Growth and characterization of AP-MOCVD iron doped titanium dioxidethin ®lms

V. Gauthiera,*, S. Bourgeoisa, P. Sibillotb, M. Maglioneb, M. Sacilottib

aLaboratoire de Recherches sur la ReÂactivite des Solides, UMR 5613 CNRS ± Universite de Bourgogne, UFR Sciences et Techniques, BP 47 870,

21 078 Dijon Cedex, FrancebLaboratoire de Physique de l'Universite de Bourgogne, UPRESA 5027 CNRS, BP 47 870, 21 078 Dijon Cedex, France

Received 23 July 1998; revised 16 September 1998; accepted 1 October 1998

Abstract

Atmospheric pressure metal organic chemical vapor deposition (AP-MOCVD) was used to prepare iron doped titanium dioxide thin ®lms.

Thin ®lms, between 40 and 150 nm thick, were deposited on Si, SiO2 and Al2O3 substrates using titanium tetra isopropoxide and ferrocene as

metal organic precursors. TiO2 iron doping was achieved in the range of 1±4 at.%. The ®lm morphology and thickness, polycrystalline

texture and doping content were studied using respectively scanning electron microscopy (SEM), X-ray diffraction (XRD) and X-ray

photoelectron spectroscopy (XPS). The in¯uence of growth temperature, deposition time, substrate type and dopant partial pressure were

studied. Electrical characterizations of the ®lms were also performed. q 1999 Elsevier Science S.A. All rights reserved.

Keywords: AP-MOCVD; Titanium oxide; Iron; Doping

1. Introduction

TiO2 has attracted a great deal of attention because of its

physical properties (high refractive index and high dielectric

constant [1]) and of its chemical behavior in photosensitive

reactions: the use of titanium dioxide has been developed

greatly since the water photo-splitting [2] and the ammonia

photocatalytic production [3] were ®rst described. Doping

titania with metal ions was considered with the aim of

improving photocatalytic properties. An extension of the

spectral response in the near-visible region was actually

observed for TiO2 doped with 3D-transition metals such

as iron [4]. However, con¯icting results were obtained

concerning the positive effects of metals doping in different

photocatalytic reactions such as, for example, the photoca-

talytic reduction of nitrogen [3,5] or the photo-oxidation of

toluene [6].

Moreover, many applications of iron doped titanium

oxide ®lms require the knowledge of the DC electrical prop-

erties namely the ®lms conductivities in the temperature

range of their potential use, that is to say near room tempera-

ture. Previous studies on the conductivity of TiO2 crystals,

especially in the rutile phase, were achieved in the sixties at

temperatures above 500 K [7±10]. They have showed that

electrodes and surface effects take fundamental part, and

that the conductivity is anisotropic. Other works should be

mentioned on mixtures of TiO2 and Fe2O3 powders: for a

mixture containing 98% TiO2 and 2% Fe2O3 the resistivity

was reduced by half [11].

The different behaviors of metal doped titania samples

could often be related to their physical properties such as

metal content, particle size (in the nanometer range or

bigger), titania crystallographic form (anatase or rutile),

etc. It was actually reported that in the case of different

metals/TiO2 catalysts, the catalyst synthesis process deter-

mines the degree of photocatalytic activity [12].

In view of the applicability of this kind of systems either

to the treatment of pollutants and wastes [13,14] or to the

conversion of light energy for electricity generation [15], it

is worthwhile preparing well characterized metal doped tita-

nia samples in order to determine the role of the metal ions

in the TiO2 matrix.

In this work, iron doped titanium dioxide thin ®lms were

prepared by atmospheric pressure chemical vapor deposi-

tion using metal organic compounds as precursors (AP-

MOCVD). Titania ®lms can be prepared by several methods

[16±19] but chemical vapor deposition belongs to the most

versatile and widely used techniques. Moreover AP-

MOCVD is a low-cost and easy to manipulate deposition

technique to obtain thin ®lms [20].

In this work, titanium tetra isopropoxide and ferrocene

Thin Solid Films 340 (1999) 175±182

0040-6090/99/$ - see front matter q 1999 Elsevier Science S.A. All rights reserved.

PII: S0040-6090(98)01469-2

* Corresponding author. Tel: 1 33-3-8039-6158; fax: 1 33-3-8039-

6132; e-mail: claudie.josse@u-bourgogne.fr.

were used respectively as titanium, oxygen and iron precur-

sors.

Thin ®lm morphologies were studied by optical micro-

scopy and by scanning electron microscopy (SEM). Infor-

mation concerning the stoichiometry, the iron content and

the possible presence of impurities were obtained by X-ray

photoelectron spectroscopy (XPS). X-ray diffraction (XRD)

experiments allowed the determination of the ®lm crystal-

lographic structure. Electrical characterizations were also

performed in order to obtain additional information on the

®lms.

All thin ®lms characterizations were studied in relation to

the different MOCVD growth parameters: the growth

temperature, the deposition time and the ferrocene iron

source temperature. Moreover, three different kinds of

substrates were used: Si, SiO2 and Al2O3.

2. Experimental

The metal organic chemical vapor deposition method

working at atmospheric pressure (AP-MOCVD) was used

to form TiO2 iron doped thin ®lms in a horizontal quartz

reactor cell. Titanium tetra isopropoxide, Ti(OC3H7)4, was

employed as both titanium and oxygen source. A previous

work has shown that the precursor oxygen content was suf®-

cient to obtain stoichiometric titanium dioxide thin ®lms

[21]. The temperature of this organometallic precursor

source was 408C.

The solid iron doping source was ferrocene, Fe(C5H5)2. It

has been chosen because near room temperature, this

precursor is relatively volatile so as to allow a good control

of the doping content [22]. This iron source was held

between 0 and 508C in a thermal bath (respective vapor

pressures were 7:9 £ 1022 and 10 Pa).

Nitrogen was used as carrier gas with ¯ow rates respec-

tively of 300 and 100 sccm (standard cubic centimeter per

minute) through the organometallic titanium and iron

sources.

The MOCVD quartz cell reactor was equipped with an

induction heated graphite susceptor on which substrates

were placed.

The ®lm growths took place between 1 and 4 h on differ-

ent substrates: Si (100), SiO2 and Al2O3 (11Å02). The

substrate temperatures were between 500 and 7008C.

Unless otherwise explained, we will mention afterwards

two kinds of layers: thin and thick layers, corresponding

respectively to 1 h and 4 h deposits.

The ®lms morphologies and thicknesses were studied by

scanning electron microscopy (SEM-Jeol 6400F). In order

to avoid electrical samples charging problems, a thin gold

overlayer was deposited on coated SiO2 and Al2O3

substrates. Energy Dispersive X-ray Spectrometry (LINK

Oxford energy dispersive analyzer) microanalyses were

used to obtain a global iron content of the TiO2 ®lms.

Thin ®lm iron content was also obtained by X-ray photo-

electron spectroscopy analyses (XPS-Riber Cameca MAC 2

using Ka Aluminum radiation) in addition to other infor-

mation concerning titanium and iron oxidation states and

possible presence of impurities. Basic operations of smooth-

ing and area calculations could be performed. Backgrounds

were removed by a Shirley routine and desummations of

complex spectra were done by ®tting with Gaussian

peaks. Quanti®cations were performed by correcting peak

areas with sensitivity factors. A differentially pumped argon

ion gun (Perkin-Elmer C04-303) was employed for the

erosion of about 20 nm of the samples.

The crystallographic phases were identi®ed by X-Ray

diffraction (Inel CPS 120) using Ka1 copper radiation.

Electrical conductivity experiments were performed

using a Keithley 236 electrometer. Four gold contacts

were formed by thermal evaporation or by sputtering and

connected by silver paint. During these conductivity experi-

ments, the gold (or silver) diffusion was negligible because

the temperatures were lower than 420 K. Only insulator

corundum and silica substrates were concerned by the

study because they give the required isolation in the resis-

tivity cell.

3. Results and discussion

3.1. SEM and XRD

Thickness and morphological characterizations of iron

doped TiO2 thin ®lms related to the different growth para-

meters were done by scanning electron microscopy (SEM).

In order to measure the ®lm thickness, a cleavage of the

samples was carried out and SEM observations were done

on the cleaved edge. Only the thickness of the ®lms depos-

ited on silicon was determined since this substrate was the

easiest to cleave.

Concerning the thicker TiO2 ®lm prepared at a growth

V. Gauthier et al. / Thin Solid Films 340 (1999) 175±182176

Fig. 1. SEM image of the cross section of a thick TiO2 ®lm grown on silicon

at 6008C.

temperature of 6008C, a 200 nm thick columnar layer is

observed (Fig. 1). For an iron doped TiO2 ®lm grown in

the same experimental conditions as the previous one, a

less columnar but much more dense layer is observed

(Fig. 2). The iron doped TiO2 ®lm thickness is 150 nm.

So, in the presence of iron, a densi®cation of the TiO2 layers

as well as a decrease of their thickness are shown.

The estimated thickness of the thinner TiO2 iron doped

®lm grown on silicon is about 40 nm.

The morphology of ®lms deposited on silicon was studied

related to the different growth parameters, namely the iron

source partial pressure, the growth temperature and the

deposition time.

Whatever the iron source pressure, thin iron doped TiO2

layers grown at 6008C have a homogeneous granularity

ranging from 10 to 20 nm.

Thin iron doped TiO2 ®lms grown at Pferrocene � 1.3 Pa

(ferrocene source temperature � 308C) exhibit irregular

grain size variation with growth temperature: for growth

temperatures of 5008C, 6008C and 7008C (Figs. 3±5), the

average grain sizes are respectively 25, 15 and 80 nm. This

feature could be connected to the anatase/rutile phase tran-

sition occurring at about 5508C for the undoped TiO2 layers

grown on silicon substrate [23].

Concerning the growth period, thicker iron doped TiO2

®lms display a much more irregular relief and coarser grains

as compared to thinner ones carried out in similar condi-

tions. A thicker doped TiO2 layer grown at 6008C and

Pferrocene � 1.3 Pa exhibits nearly ¯at regions including

hexagonal grains and rougher regions with sharper grains

(Fig. 6). After comparison of this rough zone with a thick

TiO2 deposit made in the same experimental conditions, it

seems that this region can be identi®ed as TiO2 since grains

have similar morphologies (Fig. 7).

This is con®rmed by EDX microanalysis performed for

thicker ®lms: ¯at regions include more iron compared with

rougher regions which seem principally composed of TiO2.

A back-scattered electrons study has also been done on

these samples: ¯at zones stand out clearer than rougher

zones and therefore correspond to a higher average atomic

number than rougher regions. This result is in good agree-

ment with a higher concentration of iron within the ¯at

zones.

Thus iron is not uniformly distributed in the thick layers:

a surface diffusion process is proposed as the one for the

iron localization in de®nite zones.

Concerning the thinner grown layers, iron zones localiza-

tion could not be observed by back-scattered electron

images since this element is well distributed. Moreover,

its content is not suf®cient to be detected by EDX micro-

analysis.

Thin ®lm morphology was also studied in relation to the

substrate nature: all iron doped TiO2 layers deposited on

silica have similar morphological aspects. They have homo-

geneous grain sizes and shapes, and a similar iron distribu-

tion as what has been observed on silicon. Thin and thick

iron doped TiO2 layers grown on Al2O3 at 6008C exhibit

grains of triangular pyramidal shape (Fig. 8). Within this

layer, there are no different kinds of zones: iron seems well

distributed and cannot be detected by EDX microanalysis.

In order to obtain the crystallographic structure of the

®lms, XRD was used at small incidence angle so as to

increase the ®lm volume studied. No particular structure

related to the presence of iron could be observed: only rutile

and anatase diffraction peaks are present on the spectra as it

was already observed for a titanium dioxide ®lm without

iron [23]. It should however, be noticed that the peaks full

width at half maximum (FWHM) are narrower for iron

doped layer than for a pure titanium dioxide ®lm. Calcula-

tions [24] show that the mean size of crystallites increases

with iron doping. This is in agreement with the more dense

structure shown in Fig. 2.

3.2. XPS

XPS results concerning the surface stoichiometry of iron

V. Gauthier et al. / Thin Solid Films 340 (1999) 175±182 177

Fig. 2. SEM image of the cross section of a thick iron doped TiO2 ®lm

grown on silicon at 6008C.

Fig. 3. SEM image of a thin iron doped TiO2 ®lm grown on silicon at

5008C.

doped titanium dioxide ®lms deposited on silicon are

presented here (Fig. 9) related to the different MOCVD

growth parameters. It should be noticed that deposits

performed on silica and corundum lead to XPS spectra simi-

lar to the one obtained for deposits on silicon with, however,

a shift of the global spectra towards higher binding energies

due to charging effects.

The binding energy of the Ti2p3/2 peak and the energy

difference between the Ti2p3/2 and the Ti2p1/2 peaks are of

459:2 ^ 0:1 eV and 5.7 eV, respectively. Although these

values are characteristic of titanium at the oxidation state

4, the FWHM is slightly too high (about 1.6 eV) to be

accounted for by the one presence of Ti41. In the same

experimental conditions for both excitation and detection,

the FWHM is 1.4 eV for a perfectly stoichiometric sample,

where only Ti41 is present. This fact can be accounted for by

the presence in the Ti2p3/2 peak of an additional contribution

on the low binding energy side. Indeed, the Ti2p3/2 peak

cannot be simulated by a single Gaussian: a lower oxidation

state characteristic of the presence of Ti31 is evidenced.

Therefore, iron introduction involves a slight destoichiome-

trization of the oxide surface.

In order to detect iron in the XPS spectra, a high number

of scans is necessary because of the low iron content intro-

duced so as to perform the doping of titanium dioxide. The

binding energy of the main Fe2p3/2 peak is 711:2 ^ 0:6 eV,

characteristic of iron at the oxidation state 3. A desumma-

tion of the different contributions relative to the Fe2p peak

was quite dif®cult to perform because of the weakness of the

signal. However, it has been shown that the values of the

energy differences between the Fe2p3/2 and O1s peaks are

V. Gauthier et al. / Thin Solid Films 340 (1999) 175±182178

Fig. 4. SEM image of a thin iron doped TiO2 ®lm grown on silicon at

6008C.

Fig. 5. SEM image of a thin iron doped TiO2 ®lm grown on silicon at

7008C.

Fig. 6. SEM image of a thick iron doped TiO2 ®lm grown on silicon at

6008C. (a) `¯at region'.(b) `rough region'.

not the same for the different oxidation states of iron [25].

These energy difference values are here between 180.6 and

181.8 eV, characteristic of a higher amount of Fe31

compared to Fe21. The greatest part of iron initially intro-

duced as Fe21 from the organometallic precursor Fe(C5H5)2

is thus oxidized either during the iron doped TiO2 thin ®lm

growth or during the ®lms' exposure to air.

The binding energy of the O1s peak is 530:2 ^ 0:2 eV.

The O1s peaks are asymmetrical with a shoulder on the high

binding energy side. This fact has already been observed for

titanium dioxide ®lms grown in the absence of iron [26].

This shoulder is characteristic of hydroxyl groups present on

the thin ®lm surface. It should be noticed that, after ion

(argon) bombardments, the O1s peak shoulder on iron

doped TiO2 ®lm remains constant or increases, depending

on the samples. This has to be compared to undoped TiO2

layers where the shoulder disappears after bombardment

and where the O1s peak becomes symmetric with the one

contribution of the oxygen from the oxide lattice. So, in the

case of iron doped TiO2 layers, the hydroxylation phenom-

enon is not only super®cial: it does not result only from the

sample air exposure since hydroxyl groups are also present

and probably mixed into the layer during the growth. This

fact can be related to the combination during growth of

oxygen and hydrogen coming from the two precursors

Ti(OC3H7)4 and Fe(C5H5)2.

It has been observed that whatever the iron partial pres-

sure, the iron content detected is near 1 at.% for thinner

®lms deposited at a growth temperature of 5008C. It should

be noticed that this iron content value is close to the detec-

tion limit of the XPS technique: small variations around 1

at.% should not be detected.

On the other hand, the iron content increases with the

growth temperature: a substrate temperature rise from

5008C to 7008C induces an increase of the iron content

introduced in the layer from 1 to 4 at.%. Ferrocene being

decomposed from 4508C [22], its decomposition will be all

the more ef®cient as the growth temperature exceeds 4508C.

Likewise, an increase in the hydroxyl content of the layer

from 12 to 25 at.% is also observed when the growth

temperature increases from 500 to 7008C. The enhancement

of the ferrocene decomposition leads, in the same way as for

iron, to an increase of the hydrogen content and thus to a rise

of the ability of this element to combine with oxygen

coming from Ti(OC3H7)4. This fact promotes an increase

of the hydroxyl groups content within the layer.

As far as the deposition time is concerned, iron seems

to be well distributed for the thinner layers since its con-

tent is homogeneous at the surface and within the mat-

erial: this is evidenced by the fact that the iron content

does not vary before and after an ion bombardment

which is known to erode about 20 nm, that is to say,

about one half of the grown layer. On the contrary,

concerning thicker iron doped TiO2 layers, the iron content

is different at the surface and in the bulk: during the same

kind of erosion, the iron content increases from 2 to 4 at.%.

These results con®rm EDX microanalysis where ¯at zones

richer in iron and rough zones poorer in iron are evidenced.

Therefore, the longer the growth period, the more ef®cient

is the iron diffusion and segregation within the grown

layers.

3.3. Electrical measurements

The iron doped TiO2 ®lms show high values of DC resis-

tances from 10 MV to a few TV at room temperature; in the

case of undoped TiO2 ®lms and with the same electrical

measurement procedure, the resistance cannot be measured

(R . 1015 V ).

The two kinds of substrates exhibit different current±

voltage behaviors: for silica, iron doped ®lms follow an

ohmic law up to 10 V or more. All the ®lms deposited on

Al2O3 show deviation at the same voltage. The applied

voltage was thus limited to avoid non-ohmic contributions.

It should be noticed that, for this high range of resistance,

surface effects could dominate [27]. Despite a permanent

evacuation of the resistivity cell, large discrepancies appear

V. Gauthier et al. / Thin Solid Films 340 (1999) 175±182 179

Fig. 7. SEM image of a thick TiO2 ®lm grown on silicon at 6008C.

Fig. 8. SEM image of a thick iron doped TiO2 ®lm grown on sapphire at

6008C.

for ®lms prepared with similar parameters; the absorption of

water, gases or the mentioned hydroxylation phenomenon

right after the sample processing cannot be well controlled.

Under temperature cycling between 2 1708C and 1258C,

all ®lms display the same trends: (a) the resistance R and the

temperature coef®cient dR/dT decrease with increasing

V. Gauthier et al. / Thin Solid Films 340 (1999) 175±182180

Fig. 9. XPS spectrum relative to a thick iron doped TiO2 ®lm grown on silicon at 6008C.

Fig. 10. Ferrocene vapor pressure effect on the activation energy of thin iron doped TiO2 ®lms grown on sapphire at 6008C.

temperature and the heating and cooling curves coincide;

(b) when plotted in an Arrhenius scale, the resistance curves

are nearly linear.

Both these features call for a semi-conducting like beha-

vior of the TiO2 iron doped ®lms. This is consistent with

similar observations on rutile TiO2 crystals but not on

anatase TiO2 [28] and TiO [29].

As usual in semi-conductors, the resistivity was ®tted

using an Arrhenius law R�T� � Roexp�-W=T� with a good

correlation over a range of 50 K. The activation energy W

and prefactor Ro are strongly dependent on the TiO2 iron

doped ®lm processing conditions. However, W is much

more reliable than Ro since for a given ®lm Ro is very sensi-

tive to the surface state and W is not. Ro may be increased by

a factor of 100 when unwanted pollution is present on the

®lm's surface.

Another proof of the intrinsic nature of W can be deduced

from the experimental correlation between W and the ferro-

cene source partial pressure (Fig. 10), keeping all other

parameters constant, i.e. growth temperature (6008C),

growth duration (1 h) and the substrate (Al2O3). As

evidenced in Fig. 10, the activation energy W is the same

under cooling or heating and it can be increased by a factor

of 5 when the ferrocene vapor pressure is raised from 2:6 £1022 up to 10 Pa. As already pointed out, the average grain

size was constant in all these ®lms; this means that the

activation energy variation does not depend on a grain

size effect.

This strong increase of W with the iron source pressure

(or the iron ratio) means that a purely semi-conducting

model cannot be applied to our TiO2 ®lms. When substituted

to Ti41 ions, Fe31/Fe21 impurities should lead to impurity

levels within the TiO2 band gap and the increasing amount

of iron cannot change the impurity levels W but only their

occupation. We thus conclude that some heterogeneous

interstitial iron or iron oxide is embedded between the

impurity clusters to the average distance which is decreasing

with the iron content increase.

Another alternative model for the increase of W with the

iron partial pressure is the possible occurrence of polarons at

the iron site. This possibility was raised from the compar-

ison between theoretical models and experiments on drift

mobility, Hall mobility, thermoelectric power and absorp-

tion of light [30]. At present we cannot con®rm such polaro-

nic conductivity in our ®lms, but XPS results have shown

that Fe31/Fe21 and Ti31 polaronic sites are detected in thin

®lms deposited on corundum.

For ®lms deposited on silica substrates, a similar relation

between activation energy and ferrocene vapor pressure

cannot be established. In this case, the iron segregation

impedes probably a simple correlation.

It should be noticed that the lowest measured resistances

by the four points method correspond to a (squared) resis-

tivity of about 200 V cm at room temperature. This can

restrict the electrical charge carrier exchange involved in

catalytic processes at the ®lms surface; in fact, the process

requires an exchange in a direction perpendicular to the

®lm: for one square centimeter and 40 nm thickness (1 h

deposition), the resistance will be lower than 1 mV , quite

different from the large values of R in the plane which was

observed in this work.

4. Conclusions

AP-MOCVD technique has been successfully applied for

the iron doping of TiO2 thin ®lms since this process allows

to introduce iron content over the range 1±4 at.% using

ferrocene and tetra isopropoxide as iron and titanium/

oxygen organometallic precursors.

Iron doped TiO2 thin ®lms (40±150 nm thick) have been

deposited on several substrates (silicon, silica and sapphire)

at different growth temperatures and ferrocene vapor pres-

sures during various deposition times.

Although no particular structure related to the presence of

iron could be detected by XRD technique, the iron introduc-

tion induces important modi®cations of TiO2 thin ®lms

properties:

² A densi®cation of columnar TiO2 ®lms which corre-

sponds to a decrease of XRD anatase and rutile peaks

full width at half maximum (SEM and XRD).

² A destoichiometrization of titanium dioxide: presence of

Ti31/Fe31 ions (XPS).

² An increase of the bulk hydroxyl group with respect to

undoped TiO2 thin ®lms (XPS).

² A modi®cation of the conductivity in comparison with

undoped TiO2 layers: electrical measurements on these

iron doped TiO2 ®lms have shown that the resistance may

be decreased from 1015V for undoped layers to 107V for

the highest iron concentration.

The in¯uence of different growth parameters has also

been studied:

² Growth temperature plays an important role since for an

increase of this parameter from 5008C to 7008C, an irre-

gular variation in grain size and an increase of the iron

and hydroxyl content have been observed.

² Ferrocene vapor pressure has a minor role: neither the

iron content nor the thin ®lm morphology were modi®ed

by an increase of this parameter.

² An augmentation of the deposition time promotes a grain

size increase, a modi®cation of the grain shapes and a

more pronounced relief. Moreover, the iron doped TiO2

thick ®lms grown on silicon exhibit an iron segregation

unlike for the other substrates.

Both the occurrence of Ti31/Fe31 centers evidenced by

XPS investigations and the strong variation of the ®lm resis-

tance are signatures of a possible reactivity of the ®lms. This

is of interest in view of the expected photocatalytic activity

of iron doped TiO2 layers.

V. Gauthier et al. / Thin Solid Films 340 (1999) 175±182 181

References

[1] A. Erbil, W. Braun, B.S. Kwak, J. Cryst. Growth 124 (1992) 684.

[2] A. Fujishima, K. Honda, Nature 37 (1972) 238.

[3] G.N. Schrauzer, T.D. Guth, J. Am. Chem. Soc. 99 (1977) 7189.

[4] Y. Takahashi, A. Ogiso, R. Tomoda, K. Sugiyama, H. Minoura, M.

Tsuiki, J. Chem. Soc. Faraday Trans, 1 78 (1982) 2563.

[5] A. Davies, D.L. Boucher, J.G. Edwards, in: D.C. Neckers, D.H.

Volman, G. von BuÈnau (Eds.), Advances in Photochemistry, 19,

Wiley, New York, 1995, pp. 235.

[6] J.A. Navio, M. Garcia Gomez, M.A. Pradena Adrian, J. Fuentes Mota,

in M. Guisnet et al. (Eds.), Heterogeneous Catalysis and Fine Chemi-

cals II, Elsevier, Amsterdam, 1991, p. 445.

[7] F.A. Grant, Rev Modern Phys. 31 (1959) 646.

[8] T.P. Pearsall, J. Phys. D: Appl. Phys. 3 (1970) 1837.

[9] A. Von Hippel, J. Kalnajs, W.B. Westphal, J. Phys. Chem. Solids 23

(1961) 779.

[10] S.R. Kurtz, R.G. Gordon, Thin Solid Films 147 (1987) 167.

[11] G.H. Johnson, J. Am. Ceram. Soc. 36 (1953) 97.

[12] O.A. Ileperuma, C.T.K. Thaminimulla, W.C.B. Kiridena, Solar

Energy Mater. Solar Cells 28 (1993) 335.

[13] L. Cermenati, P. Pichat, C. Guillard, A. Albini, J. Phys. Chem. 101

(1997) 2650.

[14] L. Amalric, C. Guillard, E. Blanc-Brude, P. Pichat, Wat. Res. 30

(1996) 1137.

[15] B. O'Regan, M. GraÈtzel, Nature 353 (1991) 737.

[16] Y. Sorek, R. Reisfeld, Appl. Phys. Lett. 63 (1993) 3256.

[17] P. LoÈbl, M. Huppertz, D. Mergel, Thin Solid Films 251 (1994) 72.

[18] D. Leinen, J.P. Espinos, A. Fernandez, A.R. Gonzalez-Elipe, J. Vac.

Sci Technol. A 12 (1994) 2728.

[19] G. Zhai, H. Kozuka, T. Yoko, Thin Solid Films 277 (1996) 147.

[20] A. Giovanni, G.A. Battiston, R. Gerbasi, M. Porchia, A. Mango, Thin

Solid Films 239 (1994) 186.

[21] P. Babelon, A.S. Dequiedt, H. MosteÂfa-Sba, Bourgeois, S., Sibillot,

P., Sacilotti, M, Thin Solid Films 322 (1998) 63.

[22] J.A. Long, V.G. Riggs, A.T. Macrander, W.D. Johnston, J. Cryst.

Growth 77 (1986) 42.

[23] P. Babelon, Thesis, Universite de Bourgogne, 1997.

[24] J.I. Langford, Proc. Int. Conf. on Accuracy in Powder Diffraction II,

Gaithersburg, 1992, p. 110.

[25] G. Kurbatov, E. Darque-Ceretti, M. Aucouturier, Surf. Interface Anal.

18 (1992) 811.

[26] A. Dequiedt, Rapport DEA, Universite de Bourgogne, 1996.

[27] T. Fuyuki, T. Kobayashi, H. Matsunami, J. Electrochem. Soc. Solid-

State Sci. Technol. 135 (1988) 248.

[28] L. Forro, O. Chauvet, D. Emin, L. Zuppiroli, H. Berger, F. LeÂvy, J.

Appl. Phys. 75 (1994) 633.

[29] F.J. Morin, Bell. Syst. Tech. J. 37 (1958) 1047.

[30] H. Bottger, V.V. Bryksin, Hopping Conduction in Solids, VCH,

Weinheim, ISBN 3-527-26451-5,

V. Gauthier et al. / Thin Solid Films 340 (1999) 175±182182