Effect of substrate structures on epitaxial growth and electrical properties of WO

3
thin lms deposited on 1012 and (0001) a-Al
2
O
3
surfaces
Ahmad Al Mohammad
*
Nano-Materials Labs, Physics Department, Atomic Energy Commission of Syria, 17th Nissan Street, P.O. Box 6091, Damascus, Syrian Arab Republic
a r t i c l e i n f o
Article history:
Received 3 December 2008
Received in revised form
7 April 2009
Accepted 8 April 2009
Keywords:
WO
3
thin lm
RHEED
Gas sensor
Epitaxial growth
Electrical conductivity
a b s t r a c t
The effect of surface structures of annealed 1012 and (0001) a-Al
2
O
3
substrates on epitaxial growth and
electrical properties of electron beam deposited WO
3
thin lms has been investigated. (0001) and 1012
a-Al
2
O
3
surfaces were used in (1 1) stoichiometric and reconstructed forms. The structure and the
morphology of WO
3
lms were determined by transmission electron microscopy (TEM), selected area
electron diffraction (SAED) and reection high energy electron diffraction (RHEED). Generally the lms
consist of micro-grains of monoclinic WO
3
and the (010) planes are parallel to the substrate surfaces.
Certain epitaxial relationships between WO
3
lms and the substrate surfaces were found. These
phenomena are interpreted by nucleation and growth theories in relation to a variation of the density of
surface oxygen vacancies of the a-Al
2
O
3
substrates. The electrical conductivity of the WO
3
lms was
measured as a function of annealed temperatures of the substrates. The activation energy for conduction
deduced from the Arrhenius equation is found to be dependent on the grain size and the morphology of
WO
3
lms.
2009 Elsevier Ltd. All rights reserved.
1. Introduction
Tungsten oxide (WO
3
) lms have been widely studied and used
in many applications such as electro-optical, electro-chromic,
ferroelectric and gas sensors [13]. The growth behavior of WO
3
lms on the a-Al
2
O
3
substrates was studied using several tech-
niques such as molecular beam epitaxy, reactive radio-frequency
(RF) magnetron sputtering and thermal evaporation [36].
Knowing the growth behavior of WO
3
is helpful for the fabrication
of high quality WO
3
thin lms [7]. In our previous reports, we
focused on the effect of surface structure of substrates on control-
ling the growth parameters of the WO
3
lms [5,8]. We found that
the surface structure of a-Al
2
O
3
substrates determined the prop-
erties of WO
3
lms such as the crystallographic structures, the
electrical conductivities, gas sensing and optical properties [814].
It has been shown [1416] that the increasing of temperature
during thermal treatments of a-Al
2
O
3
substrates under vacuum
reduced their surface structures. It led to the formation of a thin
aluminum-rich layer giving the excitation of aluminum surface
plasmons on a-Al
2
O
3
surfaces. The surface structures are typically
deduced from lowenergy electron diffraction (LEED) and reection
high energy electron diffraction (RHEED) experiments. The
reconstruction of clean a-Al
2
O
3
surfaces heated at high tempera-
ture has been reported including a O31 O31R9

surface struc-
ture of the (0001) substrates. (2 1) and (2 2) surface structures
have been observed by LEED on the 1012 surface. All the obser-
vations were performed after annealing in vacuum [14,15].
In this work, we investigated the morphology and the epitaxial
growth of WO
3
lms on the 1012 and (0001) a-Al
2
O
3
surfaces by
means of transmission electron microscope (TEM), selected area
electron diffraction (SAED) and RHEED. The electrical conductivity
of WO
3
lms prepared on different surface structures of a-Al
2
O
3
substrates was measured to relate the conductivity mechanisms to
the structure and morphology of WO
3
lms.
2. Experimental
The substrates consisted of thin slides of a-Al
2
O
3
, cut from an
alumina single crystal. The slides were mechanically and chemi-
cally polished so that the (0001) or 1012 planes were parallel to
the surface. Prior to the evaporation, the a-Al
2
O
3
substrates were
heated at either 1100 K for stoichiometric substrates or 2000 K for
reconstructed ones. All substrates were heated for 3 h under
a vacuum of about 10
3
Pa. Their surface structures were examined
by RHEED (Elettrorava, Italy), and atomic force microscopy, AFM
(Autoprobe CP, Park Scientic Instruments, USA). The tungsten
oxide thin lms were electron beam deposited (EBD) using
a commercial system (Elettrorava, Italy) simultaneously on all the
* Tel.: 963 11 2132580; fax: 963 11 6112289.
E-mail addresses: scientic@aec.org.sy, aalmohammad@aec.org.sy
Contents lists available at ScienceDirect
Vacuum
j ournal homepage: www. el sevi er. com/ l ocat e/ vacuum
0042-207X/$ see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.vacuum.2009.04.038
Vacuum 83 (2009) 13261332
substrates which were placed near one to other on the same
heating support and under a residual O
2
pressure about 5 10
1
Pa.
The WO
3
powder (Aldrich, w20 m, 99% purity) was compressed in
discs of 5 mm and annealed in an oxygen atmosphere at 1000 K for
1 h before using in the crucible of electron gun. During the WO
3
deposition, the a-Al
2
O
3
substrates were maintained at a deposition
temperature T
d
575 K. The vapor owcorresponds to a deposition
rate of approximately 2 nm/min and the average thickness depos-
ited is approximately 40 nm. The structure and morphology of thin
lms were studied by RHEED. Some lms were then detached from
their substrates in hydrouoric acid to be examined ex-situ by TEM
(Siemens102, A.V. is 125 KV). The conductivity is calculated from
the experimentally determined resistance R, which is monitored
during the three heating and cooling cycles. The heating cycles are
carried out in a residual pressure of oxygen about 10 Pa for
temperature at 623 K.
3. Results
The (0001) and 1012 surfaces of a-Al
2
O
3
have been chosen for
their completely different crystallographic properties, which can be
seen in Fig. 1. The (0001) surface is a basal plane of a hexagonal
alumina lattice due to the -AlOAlAlO- crystal structure. It can
be formed by close-packed hexagonal O planes, if the crystal is
ending by oxygen layer. If the upper plane is the aluminumone, the
surface is formed by Al atoms lying on trigonal oxygen sites. The
1012 surface is the more stable crystallographic plane and
consists of two atomic layers. The rst is centered rectangular
structure cutting the Al and O crystal layers. The second is formed
by zigzag lines parallel of only oxygen atoms. Before the deposition
of WO
3
lms the a-Al
2
O
3
surfaces were examined by RHEED and
AFM methods, Fig. 2. The azimuth directions of incidence RHEED
electron beam on surfaces were registered. The SAED and the
RHEED patterns of WO
3
deposited lms on all substrates exhibit
a monoclinic crystallographic structure (P21/n space group of WO
3
)
with lattice parameters: a 7.311(4) , b 7.542(6) ,
c 7.691(6) , b(a, c) 90

88. Fig. 3 is an example of the TEM

micrographs and the SAED patterns obtained from deposited WO
3
thin lms. These micrographs show regular shaped grains on
substrates while the mean grain size varies from substrate to
another. In all cases, the SAED patterns conrm that the lms have
a monoclinic structure with (010) planes, which are parallel to the
substrate surfaces. We have achieved a more accurate estimation of
the lm structure and epitaxial relationships between WO
3
lms
and a-Al
2
O
3
substrates by RHEED investigations. In general, the
deposition of the WO
3
lms results in electron diffraction patterns
with well dened elongated reections perpendicular to the
shadow line. The reection lines are of 00l, 201 and 101 types
due to large monocrystalline domains with the (010) plane parallel
to surface. These patterns are interpreted from the monoclinic
structure of WO
3
. Fig. 4 represents the principal RHEED patterns of
the WO
3
lms on the different a-Al
2
O
3
substrates and the corre-
sponded theoretical interpretation of these patterns. We deduce
that the WO
3
lms on all substrates consist of one or two domains
whose surfaces are parallel to the (010) plane. Hence the WO
3
lms
are formed of monocrystalline domains in epitaxy on the a-Al
2
O
3
supports according to the certain relationships. The main results of
the investigations of the TEM micrographs and the RHEED and the
SAED patterns are reported in Table 1.
Fig. 5 displays the conductivity s of the WO
3
lms deposited on
stoichiometric and reconstructed a-Al
2
O
3
substrates as a function
of the reciprocal temperature (in a Log s versus 10
3
/T plot). We have
observed for these conditions that during the heating or cooling
cycles, the values of conductivity have very small variations which
are attributed to experimental errors. Generally the observed
conductivity of the WO
3
lms on different a-Al
2
O
3
substrates
suggests polycrystalline n-type semiconductor behavior.
4. Discussion
The results concern the growth of WO
3
lms on (0001) and
1012 a-Al
2
O
3
surfaces carried out under the same deposition
conditions exhibit different crystallographic structure resulting
from thermal annealing of the a-Al
2
O
3
substrates. In case of 1012
surfaces, the annealing produces reconstructed (1 2) surface with
large number of oxygen vacancies, which are surface defects.
Generally these oxygen vacancies can be considered as nucleation
sites. This fact supports the hypothesis that the difference in
nucleation density is due to defects created by annealing. While the
annealing of (0001) surface gives one or more (111) aluminum
layers on the surface, which is responsible for changes in growth
mode of WO
3
lms. The crystallographic and morphologic proper-
ties of WO
3
lms, which are grown on reconstructed (0001) surface,
are resulting from Frank-Van der Merwe (layer-by-layer) growth
mode. The WO
3
lms, which grown on other kinds of substrates, are
resulting from VolmerWeber (islands or 3D growth) mode.
It is possible by the RHEED and the SAEDmethods to explore the
various orientations of the WO
3
lm by rotating the sample in its
plane or tilting the substrates holders which change the azimuth
directions of the electron beam. These changes could provide an
important information about the reciprocal lattice of the WO
3
lms.
The most important crystallographic properties are crystal
symmetry and in-plane lattice constant. Although the two-
dimensional (2D) symmetry of the substrate surface is the most
important, the three-dimensional (3D) symmetry can also be
inuential in determining the structure of the interface layer,
especially for oxide-on-oxide epitaxial growth. It has been found
that the minimization of electrostatic repulsion within a few
Fig. 1. (a) The (0001) a-Al
2
O
3
plane with the crystallographic directions and the distance between the surface aluminum atoms, (b) the 1012 a-Al
2
O
3
plane with the crystal-
lographic directions and the distances between the surface oxygen atoms.
A. Al Mohammad / Vacuum 83 (2009) 13261332 1327
interlayer spacings of the interface is a powerful driving force in
determining how perovskites nucleate on oxide substrates [17].
Lattice match refers to the quantitative comparison of in-plane
lattice parameters of the substrate and lm. The lattice mismatch
can be dened as:
f Da=a

a
film
a
sub

=a
sub
(1)
where both lattice parameters are in the growth plane. This
quantity should be as small as possible in order to reduce strain in
the lm. If a
lm
< (>)a
sub
the lm will be in tension (compression)
prior to relaxation. Strain energy accumulates rapidly with lm
thickness due to the inherent stiffness of most oxides, resulting in
mist dislocation generation, lm buckling, morphological trans-
formation from (2D) layer-by-layer to (3D) island growth, or coin-
cidence lattice formation. Which of these mechanisms is operative
depends on both the kinetics of lm growth and the relative
energetics of the different processes. Generally, the values of the
mismatches remained nearly constant giving short interface
deformations [17].
Fig. 2. AFM images of a-Al
2
O
3
surface morphologies and corresponding RHEED patterns of: 1012 stoichiometric (a, b), 1012 reconstructed (c, d), (0001) stoichiometric (e, f),
(0001) reconstructed (g, h).
A. Al Mohammad / Vacuum 83 (2009) 13261332 1328
Fig. 3. TEM images and their SAED patterns of WO
3
thin lms deposited on a-Al
2
O
3
substrates: (a) and (c) are WO
3
(010)//(1 1) of 1012 a-Al
2
O
3
and corresponding SAED patterns
with twins between grains. (b) and (d) WO
3
(010)//(1 2) of 1012 a-Al
2
O
3
and corresponding SAED patterns: selected area contains parts of three grains. The angles between
epitaxial directions of 90

or 45

. (e) and (g) WO

3
(010)//(1 1) of (0001) a-Al
2
O
3
and SAED patterns. Selected area contains parts of two grains. The angles between epitaxial
directions are 120

or 60

. (f) and (h) WO

3
(010)//(111)Al//(0001) a-Al
2
O
3
and corresponding SAED patterns.
A. Al Mohammad / Vacuum 83 (2009) 13261332 1329
We consider that these changes in properties of WO
3
lms are
mainly related to the changes in the WO
3
a-Al
2
O
3
bonding.
Different kinds of bonding can exist between an oxide and another
resulting either in chemical bonds or in Van Der Waals forces. The
kinds of these forces are according to the reactivity of the deposited
oxide and substrates. Fig. 6 illustrates the structural models for the
Fig. 4. a, c and e main RHEED patterns of WO
3
thin lms deposited on different a-Al
2
O
3
substrates obtained with rotation the substrates in their planes. b, d and f theoretical
patterns which interpret a, c and e RHEED patterns, with WO
3
zone axes.
Table 1
Main epitaxial relationships between WO
3
thin lms and a-Al
2
O
3
substrates
deduced from TEM, SAED and RHEED patterns.
Substrate
structure
Mean
dimensions
[nm]
Form of WO
3
grains
with diffracting
domains (DD)
Epitaxial relationships
(1 1) of
1012
a-Al
2
O
3
30 350 Needles with only one
DD
(010)WO
3
//1012 a-Al
2
O
3
and
[100]WO
3
//2131 a-Al
2
O
3
(1 2) of
1012
a-Al
2
O
3
28 45 Rectangular with DD1 (010)WO
3
//1012 a-Al
2
O
3
and
[100]WO
3
//2131 a-Al
2
O
3
Rectangular with DD2 Or (010)WO
3
//1012 a-Al
2
O
3
and
[100]WO
3
//2131 a-Al
2
O
3
(1 1) of
(0001)
a-Al
2
O
3
150 Small hexagonal with
DD1
(010)WO
3
//(0001)a-Al
2
O
3
and
[100]WO
3
//2131 a-Al
2
O
3
Small hexagonal with
DD2
Or (010)WO
3
//(0001) a-Al
2
O
3
and
[100]WO
3
//1012 a-Al
2
O
3
(111)Al//
(0001)
a-Al
2
O
3
600 Hexagonal with only
one DD
(010)WO
3
//(111)Al//(0001) a-
Al
2
O
3
and [100]WO
3
//1012 Al//
2131 a-Al
2
O
3
1.5 2.0 2.5 3.0 3.5
-5.5
-5.0
-4.5
-4.0 WO
3
/(-1012)-Al
2
O
3
reconstructed
WO
3
/(0001)-Al
2
O
3
reconstructed
WO
3
/(-1012)-Al
2
O
3
stoichiometric
WO
3
/(0001)-Al
2
O
3
stoichiometric
L
o
g

10
3
/T [K
-1
]
Fig. 5. Mean values of conductivity s of the stable WO
3
lms as a function of the
reciprocal temperature (in a Log s versus 10
3
/T plot) during heating and cooling cycles.
A. Al Mohammad / Vacuum 83 (2009) 13261332 1330
(010) WO
3
lms prepared on the stoichiometric and reconstructed
a-Al
2
O
3
substrates.
The activation energy of conduction, E
A
, frequently was inter-
preted using the Arrhenius equation between 300 and 900 K, see
refs. [1820]. It is obtained from the following formula:
dLogs=d1=T E
A
=k (2)
where k is Boltzmanns constant, T is anabsolute temperature ands is
the variation of conductivity. The activation energy E
A
of the WO
3
lms corresponds obviously to n-type semiconductor like behavior.
The overall results of conductivity properties of the four type WO
3
lms are grouped in Table 2. The observed conductivity of the WO
3
lms on different a-Al
2
O
3
substrates have activation energies
between 0.29 and 0.41 eV. Exclusively the conductivity of WO
3
lms
on reconstructed (0001) a-Al
2
O
3
surface becomes the largest value
andactivationenergy becomes alsothe largest value, E
A
0.49(3) eV.
The conductivity curves indicate that the samples keep the same
surface structure characterized by a high density of oxygen vacancies
[5,1719]. The surface structure destroys the bonds between the
oxygen and the tungsten atoms at grain boundary limits, which
increases the oxygen vacancies concentration and increases the
conductivity [18,19] at WO
3
on (0001) reconstructed surfaces. We
suppose that the non-stoichiometryat the surface of WO
3
considered
as n-type semiconductor originates only from oxygen vacancies of
WO
6
octahedra. This is discussed more fully in refs. [5,1719].
Finally, we can deduce the grain sizes, morphology and epitaxial
relationships of the WO
3
lms which are strongly dependant on the
surface structure of a-Al
2
O
3
substrates. TEM, SAED and RHEED
analyses show the structure to be very stable with large grains
which exhibit ripples in many areas due to stresses generated
during the growth processes. The conductivity behavior of the WO
3
thin lms with respect to annealing temperature and its interpre-
tation as being caused by non-stoichiometry is believed to originate
from surface oxygen vacancies at grain boundary edges and is
compatible with many similar reports in the literature [8,12,1820]
typically concerned with the use of WO
3
thin lms as a sensing
material for the detection of atmosphere pollutants [1822].
5. Conclusion
WO
3
thin lms were deposited simultaneously under identical
conditions on stoichiometric and reconstructed 1012 and (0001)
a-Al
2
O
3
substrates, which exhibited different surface structures
after annealing. The surfaces annealed at 1100 K are stoichiometric
and have (1 1) structure for both 1012 and (0001) surfaces. The
surfaces annealed at 2000 K are oxygen vacancies and exhibit
a (1 2) structure for 1012 surfaces and one or more (111)
aluminum layers for (0001) ones. TEM, SAED and RHEED observa-
tions showthat the grain size, grainformand epitaxial relationships
of WO
3
lms are different. On stoichiometric substrates the surface
structure plays an important role in nucleation and growth of WO
3
lms. Whereas oxygen vacancies on the 1012 reconstructed
Table 2
Activation energy of WO
3
thin lms deposited on various a-Al
2
O
3
substrates.
Substrate structure Mean dimensions of
WO
3
grains [nm]
Activation
energy (eV)
(1 1) of 1012
a-Al
2
O
3
30 350 0.29 (2)
(1 2) of 1012
a-Al
2
O
3
28 45 0.34 (2)
(1 1) of (0001)
a-Al
2
O
3
150 0.41 (2)
(111)Al//(0001)
a-Al
2
O
3
600 0.49 (3)
Fig. 6. The models of (010) WO
3
epitaxy on a-Al
2
O
3
substrate surfaces: (a) (010)WO
3
//
(0001) a-Al
2
O
3
and [101]WO
3
//[1120] a-Al
2
O
3
[101]WO
3
//[0110] a-Al
2
O
3
[101]WO
3
//[1010] a-Al
2
O
3
(b) (010)WO
3
//(111)Al//(0001) a-Al
2
O
3
and [101]WO
3
//
[011]Al//[1230] a-Al
2
O
3
(c) (010)WO
3
//(1012) a-Al
2
O
3
and [101]WO
3
//[011]Al//
[01-10] a-Al
2
O
3
(d) (010)WO
3
//(1012) a-Al
2
O
3
and [101]WO
3
//[011]Al//[01-10] a-
Al
2
O
3
[101]WO
3
//[011]Al//[1230] a-Al
2
O
3
.
A. Al Mohammad / Vacuum 83 (2009) 13261332 1331
surfaces of substrates may result in an increase of defects on
surfaces which create WO
3
lms with small grain size. In the case of
(0001) reconstructed substrates, the surface oxygen vacancies
increase until the surface consists of only one type of atomic species
and the surface aluminum layers improve the crystallographic and
morphologic properties of WO
3
lms. The observed changes in the
conductivity and activation energy E
A
of the WO
3
lms were seen to
correspond to n-type semiconductor like behavior. The changes in
electrical properties of the lms were believed to be related to the
changes in surface structure of a-Al
2
O
3
substrates.
Acknowledgements
I would like to thank Prof. I. Othman, Director General of the
Atomic Energy Commission of Syria for his support and encourage-
ment, as well as Dr. M. Kh. Rukiah and Prof. M. Kh. Sabra for valuable
discussions.
References
[1] Stankova M, Vilanovaa X, Llobet E, Caldererb J, Bittencourtc C, Pireauxc JJ, et al.
Sensors and Actuators B 2005;105:2717.
[2] Reyes LF, Saukko S, Hoel A, Lantto V, Granqvist CG. Journal of the European
Ceramic Society 2004;24:14159.
[3] Sivakumar R, Gopalakrishnan R, Jayachandran M, Sanjeeviraja C. Optical
Materials 2007;29:67987.
[4] Kanan SM, Tripp CP. Current Opinion in Solid State and Materials Science
2007;11:1927.
[5] Al Mohammad A. Physica Status Solidi A 2008;205/12:272932.
[6] Stankova M, Vilanova X, Llobet E, Calderer J, Vinaixa M, Gra` cia I, et al. Thin
Solid Films 2006;500:3028.
[7] Zhu YQ, Hu W, Hsu WK, Terrones M, Grobert N, Hare JP, et al. Chemical Physics
Letters 1999;309:32734.
[8] Al Mohammad A, Gillet M. Thin Solid Films 2002;408:302.
[9] Tagtstrom P, Jansson U. Thin Solid Films 1999;352:10713.
[10] Gillet M, Lemire C, Gillet E, Aguir K. Surface Science 2003;532/535:519.
[11] Labidi A, Jacolin C, Bendahan M, Abdelghani A, Gu erin J, Aguir K, et al. Sensors
and Actuators B 2005;106:7138.
[12] Yang YA, Yao JN. Journal of Physics and Chemistry of Solids 2000;61:64750.
[13] Nanba T, Ishikawa M, Sakai Y, Miura Y. Thin Solid Films 2003;445:17581.
[14] Gillet M, Al Mohammad A. Thin Solid Films 2000;374:134.
[15] Gillet E. Science of Ceramic Interfaces. In: Novotny J, editor; 1991. p. 439.
[16] Gillet M, Bruna JC. Surface Review & Letters 1998;5(No. 1):3259.
[17] Chambers SA. Surface Science Reports 2000;39:105.
[18] Gillet M, Delamare R, Gillet E. Thin Solid Films 2004;467:23946.
[19] Gillet M, Lemire C, Gillet E, Aguir K. Surface Science 2003:5325.
[20] Gillet M, Delamare R, Gillet E. European Physical Journal D 2005;34:2914.
[21] Li M, Altman EI, Posadas A, Ahn CH. Thin Solid Films 2004;446:23847.
[22] Di Bartolomeo E, Kaabbuathong N, DEpifanio A, Grilli ML, Traversa E, Aono H,
et al. Journal European Ceramic Society 2004;24:118790.
A. Al Mohammad / Vacuum 83 (2009) 13261332 1332