Vous êtes sur la page 1sur 6

TSF-29704; No of Pages 6

Thin Solid Films xxx (2011) xxxxxx

Contents lists available at SciVerse ScienceDirect

Thin Solid Films


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t s f

Extraction of optical properties of at and surface-textured transparent conductive


oxide lms in a broad wavelength range
J.A. Sap, O. Isabella , K. Jger, M. Zeman
Delft University of Technology, Laboratory of Photovoltaic Materials and Devices, Mekelweg 4, 2628 CD Delft, The Netherlands

a r t i c l e

i n f o

Article history:
Received 28 October 2010
Received in revised form 28 July 2011
Accepted 4 August 2011
Available online xxxx
Keywords:
Thin-lm solar cells
Transparent conductive oxide
Surface roughness
Scattering
Refractive index

a b s t r a c t
An accurate characterization method is developed to determine the refractive index of smooth and surfacetextured transparent conductive oxide (TCOs) lms. The properties are obtained from simultaneous tting of
simulated specular reectance/transmittance spectra to spectroscopic measurements for different polarizations and angles of light incidence. The simulations are based on a combination of physical models describing
dielectric function of TCO lms. Besides the refractive index also other material properties of TCO lms are
obtained, such as the band gap and free carrier absorption. A light scattering model is implemented into the
simulations to take into account the diffused part of the light scattered at randomly-textured surfaces of TCO
lms.
2011 Elsevier B.V. All rights reserved.

1. Introduction
Transparent conductive oxide (TCO) layers have become an
important part in a variety of consumer devices, such as at panel
displays and solar cells. In thin-lm silicon solar cells they serve as the
top transparent electrode and/or as intermediate layers to manipulate
light propagation in the cell. The TCO layers have to fulll several
stringent requirements, such as high optical transmission in the
spectrum of interest, low sheet resistance, temperature durability and
good chemical stability. In addition, the TCO layer has to be surfacetextured in order to enhance light absorption inside the solar cell due
to the scattering at internal rough interfaces [1]. In multi-junction
thin-lm silicon solar cells with different silicon-based absorbers the
high transparency is required in the wavelength region from 300 nm
to 1100 nm [2,3]. The accurate determination of TCOs optical
properties in such a broad wavelength region is very important. The
knowledge of the accurate properties is necessary for the optimization
of the solar cell structures using the optoelectronic device simulators
[4,5].
In order to determine the complex refractive index of the TCO lms
in a broad wavelength range an extraction method is developed in
which the simulated reectance and transmittance (R / T) spectra of
the TCO lms are tted on the spectroscopic measurements. The
simulation of the R / T spectra is based on the existing physical models
that describe the dielectric function of the TCO lms. From matching
the simulated and measured R / T spectra the model parameters can be

Corresponding author. Tel.: + 31 15 27 89546; fax: +31 5 26 22463.


E-mail address: O.Isabella@TUDelft.nl (O. Isabella).

extracted. This approach for extracting material properties has been


demonstrated in the past for different TCO materials. Mergel and Qiao
[6] obtained the properties of tin-doped indium oxide (ITO) layers by
tting only one transmittance spectrum per sample. This was
repeated for aluminium-doped zinc oxide (AZO) lms by Qiao et al.
[7]. Solieman and Aegerter [8] carried out simultaneous ts on
reectance and transmittance for ITO layers at normal incidence. In all
these cases the layers were optically at, i.e. the surface roughness
was small in comparison to the wavelength.
In this article we present an extraction method that is extended by
including a simultaneous t on a large amount of measured R / T
spectra, which enhances the accuracy of the extracted material
properties. Further a light scattering model is introduced in the
method allowing the characterization of surface-textured TCO layers.
The R / T spectra were measured with variable angle spectroscopy
(VAS) [9]. SCOUT simulation software [10] was used to simulate and
t the spectra in order to extract the material properties. Three
different TCO materials were characterized and treated with the
improved method: aluminium-doped zinc oxide (AZO) [11], tindoped indium oxide (ITO) [12] and uorine-doped tin oxide (FTO)
[13].
2. Modelling dielectric functions
The transmittance and reectance of a TCO layer can be simulated
with a set of physical models. Each model provides susceptibility as a
function of wavelength and is most accurate in a specic part of the
spectrum. Combining the separate models yields the dielectric
function of the TCO layer from which the refractive index in a broad
wavelength range can be determined. The wavelength dependent

0040-6090/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.tsf.2011.08.023

Please cite this article as: J.A. Sap, et al., Thin Solid Films (2011), doi:10.1016/j.tsf.2011.08.023

J.A. Sap et al. / Thin Solid Films xxx (2011) xxxxxx

reectance and transmittance are calculated according to this


refractive index. For TCOs, the Extended Drude model[14,15], the
O'LearyJohnsonLim (OJL) interband transitions model [16] and a
Brendel oscillator[17] are used. The scattering effects from the surface
are modelled with either a Bruggeman effective medium approach for
optically at layers or a roughness model based on scalar scattering
theory.
All physical models are effective in the entire considered
wavelength range. However, the underlying equations cause the
effects of each single model to be strong in a specic part of the
spectrum. Superposition of the physical models creates a combined
model that includes all effects that occur in the thin TCO lm. This
method of superposition has also been demonstrated by Ehrmann and
Reineke-Koch [18] in combination with ellipsometry measurements
of AZO thin lms.
2.1. Extended Drude model
The Drude model, proposed in 1900 [14,15], describes the
transport properties of electrons in materials and is used to model
free carrier absorption that is observed in the infrared (IR) part of the
spectrum. With the classical Drude formula the susceptibility of free
electrons is given by
=

p2
2

p2
2
2

+ i

p2
3 + 2





L H
cross

+
arctan

width
2

Valence band
~exp[(E-EV)/V]

EV

Conduction band
~exp[(E-EC)/C]

E0

EC

Energy

Fig. 1. Schematic representation of the density of states functions of the valence and
conduction band in the OJL interband transitions model.

The Urbach energy determines the shape of the exponential tail and is
a measure for the disorder in the material. The expressions by O'Leary
show that the DOS approaches the band edge when the Urbach
energy goes to zero. The DOS of the valence band (NV) is described by
as similar expression. These DOS functions are used to compute the
joint-density-of-states (JDOS) function with

J = NC ENV EdE

with frequency , damping factor and plasma frequency p. In the


extended Drude model the damping factor is not constant but a function
of frequency. This extension provides better results when tting doped
layers because it accounts for the effects of ionized impurity scattering
[6]. The frequency-dependent damping factor is given by
= L

Density of States

with high and low frequency damping constants H and L, crossover


frequency cross and width of the transition region, width. These four
parameters together with the plasma frequency p provide the
extended Drude model with a total of ve possible tting parameters.
2.2. O'LearyJohnsonLim (OJL) bandgap model
The empirical OJL model [16] describes the effect of interband
transitions on the dielectric function. This model is therefore most
inuential in the energy range near the bandgap energy of the TCO
material. The OJL model was originally developed for amorphous
materials. Even though TCO lms are amorphous or polycrystalline
[19], this model provided the closest tting results on the measured
spectra. In comparison, the TaucLorentz [20] and CampiCoriasso
[21] models provided worse ts. Since we are mainly interested in the
refractive indices of the TCO layers, a good t to the measured spectra
is vital. We therefore decided to work with the OJL model.
The OJL model uses expressions for the density-of-states (DOS) for
both the conduction and valence band. These DOS functions can be
represented by a parabolic function with tail states that exponentially
decay into the bandgap as schematically drawn in Fig. 1.
The DOS of the conduction band is then given by [16]
8
p

; E EC + C
EEC
>
p 3 = 2 >
>
<
2
2mC
s




3
NC =
>
2 3
1
EE


C
C
>
>
exp exp
; E<EC + C
:
2
C
2
2
where mC* represents the effective mass associated with the conduction band. EC is the disorderless band edge and C is the Urbach energy.

This JDOS function is then applied to determine the absorption


coefcient according to
2

= D J

where D2() is the optical transition matrix element [22]. The relation
for the absorption coefcient can be used to determine the imaginary
part of the refractive index through:

k =
4

This can be converted to the imaginary susceptibility. The Kramers


Kronig relations [23,24], that are valid for response functions in physical
systems, are then used to obtain the real part of this susceptibility. The
obtained complex susceptibility function can be added to those of the
other models for correct implementation of the OJL model. The OJL
expressions give also other material properties such as the Urbach
energy and the bandgap of the material E0 = EC EV.
2.3. Brendel oscillator
A Brendel oscillator [17] is added to improve the tting result in
the proximity of the bandgap. Brendel proposed a model that uses the
dielectric function of a damped harmonic oscillator with a Gaussian
distribution of resonance frequencies. This expanded version of a
harmonic oscillator can be used in the entire range of the spectrum
and improved the tting result on the measured R / T spectra. The
susceptibility is given by
h
i

x0
2p

1
2 2
dx
= p e
2
x 2 i
2

with the Gaussian distribution width and the resonance frequency


0.
2.4. Surface roughness
The effects of light scattering on specular reectance and transmittance are not contained in the dielectric function of the TCO material. To
incorporate these effects in the model a distinction is made between

Please cite this article as: J.A. Sap, et al., Thin Solid Films (2011), doi:10.1016/j.tsf.2011.08.023

J.A. Sap et al. / Thin Solid Films xxx (2011) xxxxxx

optically at layers, with a roughness much smaller than the wavelength


of the incident light, and rough layers. For optically at layers an effective
dielectric function is composed according to the Bruggeman effective
medium model [25]:
1f

p eff
h eff
+ f
= 0:
h + 2eff
p + 2eff

The effective dielectric function eff is related to the dielectric


functions of the two materials that make up the interface, h and p. The
parameter f represents the volume fraction. For higher roughness, the
decrease in specular reectance and transmittance (R / T) is described by
two relations that are based on scalar scattering theory. The etching
process to obtain rough surfaces may have caused thickness inhomogeneities along the thin lm. It is however considered that the thickness
changes smoothly along the sample in comparison with the surface
roughness such that these thickness inhomogeneities do not signicantly
inuence the R / T spectra on the measured spot. The drop in specular
reectance is then given by Eq. (9) that was proposed by Bennett and
Porteus [26]. In a similar way the drop in specular transmittance is given
by Eq. (10) that was developed by Carniglia [27].
 
 
4R cos 2
Rspec = R0 exp

 


2R 2
2
n1 cos1 n2 cos2
Tspec = T0 exp

10

In these equations, R0 and T0 are the reectance and transmittance


for an identical at layer, R is the root-mean-square (RMS) roughness
of the surface and 1 and 2 are the angles of the incident and
transmitted light respectively.
For correct implementation of Eqs. (9) and (10) in SCOUT, the
model for at layers has to be multiplied with the exponential terms
in Eqs. (5) and (6). SCOUT does not support the option to feed the
refractive index back as input for these equations. This problem is
partly circumvented by measuring the samples with the TCO layer
facing the incoming light. In this case n1 is equal to 1.0. The refractive
index of the TCO n2 is approximated as a linear function. The
parameters of this linear function are used as tting parameters. A
linear approximation is chosen because the SCOUT model only allows
the use of two tting parameters for the rough layer.
2.5. High frequency permittivity and layer thickness
The high frequency permittivity and layer thickness are also
taken into account by the simulation of the R / T spectra but are not
directly related to any of the models mentioned before. The dielectric
function of a semiconductor material as a function of energy levels out
to a constant and real value in the mid-infrared. This vertical offset can
be described with a constant dielectric function referred to as the high
frequency permittivity.
The thickness of the layer does not inuence the dielectric function
but describes the amount and distribution of the interference fringes
in the R / T spectra.
2.6. Calibration of the glass substrate
Glass is used as substrate carrier for the layers. SCOUT provides a
dielectric function of microscope slide glass over a wavelength range
of 2002000 nm. This dielectric function is composed of a number of
Brendel oscillators and is useful for rough estimations. For more
accurate characterization this standard glass model is calibrated to
match the Corning Eagle 2000 TM glass substrates that were used in the
experiments. This is done by tting the standard glass model on R / T
measurements of the Corning Eagle 2000 TM substrates.

2.7. Fitting procedure


The process of tting simulated R / T spectra to spectroscopic
measurements is schematically drawn in Fig. 2. The accuracy of the t
depends on the models used to provide the susceptibilities, i.
3. Experimental details
3.1. Sample preparation
The TCO materials that are characterized with SCOUT are AZO, ITO
and FTO. The AZO samples with a thickness of approximately 1 m are
RF-magnetron sputtered on Corning Eagle 2000 TM glass substrates.
The target that was used for sputtering contained 98 wt.% ZnO and
2 wt.% Al2O3. Randomly-textured surfaces are introduced by wet
chemical etching in a 0.5% HCl solution where the etching time
controls the roughness of the surface [11]. For AZO material, a batch of
nine samples is prepared with etching time ranging from 0 to 50 s. The
ITO sample with a thickness of approximately 500 nm is also
deposited with RF-magnetron sputtering [12]. The target used for
ITO sputtering was composed of 90 wt.% In2O3 and 10 wt.% SnO2. For
ITO no roughness is chemically induced on the surface and therefore
only optically at samples are available. The analysed FTO sample is a
standard Asahi U-type substrate where FTO is deposited using
atmospheric pressure chemical vapour deposition [13]. The thickness
of the FTO is approximately 800 nm and the RMS surface roughness
(R) is close to 40 nm. For FTO the roughness originates naturally from
the deposition process and therefore also no at FTO samples are
available. The carrier concentrations of all samples were determined
with Hall measurements and are reported in Table 1.
3.2. R / T measurements
The R / T measurements at different angles of illumination were
performed with a Perkin Elmer Lambda 950 spectrophotometer. An
accessory called Angular Reectance/Transmittance Analyzer (ARTA)
[9,28] was available to carry out the VAS measurements. The ARTA
contains automated detector and sample holder rotation stages.
Only the specular component of the transmitted or reected light is
measured, which requires the detector to be right behind the
sample for transmittance and at twice the angle of incidence of the
light for reectance measurements. R / T spectra are obtained with
ARTA over a wavelength range of 3001500 nm for AZO, ITO and FTO
at angles of incidence: 0, 15, 30, 45 and 60 and both for
p- and s-polarized light. The average is taken over positive and negative
Import
measured
spectra

Material
properties
Fit satisfying?
yes
no

R/T spectra
from Fresnels
equations

Adjust parameters of
the sub models

n=

= 1+ i
i

Fig. 2. Feedback loop for tting simulated R / T spectra on spectroscopic measurements.

Please cite this article as: J.A. Sap, et al., Thin Solid Films (2011), doi:10.1016/j.tsf.2011.08.023

J.A. Sap et al. / Thin Solid Films xxx (2011) xxxxxx

Table 1
Measured free carrier density of the AZO, ITO and FTO samples.
Material

Free carrier density


[1020 cm3]

AZO
ITO
FTO

2.9 0.3
6.5 0.4
1.6 0.2

angles of incidence to reduce misalignment errors. This gives a total of


seventeen spectra when taking into account that it is not possible to
measure reectance at 0 with ARTA and that there is no difference
between the transmittance at 0 for p- and s-polarized light.
3.3. Surface morphology characterization
Atomic force microscopy (AFM) was used to analyse the surface
morphology and measure the RMS surface roughness of the TCO lms.
An NT-MDT nTegra atomic force microscope was deployed. The
measurements were carried out in semi-contact mode at a frequency
scan of 1 line/s and resolution 512 512 pixels over areas of 5 5 m 2
and 10 10 m 2. The cantilever was a whisker type (NT-MDT NSC 05)
with a radius of curvature of 10 nm.
4. Results
For each TCO sample a simultaneous t is done on all seventeen
measured spectra. A subset of the tting results of the simulated spectra
from SCOUT on the R / T measurements is shown in Fig. 3 for the at ITO
sample at 45 angle of incidence and s-polarized light. For at layers the
Bruggeman effective medium approach is used to simulate the
scattering effects of the surface. There is a good agreement between
simulations and measurements. Tests have shown that this model is also
able to t R / T spectra of at AZO layers with comparable tting
accuracy. This model can thus be used to characterize different types of
TCO materials with at surfaces.
Similarly a subset of the results for rough TCO layers, at 45and
s-polarization, are presented in Fig. 4 for AZO and FTO where the
scattering model is used (Eqs. (9)(10)). The ts on the transmittance of
rough TCO layers are less accurate compared to the reectance. This is
mainly due to Eq. (10) that provides only limited accuracy as also
observed by Zeman et al. [29]. The loss of coherence of the light due to
scattering is in reality stronger than predicted by the model. To ensure a
good overall tting result a weight factor of 0.1 is applied to the t on the
transmittance spectra. This optimum value of 0.1 is determined
experimentally by closely monitoring the optimal t on reectance
spectra and the t on transmittance near the band gap energy and in the
NIR part of the spectrum where the scattering effects are less strong.

Fig. 3. Sub-set of tting results for at TCO layers.

Fig. 4. Sub-set of tting results for rough TCO layers.

The obtained refractive indices of the three TCO materials are


shown in Fig. 5ac. Note that for the AZO the refractive index is given
for both the at sample and the rough sample. Table 2 summarizes
other material properties used in the models obtained from the tting.
The good agreement between the obtained properties for rough and
at AZO layers is an indication that the model for rough interfaces is
working properly because the material properties are not expected to
change due to the etching process.

Fig. 5. Obtained refractive indices for (a) at and rough AZO, (b) ITO and (c) FTO samples.

Please cite this article as: J.A. Sap, et al., Thin Solid Films (2011), doi:10.1016/j.tsf.2011.08.023

J.A. Sap et al. / Thin Solid Films xxx (2011) xxxxxx

Table 2
Obtained tting parameters for the analysed AZO, ITO and FTO samples.

Bandgap energy
Urbach energy
Plasma frequency
Brendel oscillator
resonance frequency
Layer thickness
RMS roughness
Surface mix volume fraction

AZO

AZO (rough)

ITO

FTO (rough)

[eV]
[meV]
[eV]
[eV]

3.727
143.1
1.200
3.062

3.671
141.4
1.217
2.964

4.231
178.6
1.830
3.341

4.261
132.7
1.145
3.670

[nm]
[nm]
[]

1269

0.383

722.2
70.40

485.8

0.447

833.4
39.88

5. Verication
The bandgap energy of AZO was found to be approximately 3.7 eV.
This corresponds to values found in literature where the bandgap of
AZO is typically between 3.4 and 4.0 eV [7,30,31]. The same holds for
the bandgap of ITO where the obtained bandgap energy of 4.2 eV is
within the range of reported values (3.54.3 eV) [3234]. The
bandgap of FTO can be justied when looking at the transmittance
spectra in Figs. 3 and 4. The wavelength at which the transmittance of
FTO decreases to zero due to bandgap absorption is comparable to
that of ITO, which implies that the bandgap energies are comparable.
For the AZO sample this wavelength is longer implying that its
bandgap is lower than that of FTO. The obtained bandgap energies are
therefore in agreement with the shape of the measured transmittance
spectra.
Although the bandgap energies are in agreement with literature
and the shapes of the transmittance spectra, it must be pointed out
that the objective of this work is the determination of the refractive
index. For this purpose a set of models is combined with the ambition
to obtain an optimal t. The added Brendel oscillator may interfere
with the OJL bandgap model, which can cause the obtained bandgap
energies to deviate from the real values. This will, however, not
inuence the obtained refractive indices since they are directly related
to the accuracy of the t on the R / T spectra.
To quantify the accuracy and reproducibility of the TCO model the
tting was repeated ten times; each time with different and random
starting values of the tting parameters. The tting was done
completely automatic without intervention and below a deviation
threshold of 0.005 the tting was stopped. Fig. 6 presents the results
of this analysis for the most important parameters of the rough TCO
model. The standard deviation of the obtained value is for most
parameters within 5% meaning that the model is able to nd an
accurate and unique t. The model for at layers is considered to be
more accurate since it forms the basis for the rough layer model.

Fig. 7. AFM scans of (a) AZO after 10 s of etching, (b) AZO after 50 s of etching and (c)
Asahi U-type FTO. The obtained RMS roughness is 34 nm, 100 nm and 37 nm
respectively.
Fig. 6. Standard deviation of tting parameters obtained with the rough TCO model.

Please cite this article as: J.A. Sap, et al., Thin Solid Films (2011), doi:10.1016/j.tsf.2011.08.023

J.A. Sap et al. / Thin Solid Films xxx (2011) xxxxxx

The implemented model for rough interfaces allows also the characterization of randomly textured TCOs. The obtained material properties
for rough AZO lms are close to the properties of at AZO lms.
Furthermore the obtained roughness is comparable to AFM measurements from which the conclusion can be drawn that the model for rough
TCO is working properly. Besides rough AZO the model is also tested for
Asahi U-Type FTO with similar accuracy. The simultaneous t on
seventeen spectra provides a unique and accurate solution despite the
large amount of tting parameters. This addition make the composed
SCOUT interface a valuable tool for characterizing TCO lms with high
accuracy in a broad wavelength range.
References

Fig. 8. Comparison of the modelled RMS roughness with AFM measurements for all nine
AZO samples and the FTO sample. The dashed line is the ideal agreement between
measurements and modelling outcome.

With the aid of AFM measurements the RMS surface roughness


obtained with the tting was veried. Fig. 7ac shows the surface
texture of the AZO samples with an etching time of 10 and 50 s and
the FTO sample.
The RMS roughness obtained from the AFM analysis is compared
with that obtained from the models for all nine AZO samples and the
FTO sample. This comparison is presented in Fig. 8. This graph shows
that for surface roughness up to 80 nm there is a good agreement
between the modelled and measured results. This agreement also
veries the correct implementation of the model for rough surfaces.
The sample with 50 s of etching shows a larger deviation. More
research is therefore required to analyse the agreement for samples
with surface roughness larger than 80 nm.
6. Conclusions
Optical characterization with variable angle spectroscopy is an
accurate method to determine the optical properties of TCO lms.
Optical measurements were performed over a broad wavelength range
(3001500 nm) with a Lambda 950 spectrophotometer equipped with
the ARTA accessory. The measurements, performed at different angles
and polarizations, provide seventeen R / T spectra per sample. A
mathematical model is tted simultaneously on these spectra. The
basic model for optically at layers showed the possibility to
characterize AZO and ITO lms and obtain, amongst other properties,
the refractive index of these lms.

[1] M. Zeman, in: J. Poortmans, V. Archipov (Eds.), Thin Film Solar Cells: Fabrication,
Characterization and Applications, Wiley, 2006, p. 173.
[2] J. Yang, B. Yan, S. Guha, Thin Solid Films (2005) 162.
[3] M. Yamaguchi, T. Takamoto, K. Araki, N. Ekins-Daukes, Sol. Energy 79 (2005) 78.
[4] M. Burgelman, J. Verschraegen, S. Degrave, P. Nollet, Prog. Photovoltaics Res. Appl.
12 (2004) 143.
[5] M. Zeman, J. Krc, J. Mater. Res. 23 (2008) 889.
[6] D. Mergel, Z. Qiao, J. Phys. D: Appl. Phys. 35 (2002) 794.
[7] Z. Qiao, C. Agashe, D. Mergel, Thin Solid Films 496 (2006) 520.
[8] A. Solieman, M.A. Aegerter, Thin Solid Films 502 (2006) 205.
[9] P.A. van Nijnatten, Thin Solid Films 442 (2003) 74.
[10] W. Theiss, Hard and Software for Optical Spectroscopy, Dr.-Bernhard-Klein-Str.110,
D-52078 Aachen, Germany, , 2002http://www.wtheiss.com.
[11] M. Berginski, J. Hpkes, M. Schulte, G. Schpe, H. Stiebig, B. Rech, J. Appl. Phys. 101
(2007) 074903.
[12] J.B. Chu, H.B. Zhu, B. Xu, Z. Sun, Y.W. Chen, S.M. Huang, , 2008, p. 1266.
[13] K. Sato, Y. Gotoh, Y. Wakayama, Y. Hayashi, K. Adachi, N. Nishimura, Rep. Res. Lab.,
42, Asahi Glass Co. Ltd., 1992, p. 129.
[14] P. Drude, Ann. Phys. 306 (1900) 566 (in German).
[15] P. Drude, Ann. Phys. 308 (1900) 369 (in German).
[16] S.K. O'Leary, S.R. Johnson, P.K. Lim, J. Appl. Phys. 82 (1997) 3334.
[17] R. Brendel, D. Bormann, J. Appl. Phys. 71 (1992) 1.
[18] N. Ehrmann, R. Reineke-Koch, Thin Solid Films 519 (2010) 1475.
[19] T. Minami, Semicond. Sci. Technol. 20 (2005) S35.
[20] G.E. Jellison Jr., Thin Solid Films 313314 (1998) 33.
[21] D. Campi, C. Coriasso, J. Appl. Phys. 64 (1988) 4128.
[22] L. Ley, in: J.D. Joannopoulos, G. Lucovsky (Eds.), Topics of Applied Physics,
Springer, New York, 1984, p. 61.
[23] H.A. Kramers, Nature 117 (1926) 775.
[24] R.L. De Kronig, J. Opt. Soc. Am. 12 (1926) 547.
[25] D.A.G. Bruggeman, Ann. Phys. 24 (1935) 636 (in German).
[26] H.E. Bennett, J.O. Porteus, J. Opt. Soc. Am. 51 (1961) 123.
[27] C.K. Carniglia, Opt. Eng. 18 (1979) 104.
[28] K. Jger, O. Isabella, L. Zhao, M. Zeman, Phys. Status Solidi C (2010) 945.
[29] M. Zeman, R.A.C.M.M. van Swaaij, J.W. Metselaar, R.E.I. Schropp, J. Appl. Phys. 88
(2000) 6436.
[30] T.J. Coutts, D.L. Young, X. Li, Mater. Bull. 25 (2000) 58.
[31] T. Ratana, P. Amornpitoksusk, T. Ratana, S. Suwanboon, J. Alloys Compd. 470
(2009) 408.
[32] I. Hamberg, C.G. Granqvist, J. Appl. Phys. 60 (1986) R123.
[33] R.C. Hayward, D.A. Saville, I.A. Aksay, Nature 404 (2000) 56.
[34] Y. Kim, Y. Park, S.G. Ansari, J. Lee, B. Lee, H. Shin, Surf. Coat. Technol. 173 (2003)
299.

Please cite this article as: J.A. Sap, et al., Thin Solid Films (2011), doi:10.1016/j.tsf.2011.08.023

Vous aimerez peut-être aussi