In situ X-ray photoelectron spectroscopy study of the oxidation of CuGaSe2

Article abstract of DOI:10.1016/j.susc.2005.01.054

The thermal and native oxidation of CuGaSe₂ thin films was studied by in situ X-ray photoelectron spectroscopy (XPS). The special design of the XPS chamber allowed to measure XP-spectra under oxidizing gas atmospheres at pressures of up to 5 mb

Full text of DOI:10.1016/j.susc.2005.01.054

Surface Science 580 (2005) 80–94
www.elsevier.com/locate/susc
In situ X-ray photoelectron spectroscopy study
of the oxidation of CuGaSe2
a
a
a,
a
b
R. Wurz , M. Rusu , Th. Schedel-Niedrig *, M.Ch. Lux-Steiner , H. Bluhm ,
¨
b
M. Havecker , E. Kleimenov , A. Knop-Gericke , R. Schlogl
b
b
b
¨
¨
a
Hahn–Meitner Institut Berlin, Abteilung SE2, Glienicker Strasse 100, 14109 Berlin, Germany
Fritz–Haber-Institut der Max Planck Gesellschaft, Faradayweg 4–6, 14195 Berlin, Germany
b
Received 19 July 2004; accepted for publication 27 January 2005
Abstract
The thermal and native oxidation of CuGaSe2 thin films was studied by in situ X-ray photoelectron spectroscopy
(XPS). The special design of the XPS chamber allowed to measure XP-spectra under oxidizing gas atmospheres at pres-
sures of up to 5 mbar (in situ) or in ultra high vacuum (UHV). During thermal oxidation, the formation of predom-
inantly Ga2O3 and some amount of SeO2 were observed, but no copper oxides could be detected in the near surface
region of the thin films. The same oxides were found after native oxidation in air under ambient conditions. Only after
long term native oxidation for longer than 4 months Cu(OH)2 was detected. An additional sodium oxide compound
formed at the thin film surface, NaxO and Na2CO3 after thermal and native oxidation, respectively. The amount of
these sodium oxide compounds depends on the Na content on the as prepared surface. The formation of SeO2 under
humid conditions at 100 ꢀC was found to depend on the surface composition of the thin film.
ꢁ 2005 Elsevier B.V. All rights reserved.
Keywords: CuGaSe2; Sodium; Chalcopyrite; Oxidation; Photoelectron spectroscopy, polycrystalline surfaces
1. Introduction
of Eg = 1.68 eV at room temperature and optical
absorption coefficient larger than 104 cmÀ1 for
photon energies hm > 1.7 eV, CuGaSe2 is a promis-
ing absorber for thin-film solar cell devices. Up to
now energy conversion efficiencies of 9.7% and
9.5% have already been achieved for single-crystal
[1] and thin-film cells [2], respectively. A pro-
nounced degradation of ZnO/CdS/CuGaSe2 solar
cells was found, when the CuGaSe2 absorber films
CuGaSe2 belongs to the class of I–III–VI2 semi-
conducting chalcopyrites. With its high bandgap
*
Corresponding author. Tel.: +49 30 8062 2392; fax: +49 30
8062 3199.
E-mail address: schedel-niedrig@hmi.de (Th. Schedel-Nie-
drig).
0039-6028/$ - see front matter ꢁ 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2005.01.054

R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
81
were exposed to air before deposition of CdS [3].
Exposure to humid air before CdS deposition even
accelerates the degradation process in case of
Cu(In,Ga)Se2 solar cells [4]. It is concluded, that
oxidation in air deteriorates the solar cell perfor-
mance. Despite these observations, no oxidation
studies of CuGaSe2 could be found in literature.
Only CuInSe2 and Cu(In,Ga)Se2 are the most
investigated materials from the copper containing
class of chalcopyrites. Surface sensitive XPS is ap-
plied in order to study, which oxide phases form
during the oxidation process. During thermal oxi-
dation of CuInSe2 crystals and thin films Kazmer-
ski et al. [5] found formation of mainly In2O3 and
some SeO2, but no copper oxide. Further a CuxSe
transition layer was found between the oxide layer
and CuInSe2 [5]. The native oxidation of CuInSe2
takes place in the same way [6,7]. Only at higher
humidity some amount of CuxO was detected at
the surface of the thin native oxides [6]. When oxi-
dizing Cu(In,Ga)Se2 films [4,8], predominantly
Ga2O3, In2O3 and some amount of SeO2 formed,
but no CuxO was observed.
resulting gases GaCl, GaCl2, GaCl3 and H2Se then
reacted on the substrate with the Cu precursor
layer at a temperature of 450 ꢀC to CuGaSe2 (for
details see [9]). In a second annealing step the
HCl flux was reduced to allow fine tuning of the
thickness and the [Ga]/[Cu]-ratio of the resulting
CuxGaySe2 films [9]. Alternatively CuGaSe2 thin
films were grown on Mo coated soda lime glass
substrates by halogen supported chemical vapor
deposition (CVD) in an open tube system. Poly-
crystalline binary source materials Cu2Se and
Ga2Se3—synthesized from the elements—were
subjected to continuous flows of H2/I2 and H2/
HCl, respectively, and volatilized due to chemical
reactions at a source temperature of 600 ꢀC and
at a reactor pressure of 100 mbar. The resulting
gases CuI, GaClx (x = 1,2,3) and H2Se were in-
jected into the substrate zone where they were
mixed and cooled down to 500 ꢀC leading to chem-
ical reactions into the solid phase [10]. By adjust-
ing the flow rates [11] or by a two-stage process
[12] the [Ga]/[Cu]-ratio of the resulting CuxGaySe2
films can be controlled.
In this paper we will study the formation of
thermal and native oxides on CuGaSe2 thin film
surfaces with XPS. In contrast to the thermal oxi-
des we observe formation of a Cu(OH)2 phase at
the surface of thin films after native oxidation
for longer than 4 months. Details of the oxidation
process as a function of the composition of the
films will be presented. Spectra measured in gas
atmosphere will be compared with those measured
afterwards in vacuum. The results will be discussed
in the framework of a simplified model for the oxi-
dation process.
Selected samples were transported in inert Ar
gas atmosphere from the reactor to the XPS setup
to avoid oxidation and to get reference spectra of
as prepared samples. Only during mounting onto
the sample holder and transfer into the chamber,
these samples were exposed to air for about
20 min.
XPS measurements were carried out with syn-
chrotron radiation at beamline U49/2-PGM1 at
BESSY II in Berlin [13] using a high-pressure X-
ray photoelectron spectrometer that allows mea-
surements at pressures of up to 5 mbar [14]. In
addition a conventional X-ray photoelectron spec-
troscopy setup with a MgKa X-ray source was
used. In the synchrotron-based measurements the
incident photon energy for excitation was chosen
such that the kinetic energy of the emitted photo-
electrons was about 300 eV, i.e. the information
depth was comparable in all spectra. The photon
energy was calibrated by measuring the Fermi en-
ergy of the sample holder (stainless steel). The C1s
signal from adventitious carbon with a binding en-
ergy EB = 285.0 eV was also used for energy refer-
encing. The combined energy resolution of the
beamline and the electron analyzer was ꢀ0.2 eV
2. Experimental
Thin polycrystalline films of CuGaSe2 were
grown by chemical close-spaced vapor transport
(CCSVT) [9]. The soda lime glass substrates were
covered with a 750 nm thick Mo layer and a
250 nm thick Cu layer by electron beam evapora-
tion. Polycrystalline Ga2Se3—synthesized from
the elements—was subjected to a continuous flow
of H2/HCl and volatilized due to chemical reac-
tions at a source temperature of 550 ꢀC. The

82
R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
at a pass energy of 10 eV. All spectra were normal-
ized by the incident photon flux, which was mea-
sured using a photodiode with known quantum
efficiency. A linear background was subtracted
from the spectra.
to ensure that an oxide had formed. The native
oxides were grown by storing the films in air under
ambient conditions at room temperature for
several months. The bulk chemical composition
and the thickness of the films was determined by
X-ray fluorescence spectrometry (XRF) using a
Philips MagicsPRO spectrometer. Details of the
samples studied here and the various treatments
are listed in Table 1.
The thermal oxides were grown in situ in dry
oxygen at a pressure of 0.5 mbar or in humid oxy-
gen adding water vapor with a partial pressure of
0.2 mbar inside the high-pressure cell of the XPS
setup. During the XPS measurements the incident
photon beam is admitted to this cell through a sil-
icon nitride window with a thickness of 100 nm
that separates the high pressure region in the cell
from the UHV environment in the beamline. The
sample surface is placed close to a small aperture
(1 mm diameter) that is the entrance to a three-
stage differentially pumped electrostatic transfer
lens. Electrons that escape through the first aper-
ture are subsequently focussed onto two additional
apertures downstreams before reaching a conven-
tional hemispherical analyzer. This three-stage dif-
ferentially pumped electrostatic lens system allows
to measure samples at pressures of several mbar,
while the hemispherical analyzer is kept at pres-
sures of 10À8 mbar [14]. Using this high-pressure
spectrometer XPS measurements were performed
during the oxidation of the thin films in the gas
atmosphere at different oxidation temperatures
Tox, as well as afterwards in vacuum. Before start-
ing the XPS measurements in gas atmosphere, the
samples were exposed to the gas for about 30 min
3. Results
Before oxidation of as prepared thin films (sam-
ple #1 and #2) their contamination was checked,
as they were exposed to air during mounting on
the sample holder. Oxygen and carbon adsorbed
on the film surface were found to be the main con-
taminants. The peak positions of the Cu2p3/2
,
Ga3d and the Se3d5/2 level of 932.2 eV, 19.4 eV
and 53.7 eV, respectively, correspond to the litera-
ture data of as grown samples [15]. Therefore we
can assume the surface of the as prepared samples
essentially to be free of any oxide phases contain-
ing Cu, Ga or Se. This is also supported by the
Auger spectra (not shown). In Table 2 the relative
composition data of the near surface region, ob-
tained by an XPS line analysis, as well as that of
the bulk, determined by XRF, are listed. The
XPS line area analysis was done by using the core
level photoionization cross-sections from Yeh and
Table 1
CuGaSe2 samples used for XPS measurements and their oxidation process together with the oxidation temperature Tox, oxidation
time, gas atmosphere and gas pressure during oxidation
Sample
#1
[Ga]/[Cu]
Growth method
Oxide
Tox (ꢀC)
Oxidation time
Gas atmosphere
Pressure (mbar)
1.032 0.025 herCmCaSl VTT100
90 min
Humid O
0.7
0.7
0.7
2
200
300
85 min
70 min
O2
O2
#2
1.160 0.028 herCmCaSl VTT100
90 min
Humid O
70 min
0.7
0.7
2
200
O2
#1a
#2a
#3
1.032 0.025
1.160 0.028
1.079 0.026
1.079 0.026
1.033 0.025
CCSVTNative
CCSVTNative
CVD
25
25
6 months
6 months
Ambient air
Ambient air
1013
1013
Native
Native
Native
25
25
25
4 months
10 months
16 months
Ambient air
Ambient air
Ambient air
1013
1013
1013
#3a
#4
CVD
CVD
Here ‘‘humid O2’’ means 0.5 mbar O2 and 0.2 mbar H2O. The bulk [Ga]/[Cu]-ratio was determined by XRF. Samples with the same
number stem from the same substrate.

R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
83
Table 2
Surface composition of as prepared CuGaSe2 films, obtained by XPS line intensity analysis and bulk composition, determined by XRF
and HI-ERDA, compared to literature data [15]
Sample
Growth
method
Cu
(XRF)
Ga
(XRF)
Se
(XRF)
Na
(HI-ERDA)
Cu2p
(XPS)
Ga2p
(XPS)
Se3d
(XPS)
Na1s
(XPS)
a
#1
#2
CCSVT1
CCSVT1
CVD
1.03
1.16
1.98
2.16
0.04
0.04
1
1
1
1
2.64
4.34
2.22
2.86
5.84
6.53
4.00
5.43
1.75
1.36
0.2
a
[15]
[15]
1
1
1.02
1.08
2.03
2.10
0.0024b
0.0024b
CVD
0.6
All data are given relative to the Cu content, which is set to one. Relative error for XRF and XPS values is 5% and ꢁ10%, respectively.
a
The Na bulk concentration of 1 at.% [9] was related to the Cu content set to 1.
The Na bulk concentration of 0.06 at.% [15] was related to the Cu content set to 1.
b
Lindau [16] and assuming, that the product of the
analyzer transmission function and the inelastic
mean free path is constant for our spectrometer
at the utilized kinetic energies and pass energies.
All XPS data in Table 2 are given relative to the
Cu content (set to one) with a relative error of
ꢁ10%. The surface composition (XPS) of sample
#1 and #2 strongly deviates from the bulk compo-
sition (XRF) thus showing a Cu depletion. The
existence of a Cu-depleted near surface region for
stoichiometric and Ga-rich grown GuGaSe2 films
(see Table 2, [9,15]) as well as stoichiometric and
In-rich CuInSe2 films [17] is well known from liter-
ature. The bulk Na content of CCSVT thin films is
far below the detection limit of XRF. It was deter-
mined by heavy ion elastic recoil detection analysis
(HI-ERDA) to an amount of 1 at.% [9], which cor-
responds to a value of 0.04 compared to Cu (see
Table 2). Thus, the surfaces of sample #1 and #2
are Cu-depleted and enriched in Ga, Se and Na.
Cu2p3/2
(a) as prepared
(b) 4 months
Cu2+
(c) 6 months
Cu2+
(d) 16 months
CuGaSe2
935 930
945
940
Binding Energy (eV)
Fig. 1. Cu2p3/2 spectra of CuGaSe2 thin films after native
oxidation, (a) #2 as prepared, (b) #3 after 4 months, (c) #2a
after 6 months and (d) #4 after 16 months oxidation. All
spectra were measured in vacuum at 25 ꢀC, in (a) and (b) with
synchrotron radiation (hm = 1234 eV), in (c) and (d) with MgKa
radiation (hm = 1253.6 eV). The spectra were normalized to get
similar intensities for the Cu2p3/2 peak.
3.1. Native oxidation
EB = 934.8 eV connected with a satellite band that
consists of two lines at EB ꢀ 941 eV and 944 eV
(Fig. 1c and d). These satellite peaks are a charac-
teristic feature for paramagnetic Cu2+ [18]. Since
XPS is extremely sensitive to small changes in
the local binding configuration, it is possible to
distinguish between the various Cu oxidation
products that might exist after oxidation. For
instance, the Cu2p3/2 peak of Cu(OH)2 is shifted
in comparison to CuO by about 0.6 eV to
higher binding energy [19], which can be reliably
detected in the XP-spectra. From the analysis of
the spectra as plotted in Table 3, we find that the
In the following the effect of native oxidation in
air under ambient conditions is shown. The Cu2p3/2
spectra of thin films after different times of native
oxidation are plotted in Fig. 1. The as prepared
film only reveals the typical Cu2p3/2 peak of
CuGaSe2 with EB = 932.2 eV [15]. After native
oxidation for 4 months a shift to EB = 932.6 eV is
observed (Fig. 1b). A shift of the Cu2p3/2 peak to
higher binding energies was also found after ther-
mal oxidation of CuInSe2 and was attributed to
the formation of a Cu2ÀxSe phase [5]. With further
oxidation time a second structure appears at

84
R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
Table 3
XPS binding energy for the Cu2p3/2 peak, its satellite band, difference of both, kinetic energy of the Cu LVV Auger peak and Auger
parameter given in eV (with uncertainty 0.2 eV) measured at oxidized CuGaSe2 samples compared to literature data for Cu(OH)2,
CuO, Cu2O, Cu0, CuGaSe2 and CuCO3
Sample
Cu2p3/2
Satellite band
Difference
Cu(LVV)
a
#2a
#4
934.1
917.3
934.8
940.9
940.6b
941.6b
943.9
943.4b
944.0b
6.1
9.1
9.0b
10.2b
915.9a
916.3c
918.2c
917.9d
916.7c
917.4d
919.0c
918.6d
917.0f
916.3d
1850.7
1850.7
1852.0
1851.4
1849.2
1849.6
1851.5
1851.0
1849.2
1851.3
Cu(OH)2
CuO
934.4b
933.8b
933.5d
932.5b
932.2d
932.5b
932.4d
932.2e
935.0d
6.2b
7.8b
Cu2O
Cu0
CuGaSe2
CuCO3
a
Beside the CuGaSe2 related peak at Ekin = 917.2 eV the Auger spectrum contained a shoulder at 915.9 eV attributed to Cu(OH)2.
b
c
d
e
f
Ref. [19].
Ref. [20].
Ref. [21].
Ref. [15].
Ref. [24].
experimentally determined values for the binding
energy of the Cu2p3/2 line are in better agreement
with what has been reported for Cu(OH)2 rather
than to CuO, in particular with regard to the dis-
tance of the satellite peaks to the Cu2p3/2 peak.
From this observation we propose, that native oxi-
dation of copper at the surface of CuGaSe2 films
results in a Cu(OH)2 phase.
CuGaSe2
Ga3d
Ga2 O3
(a) as prepared
(b) 4 months
The corresponding Ga3d spectra after different
times of native oxidation are shown in Fig. 2.
The as prepared sample exhibits only the typical
Ga3d peak of CuGaSe2 with a binding energy
EB = 19.4 eV [15]. A shift of the line to EB =
20.1 eV is observed after native oxidation. This is
indicative of Ga2O3 in the near-surface region
(EB = 20.4 eV) [22]. The Ga L3M45M45 Auger
spectra (not shown here) also support the forma-
tion of Ga2O3. The spectra of the oxidized samples
show a peak with a kinetic energy Ekin = 1062.4 eV
characteristic for Ga2O3 [23] beside the CuGaSe2
related peak at Ekin = 1065.6 eV [24].
The native oxidation process also involves Se as
can be seen from the Se3d spectra of the films
(solid curves in Fig. 3). The spectrum of the as pre-
pared sample consists of the well known Se3d
doublet structure (Fig. 3a) due to the spin–
orbit splitting of the 3d electrons into a 3d5/2
(c) 6 months
(d) 16 months
22
24
20
18
Binding Energy(eV)
Fig. 2. Ga3d spectra of CuGaSe2 thin films after native
oxidation, (a) #2 as prepared, (b) #3 after 4 months, (c) #2a
after 6 months and (d) #4 after 16 months oxidation. All
spectra were measured in vacuum at 25 ꢀC, in (a) and (b) with
synchrotron radiation (hm = 318 eV), in (c) and (d) with MgKa
radiation (hm = 1253.6 eV). The spectra were normalized to get
similar intensities for the Ga3d peak.
(EB = 53.7 eV) and a 3d3/2 level (EB = 54.6 eV).
After an oxidation time of 4 months a peak at
EB = 58.8 eV arises that increases with increasing

R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
85
mental Selenium Se0 (EB = 55.1 eV [26]) and/or a
Cu2ÀxSe phase. For CuInSe2 it is well known, that
a Cu2ÀxSe phase forms at the interface between a
binary In2O3 oxide phase and the ternary chalco-
pyrite bulk phase [5]. Kazmerski et al. [5] found,
that the Cu2p and Se3d binding energy for the
Cu2ÀxSe phase were shifted by 0.6 eV and 0.7 eV
to higher values compared to CuInSe2, respec-
tively. Similar shifts for copper selenide compared
to CuInSe2 were also found by other groups
[27,28].
Se3d
Int. x 20
CuGaSe2
(a) as prepared
CuxSe, Se0
(b) 4 months
SeO2
Na2s
(c) 6 months
(d) 16 months
The Se3d spectrum of sample #2a (Fig. 3c)
clearly shows the Na2s peak at EB = 63.0 eV after
native oxidation. In the spectrum of the corre-
sponding as prepared sample #2 (Fig. 3a) this
Na2s peak can only be seen when enlarging the
spectrum by a factor of 20. As both spectra are
measured with different photon energies and nor-
malized to equal intensities for the Se3d peak,
the intensities of both spectra cannot be compared
directly. Nevertheless we observe an increase of the
Na2s peak after oxidation also by taking into con-
sideration the different information depths and the
different photoionization cross-sections. This in-
crease of the sodium content at the surface after
native oxidation is a hint, that sodium is also in-
volved in the oxidation process. To find out, which
sodium compound forms during oxidation, the
XPS energy values for different Na containing
compounds are listed in Table 4 besides the
64
62
60
58
56
54
52
Binding Energy(eV)
Fig. 3. Se3d spectra of CuGaSe2 thin films after native
oxidation, (a) #2 as prepared, (b) #3 after 4 months, (c) #2a
after 6 months and (d) #4 after 16 months oxidation. All
spectra were measured in vacuum at 25 ꢀC, in (a) and (b) with
synchrotron radiation (hm = 354 eV), in (c) and (d) with MgKa
radiation (hm = 1253.6 eV). The dotted lines show fit curves. The
spectra were normalized to get similar intensities for the Se3d
peak.
oxidation time and can be assigned to SeO2 [25].
Simultaneously, the Se3d peak has broadened
and can be fitted by two sets of doublets, namely
the doublet related to the CuGaSe2 phase as well
as by a second doublet shifted by 0.9 eV to higher
binding energies. This second doublet structure at
EB = 54.9 eV is attributed to the formation of ele-
Table 4
XPS binding energy values for the Na1s, Na2s, C1s and O1s peak and kinetic energy values for the Na KLL Auger peak and the
corresponding Auger parameter a for the Na compound on oxidized CuGaSe2 thin film surfaces compared to some possible sodium
compounds
Sample or compound
Na1s
EB/eV
Na(KLL)
Ekin/eV
a/eV
C1s
EB/eV
O1s
EB/eV
Na2s
EB/eV
#1, as prepared
1072.0
1071.9
1072.2
1072.0
1071.5
1071.5
1071.5
1071.3
1072.5
1070.8
1072.3
531.4
531.5
531.7
531.4
531.5
531.4
531.6
63.3
63.5
63.8
63.6
63.0
63.0
#1, thermal oxidation at 200 ꢀC
#1, thermal oxidation at 300 ꢀC
#3, native oxidation (4 months)
#1a, native oxidation (6 months)
#2a, native oxidation (6 months)
289.2
289.2
288.4
288.5
289.4
290.0
989.9
989.9
989.8
989.8
989.8
991.0
2061.4
2061.4
2061.3
2061.1
2062.3
2061.8
a
Na2CO3
NaHCO3
a
Na2Oa
Na2SeO3
Na on Cu(In,Ga)Se2
529.7
531.7
64.2
a
b
a
Values from Ref. [31].
Values for oxygen passivated Na on air exposed Cu(In,Ga)Se2 from Ref. [32].
b

86
R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
measured values for our oxidized CuGaSe2 sam-
ples. The comparison of these binding energies
suggests the formation of sodium carbonate,
Na2CO3, during native oxidation of CuGaSe2 sur-
faces. Na2CO3 was also found after native oxida-
tion of Cu(In,Ga)Se2 thin films [29,30]. We
exclude the formation of Na2SeO3, which was
found by other groups [33–35], as our measured
values better fit to the values reported for Na2CO3
(see Table 4). An estimation of the surface compo-
sition of the oxidized samples obtained by XPS
line intensity analysis is shown in Table 5. The in-
crease in the sodium content at the surface is more
pronounced for the CCSVTsamples, which also
show a higher sodium content at the as prepared
surface (see Table 2 and Fig. 3, the spectrum of
the CCSVTsample #2a in Fig. 3c clearly shows
the Na2s peak in contrast to that of the CVD sam-
ple #4 in Fig. 3d). The sample with the highest so-
dium surface content after native oxidation (#1a)
neither showed formation of Ga2O3, SeO2 nor
Cu(OH)2 (not shown). The surface composition
of this sample is close to the nominal value one ex-
pects for Na2CO3 (see Table 5). From this we con-
clude, that a protective carbonate layer of Na2CO3
formed on the surface of sample #1a, which pre-
vents further oxidation of the other elements.
We conclude, that during a 4 months native oxi-
dation process binary oxides, predominantly
Ga2O3 and some amount of SeO2 were formed.
Further we observed formation of Cu2ÀxSe and/
or Se0, but not of any copper oxide phase. Only
after oxidation for longer than 4 months we were
able to detect Cu(OH)2 in the near-surface region.
Additionally Na2CO3 was detected. The higher the
sodium content of the as prepared surface, the
higher is the amount of Na2CO3 and the lower is
the amount of the other oxides.
3.2. Thermal oxidation
We will now focus on the thermal oxidation
process studied in situ with synchrotron radiation.
After XPS measurement of its untreated as pre-
pared surface the stoichiometric sample (#1) was
oxidized in situ under humid oxidizing conditions
Table 5
Surface composition in at.% of as prepared and oxidized CuGaSe2 thin films estimated from XPS line intensity analysis compared to
nominal composition of different oxide compounds; for carbon only the XPS peak at 288.5 eV related to Na2CO3 was included; the
values for the measurements with MgKa radiation reveal a big error bar due to different information depths for the different peaks
Sample
Oxide
X-ray-source
tox/months
Cu2p
Ga2p
Se3d
Na1s
O1s
C1s
#1
#2
As prepared
As prepared
BESSY
BESSY
7.2
5.9
18.9
25.7
41.9
38.7
12.5
8.1
19.5
21.6
0.0
0.0
#1
#2
#1
#2
#1
Thermal 100 ꢀCa
Thermal 100 ꢀCa
Thermal 200 ꢀCb
Thermal 200 ꢀCb
Thermal 300 ꢀCc
BESSY
BESSY
BESSY
BESSY
BESSY
0.5
1.1
0.3
0.4
0.1
6.9
7.5
3.4
7.9
3.7
3.8
2.3
25.5
19.4
32.9
26.5
17.9
56.4
49.7
54.0
59.8
64.7
7.2
14.5
0.2
8.8
5.5
4.0
0.0
14.9
#1ad (CCSVT)
#2a (CCSVT)
#3 (CVD)
Native
Native
Native
Native
Native
MgKa
MgKa
BESSY
MgKa
MgKa
6
1.0
1.9
4.9
4.1
5.7
0.5
3.2
1.5
4.5
27.2
13.0
2.1
53.0
59.0
53.8
67.3
67.1
16.9
18.5
1.9
6
4
17.6
10.6
3.8
19.7
6.9
#3a (CVD)
#4 (CVD)
10
16
4.0
5.9
7.2
10.3
7.2
Ga2O3
40.0
25.0
60.0
50.0
50.0
50.0
Na2O/Ga2O3
Na2SeO3
Na2CO3
a
25.0
33.3
33.3
16.7
16.7
The spectra were measured in humide O2 atmosphere during thermal oxidation at 100 ꢀC.
The spectra were measured in gaseous O2 atmosphere during thermal oxidation at 200 ꢀC.
The spectra were measured at 300 ꢀC in vacuum after the thermal oxidation was finished.
In the spectra of sample #1a neither Ga2O3, SeO2 nor Cu(OH)2 was detected.
b
c
d

R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
87
(p(O2) = 0.5 mbar and p(H2O) = 0.2 mbar) at
100 ꢀC, and in dry oxygen atmosphere at 200 ꢀC
and 300 ꢀC (see Table 1). The surface concentra-
tion of the different elements was determined from
the spectra measured in vacuum at Tox after the
oxidation was finished. These surface concentra-
tions normalized to the Cu content of the as pre-
pared surface (set to one) are shown in Fig. 4.
The Cu content (squares in Fig. 4) strongly de-
creases with oxidation temperature Tox. After oxi-
dation no shift in binding energy was observed for
the Cu2p3/2 and Cu(L3VV) Auger peak (not
shown) compared to the as prepared surface. The
Se content at the surface (open triangles in Fig.
4) also decreased with Tox. Only traces of SeO2
were found in the spectra measured in vacuum
(see Figs. 7 and 8). The Ga content at the surface
(open circles in Fig. 4) increased with Tox parallel
to the oxygen content (filled triangles in Fig. 4).
The higher the oxidation temperature Tox, the
more the Ga3d peak shifts to higher binding ener-
gies reaching EB = 20.3 eV after oxidation at
300 ꢀC. This binding energy corresponds to the va-
lue reported for Ga2O3 (EB = 20.4 eV) [22] indicat-
ing the oxidation of Ga. The Ga L3M45M45 Auger
spectra (not shown here) support the formation of
Ga2O3. A feature at Ekin = 1062.5 eV characteris-
tic for Ga2O3 [23] grows up with Tox. The Na con-
tent at the surface (rhombus in Fig. 4) also
increases with Tox and then slightly decreases at
Tox = 300 ꢀC. The binding energy of EB =
1072.0 eV for the Na1s level remained unchanged
with Tox. The increase in intensity parallel to that
of oxygen suggests the formation of a sodium
oxide compound at the surface. Comparing the
measured binding energies with that of different
sodium oxide compounds (Table 4) suggests the
formation of Na2CO3 like in case of the native oxi-
des. But Na2CO3 cannot be the main sodium oxide
compound, as the surface concentration of carbon
(stars in Fig. 4) does not increase with Tox like
sodium. Further in the C1s spectra (not shown)
the peak at EB ꢀ 289.2 eV characteristic for
Na2CO3 shows the highest intensity at Tox = 100
and then decreases with increasing Tox (see Table
4). The estimated surface composition of the stoi-
chiometric sample (#1) after oxidation at
Tox = 300 ꢀC is 0.1% Cu, 14.9% Ga, 2.3% Se,
17.9% Na, 64.7% O and 0.0% C (see Table 5).
These values better fit to a mixture of Na2O and
Ga2O3 with a nominal surface composition of
25% Na, 25% Ga and 50% O. Therefore for the
thermal oxide we propose the formation of NaxO
and only small amounts of Na2CO3.
The XPS spectra of the Ga-rich sample (#2)
after thermal oxidation (not shown) reveal the
same changes at the surface as found for the stoi-
chiometric sample #1. The surface composition
after thermal oxidation at 200 ꢀC resembles that
of sample #1 (see Table 5). The Cu and Se content
at the surface exponentially decreases with Tox,
whereas the Na, Ga and O content increases
(Fig. 4). This suggests the formation of an oxide
layer consisting of NaxO and Ga2O3 covering the
CuGaSe2 film surface. In this case the O1s signal
should contain contributions from Ga2O3 and
NaxO. The O1s spectra of sample #1 after different
oxidation steps are shown in Fig. 5a. The peak
maximum shifts from EB = 531.4 eV to 531.7 eV
when increasing the oxidation temperature Tox.
The O1s spectra can be fitted by three peaks with
Gaussian profile (Fig. 5b) located at EB =
532.9 eV, 531.8 eV and 530.7 eV. The change of
peak area of these three components in depen-
dence of Tox is illustrated in Fig. 6. The component
O1s
101
100
Na1s
Ga2p
Se3d
C1s
10-1
Cu2p
0
100
200
300
TOX / °C
Fig. 4. Surface composition of the stoichiometric sample (#1)
as prepared (Tox = 25 ꢀC) and after thermal oxidation at
different temperatures Tox. The relative surface concentration
of the different elements was determined from Cu2p3/2 (square),
Ga2p (circle), Se3d (open triangle), Na1s (rhombus), O1s (filled
triangle) and C1s (star) peak areas of sample #1 measured in
vacuum at Tox after the oxidation was finished. The values are
normalized to the Cu content of the as prepared surface
(Tox = 25 ꢀC), which is set to one.

88
R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
531.8
as prepared
after oxidation at
100°C
O1s
O1s
530.7
thermal oxide
200°C
200°C
300°C
532.9
535
534
533
532
531
530
529
535
534
533 532
531
530
529
Binding Energy (eV)
(b)
Binding Energy (eV)
(a)
Fig. 5. (a) O1s spectra of the stoichiometric sample (#1) as prepared (Tox = 25 ꢀC, bold solid curve) and after thermal oxidation at
100 ꢀC (solid curve), 200 ꢀC (dashed curve) and 300 ꢀC (dotted curve) measured with synchrotron radiation (hm = 830 eV) in vacuum at
Tox after oxidation was finished. (b) Fit of the O1s spectrum measured at Tox = 200 ꢀC (dashed curve in (a)) with three Gaussian
profiles.
Na1s peak (rhombus in Fig. 4). Therefore we attri-
bute this third peak to NaxO. The binding energy
10-8
EB = 1072.0 eV for the Na1s peak of the as pre-
pared and thermally oxidized samples (see Table
4) corresponds to the value reported by Heske
et al. [32] for an oxygen passivated Na specie at
the surface of air exposed Cu(In,Ga)Se2 thin films
(see Table 4 and [32]). The formation of Na2O can-
not directly be concluded from our experiments as
we unfortunately did not measure the Na(KLL)
spectra of the thermally oxidized samples to deter-
mine the Auger parameter and as the energy values
for Na1s, Na2s and O1s peak all deviate from the
reported values for Na2O (see Table 4). The O1s
spectra of the as prepared thin film (Tox = 25 ꢀC
in Fig. 6) also contains the three different oxygen
components, revealing that formation of Ga2O3
and NaxO already takes place after exposure of
the thin film surface to ambient air for some min-
utes (during mounting on the sample holder).
From the peak area of the different components
we get quantitative information about the oxide
phases. Assuming that all Na atoms of the as pre-
pared surface of sample #1 are oxidized, we esti-
mate a composition of Na1.7O for the sodium
oxide compound, which is close to the nominal va-
lue for Na2O. From the oxygen peak related to
Ga2O3 we estimate that 7% of the Ga atoms of
the as prepared surface are oxidized.
chemisorbed OH- (532.9eV)
Ga O (531.8eV)
10-9
Nax2O3(530.7eV)
0
100
200
300
TOX (°C)
Fig. 6. Peak area of the three oxygen components (shown in
Fig. 5b) obtained by fitting the O1s spectra of the stoichiometric
sample (#1) measured in vacuum at Tox after thermal oxidation
was finished; component located at EB = 532.9 eV (triangle),
EB = 531.8 eV (circle) and EB = 530.7 eV (rhombus).
with the lowest intensity located at EB = 532.9 eV
(triangle in Fig. 6) is attributed to chemisorbed
hydroxide, OHÀ [36], as it strongly increases after
the oxidation step under humid oxygen atmo-
sphere at Tox = 100 ꢀC. The peak located at
EB = 531.8 eV (circle in Fig. 6) is connected to
Ga2O3 as its binding energy is close to the litera-
ture value of EB = 531.5 eV for Ga2O3 [37] and
as it increases monotonously with Tox like the
Ga2p peak (see Fig. 4). The area of the peak
located at EB = 530.7 eV (rhomb in Fig. 6) first
increases with Tox and then decreases at Tox
300 ꢀC. The same behavior was observed for the
=
From the exponential decrease of the Cu2p3/2
peak area (Fig. 4) it is possible to estimate the

R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
89
thickness of the complete thermal oxide layer.
˚
With an inelastic mean free path of 8 A for photo-
electrons with Ekin = 300 eV [38] we estimate a
thickness of 2 nm for the oxide layer after the oxi-
dation at 300 ꢀC.
other doublet shifted by 1.1 eV to higher binding
energies. This can be attributed to elemental Sele-
nium, Se0, which is reported to have EB(Se3d5/2) =
55.1 eV [26]. In the spectrum measured in vacuum
(solid curve in Fig. 7a) only a small amount of the
SeO2 peak and the Se0 related doublet remains
whereas the CuGaSe2 related doublet increased.
Hence, the heat treatment of the stoichiometric
sample (#1) in vacuum at 100 ꢀC strongly changes
the Se3d spectra. This is in strong contrast to the
Se3d spectra of the Ga-rich sample (#2) measured
under similar conditions (Fig. 7b). A fit of the
spectra in Fig. 7b with the line shape of the un-
treated sample reveals that it also contains the
Se0 related doublet with EB(Se3d5/2) = 54.9 eV be-
side the CuGaSe2 related doublet. But the forma-
tion of SeO2 was not observed on the surface of
the Ga-rich sample (#2) at 100 ꢀC.
Thermal oxidation at 200 ꢀC leads to the forma-
tion of SeO2 on the surface of the Ga-rich sample
(Fig. 8a) with the same amount as for the stoichi-
ometric sample (Fig. 8b). Its intensity increases
compared to the 100 ꢀC spectra and in relation
to the main peak. The spectra at 200 ꢀC also
contain the Se0 related doublet at EB(Se3d5/2) =
54.7 eV. Under vacuum conditions at 200 ꢀC both
the SeO2 peak and the shifted Se0 related doublet
strongly diminish while the CuGaSe2 related dou-
blet increases (Fig. 8b). After thermal oxidation of
As the XPS chamber allowed to measure spec-
tra under gaseous conditions, they will now be
compared with those measured under vacuum
conditions. For this purpose the spectra measured
under gaseous conditions, which show lower signal
intensity, are normalized. As the line shape of the
Ga3d spectra essentially remained unchanged all
spectra measured under gaseous conditions were
multiplied with a factor obtained by normalizing
the Ga3d spectra. After this normalizing proce-
dure the Cu2p3/2 spectra essentially remained un-
changed (not shown). The main changes were
observed in the Se3d spectra. At 100 ꢀC the Se3d
spectrum of the stoichiometric sample (#1) mea-
sured in humid oxygen differs from that measured
in vacuum (Fig. 7a). The spectrum detected in hu-
mid oxygen (bold solid curve in Fig. 7a) clearly
shows a SeO2 peak at EB = 58.9 eV. A fit of the
Se3d feature measured in humid O2 (bold curve
in Fig. 7a) with the Se3d doublet spectrum of the
untreated sample (dotted curves in Fig. 7a) reveals
that beside the CuGaSe2 related doublet at
EB(Se3d5/2) = 53.7 eV the spectrum contains an-
in humid O2
Se3d
in vacuum
Cu Se
T= 200°C
xCuGaSe2
Se3d
fit
Se0
T = 100°C
in O2
in vac.
fit
SeO2
(a) Ga-rich
Int. x 6.0
SeO2
(a) stoichiometric
Int. x 7.2
(b) stoichiometric
Int. x 3.4
T= 300°C
Int. x 2.85
Se0 CuGaSe2
(b) Ga-rich
(c) stoichiometric
60
58
56
54
52
Binding Energy (eV)
60
58
56
54
52
Binding Energy (eV)
Fig. 7. Se3d spectra measured at Tox = 100 ꢀC with synchro-
tron radiation (hm = 354 eV) in humid oxygen (bold solid
curves) and in vacuum (solid curves) after exposure to humid
oxygen of (a) stoichiometric sample #1 and (b) Ga-rich sample
#2. The dotted lines are fits for the curves in humid oxygen. The
solid curve in (b) is shifted up to better compare it to the bold
solid curve.
Fig. 8. Se3d spectra measured with synchrotron radiation
(hm = 354 eV) in O2 (bold solid curves) and in vacuum (solid
curves) after oxidation was finished of (a) Ga-rich sample #2 at
200 ꢀC and (b) stoichiometric sample #1 at 200 ꢀC and (c)
sample #1 at 300 ꢀC. The dotted lines are fits for the curves in
O2-atmosphere.

90
R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
the stoichiometric sample (#1) at 300 ꢀC no SeO2
peak appears neither in the spectrum measured
in O2 atmosphere nor in vacuum (Fig. 8c). The
spectrum measured in O2 atmosphere almost over-
laps with that acquired in vacuum. The doublet
structure of the Se3d peak becomes broader. The
spectrum measured in O2 atmosphere (bold solid
curve in Fig. 8c) can be fitted with the CuGaSe2 re-
lated doublet and another doublet shifted by
0.4 eV to higher binding energies. This shifted dou-
blet at EB(Se3d5/2) = 54.1 eV can be attributed to
Cu2ÀxSe [5].
We conclude, that thermal oxidation of
CuGaSe2 results in the formation of an oxide layer
consisting of sodium oxide and Ga2O3 at the
CuGaSe2 thin film surface. Additionally traces of
Na2CO3 are observed. In the spectra measured un-
der gaseous conditions SeO2 and Se0 are found,
which both strongly decrease during thermal
treatment in vacuum. After oxidation at 300 ꢀC a
component attributed to CuxSe was revealed. For-
mation of copper oxides was not detected.
Usually, they are covered by ultra thin layers of
hydrocarbons, oxides often also by water (OH
groups) [39].
According to the photoemission results we pro-
pose a simple model for the native oxidation of
CuGaSe2 as shown in Fig. 9. The surface of the
as prepared thin films shows a copper-depleted
surface composition of about CuGa3Se5 (Fig. 9a
and Table 2). Additionally sodium polyselenide
Na2Sex (x > 2) is assumed to be at the surface,
which is supposed to serve as Se-reservoir during
growth of CuGaSe2 [33,34]. This Na2Sex forms
during growth of CuGaSe2 by reaction of H2Se
from the reacting gas phase with Na+ diffusing
out from the soda lime glass substrate through
the Mo layer. The native oxidation already starts
during short exposure of this surface to air for
some minutes during mounting on the sample
holder. The Na2Sex is immediately oxidized to
Na2O and small amounts of Ga2O3 are formed.
During the further exposure of the surface to air
the growth of Ga2O3 proceeds and the Na2O is
transformed by the CO2 from ambient air to the
thermodynamically more stable Na2CO3. Hence
the oxidation process proceeds with the growth
of the thermodynamically most stable Na2CO3
and Ga2O3 (see Table 6 and [40]). The amount
of Na2CO3 in relation to Ga2O3 is the higher,
the higher the Na content on the as prepared sur-
face is. As the CCSVTsamples exhibit a higher Na
content at the as prepared surface (Table 2), the
Na2CO3 content after oxidation is also higher
compared to the CVD samples (Table 5). In case
of very high sodium content on the surface
4. Discussion
Generally, the driving force for the growth of
oxide layers on metal and semiconductor solid
state surfaces is the minimization of the surface
free-energy of the system, since the surface free
energies of oxides are much lower than for the
metals or semiconductors. Thus, metals and semi-
conductors are always covered by a substance
which lowers the free surface energy of the system.
CuGa3Se5, Na2Sex
Na2CO3,Ga2O3,SeO2
CuxSe, Se0
Cu(OH)2
NaxO, Ga2O3, SeO2
CuxSe, Se0
Na2CO3,Ga2O3,SeO2
CuxSe, Se0
CuGaSe2
as prepared
(a)
CuGaSe2
CuGaSe2
CuGaSe2
thermal oxid.
(d)
native oxid.
tox < 4 months
(b)
native oxid.
tox > 4 months
(c)
Fig. 9. Model for the oxidation of CuGaSe2 (a) as prepared and not exposed to air, (b) after short term native oxidation
(tox < 4 months), (c) after long term native oxidation (tox > 4 months) and (d) after thermal oxidation.

R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
91
Table 6
Standard molar enthalpy (heat) of formation DfH0 at 298.15 K in kJ/mol of different compounds [40–42]
Comp.
Na2CO3
1130.7
Ga2O3
1089.1
Na2SeO3
990.6a
Cu(OH)2
449.8
Na2O
414.2
SeO2
225.4
Cu2O
168.6
CuO
Cu2Se
65.26b
CuSe
39.5
ÀDfH0
157.3
a
From Ref. [41].
From Ref. [42].
b
Na2CO3 can even prevent oxidation of the
other elements (see #1a in Table 5). This is in
accordance with the results of Braunger et al.
[34] who found a lower amount of Ga2O3 and
In2O3 on Cu(In,Ga)Se2 layers grown on soda lime
glass compared to layers grown on Na free sub-
strates. After formation of Na2CO3 and Ga2O3
the remaining Cu and Se readjust at the interface
to form elemental Se0 and an interfacial Cu2ÀxSe
layer (Fig. 9b). Additionally, SeO2 is formed. Dur-
ing a ‘‘long term’’ oxidation process of more than
4 months the amount of this SeO2 phase increases
(Fig. 9c) and Cu2+(OHÀ)2 is formed as fourth
oxide phase. The small Cu+ ions of the bulk diffuse
from the bulk CuxSe/oxide or CuGaSe2/oxide
interface towards the surface of the already exist-
ing native oxide layer consisting of Na2CO3,
Ga2O3 and SeO2. This copper transport through
the oxide layer is mediated via cationic vacancies
either inside the bulk of the native oxide layer or
in the grain boundaries [43]. At the surface of
the native oxide these Cu+ ions react with the hy-
droxyl groups of the adsorbed water to form an
outwardly growing Cu2+(OHÀ)2 layer. According
to the higher standard molar enthalpy (heat) of
formation of Cu(OH)2 compared to SeO2 (see
Table 6) one would expect that the growth of
Cu(OH)2 already occurs during short term oxida-
tion process before SeO2 is formed, which was
not observed here for the thin films. The Cu deple-
tion of as grown film surfaces accompanied by a Se
rich surface composition favors the growth of
SeO2 instead of the formation of Cu(OH)2.
served on the as prepared surface (see Figs. 5
and 6), followed by the formation of SeO2 (Fig.
9d). By the Se3d spectra measured in gaseous
atmosphere (bold solid curves in Figs. 7 and 8) it
was possible to observe the different oxidation
states of Se. We conclude, that Se2À of the chalco-
pyrite (EB(Se3d5/2) = 53.7 eV) is stepwise oxidized
via metallic Se0 (EB(Se3d5/2) = 54.8 eV) to Se4+ of
SeO2 (EB = 58.8 eV). Thermal treatment in vac-
uum (solid curves in Figs. 7 and 8) removes the
oxide and only traces of SeO2 and Se0 could be de-
tected. This can be explained by sublimation of
SeO2 and Se0 solid state phase at T P 100 ꢀC
due to the high vapor pressures of SeO2 and
Se0.1 At 300 ꢀC the high vapor pressure of
335 mbar for SeO2 [42] even exceeds that of the
gas atmosphere with p(O2) = 0.7 mbar. Therefore,
the SeO2 related feature was not found in the spec-
tra measured at 300 ꢀC neither in gas atmosphere
nor in vacuum (Fig. 8c). Finally the Cu2ÀxSe re-
lated doublet remained in the Se3d spectra mea-
sured at 300 ꢀC (Fig. 8c).
The formation of SeO2 and Se0 was found to
depend on the surface composition and the tem-
perature. In contrast to the stoichiometric sample
#1, the growth of SeO2 is hindered on the surface
of the Ga-rich sample #2 during thermal oxidation
at 100 ꢀC (Fig. 7). At 200 ꢀC the thermal oxide of
the Ga-rich surface has the same composition like
that of the stoichiometric sample (see Table 5).
The hindered formation of SeO2 on the Ga-rich
surface can be explained by its higher Ga concen-
tration. The high amount of Ga favors the oxida-
tion of Ga before Se is oxidized, in accordance
with the enthalpy of formation (see Table 6). At
The detailed analysis of the growth mechanism
of thermal oxide layers on CuGaSe2 surfaces was
enabled by the possibility to measure PE-spectra
in situ, i.e. under gaseous conditions. The oxide
growth starts with the formation of a sodium
oxide compound NaxO and Ga2O3, the two ther-
modynamically stable compounds already ob-
1
The vapor pressure of SeO2 at 100, 200 and 300 ꢀC is
8 · 10À4, 2 and 335 mbar, respectively [42]. The vapor pressure
of Se at 100 ꢀC, 200 ꢀC and 300 ꢀC is 3 · 10À8, 2 · 10À3, and
0.3 mbar, respectively [42].

92
R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
200 ꢀC most of the Ga atoms at the surface are al-
ready oxidized so that Se is the next candidate for
oxidation and then the oxidation process is com-
parable to that of sample #1.
the main recombination path in our CuGaSe2 so-
lar cells [47], the different interface of the oxidized
layer compared to that of the virgine layer is pro-
posed to be one reason for the degradation of solar
cells. The formation of Cu(OH)2 was observed
only after native oxidation for longer than
4 months, whereas the degradation of solar cells
already occurs within few hours [3] or days [4].
Hence we exclude the Cu(OH)2 to be responsible
for the degradation of solar cells [45].
The main difference of the thermal in contrast
to the native oxidation process is the different oxi-
dation behavior of sodium and copper. Neither
Cu(OH)2 nor copper oxides have been found after
thermal oxidation. That no CuO or Cu2O formed
can be explained by the fact, that these oxides are
less stable than Ga2O3 and SeO2 (see Table 5). The
Cu depletion of thin film surfaces further inhibits
growth of copper oxides. Further the formation
of Cu(OH)2 at elevated temperatures is improba-
ble, as Cu(OH)2 is dehydrogenated at T > 100 ꢀC
[44] and can be easily reduced to Cu2O by anneal-
ing at T > 100 ꢀC [45,46]. The other difference be-
tween thermal and native oxides is, that sodium
reacts to Na2CO3 during native oxidation, whereas
on thermal oxides mainly NaxO and only small
amounts of Na2CO3 were detected. This can be ex-
plained by fact that during thermal oxidation CO2
from ambient air is missing, so that only NaxO can
be formed. Even the high amount of adsorbed
hydrocarbons found on the as prepared samples
(see Fig. 4) cannot induce formation of Na2CO3
during thermal oxidation. The hydrocarbons are
oxidized to CO2, which desorbs from the surface
at higher temperatures, explaining the decrease of
the C1s signal during thermal oxidation (stars in
Fig. 4).
In the following we want to discuss the implica-
tions of the native oxide products on solar cells. As
Na2CO3 can be easily be solved in water, it is re-
moved during the treatment of the oxidized absor-
ber in the CdS bath [29]. The formation of
Na2CO3 was the higher, the higher the Na content
on the as prepared surfaces was. The carbonate
layer can even prevent the oxidation of Ga, Se
and Cu (see #1a). Hence one can think of using
Na2CO3 as protective capping layer in solar cell
processing. This has to be proved in further exper-
iments. The oxidation products Ga2O3 and SeO2
can be solved by the ammonia of the CdS bath
[29]. The same holds for Cu(OH)2 [45]. The critical
point is, that the new interface CdS/CuGaSe2 after
removing of the native oxide layer differs from that
of a virgine film. As the interface recombination is
5. Conclusion
The novel design of our XPS setup enables us
to monitor in situ the growth of thin oxides on
CuGaSe2 thin films. The surface of as prepared
CuGaSe2 thin films is found to be copper depleted
and enriched in Ga, Se and Na compared to the
bulk composition. This favors the growth of so-
dium oxide compounds, Ga2O3 and SeO2 when
exposing the surface to oxidizing atmospheres dur-
ing thermal as well as native oxidation. The main
difference between the thermal and native oxida-
tion process is the oxidation behavior of Cu and
Na. Thermal oxides do not exhibit copper oxides,
whereas native oxides show the formation of
Cu(OH)2 after oxidation in air under ambient con-
ditions for longer than 4 months. During native
oxidation Na at the film surface reacts to Na2CO3,
whereas during thermal oxidation in oxygen atmo-
sphere mainly NaxO and only a small amount of
Na2CO3 are formed. The higher the amount of
the oxidic sodium compounds on the surface, the
lower is the amount of other oxides. The different
oxidation states of Se on the surface have been
monitored by measuring XPS under oxidizing con-
ditions. Exposure of the oxidized film surface to
ultra high vacuum at T P 100 ꢀC sublimates the
oxidation products SeO2 and Se0.
Acknowledgments
We thank Rolf Follath for technical support at
the BESSY II beamline U49/2 PGM1. Thanks to
Jo¨rg Reichardt and Alexander Grimm for per-
forming XPS measurements with MgKa radiation,

R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
93
Steiner, C. Heske, E. Umbach, J. Phys. Chem. Sol. 64
(2003) 1553.
´
to A. Meeder and D. Fuertes Marron for prepara-
tion of thin films. This work was supported by
the Bundesministerium fur Wirtschaft Deutsch-
¨
land through its CCSVT-project (contract No.
0329740B), which is gratefully acknowledged.
[16] J.J. Yeh, I. Lindau, Atom. Data Nucl. Data Tables 32
(1985) 1.
[17] D. Schmid, M. Ruckh, F. Grunwald, H.W. Schock, J.
Appl. Phys. 73 (1993) 2902.
[18] L. Fiermans, R. Hoogewijs, J. Vennik, Surf. Sci. 47 (1975)
1.
[19] N.S. McIntyre, M.G. Cook, Anal. Chem. 47 (1975)
2208.
References
[20] N.S. McIntyre, S. Sunder, D.W. Shoesmith, F.W. Stanc-
hell, J. Vac. Sci. Technol. 18 (1981) 714.
[1] M. Saad, H. Riazi-Nejad, E. Bucher, M.C. Lux Steiner, in:
1st World Conference on Photovoltaic Energy Conversion,
IEEE, Hawaii, 1994, p. 214.
[21] C.D. Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder, in:
G.E. Muilenberg (Ed.), Handbook of X-ray Photoelectron
Spectroscopy, Perkin Elmer, Eden Prairie, Minnesota,
1978.
[2] D.L. Young, J. Keane, A. Duda, J.A.M. AabuShama,
C.L. Perkins, M. Romero, R. Noufi, Prog. Photovolt.:
Res. Appl. 11 (2003) 535.
[22] G. Leonhardt, A. Berndtsson, J. Hedman, M. Klasson, R.
Nilsson, C. Nordling, Phys. Stat. Sol. 60 (1973) 241.
[23] C.D. Wagner, Discuss. Faraday Soc. 60 (1975) 291.
[24] D. Schmid, M. Ruckh, H.W. Schock, Appl. Surf. Sci. 103
(1996) 409.
[3] V. Nadenau, D. Hariskos, H.W. Schock, in: H. Ossen-
brink, P. Helm, H. Ehmann (Eds.), 14th EC Photovoltaic
Solar Energy Conference, Stephens & Associates, Barce-
lona, 1997, p. 1250.
[4] D. Hariskos, G. Bilger, D. Braunger, M. Ruckh, H.W.
Schock, Inst. Phys. Conf. Ser. No. 152: Section E: Surfaces
and Interfaces, ICTMC-11, 1997, p. 707.
[25] M.K. Bahl, R.L. Watson, K.J. Irgolic, Anal. Chem. 51
(1979) 466.
[26] M.K. Bahl, R.L. Watson, K.J. Irgolic, J. Chem. Phys. 72
(1980) 4069.
[5] L.L. Kazmerski, O. Jamjoum, P.J. Ireland, S.K. Deb, J.
Vac. Sci. Technol. 19 (1981) 467.
[27] D. Cahen, P.J. Ireland, L.L. Kazmerski, F.A. Thiel, J.
Appl. Phys. 57 (1985) 4761.
[6] L.L. Kazmerski, O. Jamjoum, J.F. Wagner, P.J. Ireland,
K.J. Bachmann, J. Vac. Sci. Technol. A 1 (1983) 668.
[7] A.J. Nelson, S. Gebhard, L.L. Kazmerski, E. Colavita, M.
Engelhardt, H. Ho¨chst, Appl. Phys. Lett. 57 (1990) 1428.
[28] E.P. Domashevskaya, V.V. Gorbachev, V.A. Terekhov,
V.M. Kashkarov, E.V. Panfilova, A.V. Shchukarev, J.
Electron Spectrosc. Relat. Phenom. 114–116 (2001) 901.
[29] A. Kylner, J. Electrochem. Soc. 146 (1999) 1816.
[30] M. Ruckh, D. Schmid, M. Kaiser, R. Scha¨ffler, T.R.
Walther, H.W. Schock, Sol. Energy Mater. Sol. Cells
41/42 (1996) 335.
¨
[8] I. Dirnstorfer, W. Burkhardt, W. Kriegseis, I. Osterreicher,
H. Alves, D.M. Hofmann, O. Ka, A. Polity, B.K. Meyer,
B. Braunger, Thin Solid Films 361–362 (2000) 400.
´
[9] M. Rusu, S. Wiesner, D. Fuertes Marron, A. Meeder, S.
Doka, W. Bohne, S. Lindner, Th. Schedel-Niedrig, Ch.
Giesen, M. Heuken, M.Ch. Lux-Steiner, Thin Solid Films
451 (2004) 556.
[31] J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben,
Handbook of X-ray Photoelectron Spectroscopy, Perkin-
Elmer Corp., Eden Prairie, Minnesota, 1992.
[32] C. Heske, G. Richter, Z. Chen, R. Fink, E. Umbach, W.
Riedl, F. Karg, J. Appl. Phys. 82 (1997) 2411.
[33] M.R. Balboul, U. Rau, G. Bilger, M. Schmidt, H.W.
Schock, J.H. Werner, J. Vac. Sci. Technol. A 20 (2002)
1247.
[10] D. Fischer, N. Meyer, M. Kuczmik, M. Beck, A. Ja¨ger-
Waldau, M.Ch. Lux-Steiner, Sol. Energy Mat. Sol. Cells
67 (2001) 105.
[11] D. Fischer, T. Dylla, N. Meyer, M.E. Beck, A. Ja¨ger-
Waldau, M.Ch. Lux-Steiner, Thin Solid Films 387 (2001)
63.
[34] D. Braunger, D. Hariskos, G. Bilger, U. Rau, H.W.
Schock, Thin Solid Films 361 (2000) 161.
[35] D.W. Niles, K. Ramanathan, F. Hasoon, R. Noufi, B.J.
Tielsch, J.E. Fulghum, J. Vac. Sci. Technol. A 15 (1997)
3044.
´
[12] D. Fuertes Marron, A. Meeder, R. Wurz, S.M. Babu, Th.
¨
Schedel-Niedrig, M.Ch. Lux-Steiner, PV in Europe, from
PV technology to energy solutions, Conference Proceed-
ings, Rome, Italy, 2002, p. 421, published by WIP Munich
and ETA Florence.
[36] D.Y. Zemlyanov, E. Savinova, A. Scheybal, K. Doblhofer,
R. Schlo¨gl, Surf. Sci. 418 (1998) 441.
[13] K.J.S. Sawhney, F. Senf, W. Gudat, Nucl. Instrum. Meth.
A 467 (2001) 466.
[37] H. Iwakuro, C. Tatsuyama, S. Ichimura, Jpn. J. Appl.
Phys. 21 (1982) 94.
[14] H. Bluhm, M. Ha¨vecker, A. Knop-Gericke, E. Kleimenov,
R. Schlo¨gl, D. Teschner, V.I. Bukhtiyarov, D.F. Ogletree,
M. Salmeron, J. Phys. Chem. B, in press;
[38] S. Tanuma, C.J. Powell, D.R. Penn, Surf. Interface Anal.
17 (1991) 911.
D.F. Ogletree, H. Bluhm, G. Lebedev, C.S. Fadley, Z.
Hussain, M. Salmeron, Rev. Sci. Instrum. 73 (2002) 3872.
[39] A. Zangwill, Physics at surfaces, Cambridge University
Press, 1983;
´
[15] A. Meeder, L. Weinhardt, R. Stresing, D. Fuertes Marron,
R. Wurz, S.M. Babu, T. Schedel-Niedrig, M.Ch. Lux-
G.A. Somorjai, Chemistry in Two Dimensions: Surfaces,
Cornell University Press, Ithaca, 1981.
¨

94
R. Wu¨rz et al. / Surface Science 580 (2005) 80–94
[40] J.D. Cox, D.D. Wagman, V.A. Medvedev, in: D.R. Lide
(Ed.), CRC Handbook of Chemistry and Physics 2001–
2002: A Ready-reference Book of Chemical and Physical
Data, 82nd ed., CRC Press, Boca Raton, FL, USA, 2001.
[41] R.J. Meyer, E.H. Erich Pietsch, Gmelins Handbuch der
anorganischen Chemie, Natrium, Erga¨nzungsband 3, Sys-
tem Nr. 21, Verlag Chemie, Weinheim/Bergstraße, 1966.
[42] O. Knacke, O. Kubaschewski, K. Hesselmann, Thermo-
chemical Properties of Inorganic Substances, Springer,
New York, 1991.
[43] H. Schmalzried, Chemical Kinetics of Solids, VCH Wein-
heim, 1995, p. 7 (Chapter 6).
[44] E.H. Erich Pietsch, Gmelins Handbuch der anorganischen
Chemie, Kupfer Teil B, System Nr. 60, Verlag Chemie,
Weinheim/Bergstraße, 1958.
´
[45] R. Wurz, A. Meeder, D. Fuertes Marron, Th. Schedel-
¨
Niedrig, H. Bluhm, K. Lips, Phys. Rev. B 70 (2004)
205321.
[46] T. Robert, M. Bartel, G. Offergeld, Surf. Sci. 33 (1972) 123.
[47] M. Rusu et al., Thin Solid Films, in press.