Anda di halaman 1dari 7

The role of surface oxygen vacancies upon WO

3
conductivity
M. Gillet
*
, C. Lemire, E. Gillet, K. Aguir
L2MP-UMR CNRS 6137, Facult ee des Sciences et Techniques de Saint J eer^ oome Case 261, Universit eedAix-Marseille,
52 Ave. Escadrille Normandie Niemen, 13397 Marseille Cedex 20, France
Abstract
Tungsten oxide thin lms have been prepared either by reactive sputtering or by thermal evaporation on SiO
2
or
Al
2
O
3
and stabilized by annealing in dry air. The morphology and the surface structure have been investigated by
reection high energy electron diraction and atomic force microscopy. The structure of the tungsten oxide is
monoclinic and the top surface is preferentially (0 0 1) oxygen terminated plane with many oxygen vacancies, the density
of which depends on the partial oxygen pressure. The resistivity of the WO
3
thin lms has been investigated as a
function of temperature in various atmospheres. The activation energy for conduction deduced from the Arrhenius
equation is found to depend on oxygen partial pressure. We interpret this behaviour with a defect band model which
supposes that surface oxygen vacancies introduce donor levels in the gap and free electrons are produced by thermal
activation. When the surface oxygen vacancy concentration increases the donor orbitals overlap and lead to the for-
mation of a band which lessens the gap resulting in a decrease of the activation energy which can be inhibited for high
vacancy concentration.
2003 Elsevier Science B.V. All rights reserved.
Keywords: Tungsten oxide; Oxygen; Surface structure, morphology, roughness, and topography
1. Introduction
Technological and scientic interest on tungsten
oxide have stimulated many works in the past few
years [15]. In particular WO
3
presents a variety of
interesting properties: this oxide is well known as
electrochromical material [68] and recently many
investigations have been performed for testing its
sensing properties [914]. Oxidation or reduction
reactions of gas species on the oxide surface cause
changes in the lm conductivity. However the
mechanisms of the involved phenomena are not
yet fully understood.
In the present work we study the surface
structure and the conductivity of thin tungsten
lms. The WO
3
lms were grown either by reactive
sputtering of W target on SiO
2
/Si substrates or by
thermal evaporation of WO
3
powder on a-Al
2
O
3
(0 0 0 1) substrates. The morphology and the sur-
face structure were investigated by reection high
energy electron diraction (RHEED) and atomic
force microscopy (AFM). The electrical conduc-
tivity versus temperature was measured in air (at-
mospheric and reduced pressure) and in vacuum
with the aim to relate the conduction mechanisms
to the surface structure of WO
3
thin lms.
*
Corresponding author. Tel.: +33-4-91-28-83-72; fax: +33-4-
91-28-87-72.
E-mail address: marcel.gillet@l2mp.u-3mrs.fr (M. Gillet).
0039-6028/03/$ - see front matter 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0039-6028(03)00477-1
Surface Science 532535 (2003) 519525
www.elsevier.com/locate/susc
2. Experimental
Tungsten oxide thin lms were prepared either
by sputtering or by thermal evaporation. The lms
obtained by sputtering were grown on SiO
2
/Si
substrates by reactive RF magnetron sputtering of
pure tungsten target by means of an argon plasma
as carrier and oxygen as reactive gas. The experi-
mental setup has been described in [15]. The lms
deposited by thermal evaporation were grown on
(0 0 0 1) a-Al
2
O
3
substrates at a substrate temper-
ature ranging from 20 to 300 C. Details on the
tungsten oxide lm growth were given elsewhere
[16,17]. The advantage to use the thermal deposi-
tion on a crystalline substrate as a-Al
2
O
3
is to
produce large and oriented tungsten oxide grains
which are convenient for surface analysis with
RHEED and AFM techniques. The AFM tech-
nique allows us to characterize the WO
3
thin lm
morphology specially the grain size which is given
by the mean grain diameter estimated and the
mean roughness estimated to 1 nm. The experi-
ments on WO
3
(annealings or electrical measure-
ments) performed in air are carried out in a glove
box where the humidity is maintained around
10%.
The experimental setup for conductivity mea-
surements has been reported elsewhere [18]. The
conductivity variations are investigated as a func-
tion of temperature in dry air with two dierent
pressures (low pressure: P 10 mbar and atmo-
spheric pressure: P 1 bar) and in oxygen (P 1
bar, purity: 99,995). The WO
3
lm undergoes cy-
cles of heating in a range 20450 C where the
temperature increases and decreases step by step of
20 C and after each step the temperature is sta-
bilized during 10 min before the resistance mea-
surement.
3. Results
3.1. Morphology and surface structure of thin WO
3
lms
The grain size of the WO
3
thin lms depends on
the method and conditions of deposition. The
sputtered lms on SiO
2
/Si substrates have a mean
grain size which depends on the partial oxygen
pressure (P
O
2
) in the sputtering plasma. The grain
size decreases signicantly with increasing P
O
2
[19].
Fig. 1a is an AFM image of a WO
3
thin lm
sputtered on SiO
2
/Si substrate maintained at
T
D
300 C during deposition with P
O
2
=P
Ar
1.
The lm appears as nanostructured with grains of
a mean size about 30 nm and separated by well
dened grain boundaries. When annealed in air at
Fig. 1. AFM image of a sputtered WO
3
thin lm: (a) as de-
posited (T
D
300 C), (b) after annealing in air (T
A
400 C,
t
A
60 mn).
520 M. Gillet et al. / Surface Science 532535 (2003) 519525
a temperature T
A
400 C during t
A
1 h. The
mean grain size increases slightly, as shown in Fig.
1b which represents the AFM image of the an-
nealed WO
3
lm with a mean grain size of 40 nm.
With further annealings the morphology does not
change and the grain size is stabilized.
Fig. 2 is an AFM image of a thermal deposited
WO
3
thin lm grown on a-Al
2
O
3
(0 0 0 1) at a
deposition temperature T
D
300 C and annealed
at 400 C in air during 1 h. This lm exhibits larger
grains (mean size 400 nm) as compared to the
sputtered lms. But each grain is clearly composed
of smaller grains with variable size and generally
elongated in three specic directions according to
the hexagonal symmetry of the Al
2
O
3
substrate
[20].
Electron diraction analysis reveals that the
WO
3
thin lms prepared either by reactive sput-
tering or thermal evaporation and stabilized by
annealing in air exhibit a monoclinic structure
(a 7:29

AA, b 7:53

AA, ba; c 9091
0
. Fig. 3 is
a RHEED pattern with an electron beam parallel
to (0 0 1) WO
3
. The reexions are elongated in the
[0 0 1] WO
3
direction showing that the grains have
[0 0 1] at top terraces which are suitable for AFM
analyses. Fig. 4 is an AFM image observed on the
(1 0 0) surface of a sputtered WO
3
lm and an-
nealed in dry air. This image exhibits parallel lines
imaging [1 1 0] atomic rows separated by 2.7

AA.
The AFM observations show that the surface has
a large number of points defects, the density of
which is variable depending on the history of the
sample. By using Electron Energy loss spectro-
scopy analyses with a primary energy of 100 eV, the
formation of oxygen vacancies on the surface was
evidenced after heating treatments under vacuum
Fig. 2. AFM image of a thermal deposited WO
3
thin lm
grown on a (0 0 0 1) a-Al
2
O
3
substrate (T
D
300 C) and an-
nealed in air (T
A
400 C, t
A
60 mn).
Fig. 3. RHEED pattern from the thermal deposited WO
3
lm
of Fig. 2.
Fig. 4. AFM image of a (1 0 0) WO
3
surface.
M. Gillet et al. / Surface Science 532535 (2003) 519525 521
[21]. It is possible to observe ordering of these
vacancies if the sample is annealed in vacuum
(T
A
400 C, t
A
30 mn, P 10
8
mbar) and
analysed immediately after removed from vacuum
in air. In this case we have evidenced on some
parts of the sample surface a c(2 2) structure.
Such structures have been also observed by scan-
ning tunnelling microscopy which has revealed a
wide variety of surface reconstructions [2225].
These observations indicate that the WO
3
thin
lms are formed of grains whose top surface is
predominately terminated by an oxygen plane with
variable concentration of oxygen vacancies which
can reach 50% corresponding the c(2 2) struc-
tures. This surface structure appears to be rela-
tively stable because it survives to air transfer.
3.2. Electrical properties
Fig. 5 shows typical resistivity q trends versus
reciprocal temperature (in a Logq f 1000=T
representation).
First the WO
3
lm is heated under a low air
pressure (P 10 mbar). During the rst heating
cycles the resistivity varies in larges values and
progressively stabilizes. The curve 1 represents the
resistivity variation after ve heating cycles. These
variations follow an Arrhenius law. Then the
electrical characterization chamber is lled with
dry air up to the atmospheric pressure and the
resistivity measured as previously. The resistivity
follows the linear curve 2 the slope of which is
larger than the slope of the resistivity curve 1 ob-
tained in a low pressure.
In the next experiment, the electrical charac-
terization chamber is evacuated and lled with
oxygen up to a pressure P
O
2
1 bar. The results of
resistivity measurements in oxygen are represented
by the curves 3 and 4 for increasing and decreasing
temperature respectively. The curve 3 presents a
linear Arrhenius part for low temperature (up to
approximatively 200 C), with the same slope than
the curve 1 and for T > 200 C the resistivity in-
creases. The second part of the cycle for decreasing
temperature (curve 4) can be approximated with a
linear Arrhenius with a characteristic slope value
larger than the slope of the resistivity curve ob-
tained in air and in vacuum.
Fig. 5. Graphs of the resistivity versus reciprocal temperature: Logq f 1000=T obtained under dierent oxygen partial pressures:
curve 1: in air, P 10 mbar; curve 2: in air, P 1 bar; curves 3 and 4: in oxygen, P 1 bar.
522 M. Gillet et al. / Surface Science 532535 (2003) 519525
4. Interpretation of the results
The resistivity of WO
3
thin lms against the
reciprocal temperature are represented by Arrhe-
nius curves, the slopes of which depend on
the sample environment conditions. The variation
of conductivity r with temperature is frequently
interpreted using the Arrhenius equation r
r
0
expE
r
=kT, where r 1=q (q resistivity)
and E
r
the activation energy for electrical
conduction and k the Boltzmann constant. The
activation energy deduced from this equation
corresponds to the slopes of the Arrhenius curves.
Table 1 gives the values of E
r
for dierent oxygen
partial pressures. The data are in agreement with
the already published results: the activation energy
for conduction decreases with the oxygen partial
pressure as found in several studies [2630].
We suppose that the conductivity of tungsten
oxide lms is governed by the non-stoichiometry
of WO
3
considered as an n-type semiconductor
and we assume that the non stoichiometry origi-
nates only from oxygen vacancies. The creation of
oxygen vacancies in WO
3
may be expressed by a
chemical equilibrium
O
x
o
V
x
o

1
2
O
2

g
where O
x
o
represents the neutral oxygen atom in an
oxide site and V
x
o
represents the neutral oxygen
vacancy with two trapped electrons which gives an
donor level in the gap. With increasing tempera-
ture the donors are successively ionized with an
activation energy E
a
generating free electrons e in
the conduction band. This process is expressed by
V
x
o
V
2
o
2e
where V
2
o
represents an oxygen vacancy doubly
ionized and the total equilibrium for carrier gen-
eration is
O
x
o
V
2
o
2e
1
2
O
2

g
Assuming that the conductivity r is proportional
to the concentration of electrons [e] with V
2
o

e=2 we obtain the relationship r / P
1=6
O
2
e
E
a
=kT
.
So the conductivity depends both on the oxygen
partial pressure and on the thermal activation
energy E
a
for carrier generation. We assume for
simplicity that E
a
E
r
supposing that the con-
ductivity is mainly due to the carrier generation
originated from the oxygen vacancy formation at
the WO
3
surface. In fact there are many other
parameters to take into account, in particular the
grain size with the barrier at the grain boundaries
should play an important role in the conduction
mechanism. However we have noticed that WO
3
lms with dierent grain sizes exhibit similar
conductivity behaviour and we suppose that the
most important phenomenom in our experimental
conditions is the oxygen vacancy formation.
The most important result is that the activation
energy tends to decrease with the partial oxygen
pressure and gradually reaches a very low value in
vacuum. Such a behaviour for semiconducting
material have been already found with doped ox-
ides [31,32], it was explained considering that di-
lute defects give sharp donor levels and to generate
extra carriers from these donors needs an activa-
tion energy E
a
[33]. When the defect levels con-
centration increases, the donor orbitals overlap
and lead to the formation of band which lessens
the gap required for carrier ionization. For high
defect concentration the defect band broadens
suciently so that the gap disappears. Such a
defect band model can be applied to the con-
ductivity results that we obtain on WO
3
where the
defects are surface oxygen vacancies. When the
oxygen pressure decreases the density of oxygen
vacancies, increases up to a c(2 2) structure for
some parts of the sample. So according to the
model of defect band the activation energy for
electron generation decreases as the density of
vacancies increases with decreasing oxygen partial
pressure and drop to zero when the WO
3
sample is
annealed in vacuum.
This c(2 2) surface structure seems to be rel-
atively stable. This observation explains the con-
ductivity behaviour in oxygen (curve 3Fig. 5)
observed with the WO
3
sample previously heated
in vacuum and supposed to have some parts with a
Table 1
Activation energies deduced from conductivity measurements
under dierent oxygen pressures
Conditions Air (10 mbar) Air (1 bar) O
2
(1 bar)
E
r
(eV) 0.19 0.28 0.42
M. Gillet et al. / Surface Science 532535 (2003) 519525 523
c(2 2) surface structure. When the temperature
increases up to 200 C the conductivity curve ex-
hibits the same slope that the conductivity curve 1
obtained in low oxygen pressure indicating that
the sample keep the same surface structure char-
acterized by a high density of oxygen vacancies.
For a temperature higher than 200 C this surface
structure is destroyed by oxygen adsorption which
reduced the oxygen vacancy concentration and
decreases the conductivity. The new surface gives
the resistivity curve 4 with an activation energy
(E
a
0:4 eV) characteristic of the vacancy con-
centration resulting of the new oxygen pressure.
We have observed that during the rst heatings
the resistivity value undergoes large variations
and it is tempting to attribute this eect either to
structural modications [17] or to water interac-
tion with the WO
3
surface. It has been shown [34]
that H
2
O molecularly adsorbed on an oxidized
W(1 0 0) surface desorbs by heating at room tem-
perature and a fraction of H
2
O molecules disso-
ciates resulting in OH

and O

species which
remain stable on the surface at a temperature 150
C, so it is likely that, in our experiments, inter-
action with H
2
O is low but it is possible that for
low temperature annealings (T < 200 C) the sur-
face composition and the density of oxygen va-
cancies be aected by the water vapour when
experiments are performed in air.
5. Conclusion
We have interpreted the results on resistivity
measurements and the variations of the activation
energy for conduction by the formation of surface
oxygen vacancies, the concentration of which is
governed by the partial oxygen pressure. The for-
mation of vacancies generate defect levels in the
gap, near the valence band and with increasing
temperature, these defects levels are ionized and
produce free electrons which participate to the
conduction. When the vacancy concentration is
low the defects give sharp donor levels and need
activation energy E
a
to be ionized. If defect con-
centration increases, the donor orbitals overlap
and lead to the formation of defect band. This
defect band model allows to explain the de-
crease of the activation energy E
a
and in some
conditions the disappearance of the gap.
References
[1] J.S.E.M. Svensson, C.G. Granqvist, Sol. Energ. Mater. 11
(1984) 29.
[2] C.G. Granqvist, Sol. Energ. Mater. 60 (2000) 201.
[3] J.L. Solis, S. Saukko, L. Kish, C.G. Granqvist, V. Lantto,
Thin Solid Films 391 (2001) 255.
[4] P.V. Ashirt, Thin Solid Films 385 (2001) 81.
[5] A. Antonia, T. Polichetti, M.L. Addonizio, S. Aprea, C.
Miraniri, A. Rubino, Thin Solid Films 354 (1999) 73.
[6] D. Davazoglou, A. Donnadieu, Thin Solid Films 147
(1987) 131.
[7] O. Bohnke, C. Bohnke, A. Donnadieu, D. Davazoglou, J.
Appl. Electrochem. 18 (1998) 447.
[8] T. Maruyama, T. Kanagawa, J. Electrochem. Soc. 141
(1994) 1021.
[9] C. Cantalini, H.T. Sun, M. Faccio, M. Pelino, S. Santucci,
L. Lozzi, M. Passacantando, Sensor. Actuator. B 31 (1996)
81.
[10] M. Akiyama, Z. Zhang, J. Tamaki, N. Miura, N. Yam-
azoe, T. Harada, Sensor. Actuator. B 14 (1993) 619.
[11] D. Manno, A. Serra, M. Di Giulio, G. Micocci, A. Tepore,
Thin Solid Films 324 (1998) 4451.
[12] M. Penza, M.A. Tagliente, L. Mirenghi, C. Gerardi, C.
Martucci, G. Cassano, Sensor. Actuator. B 50 (1998) 918.
[13] A. Agrawal, H. Habibi, Thin Solid Films 169 (1989) 257.
[14] H. Meixner, J. Gerblinger, U. Lampe, M. Fleisher, Sensor.
Actuator. B 23 (1995) 119.
[15] D.B.B. Lollman, K. Aguir, B. Roumiguieeres, H. Carchano,
Diam. Relat. Mater. 6 (1997) 1568.
[16] M. Gillet, J.C. Bruna, Surf. Rev. Lett. 5 (1) (1998) 325.
[17] A. Al Mohammad, M. Gillet, Thin Solid Films 408 (2002)
302.
[18] K. Aguir, C. Lemire, D.B.B. Lollman, Sensor. Actuator. B
84 (2002) 1.
[19] C. Lemire, D.B.B. Lollman, A. Al Mohammad, E. Gillet,
K. Aguir, Sensor. Actuator. B 84 (2002) 43.
[20] M. Gillet, A. Al Mohammad, C. Lemire, Thin Solid Films
410 (2002) 194.
[21] C. Lemire, Elaboration and caracteerisation de lms minces
de WO3 en vue de leur application comme capteur de gaz,
Theese UniversiteeAix-Marseille, 2001.
[22] F.H. Jones, K. Rawlings, J.S. Foord, P.A. Cox, R.G.
Edgell, J.B. Pethica, B.M.R. Wanklyn, Phys. Rev. B 52
(1995) R14392.
[23] F.H. Jones, K. Rawlings, J.S. Foord, P.A. Cox, R.G.
Edgell, J.B. Pethica, B.M.R. Wanklyn, S.C. Parker, P.M.
Olivier, Surf. Sci. 359 (1996) 107.
[24] S.C. Parker, P.M. Olivier, R.G. Edgell, F.H. Jones, J.
Chem. Soc. Faraday Trans. 92 (1996) 2049.
[25] F.H. Jones, R.A. Dixon, A. Brown, Surf. Sci. 369 (1966)
343350.
524 M. Gillet et al. / Surface Science 532535 (2003) 519525
[26] M.G. Hutchins, N.A. Kamel, N. Elkadry, A.A. Ramadam,
K. Abdel-Hady, Phys. Stat. Sol. A 175 (1999) 991.
[27] H. Kaneko, K. Miyake, Y. Teramoto, J. Appl. Phys. 53 (4)
(1982) 3070.
[28] J.P. Bonnet, J.F. Marucco, M. Onillon, P. Hagenmuller, J.
Solid State Chem. 40 (1981) 270.
[29] C. Scott, Moulzolf, Sun-an Ding, R.J. Lad, Sensor.
Actuator. B 77 (2001) 375.
[30] J. Novotny, S. Sloma, in: J. Novotny, L.C. Dufour (Eds.),
Surface and Near Surface Chemistry of Oxide Materials,
Elsevier Science Publishers, 1988.
[31] L.R. Friedman, D.P. Tunstall, The Metal Nonmetal
Transitions in Disordered Systems, Scottish Universities
Summer School in Physic Publications, Edinburgh, 1974.
[32] A. Hamnett, J.B. Goodenough, in: O. Madelung (Ed.),
Binary Transition Metal Oxides, Landolt-Bornstein (New
Series III), vol. 17g, Springer, Berlin, 1984.
[33] P.A. Cox, Transition Metal Oxides, International Series of
Monographs on Chemistry, Clarendon Press, Oxford,
1995.
[34] M. Akbulut, N.J. Sack, T.E. Madey, Surf. Sci. 351 (1966)
209.
M. Gillet et al. / Surface Science 532535 (2003) 519525 525

Anda mungkin juga menyukai