Recent studies on In2S3 containing oxygen thin films

Article abstract of DOI:10.1016/S0038-1098(02)00099-6

Attention has recently been paid to In₂S₃ thin films because of their potential application as buffer layer in CIGS-based solar cells. In this paper, the recent studies performed on such thin films deposited by chemical bath deposition (CBD) and physical vacuum deposition (PVD) are reported and compared. The main part of the experimental results exposed concerns the wide band gap PVD deposited β-In₂S₃ thin films. The influence of the synthesis conditions on the physico-chemical, optical, and electrical properties are reported and discussed in the paper. The oxygen present in the CBD and PVD films has been found to be the origin of their optical properties, which make them a good candidate to substitute CdS in thin films solar cells as buffer layer. The films have a n-type electrical conductivity and their optical band gap is about 2.8 eV.

Full text of DOI:10.1016/S0038-1098(02)00099-6

PERGAMON
Solid State Communications 122 (2002) 445±450
www.elsevier.com/locate/ssc
Recent studies on In S containing oxygen thin ®lms
2
3
a
N. Barreau , J.C. Bern eÁ de , H. El Maliki , S. Marsillac , X. Castel , J. Pinel
a,
a
a
b
b
*
a
LPSE-FSTN, Equipe couches minces et materiaux nouveaux, Universite de Nantes, 2, Rue de la Houssini eÁ re, BP 92208,
44322 Nantes Cedex 3, France
URER, Groupe de Micro e lectronique et Visualisation, Universit e de Rennes I, Campus de Beaulieu, 35042 Rennes Cedex, France
b
Received 21 June 2001; received in revised form 4 February 2002; accepted 13 February 2002 by P. Burlet
Abstract
Attention has recently beenpaid to In
2 3
S thin ®lms because of their potentialapplication asbuffer layer in CIGS-based solarcells. In
this paper, the recent studies performed on such thin ®lms deposited by chemical bath deposition (CBD) and physical vacuum
deposition (PVD) are reported and compared. The main part of the experimental results exposed concerns the wide band gap PVD
deposited b-In S thin ®lms. The in¯uence of the synthesis conditions on the physico-chemical, optical, and electrical properties are
2
3
reported and discussed in the paper. The oxygen present in the CBD and PVD ®lms has been found to be the origin of their optical
properties, which make them a good candidate to substitute CdS in thin ®lms solar cells as buffer layer. The ®lms have a n-type
electrical conductivity and their optical band gap is about 2.8 eV. q 2002 Elsevier Science Ltd. All rights reserved.
PACS: 71.20.Nr; 74.25.Gz; 78.20. 2 e
Keywords: A. Indium sulphide; A. Thin ®lms; D. Optical properties; D. Structural properties
1
. Introduction
cadmium, involves the effort of many research teams [4±
]. Some very promising results have been obtained with
7
Chalcogenide based solar cells with high ef®ciencies
.17%) have been achieved by several groups [1±3]. The
ZnS deposited by CBD [6]. However, the CBD technique
itself constitutes a bottleneck in the deposition process and a
whole process using only physical vacuum deposition
(PVD) technique should be, ®nely, preferable for industrial
development.
(
absorbing layer in these devices is a p-Cu(In,Ga)Se of
2
about 2±3 mm thickness, which is crystallized in the
chalcopyrite structure. This ®lm is obtained by either co-
evaporation or selenization of metal layers sequentially
evaporated. The junction necessary for the occurrence of
the photovoltaic effect is typically fabricated using a CdS
buffer layer deposited by chemical bath deposition (CBD).
Not only these ®lms are n-type but also quite resistive,
which allows to avoid the formation of shunt path between
the Mo under electrode and the transparent conductive oxide
of the upper electrode. If the use of CdS is very ef®cient in
achieving performing cells, it is undesirable from the
environmental safety point of view because of the cadmium.
Moreover, its band gap value (2.48 eV) induces some
optical absorption loss at short wavelengths. Therefore,
the research of a wider band gap semiconductor without
In the present paper we have focused our interest on a n-
type semiconductor thin ®lm, which can be synthesized
either by CBD or PVD, the b-In S containing oxygen.
2
3
This material corresponds to the speci®cation of a good
solar cell buffer layer. More speci®cally, many works
have been reported on CBD-In S [2,8,9]. In this work, the
2
3
ability to grow In S thin ®lms by CVD and PVD techniques
3
2
is demonstrated and their properties are compared. Since
many works have not been reported on the PVD technique
we mainly focus our interest on the PVD ®lms.
2. Experimental
2
.1. Deposition technique
* Corresponding author. Tel.: 133-2-51-12-55-30; fax: 133-2-
1-12-55-28.
5
E-mail address: jean-christian.bernede@physique.univ-
nantes.fr (J.C. Bern eÁ de).
In order to compare the quality of the different ®lms, we
have deposited In S by CBD and PVD.
2 3
0038-1098/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S0038-1098(02)00099-6

446
N. Barreau et al. / Solid State Communications 122 (2002) 445±450
2
.1.1. Chemical bath deposition conditions
In thin ®lms were grown by CBD on SnO
bare soda lime glass substrates. SnO (TCO) ®lms were
and qualitative) of the ®lms were obtained from X-ray
2
photoelectron spectroscopy (XPS) measurements. Electron
S
2 3
2
coated or
2
micropobe analysis (PGT-IMIX PTS model) was used to
1
provided by SOLEMS. The substrate were sonicated in
alcohol and rinsed in boiling distilled water. The CBD-
determine the chemical composition of the ®lms.
The optical measurements were done at room temperature
using a spectrophotometer CARY and the transmittance is
2 3
In S ®lms were grown in a thermostated bath whose
3
temperature was 345 K. The substrates were clamped verti-
cally with plastic clamps and kept immersed in the stirred
solution. The deposition takes place in a aqueous solution
measured in the wavelength interval of 300±800 nm.
Infra red spectra were recorded on Perkin Elmer 1600
Fourier transform infrared (FTIR) spectrometer in the
2
1
21
range 400±4000 cm . The resolution of the instrument
containing indium(III) chloride ꢀ‰InCl Š ˆ 0:01 mol l
†
3
21
21
was 4 cm .
and thioacetamide ꢀ‰CH CSNH Š ˆ 0:15 mol l †: After
3
2
deposition, the samples were removed from the bath and
sonically washed in distilled water until heterogeneities
are removed. Then they were dried in air.
3. Experimental results
Since CBD-In S ®lms have been already studied [12,13],
2
3
2
.1.2. Physical vapor deposition process
Thin ®lm obtention process: the technique used will only
the main part of the experimental works concerns the PVD
process, while results obtained on CBD ®lms are presented
at the end of this section.
be shortly recalled since it has already been described else-
where [10,11].
The thin ®lm obtention process can be divided into two
steps, the deposition of the constituents onto substrates and
the post-deposition annealing in ¯owing argon of the
samples evaporated.
2 3
3.1. PVD-In S ®lms
The ®lms obtained from the PVD process have been
synthesized in two steps. The ®rst one is the deposition of
the ®lms by thermal evaporation and the second is the post-
deposition annealing.
The deposition of the constituents is carried out by vacuum
24
thermal evaporation, at a gas pressure of 5 £ 10 Pa, using a
tungsten crucible for indium and a `Laboratory made' Pyrex
crucible for the sulfur. The purities of indium and sulfur used
are 99.99 and 99.98%, respectively. The ®lms were deposited
on soda lime glasses. All the substrates were cleaned chemi-
cally before evaporation. In order to obtain the ®lms, thin
indium and sulfur layers were sequentially deposited, their
thickness and evaporation rate being controlled in situ by an
hf quartz monitor. The thickness ratio has been experimentally
determined to ®nally have homogenous ®lms with an atomic
ratio at.%S/at.%In $ 3.
As described earlier, the synthesis process of the PVD
deposited thin ®lms consists in the annealing, under constant
argon ¯ow, of structures composed of thin layers of indium
and sulfur superposed. The properties of the thin ®lms
obtained have been found to strongly depend on the anneal-
ing conditions; therefore, the annealing temperature and
duration have been used as parameters. The in¯uence of
the temperature rate increase has also been studied.
At ®rst, it must be speci®ed that, whatever may be the
experimental conditions used in the present work, all the
The structures obtained from this process were then
annealed in a tubular oven for 30 min under a constant
®
lms have a n-type conductivity. Moreover, it has been
3
21
shown by SEM [11] that the ®lms are very homogeneous
and their coverage ef®ciency is excellent.
argon ¯ow of 0.6 dm min . The ®lms have been annealed
at ®ve different temperatures 523, 573, 623, 673, and 723 K.
The heating time and the temperature rate increase have
been varied as discussed in the manuscript. The samples
studied in this paper were about 100 nm thick.
For the ®rst approach the annealing duration, with refer-
ence to earlier works, has been chosen to be half an hour
[
14].
The annealing temperature has been shown [11] very
important in the crystalline properties of the ®lms. This
temperature has been varied in steps of 50 K between 523
and 723 K. The XRD diagram obtained from each ®lm are
shown in Fig. 1. It can be seen that there is a `threshold'
temperature at 623 K. Below this temperature the crystal-
lization is only initialized while above this temperature
the ®lms are wholly crystallized in the quadratic structure
2
.2. Experimental characterization
The obtained ®lms were analyzed by X-ray diffraction
XRD) to investigate the compound and structure formation;
(
a Siemens X-ray goniometer with Cu Ka line was used in
the study.
For observing the surface and cross-section morphology
of the ®lms, a JEOL-type JSM 6400F scanning electron
2 3
of b-In S (JCPDS Data 25-390), and only the grain size
1
microscope (SEM) was used.
The surface and depth pro®ling composition (quantitative
varies. It can be observed that this threshold concerns not
only the crystalline properties of the ®lms, but also their
1
2
Service commun de microscopie electronique aÁ balayage et de
microanalyse de l'Universit e de Nantes.
Universit e de NantesÐCNRS.
3
Laboratoire de Physique CristallineÐIMN.

N. Barreau et al. / Solid State Communications 122 (2002) 445±450
447
Fig. 2. Evolution of the optical absorbance versus the wavelength of
the PVD-®lms annealed at different temperatures, T , and of ®lms of
pure b-In
Fig. 1. XRD diagrams of PVD-In
2 3
S thin ®lms synthesized at 523,
a
573, 623, 673, and 723 K.
2 3
S .
optical properties. Fig. 2 shows the evolution of the absor-
bance of a pure b-In S thin ®lm, the increase in the optical
band gap of PVD ®lms with the annealing temperature until
it is 623 K, whereas for higher annealing temperatures
absorption threshold is stable.
2
3
and sodium contamination of the ®lms. The difference
between the oxygen involved in surface contamination
and that introduced in the ®lms during annealing is clearly
shown in Fig. 4. After annealing the main part of the oxygen,
in the ®lms, corresponds to oxygen bonded to indium. This
contribution increases relatively after surface etching, which
eliminates preferentially weakly bonded atoms, i.e. con-
tamination. The increase in annealing temperature increases
the crystalline properties of the ®lms. However, above
673 K, there is a saturation effect, while the Na and O super-
®cial concentration increases strongly. Therefore, it appears
The composition of the ®lms has been studied by EMPA
and XPS before and after annealing. As far as the results of
the EMPA study on the annealed ®lms are concerned, the
atomic ratio of sulfur and indium (Fig. 7) has only been
studied because the thin thickness of the ®lms (100 nm)
induces the detection of the elements of the substrate,
which disturbs the quanti®cation of oxygen and sodium.
The quantitative XPS analyses have shown that the ®lms
after annealing contain oxygen and that there is some
sodium not only at the surface but also in the bulk of the
that the optimum annealing temperature, T , is 673 K. So we
a
have proceeded to different annealing at T ˆ 673 K, using
a
the annealing time or the temperature rate increase as para-
meter. In the ®rst case, the annealing time, t , was varied
®
lms. It should be observed that before annealing, there is
a
2
between 0 and 45 min, the heating rate being 4.4 K min ,
1
neither oxygen nor sodium contamination of the ®lms, while
there is some excess sulfur present. We have performed
some more characterizations of these unheated samples.
The visualization of the cross-section (in the back scattered
mode) of the ®lms before annealing testify clearly that even
if, as shown by the XRD diagrams, the ®lms are amorphous,
there is an `intimate' interdiffusion between the In and S
sequentially deposited layers (not shown). There is not only
a physical interdiffusion but also a chemical interaction
between indium and sulfur as shown by XPS (Fig. 3(a)).
The XPS analyses allow the study of the chemical state of
different elements. If the binding energy of the In 3d peak is
not highly sensitive to the binding state of indium atom, that
of sulfur testi®es clearly that even before annealing there is,
at least, a partial bond formation between S and In. The
i.e. the time necessary to achieve T was 90 min, while the
a
cooling time was 150 min. The thickness of all the ®lms
being the same, the physico-chemical characterization of
these samples shows that the XRD spectra are similar
(Ipeak, FWHM), the results of the quantitative analysis are
also similar with a small sulfur de®ciency as usually observed
in the ®lms, the micrographs of the surface are also alike, and
more surprising, the band gap also.
When the heating rate is used as a parameter, with
T ˆ 673 K, t ˆ 30 min, the main new information is
a
a
obtained with transmission measurements. Fig. 5 shows
that when the heating rate decreases, mainly from 6.7 to
2
1
4.4 K min there is a blue shift of the absorption threshold.
We have also proceeded to some IR absorption measure-
ments and the presence of hydroxyl radicals in the ®lms
after annealing is probable, as shown in Fig. 6.
binding energy of the S 2p ray is indeed 162 eV in In
[
2 3
S
15] while it is 164 eV [16] for sulfur alone. The XPS
depth pro®le has con®rmed the In and S interdiffusion. It
should be also noted that the oxygen present at the surface
of the ®lm is not bonded to indium (530 eV) or sulfur (there
is no sulfur oxide peak), but it is only present as a contaminant
physically adsorbed (Fig. 3(b)) (535 eV) [16].
3
.2. CBD ®lms
The ®lms are about 30±50 nm thick after a deposition
time of about 25 min and are amorphous. XPS measure-
ments have been performed on such samples. They have
After annealing, as shown earlier, there is some oxygen

448
N. Barreau et al. / Solid State Communications 122 (2002) 445±450
there will be some sulfur escape during annealing under
pressure atmosphere.
The argon used during the annealing is the argon Nertal (O
ppm, H O: 3 ppm), i.e. there is a small oxygen contaminant
2
:
3
2
pressure during the crystallization process. It should be also
is 2957 kJ mol
21
noted that the `free enthalpy' of In
2
O
3
,
21
while that of In S is 2414 kJ mol at 298 K [17].
2
3
Therefore, there is a competition between sulfur escape
from the ®lm, and sulfur and oxygen reactivity with indium.
The overall process is controlled, not only by thermo-
dynamics but also by chemical kinetics.
Before any interpretation, it is very informative to plot the
evolution of the atomic ratio of S/In with the annealing
temperature. Fig. 7 shows that the shape of the two curves
obtained from EMPA and XPS is the same, there is only a
translation effect related to the fact that XPS measurements,
in order to avoid surface contamination effect, have been
performed after 1 min of etching, which minimizes the S
atomic concentration as its etching rate is higher than that
of indium.
a
When the annealing temperature, T , is smaller than 623 K,
Fig. 3. (a) Evolution of the S 2p peak with the etching time of a
PVD-®lm before annealing. (b) Evolution of the O 1s peak with the
etching time of a PVD-®lm before annealing.
the oxygen contamination is small, however, as shown by
XRD the ®lms are poorly crystallized i.e. many In±S bonds
are weak, and the sulfur escape from the ®lm is important and
overpass the expected boundary corresponding to In
end of the annealing process, we have b-In poorly crystal-
lized ®lms slightly contaminated with oxygen.
2 3
S . At the
shown that, after an etching of 1 min, in order to avoid the
surface contamination, there is 37 at.% of In, 18 at.% of S,
2 3
S
and 45 at.% of O, i.e. we have nearly In
2
(S
x
y
,O )
3
. It should
When the annealing temperature increases up to 623 K,
the crystalline properties of the ®lms increase, i.e. the
chemical reaction between the constituents is stronger, the
In±S bonds are stable and the atomic ratio of S/In increases.
At the same time, since the heating temperature is higher,
the oxygen reactivity increases and the oxygen contamina-
tion is higher. The sodium contamination, which originates
from the soda lime glass substrate increases. This last
phenomenon has been described earlier in the case of
be observed that, here also, the main contribution of the
oxygen corresponds to In±O bonds. The presence of the
OHis attested by IR measurements. The optical band gap
of these ®lms, which are n-type, has also been measured to
be about 2.90 eV.
All the obtained results and those reported in the literature
clearly show that CBD In S ®lms do not correspond to
2
3
2 3
pure In S . There is possibly a mixture of oxide and
hydroxyl but is dif®cult to attribute clearly the signal to
either contribution.
ternary chalcopyrites such as CuInSe [18,19]. However, it
2
should be noted that presently the sodium remains all over
the thickness of the ®lms, while it accumulates at the surface
in the case of the ternary compounds.
4
. Discussion
We have shown that probably some hydroxyl radicals are
bonded to indium in the ®lms. Contrarily to oxygen, which
First of all we will describe the reaction process involved
can be substituted by sulfur is a In
bonds stop the crystallite growth. This effect can explain
that when T is increased above 673 K there is no improve-
2 3
S crystalline, the OH
in the growth of the PVD In
oxygen/hydroxyl on the optical properties of the In
S
2 3
®lms, then the in¯uence of
®lms
S
2 3
a
will be discussed and ®nally some interpretation on the
variation of the electrical properties with the temperature
ment in the crystallization of the ®lms, although the grains
remain quite small (,100 nm). The only effect of the
increase in temperature is an increase in the oxygen con-
tamination, mainly at the surface of the ®lms.
a
of annealing, T , will be proposed.
Before annealing, it has been shown that there is some
initialization of the In±S chemical reaction. However, the
The growth of CBD ®lms is not discussed here since it has
been extensively studied by other authors [2,8,12,13].
After the discussion of the crystallization process, some
proposition could be performed to explain the band gap
value of the ®lms. First of all, it should be noted that the
®
lms are amorphous, they contain a large amount of excess
sulfur. Therefore, the ®lms consist in In±S bonds randomly
distributed in an amorphous sulfur matrix. The density of
dangling bonds and defaults is very high in such hetero-
geneous materials. It should be noted that the vapor pressure
of sulfur is very high, 1 Pa at 606 K and 10 Pa at 739 K, and
2 3
band gap value of b-In S is 2 eV [20]. As far as the PVD b-
In S ®lms are concerned, the values deduced from the
2 3

N. Barreau et al. / Solid State Communications 122 (2002) 445±450
449
Fig. 4. Evolution of the O 1s peak with the etching time of a PVD-
lm after annealing.
Fig. 6. IR absorption spectrum of a PVD-In
substrate for comparison.
2
S
3
®lm and of its naked
®
transmittance are about 2.9 eV [21]. It can be observed that,
when oxygen is introduced into the ®lms, the band gap
values are far higher than that of the single crystals. In the
case of CBD thin ®lms, systematically broadened gaps
Some others attribute the gap broadening to excess sulfur
[24] but our ®lms are sulfur de®cient. Since it has been
shown that some In±O bands are present in the ®lms, the
®lms can behave like an alloy as described by Hill [25].
If A and B are two semiconductor compounds, the gap
variation, when xA is present in (1 2 x)B, is given, in the
case of often used linear model, by:
(
.2.8 eV) are obtained.
2 3
Many studies have been made on CBD-In S ®lms, they
have shown that the deposited ®lms are in fact indium
hydroxyl sul®de thin ®lms. The XPS studies of these ®lms
Y ˆ E 2 ꢀE 2 E †x
ꢀ1†
B
B
A
[
13] show that there is some oxygen and some hydroxyl
bonded to indium. The CBD ®lms are sulfur de®cient and
there are more oxygens than sulfurs in the ®lms, with this
oxygen compensating the sulfur de®ciency. Therefore, XPS
characterization proves that ®lms are mainly composed of
indium hydroxyl sul®de, indium oxide, and indium sul®de.
The dif®culty is that the amount of hydroxide ion could vary
strongly from one sample to another without any coherent
relation to the deposition conditions.
where E ˆ E
ˆ 2:1 eV optical band gap value of
B
gIn2S3
In
In
2
S
3
, E ˆ E
ˆ 3:5 eV optical band gap value of
A
gIn2O3
2
O
3
, Y optical band gap value of the alloy A
Using Eq. (1), even if the most favorable condition is
x (12x)
B
.
taken with x ˆ 0:1; the value calculated is Y < 2:24 eV.
This value is far smaller than those obtained experimentally.
Hill [25] has shown that it is usual to describe the energy
gap of an alloy A
gap E and E as:
x
B
12x in terms of pure compound energy
When the PVD ®lms are concerned, different contributions
can be invoked to justify the broadening effect of the band
gap. Some authors have invoked a grain size effect [22]. A
blue shift effect of the optical properties can be induced by
very small grain size. As discussed by Yoshida et al. [23], to
be signi®cant the grain size should be 8 nm or less, while the
grain size of our ®lm estimated from the FWHM of the XRD
peak and visualized by SEM is about 20±30 nm, therefore
such quantization effect should be very small in our ®lms.
A
B
2
E
ꢀ
ˆ E 1 ꢀE 2 E 2 b†x 1 bx
ꢀ2†
x†
B
A
B
where b is the bowing parameter which induces a difference
between the linear average. If the linear evolution is noted
Elin and the evolution with the bowing parameter is noted
EHill, the difference DE between these two evolutions is
2
DE ˆ E 2 EHill ˆ bx 2 bx : As b $ 0 and x [ ‰0; 1Š;
lin
the difference DE is always negative. Therefore, this
Fig. 5. Evolution of the optical transmittance of PVD-®lms annealed
during 30 min at 673 K, the temperature rate increase being 3.3, 4.4,
Fig. 7. Variation of the atomic ratio S/In in the PVD-In
their annealing temperature.
2 3
S ®lms with
2
1
and 6.7 K min
.

450
N. Barreau et al. / Solid State Communications 122 (2002) 445±450
evolution, which is under linear, cannot a fortiori explains
2 3
totally the anomalously broad optical band gap of our In S
5. Conclusion
compounds, and hence another contribution should be taken
into account. We have shown that not only oxygen but also
Na is present in the ®lms. The substitution of indium by
We have shown that by using appropriate annealing
conditions PVD-In S ®lms can be grown with similar prop-
2
3
2 3
erties to that of CBD-In S ®lms. They have n-type conduc-
sodium in the In
gap [26], however, EMPA study has shown that the atomic
ratio %In/(%S 1 %O) is about 0.66. For the b-In crystal-
lizing under a defect spinel form, one can think that sodium
could be intercalated in the cation vacancies, however, such
a compound has never been found.
2
S
3
®lms should increase also the band
tivity, strong adherence to the substrate, high coverage
ef®ciency and a gap value of about 2:8±2:9 eV: Therefore,
they can be used as buffer layer in a whole PVD process of
thin ®lm solar cells. More generally it is very important to
understand, that, in the present ®lms, the oxygen contami-
nation has a positive effect on properties of the ®lm, to
obtain layers with properties, which match very well with
the speci®cations of solar cell buffer layers. Therefore,
sometimes, it is important going off the beaten tracks of
the knowledge which states that purity is a guarantee for
performing layer.
2 3
S
2 3
When the CBD-In S ®lms are concerned, the oxygen and
hydroxyl contents are far higher to justify the value of the
gap. In the case of PVD ®lms, if all the effects discussed
earlier can contribute to the gap broadening, the oxygen
atomic percentage is too small and the grain size too large
to justify the broadening effect alone.
Calculations have been performed using a scalar-
relativistic version of the k-space TBLMTO model with
the atomic-sphere approximation (ASA). The simulation
References
2 3
of oxygen introduction in the b-In S structure shows that
[
[
1] H.W. Schock, F. P®stterer, Renewable Energy World March±
April (2001) 75.
the oxygen states are not located near the band gap, however
oxygen/hydroxyl can modify the crystal ®eld with crystal
structure distortions and small sulfur position variations.
Such perturbations have been proved to induce large band
gap increasing in the case of other sulfur compounds and
calculations introducing such small deviations on the sulfur
bond length are under way, and therefore this new and excit-
ing fundamental ®eld remains open [27].
2] L. Stolt, et al., 13th European Photovoltaic Solar Energy
Conference, 23±27 October, 1995, p. 1451.
[3] M.A. Contreras, et al., Progr. Photov. Res. Appl. 7 (1999) 311.
[4] A. Ennaoui, et al., Solar Energy Mater. Solar Cells 67 (2001) 31.
[5] K. Kushiya, et al., Solar Energy Mater. Solar Cells 67 (2001) 11.
[6] T. Nakada, et al., Solar Energy Mater. Solar Cells 67 (2001) 255.
[7] A. Shimizu, et al., Thin Solid Films 361±362 (2000) 193.
[8] D. Hariskos, et al., Solar Energy Mater. Solar Cells 41/42
(1996) 345.
If we compare the CBD samples to the PVD ®lms, the
oxygen and hydroxyl contamination is smaller in the second
case, while the band gap values are similar. The major differ-
ence between the two kinds of ®lms consists in their crystal-
line properties, the crystallinity of PVD ®lms is far better than
that of CBD. The crystal ®eld of the ®rst must be more
perturbed than the latter, which can explain the greater in¯u-
ence of oxygen substituted by sulfur on the electronic structure
of the material. In addition to oxygen contaminants, there is
also sodium in the PVD ®lms. The in¯uence of such impurity
on the optical band gap of materials has never been reported in
the literature and is actually studied in our laboratory.
[9] D. Braunger, et al., Solar Energy Mater. Solar Cells 40 (1996) 97.
[
[
[
[
[
[
[
10] J.C. Bern eÁ de, N. Barreau, French Patent No. 7322.
11] N. Barreau, et al., Vacuum 56 (2000) 101.
12] C.D. Lockhande, et al., Thin Solid Films 340 (1999) 18.
13] R. Bayon, et al., Thin Solid Films 353 (1999) 100.
14] E. Gourmelon, et al., Vacuum 48 (1997) 509.
15] S.-H. Yu, et al., J. Am. Ceram. Soc. 82 (2) (1999) 457.
16] J.F. Moulder, et al., Handbook of X-ray Photoelectron Spec-
troscopy, J. Chastain (Ed.), Perkin-Elmer Corporation, 1992,
Minnesota, USA.
[17] J.F. Guillemoles, Thin Solid Films 361±362 (2000) 338.
[18] A. Rockett, et al., Thin Solid Films 372 (2000) 212.
The predominance of the heating rate on the properties of
the ®lms is clearly shown in Fig. 7, while the in¯uence of
the annealing duration is small [21]. When heating rate
increases, the gap is not as broad as discussed earlier because
of the fast crystallization. For a given heating rate, the heating
time itself in not important. When the heating rate is quick
enough, the In±S bonds are consolidated rapidly, before all
the excess sulfur escapes, which will prevent large oxygen
[
[
[
[
[
[
[
19] D. Braunger, et al., Thin Solid Films 361±362 (2000) 161.
20] W. Rehwald, G. Harbeke, J. Phys. Chem. Solids 26 (1965) 1309.
21] N. Barreau, et al., Phys. Status Solidi A 184 (1) (2001) 179.
22] E.B. Yous®, et al., Thin Solid Films 361±362 (2000) 183.
23] T. Yoshida, et al., Electrochem. Soc. Proc. 97±20 (1997) 37.
24] W.-T. Kim, C.-D. Kim, J. Appl. Phys. 60 (7) (1986) 2631.
25] R. Hill, J. Phys. C: Solid State Phys. 7 (1974) 521.
[26] R. Mauricot, Thesis of Doctorat, University of Nantes
(France), ISITEM, 1995.
contamination and the band gap is smaller. When T is
a
achieved, the bonds are stabilized, the oxygen contamination
does not increase and the band gap is stable.
[27] R. Robles, et al., Opt. Mater. (2001).