.

Thin Solid Films 392 2001 338344

Photocatalytic TiO thin lm deposited onto glass by DC
2
magnetron sputtering
Satoshi Takeda
a,U
, Susumu Suzuki
a
, Hidefumi Odaka
a
, Hideo Hosono
b
a
Research Center, Asahi Glass Co., Ltd, 1150 Hazawa-cho, Kanagawa-ku, Yokohama 221-8755, Japan
b
Materials and Structures Laboratory, Tokyo Institute of Technology, 4259 Nagatsuda-cho, Midori-ku, Yokohama 226-8503, Japan
Abstract
A high performance photocatalytic TiO thin lm was successfully obtained by reactive DC magnetron sputtering. The lm was
2
deposited onto SiO -coated glass at a substrate temperature of 220C using a titanium metal target in O 100% atmosphere. The
2 2
lm showed good uniformity of thickness in a large area with the optical transmittance of ;80% in the visible region. The
. decomposition ability of acetaldehyde CH CHO of the lm under UV irradiation was almost the same as that of the
3
solgel-derived TiO thin lm but the sputtered lm showed a much higher mechanical durability. The characterization of the
2
lms was carried out using XRD, SEM, AFM, XPS and SIMS, and the electronic structures of the lms were calculated using a
rst-principle calculation method based on the density functional theory. It was found that the amount of incorporated
18
O into
the lm was larger for the lms with lower photocatalytic activity when the lms were annealed in
18
O rN atmosphere. This
2 2
result indicates that the amount of oxygen vacancies, which were occupied by incorporated
18
O, was larger for the lms with
lower photocatalytic activity. Furthermore, the introduction of structural defects associated with oxygen vacancies was found to
create some energy levels around the mid-gap, indicating that they could work as recombination centers of photo-induced holes
and electrons, causing the decrease in photocatalytic activity. Therefore, the decrease in the structural defects associated with
oxygen vacancies is important for improving the photocatalytic activity of the lms. 2001 Elsevier Science B.V. All rights
reserved.
Keywords: TiO ; Photocatalytic activity; Sputtering; Oxygen vacancy
2
1. Introduction
.
Titanium oxide TiO thin lms are widely used in
2
various elds such as optical and protective coatings
and optical bers because of their excellent chemical
stability, mechanical hardness and optical transmit-
tance with high refractive index. Recently, TiO thin
2
lms have become the most promising materials in
environmental cleaning such as photocatalytic purier
w x w x
1,2 and photochemical solar cells 3,4 . Many re-
searchers have focused on the application of TiO
2
photocatalyst to purication and treatment of air and
U
Corresponding author. Tel.: q81-45-374-8794; fax: q81-45-374-
8892.
. E-mail address: [email protected] S. Takeda .
water, e.g. through the photolysis of organics and toxic
w x
gases 2 .
TiO thin lms have been prepared by a variety of
2
w x
deposition techniques such as solgel processes 5 ,
w x w x
chemical vapor deposition 6,7 , evaporation 8 , various
w x
sputtering depositions 911 , and ion beam-assisted
w x
processes 12 . However, most of the photocatalytic
TiO thin lms used in the market are prepared by wet
2
process such as a solgel method. Although the lms
show excellent photocatalytic activity, the mechanical
durability is not enough for practical uses such as
architectural or automotive glasses. In addition, the
uniformity of the thickness in a large area is poor.
These points are not desirable for architectural or
automotive applications. Furthermore, a heating
process at approximately 500600C is indispensable in
the method in order to decompose metalorganic,
0040-6090r01r$ - see front matter 2001 Elsevier Science B.V. All rights reserved.
. PII: S 0 0 4 0 - 6 0 9 0 0 1 0 1 0 5 4 - 9
( ) S. Takeda et al. rThin Solid Films 392 2001 338344 339
causing a limitation of the use of non-refractory sub-
strates.
In the present study, the sputtering method was
applied to obtain a photocatalytic TiO thin lm on
2
w x
glass. Part of this work has recently been reported 13 .
Sputtering methods are widely used in industrial
products because the high quality lms; high density,
high adhesion, high hardness, etc., can be obtained at
low substrate temperature with good uniformity of the
lm thickness in a large area. Unfortunately, the pho-
tocatalytic activity of the sputtered lm used for archi-
tectural glass is much lower than that of the lms
prepared by the wet process. Weinberger and Gerber
w x
14 reported that a photocatalytic TiO lm with a
2
thickness of ;10 m was obtained by reactive RF
magnetron sputtering, and that the lm showed high
photocatalytic activity because of its porous columnar
morphology, resulting in a high surface area. However,
the optical transmittance of the lm may be low due to
the large thickness. To our knowledge, a high perfor-
mance photocatalytic TiO thin lm with both a high
2
mechanical durability and transparency in the visible
region has not been reported so far.
Here, the TiO thin lm was deposited by DC mag-
2
netron sputtering onto glass at a substrate temperature
of 220C using a titanium metal target in a 100% O
2
atmosphere. From an industrial point of view, DC
sputtering is desirable for the large area coating be-
cause it is more simple technology than RF sputtering
and it gives a higher deposition rate. The characteriza-
tion of the lm was carried out using X-ray diffraction
. .
XRD , scanning electron microscope SEM , atomic
.
force microscopy AFM and secondary ion mass spec-
.
trometry SIMS . The electronic structure of the lm
was calculating using a rst-principle calculation
method based on the density functional theory. The
photocatalytic performance of the lm was evaluated
by the amount of decomposition of acetaldehyde
.
CH CHO as a function of UV irradiation time. From
3
the results obtained, the relationship between the pho-
tocatalytic performance and the structure of the lm
was discussed.
2. Experimental details
2.1. Sample preparation
SiO coated sodalimesilica glass was used as a
2
substrate. The SiO lm was deposited onto the glass
2
by reactive RF magnetron sputtering with a thickness
of ;50 nm in order to suppress alkali migration from
the glass substrate. TiO thin lms were deposited by
2
reactive DC magnetron sputtering with a thickness of
;200 nm. The sputtering conditions are listed in Table
1. The base pressure in the coating chamber was less
than 1.3=10
y3
Pa. The target was a titanium metal
Table 1
The sputtering conditions of TiO thin lms
2
Substrate Target Sputtering Sputtering
. . temperature T gas pressure P
s
. . C Pa
Room temperature Ti O s100% 0.4
2
140 Ti O s100% 0.4
2
220 Ti O s100% 2.0
2
.
purity 99.9% of 10 cm diameter. The deposition rate
of the lm was ;4 nmrmin. The discharge during the
deposition was stable.
The TiO thin lm was also prepared by a solgel
2
w . x
method. Titanium n-butoxide Ti OC H , TBT was
4 9 4
used as a precursor. TBT was dissolved in ethanol and
acetylacetone. The SiO -coated glass was coated with
2
this solution by the spin-coating process and then
heated to 550C for 30 min in an electric furnace.
2.2. Film properties
The photocatalytic performance of the lms was
evaluated by measuring the amount of decomposed
.
acetaldehyde CH CHO by UV irradiation. UV light
3

was irradiated using a black light central wavelength

.
2
;352 nm with ;2 mWrcm . The experimental appa-
ratus is schematically illustrated in Fig. 1. The TiO
2
lm coated on a 45=100-mm
2
substrate was set in a
Pyrex vessel, and CH CHO solution was injected in the
3
vessel. The injected CH CHO was vaporized immedi-
3
ately, and the concentration of the CH CHO vapor
3
was adjusted at 750"10 ppm. Thereafter, the change
in CH CHO concentration was measured using a gas
3
.
detector GASTEC as a function of UV irradiation
time. The accuracy of the measurements was "10 ppm.
The mechanical durability of the lms was evaluated
.
by a Taber abrasion tester Teledyne Taber 503 . The
tests were carried out using CS-10F abrasion wheels
loaded with 500 g. The degree of degradation was
evaluated by the change in the haze value. The haze

value, H, which is dened by T rT =100% T , scat-

d t d
.
tered light; T , transmitted light . The haze value
t
Fig. 1. Schematic illustration of experimental apparatus for the mea-
surement of photocatalytic activity.
( ) S. Takeda et al. rThin Solid Films 392 2001 338344 340
. .
change, H which is dened by H N yH 0 ; N de-
notes the number of cycles.
The optical transmission spectra of the lms were
measured at room temperature in air using a dual
.
beam spectrometer Shimazu UV3100 . Observation of
surface morphology and the quantitative analysis of the
surface roughness of the lms were performed using
. .
SEM Hitachi S-900 and AFM Seiko SPI 3700 , re-
spectively. Crystalline phases of the lms were identi-
.
ed by the glancing-angle XRD Rigaku Rint-2000
using Cu-K radiation operated at 50 kV200 mA.
The incident angle was kept at 0.5.
The stoichiometry of the lms was rstly determined
.
by X-ray photoelectron spectroscopy XPS . However,
no difference in OrTi ratio among the lms was
observed. In order to obtain more accurate information
about the stoichiometry, SIMS analyses with oxygen

18
.
tracer O gas were performed. The SIMS is widely
used in various elds such as semiconductors to obtain
detailed information for in-depth distribution of impu-
rities or dopants in the materials because of its excel-
lent sensitivity and high-depth resolution compared
with XPS. In this analysis, the lms were heat-treated
at 500C for 1 h in
18
O rN s1r4 atmosphere, and
2 2
thereafter the amount of incorporated
18
O was mea-
.
q
sured by SIMS PHI Adept1010 . The Cs primary ion
beam was operated at 5 keV, 200 nA and rastered on
the area of 300=300 m
2
. The angle of incidence was
60 to the normal of the sample surface. The charge
neutralization was accomplished using an electron ood
gun.
The electronic structure of the lms was calculated
using a rst-principle calculation method based on the
w x
density functional theory 15,16 . The calculations were
w x
performed using VASP programs 1720 . Two types of
oxygen vacancy in anatase TiO were simulated by
2
.
2=2=1 supercell model containing 48 atoms. One
was that a single oxygen atom was removed from a Ti

atom in a perfect anatase crystal single oxygen vacancy

.
model , causing the decrease in the coordination num-
.
bers of oxygen around the titanium Ti atom in the

supercell there are three vefold-coordinated Ti

.
atoms . The other was that two oxygen atoms were

removed from a certain Ti atom double oxygen vacan-

.
cies model , resulting in the formation of two fourfold-
coordinated and four vefold-coordinated Ti atoms in
Fig. 2. Change in concentration of CH CHO as a function of UV
3
irradiated time by the lms obtained by solgel and sputtering
method.
the supercell. The geometries of those vacancies were
optimized by a conjugated gradient technique in a
minimization of the KohnSham energy functional.
w x
Ultra soft Vanderbilt-type pseudo-potentials 21 sup-
w x
plied by Kresse and Hafner 22 were used, where Ti 3s,
3p, 3d and 4s states and oxygen 2s and 2p states were
treated as valence electrons. A cut-off of 450 eV was
used for the valence electron wave functions. The total
energy was calculated using a local density approxima-
tion for the exchange and correlation energy with the
Ceperley and Alder form of the exchangecorrelation
w x
potential 23 . Density of states were obtained by
smearing eigenvalues of the point.
3. Results
3.1. Photocatalytic performance
Fig. 2 shows the change in concentration of CH CHO
3
by TiO thin lms as a function of UV irradiation time.
2
It is found that the decomposition ability of CH CHO
3
for the sputtered lm deposited at room temperature
.
T sr.t. is the same as that of the solgel lm without
s
UV irradiation, indicating that the photocatalytic activ-
ity of T sr.t. is almost zero. The decomposition ability
s
Table 2
Mechanical durability of the solgel-derived TiO thin lm and the sputtered TiO thin lm deposited at T s220C, Ps2.0 Pa
2 2 s
. . . Sample H Ns100 H Ns200 H Ns300
. . . % % %
Sputtered TiO 4.4 7.2 9.5
2
. T s220C, Ps2.0 Pa
s
Solgel-derived TiO Delaminated Delaminated Delaminated
2
( ) S. Takeda et al. rThin Solid Films 392 2001 338344 341
Fig. 3. Transmission spectra of the lms obtained by solgel and
sputtering method.
increases with increasing T . However, the ability of the
s
lm deposited at T s140C is still lower than that of
s
the solgel lm. The ability of the lm deposited at
T s220C signicantly increases, and its ability is al-
s
most the same as that of the solgel lm.
3.2. Mechanical durability
Table 2 shows the results of the Taber test for the
solgel lm and the sputtered lm of T s220C. It is
s
seen that the solgel lm is delaminated after the
rubbing of 100 cycles. On the other hand, the sputtered
lm is not delaminated by the rubbing of 300 cycles.
This result indicates that the mechanical durability of
the sputtered lms is much higher than that of the
solgel lm although the photocatalytic activity of the
lms is almost the same.
3.3. Characterization
Fig. 3 shows the transmission spectra of the solgel
lm and the sputtered lm of T s220C. The average
s
transmittance of the sputtered lm is ;80% in the
visible region. The uniformity of thickness for the sput-
tered lm in a large area was also better than that of
the solgel lm. These results indicate that the lm has
a possibility to use it for architectural or automotive
glasses.
Fig. 4 shows the SEM photographs of the solgel
lm and the sputtered lms deposited at different T .
s
The surface morphology of the sputtered lm is clearly
different with and without the substrate heating. The
grain size of the crystalline increases with increasing T .
s
This is because Ti particles sputtered from the target
effectively react with oxygen atoms on the substrate
due to a thermal effect, promoting the crystalline
growth. On the other hand, the surface morphology is
found to be different between the sputtered lm and
the solgel lm. No signicant relationship was
observed between the decomposition ability of
Fig. 4. SEM images of the lms obtained by solgel and sputtering
method.
.
CH CHO and the root-mean-square roughness r.m.s.
3
estimated from the AFM measurements for the sput-
tered and solgel lm surfaces.
Fig. 5 shows the 2 X-ray diffraction patterns of the
solgel lm and the sputtered lms. There is no peak
for the sputtered lm of T sr.t., indicating that the
s
lm is amorphous. On the other hand, several peaks
are clearly observed for the sputtered lms of T s140
s
and 220C although the photocatalytic activity is dif-
ferent between them, as shown in Fig. 3. These peaks
Fig. 5. 2 X-ray diffraction patterns for the lms obtained by
solgel methods and sputtering methods.
( ) S. Takeda et al. rThin Solid Films 392 2001 338344 342
could be assigned to the anatase structure of TiO
2
without other phases. The XRD pattern of the solgel
lm is almost the same as that of the sputtered lm
with substrate heating.
Fig. 6 shows the SIMS depth proles for the lms
heat-treated in
18
O rN atmosphere. Before the heat
2 2
treatment, the secondary ion intensity of
133
Cs
18
O
q
was less than 10
2
counts for all the lms. After the
heat treatment, the intensity of
133
Cs
18
O
q
increases
for all the lms although no signicant change is
observed for the intensity of
133
Cs
16
O
q
before and
after the heat treatment, indicating that
18
O is incor-
porated into the lms. The amount of incorporated
18
O
is approximately two orders of magnitude larger than
that before the heat treatment. It is found that the
amount of incorporated
18
O is different among the
lms, and that the amount of incorporated
18
O is larger
for the lm with lower decomposition ability of
CH CHO than that of the lm with higher ability. The
3
incorporation of
18
O was not observed for a single
crystal of rutile TiO , which is a stoichiometiric TiO ,
2 2
indicating that
16
O
18
O isotopic exchange reaction is
negligible during the heat treatment. These results
suggest that the incorporated
18
O occupies an oxygen
vacancy site, and that the amount of incorporated
18
O
Fig. 6. SIMS depth proles for the lms heat-treated at 500C for 1 h
in
18
O atmosphere.
2
may represent the amount of oxygen vacancies in the
lm. Namely, the decomposition ability of CH CHO,
3
the photocatalytic activity, of the lm is seen to de-
crease with increasing the oxygen vacancies in the lm.
. .
Fig. 7 shows the density of states DOS for a perfect
. .
model; b single oxygen vacancy model; and c double
oxygen vacancies model obtained from the rst-princi-
ple calculations. It is found that no energy levels ap-
. . Fig. 7. Density of states obtained from rst-principle calculations based on the density functional theory for a perfect model; b single oxygen
. vacancy model; and c double oxygen vacancies model.
( ) S. Takeda et al. rThin Solid Films 392 2001 338344 343
pear at the mid-gap when the single oxygen atom is
simply removed from a perfect anatase crystal, as shown
in Fig. 7b. On the other hand, some energy levels
pointed out by an arrow are clearly observed at the
mid-gap in the double oxygen vacancies model, as
.
shown in Fig. 7 c .
4. Discussion
As mentioned before, there was no relationship
between the photocatalytic performance and the sur-
.
face roughness r.m.s. of the lms. This means that the
surface roughness is not a major factor governing the
photocatalytic performance of the lms. As shown in
Fig. 6, the amount of incorporated
18
O is larger for the
lm with lower photocatalytic activity than that of the
lm with higher photocatalytic activity. This result indi-
cates that the amount of incorporated
18
O may repre-
sent the amount of oxygen vacancies in the lm, and
that the photocatalytic activity decreases with increas-
ing oxygen vacancies in the lms, as mentioned above.
w x
According to Takeuchi 24 , the generation of oxygen
vacancies of the sputtered TiO lms was related to
2
the reduction of TiO during the sputtering process.
2
In order to investigate how the oxygen vacancy af-
fects the photocatalytic activity of the lm, the elec-
tronic structure of the lm was calculated using a
rst-principle calculation method. As shown in Fig. 7b,
no energy level appears at the mid-gap when a single
oxygen atom is simply removed from a perfect anatase
crystal. On the other hand, some energy levels are
clearly observed around the mid-gap in the double
oxygen vacancies model, as shown in Fig. 7c. These
results suggest that the increase in oxygen vacancies,
which causes the locally decrease in the coordination
numbers of oxygen around the Ti atom from 6 to 45,
can create some energy levels around the mid-gaps.
It is known that TiO is a semiconductor with a band
2
gap of ;3.0 eV, and that UV light with wavelengths
shorter than ;400 nm can excite pairs of electrons
and holes. The photo-generated electrons react with
.
molecular oxygen O to produce superoxide radical
2

y
.
anions O , and the photo-generated holes react with
2
.
water to produce hydroxyl OH radicals. These reac-
tive radicals then work together to decompose organic
materials. If an energy level is present at the band-gap,
the photo-generated electrons and holes are con-
sidered to be recombinated. Consequently, the
probability of the formation of O
y
and OH radicals
2
decreases, leading to the decrease in photocatalytic
activity.
Taking this into consideration, the formation of some
energy levels around the mid-gap, as shown in Fig. 7c,
can cause the decrease in photocatalytic activity be-
cause they can work as recombination centers of
photo-generated holes and electrons. The change in
coordination structure of Ti is induced by the structural
defects associated with oxygen vacancies. Namely, there
is a possibility that the sputtered TiO thin lm has
2
many more structural defects associated with oxygen
vacancies than a perfect crystal, resulting in the de-
crease in the coordination numbers of oxygen around a
Ti atom. Therefore, it is important to decrease the
structural defects associated with oxygen vacancies in
the lm for improving the photocatalytic activity.
w x
Takeuchi 24 reported that the structural defects asso-
ciated with oxygen vacancies of the sputtered TiO lm
2
decreased with increasing the oxygen concentration in
the sputtering gas. This result suggests that controlling
the oxygen concentration during the sputtering process
is a key parameter for improving the photocatalytic
activity.
Furthermore, the decomposition ability of CH CHO
3
is higher for the crystallized lm than that of the
amorphous lm, as shown in Figs. 2 and 5. It is known
that there are some energy levels at the band-gap for
the amorphous lm because of its structural disorder.
This means that the photo-generated electrons and
holes are recombinated, as mentioned above. Conse-
quently, the photocatalytic activity of the amorphous
lm decreases compared with that of the crystallized
lm. In addition, the crystalline structure of the crystal-
lized lm is found to be the anatase form. The anatase
form is known to exhibit excellent photocatalytic activ-
w x
ity compared with the rutile form 25 . Namely, the
formation of anatase form is also important for improv-
ing the photocatalytic activity of the lm.
5. Conclusions
A photocatalytic TiO thin lm with high mechanical
2
durability and transparency in the visible region has
successfully been obtained by reactive DC magnetron
sputtering. The lm showed good uniformity of thick-
ness over a large area with the optical transmittance of
;80% in the visible region. The mechanical durability
of the lm was much higher than that of solgel-
derived TiO thin lms. The photocatalytic activity of
2
the lm was almost the same as that of the solgel
lms. The lm was characterized by various analytical
methods. The electronic structure of the lm was also
calculated using a rst-principle calculation method. It
was found that the structural defects associated with
the oxygen vacancies and the crystalline structure
played an important role for the photocatalytic perfor-
mance of the lms. Therefore, controlling these factors
is important for improving the photocatalytic activity of
the lms.
( ) S. Takeda et al. rThin Solid Films 392 2001 338344 344
References
w x 1 I. Sopyan, S. Murasawa, K. Hashimoto, A. Fujishima, in: D.E.
. Ollis, H. Al-Ekabi Eds. , Photocatalytic Purication and Treat-
ment of Water and Air, Elsevier, Amsterdam, 1993, p. 747.
w x .
st
2 D.E. Ollis, H. Al-Ekabi Eds. , Proc. 1 Int. Conf. TiO Photo-
2
catalytic Purication and Treatment of Water and Air, Else-
vier, Amsterdam, 1993, p. 747.
w x . 3 A. Fujishima, K. Honda, Nature 238 1972 37.
w x . 4 M. Graetzel, Comments Inorg. Chem. 12 1991 93.
w x 5 T. Watanabe, A. Kitamura, E. Kojima, C. Nakayama, K.
. Hashimoto, A. Fujishima, Chem. Lett. 1994 723.
w x 6 G.A. Battiston, R. Gerbasi, M. Porchia, A. Marigo, Thin Solid
. Films 239 1994 186.
w x . 7 L.M. Williams, D.W. Hess, J. Vac. Sci. Technol. A1 1983
1810.
w x 8 T. Fujii, N. Sakata, J. Takada, Y. Miura, Y. Daitoh, J. Mater.
. Res. 9 1994 1468.
w x 9 H. Tang, K. Prasad, R. Sanjines, P.E. Schmid, F. Levy, J. Appl.
. Phys. 75 1994 2042.
w x 10 S. Ben Amor, G. Baud, J.P. Besse, M. Jacqet, Thin Solid Films
. 293 1997 163.
w x 11 N. Martin, C. Rousselot, D. Rondot, F. Palmino, Thin Solid
. Films 300 1997 113.
w x . 12 M. Gilo, N. Croitoru, Thin Solid Films 283 1996 84.
w x . . 13 S. Suzuki, S. Takeda, Sputtering Plasma Process. 14 3 1999
. 33 in Japanese .
w x . 14 B.R. Weinberger, R.B. Garber, Appl. Phys. Lett. 66 1995
2409.
w x . 15 P. Hohenberg, W. Kohn, Phys. Rev. 136 1964 B864.
w x . 16 W. Kohn, L.J. Sham, Phys. Rev. 140 1965 A1133.
w x . 17 G. Kresse, J. Hafner, Phys. Rev. B 47 1993 58.
w x 18 G. Kresse, Thesis, Technische Universituat at Wien, 1993.
w x . 19 G. Kresse, J. Furthmuuller, Compt. Mat. Sci. 6 1996 15.
w x . 20 G. Kresse, J. Furthmuuller, Phys. Rev. B. 54 1996 11169.
w x . 21 D. Vanderbilt, Phys. Rev. B 41 1990 7892.
w x . 22 G. Kresse, J. Hafner, J. Phys. Condens. Mater. 6 1994 8245.
w x . 23 D.M. Ceperley, B.J. Alder, Phys. Rev. Lett. 45 1980 566.
w x . . 24 M. Takeuchi, Phys. Stat. Solid a 55 1979 653.
w x . 25 S.N. Frank, A.J. Bard, J. Am. Chem. Soc. 99 1977 303.