxygen
bears more negative charge than carbon (because the
electronegativity of oxygen is larger), the recombination barrier
for O−H dissociation is also lower. In addition, the O−H
dissociation does not introduce much geometric rotation, while
C−H dissociation results in the rotation of the carbon center
from tetrahedral to planar. We have checked the intermediate
structures on the reaction path and found that the bond
breaking and rotation are simultaneousthere is not an
independent barrier for the rotation. This geometrical effect
also contributes to a larger barrier for C−H dissociation/
recombination than for O−H dissociation/recombination.
Almeida et al.43 have proposed that the hydration of the
surface may influence adsorption/desorption enthalpies of
other molecules. Our calculations confirm their result, but the
influence is quite small. For example, on our 4 × 2 slab, when
one water molecule is adsorbed, the adsorption energy is
calculated to be 0.71 eV, while if the slab is preadsorbed by one
water molecule, the adsorption energy of a new water molecule
is 0.68 eV. The adsorption energies also decrease with the
increasing coverage of methanol. The average adsorption
energies of methanol for 1/8, 1/4, and 1/2 ML surface
coverage are 0.78, 0.75, 0.73 eV, respectively. Therefore, the
results reported here should be unaffected by modest surface
hydration.
step makes recombination facile, thus preventing efficient water
splitting. These results suggest that efficient catalysts for water
splitting require a high barrier to recombination. In other
words, the oxygen atom must be sequestered in a way that will
inhibit recombination, just as the oxygen of methanol is
sequestered by becoming part of a stable formaldehyde
product. The results of the present study clearly show that
surface dynamics are very important in understanding photo-
catalytic chemistry on surfaces.
ASSOCIATED CONTENT
* Supporting Information
■
S
Effect of BBO-H(D) on the CD3OH TPD spectra. TPD
spectra for CD3OD versus CD3OH. Optimized transition states
and dissociated structures for stepwise dissociation of CH3OH
and H2O, from theory. This material is available free of charge
via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
fanhj@dicp.ac.cn, tminton@montana.edu, xmyang@dicp.ac.cn
■
Notes
The authors declare no competing financial interest.
§Also with School of Physics and Optoelectric Engineering,
Dalian University of Technology, Dalian, Liaoning 59717,
China.
∥Chinese Academy of Sciences Visiting Senior Scientist.
ACKNOWLEDGMENTS
■
This work was supported by the Chinese Academy of Sciences,
National Science Foundation of China, and the Chinese
Ministry of Science and Technology. H.F. acknowledges
support from the National Natural Science Foundation of
China (No. 21173212). T.K.M. is grateful for support from the
Chinese Academy of Sciences Visiting Professorships for Senior
International Scientists.
REFERENCES
■
(1) Henderson, M. A. Surf. Sci. Rep. 2011, 66, 185−297.
(2) Fujishima, A.; Honda, K. Nature 1972, 238, 37−38.
(3) Gratzel, M. Nature 2001, 414, 338−344.
(4) Wahlstrom, E.; Vestergaard, E. K.; Schaub, R.; Rønnau, A.;
̈
Vestergaard, M.; Læsgaard, E.; Stensgaard, I.; Besenbacher, F. Science
2004, 303, 511−513.
(5) Fox, M. A.; Dulay, M. T. Chem. Rev. 1993, 93, 341−357.
(6) Linsebigler, A. L.; Lu, G.; Yates, J. T., Jr. Chem. Rev. 1995, 95,
735−758.
(7) Khan, S. U. M.; Al-Shahry, M.; Ingler, W. B., Jr. Science 2002, 297,
2243−2245.
(8) Wang, R.; Hashimoto, K; Fujishima, A.; Chikuni, M.; Kojima, E.;
Kitamura, A.; Shimohigoshi, M.; Watanabe, T. Nature 1997, 388,
431−432.
(9) Kamat, P. V. Chem. Rev. 1993, 93, 267−300.
(10) Zhang, Z.; Bondarchuk, O.; White, J. M.; Kay, B. D.; Dohnalek,
Z. J. Am. Chem. Soc. 2006, 128, 4198−4199.
(11) Chen, X.; Liu, L.; Yu, P. Y.; Mao, S. S. Science 2011, 331, 746−
750.
CONCLUSION
■
In this combined experimental/theoretical study, we have
investigated the photocatalytic dissociation of methanol and
water on a TiO2(110) surface. A clear physical picture has
emerged for stepwise dissociation of two hydrogen atoms. The
high reverse barrier of the second C−H(D) dissociation step of
methanol inhibits recombination and is key for efficient
methanol dissociation. In the case of water, however, the
much lower reverse barrier of the second O−H dissociation
(12) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W.
Chem. Rev. 1995, 95, 69−96.
(13) Ollis, D. F.; Al-Ekabi, H. Photocatalytic Purification and
Treatment of Water and Air; Elsevier: Amsterdam, 1993.
(14) Sato, S.; White, J. M. Chem. Phys. Lett. 1980, 72, 83−86.
(15) Kawai, T.; Sakata, T. J. Chem. Soc., Chem. Commun. 1980, 24,
694−695.
(16) Diebold, U. Surf. Sci. Rep. 2003, 48, 53−229.
13372
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373

Journal of the American Chemical Society
Article
(17) Bates, S. P.; Gillan, M. J.; Kresse, G. J. Phys. Chem. B 1998, 102,
2017−2026.
(18) Henderson, M. A.; Otero-Tapia, S.; Castro, M. E. Faraday
Discuss. 1999, 114, 313−329.
(19) Pang, C. L.; Lindsay, R.; Thornton, G. Chem. Soc. Rev. 2008, 37,
2328−2353.
(20) Onda, K.; Li, B.; Zhao, J.; Petek, H. Surf. Sci. 2005, 593, 32−37.
(21) Li, B.; Zhao, J.; Onda, K.; Jordan, K. D.; Yang, J.; Petek, H.
Science 2006, 311, 1436−1440.
(22) Zhou, C.; Ren, Z.; Tan, S.; Ma, Z.; Mao, X.; Dai, D.; Fan, H.;
Yang, X.; LaRue, J.; Cooper, R.; Wodtke, A. M.; Wang, Z.; Li, Z.;
Wang, B.; Yang, J.; Hou, J. Chem. Sci. 2010, 1, 575−580.
(23) Shen, M; Henderson, M. A. J. Phys. Chem. Lett. 2011, 2, 2707−
2710.
(24) Zehr, R. T.; Henderson, M. A. Surf. Sci. 2008, 602, 1507−1516.
(25) Kresse, G.; Hafner, J. Phys. Rev. B 1993, 47, 558−561.
(26) Kresse, G.; Furthmuller, J. Phys. Rev. B. 1996, 54, 11169−11186.
(27) Blochl, P. E. Phys. Rev. B 1994, 50, 17953−17979.
(28) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77,
3865−3868.
(29) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188−5192.
(30) Henkelman, G.; Uberuaga, B. P.; Jonsson, H. J. Chem. Phys.
2000, 113, 9901−9904.
(31) Henkelman, G.; Jonsson, H. J. Chem. Phys. 2000, 113, 9978−
9985.
(32) Sanchez de Armas, R.; Oviedo, J.; San Miguel, M. A.; Sanz, J. F.
J. Phys. Chem. C 2007, 111, 10023−10028.
(33) Kowalski, P. M.; Meyer, B.; Marx, D. Phys. Rev. B 2009, 79,
115410.
(34) Stefanovich, E. V.; Truong, T. N. Chem. Phys. Lett. 1999, 299,
623−629.
(35) Zhao, J.; Yang, J.; Petek, H. Phys. Rev. B 2009, 80, 235416.
(36) Lu, G.; Linsebigler, A.; Yates, J. T., Jr. J. Phys. Chem. 1994, 98,
11733−11738.
(37) Sanchez de Armas, R.; Oviedo, J.; San Miguel, M. A.; Sanz, J. F.
J. Phys. Chem. C 2007, 111, 10023−10028.
(38) Yin, X.-L.; Calatayud, M.; Qiu, H.; Wang, Y.; Birkner, A.; Minot,
C.; Woll, Ch. Chem. Phys. Chem. 2008, 9, 253−256.
̈
(39) Perdew, J.; Ernzerhof, M.; Burke, K. J. Chem. Phys. 1996, 105,
9982−9985.
(40) Oviedo, J.; Sanchez de Armas, R.; San Miguel, M. A.; Sanz, J. F.
J. Phys. Chem. C 2008, 112, 17737−17740.
(41) Sanchez, V. M.; Cojulun, J. A.; Scherlis, D. A. J. Phys. Chem. C
2010, 114, 11522−11526.
(42) Tilocca, A.; Selloni, A. Chem. Phys. Chem. 2005, 6, 1911−1916.
(43) Almeida, A. R.; Moulijn., J. A.; Mul, G. J. Phys. Chem. C 2008,
112, 1552−156.
13373
dx.doi.org/10.1021/ja304049x | J. Am. Chem. Soc. 2012, 134, 13366−13373