Solar Energy Materials & Solar Cells

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Solar Energy Materials & Solar Cells 95 (2011) 14521462

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells


journal homepage: www.elsevier.com/locate/solmat

Analysis of Cu(In,Ga)(S,Se)2 thin-lm solar cells by means of


electron microscopy
D. Abou-Ras a,n, J. Dietrich b, J. Kavalakkatt a, M. Nichterwitz a, S.S. Schmidt a, C.T. Koch c, R. Caballero a,
J. Klaer a, T. Rissom a
a

Helmholtz-Zentrum Berin f
ur Materialien und Energie, Hahn-Meitner-Platz 1, 14109 Berlin, Germany
Department of Semiconductor Devices, TU Berlin, Einsteinufer 19, Sekr. E2, 10587 Berlin, Germany
c
Max-Planck Institute for Metals Research, Heisenbergstr. 3, 70569 Stuttgart, Germany
b

a r t i c l e i n f o

a b s t r a c t

Available online 16 December 2010

The present work gives an overview of how electron microscopy and its related techniques are used to
analyze individual layers and their interfaces in Cu(In,Ga)(S,Se)2 thin-lm solar cells. Imaging of samples
can be performed at scales of down to the (sub)angstroms range. At similar spatial resolutions,
information on composition can be gathered by means of energy-dispersive X-ray spectroscopy (EDX)
and on spatial distributions of electrostatic Coulomb potentials in the specimen by applying electron
holography. Microstructural and compositional properties as well as charge-carrier collection and
radiative recombination behavior of the individual layers are accessible by use of electron backscatter
diffraction, EDX, electron-beam-induced current (EBIC) and cathodoluminescence measurements,
available in scanning electron microscopy. The present contribution gives an overview of the various
scanning and transmission electron microscopy techniques applied on Cu(In,Ga)(S,Se)2 thin-lm solar
cells, examples from case studies, and also demonstrates how these techniques may be combined in order
to improve the analysis. Particularly, EBIC results show a reduced charge-carrier collection at Cu(In,Ga)Se2
grain boundaries, while no indication was found for a charge accumulation at the grain boundaries by
electron holography.
& 2010 Elsevier B.V. All rights reserved.

Keywords:
Cu(In,Ga)(S,Se)2 solar cells
Electron microscopy
Electron backscatter diffraction
Electron-beam-induced current
Electron energy-loss spectroscopy
Electron holography

1. Introduction

2. Experimental details

Electron microscopy analyses have been performed on


Cu(In,Ga)(S,Se)2 thin-lm solar cells ever since the beginning of
their research and development more than 30 years ago. Apart from
providing images down to the angstroms range of the thin-lm
stacks with total thicknesses of few micrometers, electron microscopy supplies a versatile collection of various analysis methods for
studying microstructures and compositions as well as of electrical
and optoelectronic properties.
The present work intends to give an overview of some of these
methods, describing how they are applied to provide insight into
the physics of Cu(In,Ga)(S,Se)2 thin-lm solar cells. Particularly,
the benet of the combination of various techniques on
identical positions is highlighted when studying structural
defects and compositional gradients or also grain boundaries in
Cu(In,Ga)(S,Se)2 thin-lm solar cells.

Co-evaporated Cu(In,Ga)Se2 and CuInS2 produced by a rapid


thermal process (RTP) were deposited on Mo-coated soda-lime
glass substrates. For the Cu(In,Ga)Se2 layers, a multi-stage
process [1] was applied, in which rst InGaSe is evaporated,
then CuSe until the layer exhibits a concentration ratio
of [Cu]/([In]+[Ga]) 41, and nally again InGaSe until
[Cu]/([In]+ [Ga])o1. In the RTP [2], CuIn metal precursors were
sputtered on Mo/glass and then transformed into CuInS2 in a sulfur
atmosphere. Since the Cu concentration exceeded the one of
In ([Cu]/[In]41) in the precursors, CuS formed on top of CuInS2,
which was removed later by use of diluted KCN, prior to further
processing.
The solar cells were completed by chemical-bath-deposited CdS
buffers, a sputtered i-ZnO/ZnO:Al bilayer as window layers and a
NiAl metal grids for enhancing the current collection. Table 1 gives
an overview of the samples studied for the present work, including
their compositions and photovoltaic parameters.
Cross-section specimens for scanning electron microscopy
(SEM) were prepared by gluing two stripes of solar cells face-toface together using epoxy glue, cutting stripes from these stacks,
then by polishing the cross-sections of these stripes mechanically

Corresponding author.
E-mail address: daniel.abou-ras@helmholtz-berlin.de (D. Abou-Ras).

0927-0248/$ - see front matter & 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.solmat.2010.11.008

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

1453

Table 1
Compositional ratios and photovoltaic parameters (open-circuit voltage Voc, short-circuit current density jsc, ll factor FF, and solar conversion efciency Z) of Cu(In,Ga)Se2
(CIGSe), CuGaSe2 (CGSe), and CuInS2 (CIS) thin-lm solar cells studied for the present work.
jsc (mA/cm2)

Z (%)

Sample

[Cu]/([Ga] +[In])

[Ga]/([Ga] + [In])

Voc (mV)

CIGSe-1
CIGSe-2
CIGSe-3
CIGSe-4
CGSe-1
CIS-1
CIS-2

0.86
0.92
0.83
0.86
0.80
1
1

0.33
0.33
0.28
0.37
1

633

34.3

70

15.2

700
674
704
702
692

31.0
31.1
32.6
13.1
21.4

76
71
72
55
68

16.4
14.8
16.6
5.1
10.1

694

20.8

66

10.1

and using Ar-ion beams. Very thin (45 nm nominally) graphite


layers were deposited on top of these cross-sections in order to
reduce drift during the measurements. Backside Cu(In,Ga)Se2
layers were exposed by lifting off Cu(In,Ga)Se2/CdS/i-ZnO/ZnO:Al
stacks from the Mo/glass substrates, after the stacks had been glued
to a SEM holder by use of Ag epoxy glue. Very thin graphite layers
were deposited on the ZnO:Al layers prior to gluing and also on the
exposed Cu(In,Ga)Se2 layers in order to inhibit Ag diffusion into the
solar-cell stack and to reduce drift during the measurements.
Cross-section specimens for transmission electron microscopy
(TEM) were prepared by gluing two stripes of solar cells face-toface using epoxy glue, cutting stripes from these stacks, then by
polishing the cross-sections of these stripes mechanically and
gluing them to Mo rings. The polished cross-sections were Ar-ionmilled until they were transparent for the electron beam.
Electron-backscatter diffraction (EBSD), energy-dispersive Xray spectrometry (EDX), electron-beam-induced current (EBIC) and
cathodoluminescence (CL) measurements were performed using a
LEO 1530 GEMINI scanning electron microscope equipped with an
Oxford Instruments HKL Nordlys II EBSD camera (acquisition and
evaluation software FastAcquistion/Channel5), a Thermo Noran Xray silicon-drift detector (acquisition and evaluation software
Noran System Seven), a FEMTO DLPCA-200 impedance amplier,
and an EMSystems CL system (acquisition software Proscan Spectra
Interpreter) with a Hamamatsu photomultiplier tube R3896.
TEM images and EDX maps were acquired using a Zeiss Libra
200FE transmission electron microscope. Details on the reconstruction of the amplitude and phase of the exit-plane wave
function from focus series acquired by using the same microscope,
also called inline electron holography, can be found in Refs. [3,4].

3. Results and discussions


3.1. Analyses of microstructures, compositions and optoelectronic
properties
3.1.1. Scanning electron microscopy
3.1.1.1. Electron backscatter diffraction. A straightforward method
in order to obtain information of microstructural details of a
Cu(In,Ga)(S,Se)2 thin lm, such as its average grain size, is to
fracture the solar cell (particularly for those on glass substrates)
and image the cross-section. The average grain size, e.g., may be
estimated from such images under the assumption that the thin
lm fractures along the grain boundaries, which needs not to be
the case.
A more precise method for measuring average grain sizes is
electron-backscatter diffraction [5,6] on polished cross-section
specimens or on chemically etched surfaces of thin lms. Such
surfaces of Cu(In,Ga)(S,Se)2 layers may also be exposed by lifting off
Cu(In,Ga)(S,Se)2/CdS/ZnO stacks from the Mo/glass substrates. The
careful surface preparation of specimens for EBSD is important
since this technique is surface sensitive with information depths in

FF (%)

Cu(In,Ga)(S,Se)2 of only about 2030 nm. Fractured cross-section


samples allow for pointwise EBSD measurements, but contiguous
EBSD maps are impeded by shadowing of emitted backscattered
electrons by the surface roughnesses. EBSD is based on the
diffraction of backscattered electrons at various families of lattice
planes, upon irradiation of a crystal in a polycrystalline specimen
by an electron beam. Even for beam energies as high as 20 keV, with
penetration depths of about 2 mm in Cu(In,Ga)(S,Se)2, most of the
electrons backscattered from this volume are reabsorbed by the
surrounding material. The small information depths for EBSD have
as consequence a high spatial resolution of about 2030 nm.
Upon electron irradiation, backscattered electrons diffracted at
a certain family of lattice planes in a crystal emerge from the
specimen. The ensemble of the various diffraction bands upon
diffraction at the various families of lattice planes in a crystal, the
EBSD pattern, contains information not only on the symmetry of
the crystal (and thus allows for the estimate of its phase) but also on
the crystal orientation with respect to a dened reference coordinate system. Therefore, by the acquisition of EBSD patterns by use
of a video camera during the scanning of the electron beam across a
given specimen area, EBSD maps containing local crystal orientations and symmetries are obtained. These maps allow for quantifying average grain sizes, local and integral textures as well as give
information on positions and misorientations of grain boundaries.
Further details may be found in Ref. [7], and examples of EBSD maps
will be demonstrated further below in section 3.1.1.3.
3.1.1.2. Energy-dispersive X-ray spectrometry. Phase analysis by
means of EBSD alone remains often ambiguous, owing to many
phases having similar crystal symmetries which may not be distinguished by EBSD. In order to improve the quality of the analysis,
elemental distributions in a given sample may be studied by means
of EDX. The combination of EBSD and EDX allows in most cases for
unambiguous phase analysis [8].
It is often argued that for EDX measurements in SEM, the spatial
resolution is too small for the analysis of elemental distributions in
the thin lms of a Cu(In,Ga)(S,Se)2 solar-cell stack. This is, however,
not true. The spatial resolution of an EDX measurement is limited
by two quantities: the acceleration voltage and the energy of the Xrays which are evaluated. The volume of excitation is dened by the
acceleration voltage; the smaller this volume, the higher the spatial
resolution. Also, the smaller the energy of the generated characteristic X-rays, the shorter is their mean-free path, and the higher
again the spatial resolution of the EDX measurement. It has been
shown [9] that spatial resolutions of down to 100 nm and below
may be obtained in SEMEDX measurements by reducing the
acceleration voltage down to few kV.
As an example, Fig. 1 shows an elemental-distribution map,
superimposed on a secondary-electron image, acquired on a crosssection specimen from a ZnO/CdS/CuInS2/Mo/glass solar cell (CIS-1,
see Table 1). The map is composed of the individual maps containing Zn-L, Cd-L, Cu-L and Mo-L signals. The acceleration voltage for
the acquisition of these maps was 7 kV. It is apparent that the

1454

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

Fig. 1. EDX elemental distribution map superimposed on a SEM image, acquired on


a cross-section specimen from a ZnO/CdS/CuInS2/Mo/glass solar cell (CIS-1, see
Table 1). The map is composed of individual Zn-L, Cd-L, Cu-L and Mo-L
distribution maps.

signals from the thin CdS layer (30 nm thick) can be clearly
distinguished from those attributed to the ZnO and CuInS2 layers.
The thickness of the CdS layer appears larger in the EDX map than
its real value because of the lateral extension of the excitation
volume at 7 kV.
It is also shown by the EDX elemental-distribution map in Fig. 1
that modern evaluation software suites are able to deconvolute
successfully overlapping X-ray peaks in the EDX spectrum, e.g., the
Cu-L and Zn-L, and also the S-K (not shown here) and Mo-L lines.
Because of the rather strong roughness of the CuInS2 layer, the
electron beam impinging on the CdS layer excites also Zn signals
from ZnO. This is why the CdS layer does not appear contiguous,
and also why Zn seems to be present at positions in between the
rough ZnO layer.
3.1.1.3. Correlation between electron backscatter diffraction and
cathodoluminescence. Apart from EDX for phase analyses, EBSD
may also be combined with other SEM techniques. Fig. 2 shows
cross-sectional SEM and panchromatic CL images as well as EBSD
orientation-distribution and pattern-quality maps from the
identical position of a ZnO/CdS/CuGaSe2/Mo/glass solar-cell stack
(sample CGSe-1, see Table 1). The local orientations in Fig. 2 b are
given by colors, see the legend. In the pattern-quality map shown in
Figs. 2d, S3 grain boundaries are highlighted by red lines (please
refer to Ref. [7] and to the references therein for explanations
of the S value and further details on S3 grain boundaries in
Cu(In,Ga)(S,Se)2 thin lms).
It is important to point out that the EBSD maps in Fig. 2(b)(d)
are different representations of the identical EBSD data. Apart from
local orientations recorded in each pixel of the EBSD maps
(Fig. 2(b)), the pattern quality can be determined by extracting
proles across the diffraction bands in the EBSD pattern and
relating the slope of the (ideally rectangular) prole to a gray
value. At grain boundaries, low gray values are obtained since
superimposing EBSD patterns from neighboring grains are
recorded, overall leading to a poor pattern quality. Since the
orientations of two adjacent grains are known, their misorientations (usually expressed by a rotation about a crystal axis through
an angle, transforming the point lattice of the one grain into that of
the other) can be determined.
S3 grain boundaries in Cu(In,Ga)(S,Se)2 (Fig. 2(d)) exhibit specic
misorientations [7] and are generally of high symmetry (low density
of crystal defects). About 50 % (and above) of the grain boundaries in

Fig. 2. SEM image (a), EBSD orientation distribution (b) and pattern-quality maps
(c), also with S3 grain boundaries highlighted by red lines (d) as well as a
panchromatic CL image (e) all acquired at the identical position of a ZnO/CdS/
CuGaSe2/Mo/glass cross-section specimen (CGSe-1, see Table 1). (For interpretation
of the references to color in this gure legend, the reader is referred to the web
version of this article.)

Cu(In,Ga)(S,Se)2 thin lms produced by co-evaporation from the


elements or by an RTP are counted among this type [12]. This nding
may be explained by assuming that most of the S3 grain boundaries
have a twin constellation, where twin boundaries are special cases
of stacking faults. The stacking-fault formation enthalpies in
Cu(In,Ga) Se2 were shown to be considerably small [10]. Thus, the
formation of these planar defects represents an appropriate mechanism in order to reduce the strain present in the Cu(In,Ga)(S,Se)2
thin lms during growth and probably also during the coolingdown stage.
The panchromatic CL image (Fig. 2(e)) exhibits an intensity
distribution which can be attributed to the microstructure represented by the EBSD pattern-quality map (Fig. 2(c)). It appears that
the CL signal intensities are higher from certain grains as compared
with those from others. By comparison of the EBSD orientationdistribution map (Fig. 2(b)) and the CL image (Fig. 2(e)), it can be
excluded that the CL intensity is inuenced by the local orientation.
Recently, it was shown [11] that the CL intensity emitted from
Cu(In,Ga)(S,Se)2 may be considerably inuenced by the presence of
linear or planar crystal defects, probably of dislocations. Also, the
distribution of the CL signal intensity across the polycrystalline
CuGaSe2 layer may be affected by slight variations in the local
doping since the radiative recombination rate depends on the
charge-carrier concentration.
It was also demonstrated [11,12] that S3 (twin) grain boundaries in Cu(In,Ga)(S,Se)2 do not inuence considerably CL intensities, i.e., the radiative recombination is similar as in the grain
interiors, in contrast to non-S3 grain boundaries, where decreased
CL signal intensities were found. These results can be also
conrmed at few positions in Fig. 2(e); but overall, the signal to
noise ratio in this CL image (Fig. 2(e)) is too low for extensive
analysis of the CL-signal changes at CuGaSe2 grain boundaries.
Further analyses of grain boundaries in Cu(In,Ga)(S,Se)2 thin lms
are reported further below in section 3.2.
Generally, the CL signals are emitted upon generation of
electronhole pairs by the incident electron beam, diffusion of
these charge carriers and their radiative recombination. That is, the
spatial resolutions of CL images depend on the diffusion lengths
and thus on the lifetimes of the minority charge carriers. These
resolutions can be as low as 50100 nm and are affected by the

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

1455

temperature of the specimen. The CL intensities do not only depend


on the fraction of radiative recombination in the total recombination of charge carriers, but also on the generation of electronhole
pairs, which may be inuenced by the band-gap energy of
semiconductors or the absorption coefcients of the materials
under investigation. When CL spectra acquired in each measuring
point of a map are compared, changes in phases may be detected as
well as point-defect densities and their energy positions. Timeresolved CL with pulsed incident electron beams can provide access
to lifetimes of charge carriers.
3.1.2. Analyses of structural defects and elemental distributions by
transmission electron microscopy techniques
Due to its high spatial resolution, TEM is a powerful tool to
characterize crystal defects in polycrystalline semiconductors [13].
In particular, it is used to identify and distinguish various types of
crystal defects such as dislocations, stacking faults and also
antiphase or grain boundaries. Among these types, dislocations
play an important role in semiconductors because of their
electrically active centers in the crystal structure, generated by
dangling bonds and displacement elds [14].
The characterization of dislocation networks includes the
determination of their densities and crystal orientations. A dis!
location is dened by its line direction and its Burgers vector b ,
standing for the resultant displacement of the lattice. The displacement eld around the dislocation changes the Bragg diffraction condition, which leads to a change in the diffraction contrast.
This is why dislocations appear as black lines in TEM bright-eld
(BF) images, as in the Cu(In,Ga)Se2 layer in Fig. 3(a), where a
BF-TEM cross-section image from a ZnO/CdS/Cu(In,Ga)Se2/Mo/
glass stack (CIGSe-1, see Table 1) is given. In addition, also stacking
faults and microtwins are visible (however, they cannot be
distinguished from one another in this image, only by highresolution imaging). These structural defects are highlighted by
blue and red lines in Fig. 3(b).
In contrast to BF mode, in dark-eld (DF) TEM, only that area
appears bright which satises the Bragg condition of the beam,
selected by an aperture applied in the back-focal plane of the
objective lens. Fig. 3(c) shows a DF-TEM image from an area in the
Cu(In,Ga)Se2 layer (identical specimen as for Fig. 3(a) and (b)), with
dislocations visible as narrow bright lines on dark background. The
DF conditions were selected in a way that only the surrounded
displacement eld satises the Bragg condition in contrast to the
orientation of the grain. Furthermore, one can use special DF
analyses to determine the dislocation parameters such as the
burgers vector and the line direction as well as the dissociation
width of partial dislocations [13]. TEM is thus able to provide a
detailed structural characterization of one- and two-dimensional
lattice defects.
TEM analyses of various Cu(In,Ga)(S,Se)2 layers deposited by coevaporation from the elements or by means of an RTP show a
considerably high relative frequency of twin boundaries and
stacking faults, as already found by means of EBSD (Section
3.1.1.3). It is not yet clear what the formation mechanisms of
stacking faults and microtwins are, and whether they are more
preferentially formed than the dislocation networks shown in
Fig. 3. This issue is currently under investigation.
In addition to microstructural studies, compositional analyses
may be performed using EDX or also by energy electron energy-loss
spectroscopy (EELS) in TEM. While EDX has been applied to a large
extent in investigations on Cu(In,Ga)(S,Se)2 thin-lms and corresponding solar cells, only few reports can be found in the literature
on EELS analyses [1517]. This is mainly due to the fact that the
corresponding loss edges for Cu, Ga and Se, used for elemental
imaging, are positioned at rather high energy losses ranging from

Fig. 3. (a) Cross-sectional BF-TEM image of a ZnO/CdS/Cu(In,Ga)Se2/Mo/ glass solarcell stack (CIGSe-1, see Table 1); (b) identical image, with the stacking faults and
microtwins highlighted by red as well as the dislocations by blue lines. Figures a and
b are two BF-TEM images stitched together (cf. vertical line). (c) A DF image from the
identical Cu(In,Ga)Se2 layer shown in (a) and (b), with dislocations visible as white
lines within a Cu(In,Ga)Se2 grain.

about 930 to 1500 eV, leading to low signal intensities and thus to
low signal-to-noise ratios and overall unfavorable conditions with
respect to EDX.

1456

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

Due to the very small thicknesses of the TEM specimens studied


(50100 nm), the extension of the generation volume of the
electron beam for an EDX measurement in TEM is one or two
orders of magnitude smaller compared with EDX in SEM (Section
3.1.1.2), which leads to a much higher spatial resolution of few
nanometers (nowadays, by use of aberration-corrected microscopes, values of down to 0.40.6 nm have been realized [18]).
Disadvantages of the small specimen thicknesses are the generally
poor measurement statistics and the sensitivity to thickness
variations.
Fig. 4 is composed of the EDX elemental distribution maps using
Zn-K, Cd-L, Ga-K and Mo-K signals, superposing a scanning TEM
(STEM) image, from a ZnO/CdS/Cu(In,Ga)Se2/Mo/glass solar-cell
stack (CIGSe-1, see Table 1). The position of the CdS layer is much
better visible than from the maps in Fig. 1. Also, the Ga signal
exhibits an intensity gradient across the Cu(In,Ga)Se2 layer, with
local maxima close to the interfaces to the Mo and CdS layers. Such
elemental distribution maps are essential for the research and
development of Cu(In,Ga)Se2 solar cells, giving information on local
compositions and interdiffusions down to few nanometers (in
uncorrected microscopes).
A further example is shown in Fig. 5. The EDX linescans given in
there show Cu, In, Ga and Cd distribution proles, extracted from
corresponding Cu-K, In-L and Cd-L distribution maps acquired on a
CuInS2/CdS interface. It is clearly visible that the In signal decreases
more gradually than the Cu signal, where the drop-off of the Cu
signal is shifted towards the CdS signal, suggesting Cu diffusion
from CuInS2 into CdS. Also, the Cu signal is still enhanced where the
In signal intensity is already decreased to zero. Indeed, such a
diffusion has been found by various authors [19,20] for CBD-CdS
buffer layers grown on Cu(In,Ga)Se2.

3.2. Grain-boundary physics in Cu(In,Ga)(S,Se)2 thin lms


Although there is a high density of grain boundaries in polycrystalline Cu(In,Ga)Se2 layers, solar-conversion efciencies of up
to more than 20 % have been achieved [21]. There are various
models, simulations and experimental results treating the inuence of grain boundaries and explaining why they may not have a
detrimental effect on charge-carrier collection and thus on the
device performance [2234]. Some of these models and results
[22,28,29] even suggest that grain boundaries have a positive effect
originating from charged or neutral hole or electron barriers

Fig. 4. EDX elemental distribution maps composed by use of Zn-K, Cd-L, Ga-K and
Mo-K signals, superposing a STEM image, acquired on a ZnO/CdS/Cu(In,Ga)Se2/Mo/
glass solar-cell stack (CIGSe-1, see Table 1).

Fig. 5. EDX elemental distribution proles, extracted from corresponding EDX maps
across an CuInS2/CdS interface (CIGSe-12, see Table 1). The net counts were
normalized to the maximum intensity of the Cd-L signal, for easier comparison.
The net counts are given as solid symbols, the solid lines are smoothed data given as
guide for the eye. Apparently, the In-L decreases more gradually than the Cu-K signal,
indicating Cu diffusion CuInS2 from into CdS. The Cd-L signal is not zero in the CuInS2,
but it remains ambiguous because of possible superposition with the In-L signal.

present in their proximity, which would impede recombination


of electrons and holes.
The following subsections describe approaches for exploring the
effects of grain boundaries in Cu(In,Ga)Se2 layers on charge-carrier
collection. It is important to note that the corresponding electron
microscopy results are rather related to grain-boundary properties
in the volume of these Cu(In,Ga)Se2 layers, in contrast to scanning
probe techniques analyzing the ones at their surfaces. While at the
grain boundaries on the surface, charge accumulation is probable
[24,26], this is not necessarily the case for grain boundaries in the
volume [35], as we will also show further below.
3.2.1. Combination of electron-beam-induced current and electron
backscatter diffraction measurements
In this section, electron-beam induced current (EBIC) measurements and EBSD in the cross-section and backside congurations
are presented as capable techniques in order to gain insight
into charge-carrier collection properties of grain boundaries in
Cu(In,Ga)Se2 layers of thin-lm solar cells. Due to the relatively
small extension of the generation volume of the electron beam as
compared with generation by light, EBIC is an adequate tool to
study charge-carrier collection spatially resolved, while EBSD
serves to localize and classify grain boundaries [7]. It can be shown
(see below) that effects of single grain boundaries in standard
Cu(In,Ga)Se2 absorber layers with an average grain size and an
electron diffusion length in the mm range can be identied by
means of EBIC when using low electron beam energies, ensuring a
small extension of the generation volume and Green penetration
depth [36] (e.g., about 200 nm for 5 keV).
Problems in the interpretation of EBIC results may arise due to
the fact that the data is only accessible in two dimensions, whereas
charge-carrier collection is a process taking place in three dimensions. Also, there are additional surfaces present, where enhanced
recombination or charge accumulation is possible.
3.2.1.1. Introduction to EBIC. Fig. 6 shows schematically the setup of
an EBIC experiment. The solar-cell sample is irradiated by a focused
electron beam, and charge carriers are generated in a dened region
around the position of irradiation. Empirical expressions for the

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

1457

description of the lateral and depth-dependent generation proles


are given in Refs. [36,37]. The smaller the beam energy Eb, the
narrower is the lateral generation prole (see Fig. 6). Thus, for
identifying effects of individual grain boundaries, it is necessary to
use low beam energies. However, there is a trade off between spatial
resolution and the inuence of surface effects, which has to be taken
into account. Therefore, beam energies between about 4 and 10 kV
are recommended for average grain sizes larger than 500 nm.
Most relevant in order to quantify charge-carrier collection
!
processes is the collection function fc x , which stands for the
!
probability of a charge carrier generated at position x to be
collected. The measured EBIC signal can be expressed as a
!
convolution of the collection function fc x and the generation
!
!!
function g x , a of the electron beam ( a is the impinging position
of the electron beam) [38]
Z
!
! !! !
1
I a
fc x g x , a d x
!
A solution for the collection function fc x for the quasi-neutral
region (QNR) of the absorber layer under low injection conditions
can be gained from the following differential equation, which was
derived from the continuity equation via a reciprocity theorem [38]
!
!
!
! fc x
DDfc x m E !fc x 
0

where D is the minority charge carrier (i.e., electron) diffusion


!
constant, m the electron mobility, E the electrical eld and t the
electron lifetime. If translation invariance in the directions parallel
to the pn junction is assumed, a one-dimensional equation can be
used. The effects of grain boundaries are not covered in this
approach; for simplication, it is necessary to neglect them at this
point. As a boundary condition, it is assumed that the collection
probability at the edge of the space-charge region (SCR) is 1
(fc(xSCR)1) and that fc(x)-0 for x-N for an innite semiconduc!
tor layer. In this case, the solution for the collection function fc x is
a simple exponential function fc(x)exp(  x/L), where L is the
electron diffusion length. Assuming a nite semiconductor limited
by a back contact at position xBC, the second boundary condition
changes to dfc =dxxBC SBC =Dfc xBC , where SBC is the electron
recombination velocity at the back contact. A solution for fc (x) is
fc x

Fig. 6. Schematic representation of EBIC setups for backside (a) and cross-section
(c) congurations. The depth dependent (b) and lateral (d) generation proles
in a Cu(In,Ga)Se2 layer as well as the corresponding, calculated collection functions
fc in Cu(In,Ga)Se2 solar cells are also shown. fc is assumed to be 0 in the n-type layers,
1 in the SCR and is distributed according to Eq. (3) in the QNR of the Cu(In,Ga)
Se2 layer.

1=LcoshxxBC =LSBC =DsinhxxBC =L


SBC =D sinhxBC xSCR =L 1=LcoshxBC xSCR =L

In this study, we present EBIC measurements performed on the


backside of a Cu(In,Ga)Se2/CdS/ZnO stack lifted off from the Mo/
glass substrate (Fig. 6(a)), and on a polished cross-section of the
ZnO/CdS/Cu(In,Ga)Se2/Mo/glass solar-cell stack. In Fig. 6(b) and (d)
(red curves), exemplary collection functions of a Cu(In,Ga)Se2
thin-lm solar cell are shown. fc(x) is assumed to be 0 in the
window and buffer layers, and 1 in the SCR. The diffusion length
and the recombination velocity at the back contact are L 500 nm
and SBC 1  105 cm/s.
In former studies, EBIC results on Cu(In,Ga)Se2 solar cells with CdS
buffer layers were found to deviate from the theoretical expectations
given above. This deviation was explained [39] by assuming that
electrical defects and in consequence charges accumulate at the CdS/
Cu(In,Ga)Se2 interface, causing a conguration where the collection
function is not independent of the generation function. This has to be
taken into account when evaluating EBIC data.
3.2.1.2. EBIC signals at grain boundaries. In case of enhanced
recombination at a grain boundary, a local minimum in the EBIC
signal across the grain boundary is expected, because charge

1458

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

carriers generated at or close to the grain boundary exhibit a


shorter effective lifetime and are more likely to recombine than
charge carriers generated in the interior of the Cu(In,Ga)Se2 grain.
The equations given in Refs. [40,41] provide an expression for a
reduced, effective diffusion length measured in the proximity of the
corresponding grain boundary. The shape of an EBIC linescan
perpendicular to a grain boundary can be calculated from these
equations, and by comparing experimental data with the simulated
curves, an estimate of the recombination velocity at the grain
boundary, Sgb, is possible. Fig. 7 shows simulated EBIC proles
perpendicular to a grain boundary in the cross-section conguration assuming different recombination velocities (for details see
Ref. [27]). These simulations also show that the spatial resolution of
EBIC is sufcient to resolve effects of individual grain boundaries
when using low beam energies of around 5 keV and for average
grain sizes in the range of 0.5 mm or larger. When using higher
beam energies (e.g. 15 keV) and in case average grain sizes are
signicantly smaller, the effects of various grain boundaries
superimpose and the EBIC signal smears out.
In case of a neutral hole barrier at a grain boundary [22], reduced
recombination of electrons and therefore enhanced charge-carrier
collection is expected. The widths of the resulting EBIC maxima at
such grain boundaries depends on the electron-beam energy Eb
applied and on the electron diffusion-length L in Cu(In,Ga)Se2 as in
case of enhanced recombination.

If a grain boundary is positively charged, a SCR develops, which


may assist in charge-carrier collection by separating generated
electrons and holes [28]. Depending on the doping density of the
Cu(In,Ga)Se2 layer, this separation may lead to either a maximum
in the EBIC linescan in the proximity of the grain boundary (broader

Fig. 7. Simulated normalized EBIC proles perpendicular to a grain boundary of the


Cu(In,Ga)Se2 layer. The current values were extracted from simulated EBIC line
proles perpendicular to the pn-junction in a distance of 200 nm to the edge of the
SCR in the QNR of the Cu(In,Ga)Se2 layer. The corresponding effective diffusion
lengths for the different recombination velocities Sgb, the different distances to the
grain boundary and the different electron-beam energies were calculated from the
equations given in Refs. [40,41].

Fig. 8. SEM image (a), EBSD orientation-distribution map (b) with local orientations
given by colors (see legend), and EBIC image acquired at 8 keV (c) on the identical
sample position of a polished cross-section of a Cu(In,Ga)Se2 thin-lm solar cell
(CIGSe-2, see Table 1). The white arrows indicate the position of the extracted EBIC
proles shown in (d). When comparing the experimental data with the simulations
(d), a recombination velocity of 3  104 cm/s can be estimated for this grain
boundary. For details, see Ref. [27].

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

1459

than in the case of the neutral hole barrier) or a uniform EBIC signal
(see also Ref. [27]).
The results shown in the following are not considered representative for all grain boundaries in Cu(In,Ga)Se2 layers; the
intention here is rather to demonstrate how to apply EBSD and
EBIC on identical sample positions.
3.2.1.3. Cross-sectional electron-beam-induced current measurements. Fig. 8 shows SEM and EBIC images as well as an EBSD
orientation-distribution map from the identical position of the
polished cross-section of a ZnO/CdS/Cu(In,Ga)Se2/Mo/glass solar
cell (CIGSe-2, see Table 1). The white arrow indicates the position of
the extracted EBIC prole perpendicular to a grain boundary, which
is shown in Fig. 8(d). The current values of these proles were
determined by rst extracting EBIC proles perpendicular to the
pn junction at different distances to the grain boundary. Then, the
position of the edge of the SCR of each of these EBIC proles was
dened and the current value of this position was set to 1. For the
resulting normalized EBIC prole perpendicular to the grain
boundary and parallel to the pn junction (Fig. 8d), the current
values were extracted at a distance of 200 nm from the edge of the
SCR in the QNR of the Cu(In,Ga)Se2 layer and normalized to the
plateau on the right hand side, which was dened to be 1. For
further details, see Ref. [27]. This procedure has to be carried out
since the model explained above takes into account only differences in the EBIC prole perpendicular to the pn junction in
the QNR.
The resulting EBIC prole exhibits a local minimum at the
position of this grain boundary. A minimum was also found for
higher beam energies Eb (see Ref. [27]). When comparing the
decrease of the EBIC signal intensity with the simulated curves [27],
a recombination velocity of 3  104 cm/s was estimated for this
grain boundary. This is a rather low value.
3.2.1.4. Backside electron-beam-induced current measurements. By
lifting off the glass substrate and the Mo layer the Cu(In.Ga)Se2
solar cell is accessible from the backside of the absorber layer.
Again, EBSD and EBIC measurements were performed on the
identical sample position (see Fig. 9). The EBIC distribution prole
given in Fig. 9(d) is an average of 30 individual proles and shows a
minimum (decrease of 5 rel.%) at the grain-boundary position.
Similarly as for the analyses in cross-section conguration (Fig. 8),
signicant changes in the EBIC signal have only been found at nonS3 grain boundaries. However, the simple, one-dimensional analytical model applied for the EBIC proles in the cross-section
conguration was unable to describe satisfactorily the proles
extracted from the backside EBIC images.
It may be noted that EBSD and EBIC results also from other
backside Cu(In,Ga)Se2/CdS/ZnO samples with varying Ga and Cu
concentrations [42] showed a reduced short-circuit current measured upon electron-beam irradiation on non-S3 grain boundaries.
In general, it was found that the EBIC signal does not change
signicantly at the positions of S3 (twin) grain boundaries,
whereas there are different effects observed at randomly oriented
grain boundaries of Cu(In,Ga)Se2 absorber layers. The results
shown above are only examples and serve to demonstrate the
experimental approach. For a general statement about the inuence of grain boundaries more data needs to be collected. Additionally, two- and three-dimensional numerical device modelling
is needed and will be applied for future studies.
3.2.2. Grain boundaries analysed by inline electron holography
Inline electron holography in TEM [3,43] is a valuable tool to
investigate the electromagnetic potentials at interfaces on a sub
nanometer scale. In the present section, we apply this technique for

Fig. 9. SEM (a) and EBIC images (b) as well as a EBSD orientation-distribution map
(c) from the identical sample position on the backside of Cu(In,Ga)Se2/CdS/ZnO stack
(prepared from CIGSe-3, see Table 1). The white arrows in (b) and (c) indicate the
position of the extracted EBIC prole (d) perpendicular to a grain boundary at EB
8 keV. IEBIC/IB is the ratio of measured EBIC and electron-beam currents.

1460

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

studying the behavior of the Coulomb potential at grain boundaries


in polycrystalline Cu(In,Ga)Se2 and CuInS2 solar cell absorbers.
In TEM, the electron beam can be described by a plane wave
function that impinges on a specimen. The interaction with the
specimen then alters the amplitude and the phase of the wave
function. The phase of the wave function at the exit plane of the
specimen contains information about the electromagnetic potential within the specimen. However, this information cannot be
utilized in conventional transmission electron microscopy since for
this microscope mode, only the intensity distribution is recorded,
which is given by the squared modulus of the image wave function
that arises from the transfer of the exit plane wave function by the
microscopic system.
Thus, all phase information is lost. However, the phase j(x,y) as
function of the local position (x,y) within the exit plane may be
retrieved by using holographic methods. We applied inline electron
holography (see Ref. [44] for a comparison with off-axis electron
holography). I.e., we acquired a through-focal series of images,
from which the relative phase distribution Dj(x,y) with respect to
an unknown reference phase value was reconstructed [3].
In the following, we will apply the absorbing phase object
approximation, in which multiple elastic scattering and also inelastic scattering out of the elastic channel are treated to all orders, but
the curvature of the Ewald sphere is effectively neglected. This
approximation is therefore suitable for zero-loss, energy-ltered
images at medium resolution as were acquired for this work. The
phase of the impinging electron plane wave is affected by the
interaction with the specimen, because the wavelength of the wave
function depends on the local electrostatic and vector potential
within the specimen. Assuming that there is no magnetic ux within
the specimen, the wavelength l is given by [45]
h

l p ,
2m0 eU V1 eU V=2m0 c2

where h is Plancks constant, m0 and e are the electron rest mass and
charge, U the acceleration voltage of the transmission electron
microscope and V the electrostatic potential which the electrons
from the incident beam experience when they travel through the
specimen. It contains contributions from scattering atoms (or ions)
and free charge carriers within the sample and from additional
electric elds. At the exit plane of the TEM specimen, the change in
phase j(x,y) due to the electrostatic potential V(x,y,z) within the
specimen, which is assumed to be small compared with the kinetic
energy of the beam electrons, is then given by [45]
Z
jx,y s Vx,y,zdz sUtx,yVe x,y,
4
where s(U) is the interaction constant, Ve(x,y) the electrostatic
potential averaged along the z-axis perpendicular to the specimen
surface and t(x,y) the local specimen thickness. Note that Ve(x,y) is
the electrostatic potential and not the potential energy. For neutral
solids without magnetic ux, Ve(x,y) is positive, i.e., the electrons see
an attractive potential within the solid. By means of inline electron
holography, we cannot determine the absolute phase shift induced
by interaction with the specimen but only relative phase shifts
Dj(x,y) originating from relative potential differences DVe(x,y) at
various specimen positions.
For the acquisition of the through-focal series, we tilted the
specimen, until the planes of the grain boundaries were parallel to
the incident electron beam. We reconstructed gray-value images of
the relative phase shift Dj(x,y) at the grain boundaries by applying
an algorithm developed by Koch [4]. Fig. 10 exemplarily shows
such a gray-value image of the reconstructed phase Dj(x,y) at a
non-twin (random, probably also non-S3) grain boundary within
a polycrystalline CuInS2 absorber.

Fig. 10. Gray-value image of the reconstructed phase difference Dj(x,y), acquired at
a random (non-S3) grain boundary (GB) within a CuInS2 absorber (CIS-2, see
Table 1). Individual proles were extracted across grain boundaries, from areas
similar to that depicted by the black frame.

While the grain interiors exhibit a negligible phase shifts, the


phase shifts at the grain boundaries are signicantly larger. From
gray-value images of the phase difference Dj(x,y) we extracted
proles across grain boundaries for Cu(In,Ga)Se2 and CuInS2 solarcell absorbers (CIGSe-4 and CIS-2, see Table 1). The black frame in
Fig. 10 denotes the region within which about 50 proles were
averaged parallel. From these proles, we calculated the relative
electrostatic potential DVe(x,y) by applying Eq. (4).
For the measurement of the specimen thicknesses t(x,y), we
applied energy-ltered TEM. We determined maps of the specimen
thickness t(x,y) in units of the inelastic mean free path lmean of
electrons within the specimen by taking the natural logarithm of
the ratio between an unltered and a zero-loss ltered image [46]
t(x,y)/lmean ln(It(x,y)/I0(x,y)). It(x,y) is the intensity distribution
of an unltered image, and I0(x,y) the intensity distribution of a
zero-loss ltered image, to which only unscattered or elastically
scattered electrons contribute. lmean can be estimated by use of an
algorithm given in Ref. [46].
Fig. 11 depicts the resulting electrostatic proles (averaged)
across a twin (S3) and a random (probably non-S3) grain boundary
in a Cu(In,Ga)Se2 layer as well as a random grain boundary within a
CuInS2 thin lm (CIGSe-4 and CIS-2, see Table 1). The potential
wells at twin boundaries are often indistinguishable from the noise
in electrostatic potential proles (about 100 mV) and were found to
exhibit depths of about 200 mV at most. This indicates, correspondingly to the EBIC results, not any substantial changes of the
electrostatic potential at this type of grain boundary.
We measured only potential wells (and not any barriers) at grain
boundaries in various Cu(In,Ga)Se2 and CuInS2 thin lms in
completed solar cells. While the full width at half minimum
(FWHM) of all potential wells was 12 nm, the potential-well
depths ranged from 200 mV for the twin grain boundary in
Cu(In,Ga)Se2 to 1.5 V for random grain boundaries in CuInS2. In
general, from studies on various other Cu(In,Ga)Se2 thin lms,
values between 1 and 3 V were obtained.

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

1461

At grain boundaries, EBSD maps together with EBIC and CL


measurements performed on identical positions give information
on the corresponding charge-carrier collection and radiative
recombination. For non-S3 grain boundaries in Cu(In,Ga)(S,Se)2
thin lms, reduced EBIC and CL signals were found, with estimated
recombination velocities of few 104 cm/s. We would like to note
that in few samples deviations from these ndings were also
obtained. In contrast, not any considerable electronic activity was
detected at S3 (twin) boundaries. This result was conrmed by
inline electron holography, by which only at non-S3 grain boundaries in Cu(In,Ga)(S,Se)2 thin lms, Coulomb-potential wells with
widths of 12 nm and depths of 13 V were measured. Overall,
the results presented here indicate an atomic reconstruction
via neutral defects or defect complexes at grain boundaries in
Cu(In,Ga)(S,Se)2, which only slightly reduce the short-circuit current.

Acknowledgements
Fig. 11. Extracted electrostatic potential proles across grain boundaries in
Cu(In,Ga)Se2 and CuInS2 absorbers (CIGSe-4 and CIS-2, see Table 1). These proles
are averages of 50 individual proles extracted from areas similar to the black frame
given in Fig. 10.

There are several possible reasons for the presence of potential


wells at grain boundaries. First, a negative charge at the grainboundary core may be considered, which is screened by positive
charges in the grain interiors. For typical doping concentrations of
10161017 cm  3 in Cu(In,Ga)Se2 and CuInS2, the corresponding
Debye lengths can be calculated to be between 12 and 44 nm at
room temperature. However, the FWHM of the potential wells is
much smaller than these values. With the reasonable assumption
that the electron beam does not generate a free-charge carrier
density exceeding that of the doping (the electron-beam intensity
was reduced accordingly), the potential-well shape cannot be
explained only by the screening of negative charges located at
the grain-boundary cores.
As second possibility, a space-charge region screening a positive
charge at the grain boundary core is possible, but would result in a
potential barrier and thus, does not agree with our results.
However, we cannot exclude from our measurements that a
potential barrier smaller than about 100 meV is present, which
may be caused by a positive charge at the grain boundary. The noise
of the electrostatic potential signals is in about the same range.
It seems therefore likely that the main contribution to the
potential wells obtained by inline electron holography are not
charge accumulations but rather changes of crystal structure and
composition at Cu(In,Ga)(S,Se)2 grain boundaries. This issue is
currently under investigation in order to provide evidence for
these compositional changes. It is also to be claried what the
link of these reconstructed grain boundaries are to the reduced
short-circuit current measured by means of EBIC.

4. Conclusions
In the present work, various SEM and TEM techniques are
reported which may be applied for the analysis of Cu(In,Ga)(S,Se)2
solar cells. It was shown that each technique gives important
information on microstructure, composition as well as on electrical
and optoelectronic properties, but that combinations of electronmicroscopy methods improve considerably the scientic quality of
the results. EBSD may be combined with EDX for unambiguous
phase analysis. In TEM, BF and DF imaging reveal linear or planar
defects, and compositional gradients in Cu(In,Ga)Se2 layers as well
as interdiffusion between Cu(In,Ga)Se2 and buffer layers may be
studied by means of EDX.

The authors would like to thank T. Unold, H.-W. Schock, and C.


Boit for fruitful discussions and continuous support. Special thanks

are due to B. Bunn, C. Kelch, M. Kirsch, and T. Munchenberg


for help
in sample preparation, as well as to J. Bundesmann, H. Stapel and M.
Wollgarten for scientic and technical assistance with the electron
microscopes. The authors would like to acknowledge funding from
the German ministry for environment (BMU) under contract
#0327559H (German-Israeli project ProGraB) as well as from the
German ministry for Education and Research (BMBF) under contract #033F0359C (project GRACIS).
References
[1] C.A. Kaufmann, R. Caballero, T. Unold, R. Hesse, R. Klenk, S. Schorr,
M. Nichterwitz, H.-W. Schock, Depth proling of Cu(In,Ga)Se2 thin lms
grown at low temperatures, Sol. Energy Mat. Sol. Cells 93 (2009) 859863.

[2] K. Siemer, J. Klaer, I. Luck, J. Bruns, R. Klenk, D. Braunig,


Efcient CuInS2 solar
cells from a rapid thermal process (RTP), Sol. Energy Mat. Sol. Cells 67 (2001)
159166.

[3] S. Bhattacharyya, C.T. Koch, M. Ruhle,


Projected potential proles across
interfaces obtained by reconstructing the exit face wave function from through
focal series, Ultramicroscopy 106 (2006) 525538.
[4] C.T. Koch, A ux-preserving non-linear inline holography reconstruction
algorithm for partially coherent electrons, Ultramicroscopy 108 (2008)
141150.
[5] A.J. Schwartz, M. Kumar, B.L. Adams, D.P. Field (Eds.), Electron backscatter
diffraction in materials science, Springer, New York, , 2009.
[6] D. Abou-Ras, R. Caballero, J. Kavalakkatt, M. Nichterwitz, T. Unold, H.W. Schock,

S. Bucheler,
A.N. Tiwari, Electron backscatter diffraction: exploring the microstructure in Cu(In,Ga)(S,Se)2 and CdTe thin-lm solar cells, in: Conference
Records of the 35th IEEE Photovoltaics Specialists Conference, Honolulu,
Hawaii, June 2025, 2010, IEEE, Piscataway, NJ, pp. 418423.
[7] D. Abou-Ras, S. Schorr, H.W. Schock, Grain-size distributions and grain
boundaries of chalcopyrite-type thin lms, J. Appl. Cryst. 40 (2007) 841848.
[8] D. Abou-Ras, M. Nichterwitz, R. Caballero, C.A. Kaufmann, T. Unold, S. Schorr, R.
Scheer, J. Klaer, H.-W. Schock, (Enhanced) insight in the microstructure and
composition of chalcopyrite-type thin lm solar cells, in: Proceedings of the
22nd European Photovoltaic Solar Energy Conference and Exhibition, Milan,
Italy, September 37, 2007, WIP Munich, 2007, pp. 19111914.
[9] I. Barkshire, P. Karduck, W.P. Rehbach, S. Richter, High-spatial-resolution lowenergy electron beam X-ray microanalysis, Mikrochim. Acta 132 (2000)
113128.
[10] N.I. Medvedeva, E.V. Shalaeva, M.V. Kuznetsov, M.V. Yakushev, First-principles
study of deformation behavior and structural defects in CuInSe2 and Cu(In,Ga)Se2,
Phys. Rev. B 73 (2006) 035207-1035207-6.
[11] D. Abou-Ras, U. Jahn, M. Nichterwitz, T. Unold, J. Klaer, H.-W. Schock, Combined
electron backscatter diffraction and cathodoluminescence measurements on
CuInS2/Mo/glass stacks and CuInS2 thin-lm solar cells, J. Appl. Phys. 107
(2010) 014311-1014311-8.

[12] D. Abou-Ras, C.T. Koch, V. Kustner,


P.A. van Aken, U. Jahn, M.A. Contreras,
R. Caballero, C.A. Kaufmann, R. Scheer, T. Unold, H.-W. Schock, Grain-boundary
types in chalcopyrite-type thin lms and their correlations with lm texture
and electrical properties, Thin Solid Films 517 (2009) 25452549.
[13] C.J. Kiely, R.C. Pond, G. Kenshole, A. Rockett, A. TEM, study of the crystallography and defect structures of single crystal and polycrystalline copper
indium diselenide, Philos. Mag. A 63 (1991) 12491273.

1462

D. Abou-Ras et al. / Solar Energy Materials & Solar Cells 95 (2011) 14521462

[14] D.C. Holt, B.G. Yacobi, Extended defects in semiconductors, chapter 5: The
electrical, optical and device effects of dislocations and grain boundaries,
Cambridge University Press, 2007.
[15] R. Saez-Araoz, D. Abou-Ras, T.P. Niesen, A. Neisser, K. Wilchelmi, M.Ch.
Lux-Steiner, A. Ennaoui, In situ monitoring the growth of thin-lm ZnS/Zn(S,O)
bilayer on Cu-chalcopyrite for high performance thin lm solar cells, Thin Solid
Films 517 (2009) 23002304.

[16] T. Torndahl,
E. Coronel, A. Hultqvist, C. Platzer-Bjorkman,
K. Leifer, M. Edoff, The
effect of Zn1-xMgxO buffer layer deposition temperature on Cu(In,Ga)Se2 solar
cells: a study of the buffer/absorber interface, Prog. Photovolt.: Res. Appl. 17
(2009) 115125.

[17] C. Camus, D. Abou-Ras, N.A. Allsop, S.E. Gledhill, T. Kohler,


J. Rappich,
I. Lauermann, M.C. Lux-Steiner, C.-H. Fischer, Formation of CuInS2carbon
multilayers in the spray ILGAR process, Phys. Status Solidi A 207 (1) (2010)
129131.
[18] M. Watanabe, D.W. Ackland, A. Burrows, C.J. Kiely, D.B. Williams, O.L. Krivanek,
N. Dellby, M.F. Murtt, Z. Szilagyi, Improvements in the X-ray analytical
capabilities of a scanning transmission electron microscope by sphericalaberration correction, Microsc. Microanal. 12 (2006) 515526.
[19] T. Nakada, A. Kunioka, Direct evidence of Cd diffusion into Cu(In,Ga)Se2 thin
lms during chemical-bath deposition process of CdS lms, Appl. Phys. Lett. 74
(17) (1999) 24442446.
[20] D. Abou-Ras, G. Kostorz, A. Romeo, D. Rudmann, A.N. Tiwari, Structural and
chemical investigations of CBD- and PVD-CdS buffer layers and interfaces in
Cu(In,Ga)Se2-based thin lm solar cells, Thin Solid Films 480-481 (2005)
118123.

[21] M. Powalla, P. Jackson, E. Lotter, D. Hariskos, S. Paetel, R. Wurz,


R. Menner,
W. Wischmann, unpublished.
[22] C. Persson, A. Zunger, Anomalous grain boundary physics in polycrystalline
CuInSe2: the existence of a hole barrier, Phys. Rev. Lett. 91 (26) (2003)
266401-1266401-4.
[23] J.Y.W. Seto, The electrical properties of polycrystalline silicon lms, J. Appl.
Phys. 46 (12) (1975) 52475254.
[24] S. Sadewasser, Th. Glatzel, S. Schuler, S. Nishiwaki, R. Kaigawa, M. Ch., LuxSteiner, Kelvin probe force microscopy for the nano scale characterization of
chalcopyrite solar cell materials and devices, Thin Solid Films 431-432 (2003)
257261.
[25] D. Fuertes Marron, S. Sadewasser, A. Meeder, T. Glatzel, M.Ch. Lux-Steiner,
Electrical activity at grain boundaries of Cu(In,Ga)Se2 thin lms, Phys. Rev. B 71
(2005) 033306-1-4.
[26] C.S. Jiang, R. Nou, J.A. AbuShama, K. Ramanathan, H.R. Moutinho, J. Pankow,
M.M. Al-Jassim, Local built-in potential on grain boundary of Cu(In,Ga)Se2 thin
lms, Appl. Phys. Lett. 84 (18) (2004) 34773479.
[27] M. Nichterwitz, D. Abou-Ras, K. Sakurai, J. Bundesmann, T. Unold, R. Scheer,
H.W. Schock, Inuence of grain boundaries on current collection in
Cu(In,Ga)Se2 thin-lm solar cells, Thin Solid Films 517 (2009) 25542557.
[28] D. Azulay, O. Millo, I. Balberg, H.W. Schock, I. Visoly- Fisher, D. Cahen, Current
routes in polycrystalline CuInSe2 and Cu(In,Ga)Se2 lms, Sol. Energy Mater. Sol.
Cells 91 (2007) 8590.

[29] M. Kawamura, T. Yamada, N. Suyama, A. Yamada, M. Konagai, Grain boundary


evaluation of Cu(In1-xGax)Se2 solar cells, Jpn. J. Appl. Phys. 49 (2010)
062301-1-3.
[30] K. Taretto, U. Rau, J.H. Werner, Numerical simulation of grain boundary effects
in Cu(In,Ga)Se2 thin-lm solar cells, Thin Solid Films 480-481 (2005) 812.
[31] K. Taretto, U. Rau, Can grain boundaries improve the performance of
Cu(In,Ga)Se2 solar cells?, in: T. Gessert, S. Marsillac, T. Wada, K. Durose,
C. Heske (Eds.), Thin-Film Compound Semiconductor Photovoltaics 2007,
Warrendale, PA, 2007, p. Y09-01-1-6 Mater. Res. Soc. Symp. Proc. 1012.
[32] M. Gloeckler, W.K. Metzger, J.R. Sites, Simulation of polycrystalline
Cu(In,Ga)Se2 solar Cells in two dimensions, in: W. Shafarman, T. Gessert,
S. Niki, S. Siebentritt (Eds.), Thin-Film Compound Semiconductor Photovoltaics, F14, Warrendale, PA, 2005, p. 1-1-6 Mat. Res. Soc. Symp. Proc. 865.
[33] M. Gloeckler, J.R. Sites, W.K. Metzger, Grain-boundary recombination in
Cu(In,Ga)Se2 solar cells, J. Appl. Phys. 98 (2005) 113704-1113704-10.
[34] W.K. Metzger, M. Gloeckler, The impact of charged grain boundaries on
thin-lm solar cells and characterization, J. Appl. Phys. 98 (2005)
063701-1063701-10.
[35] S. Sadewasser, D. Abou-Ras, D. Azulay, R. Baier, I. Balberg, D. Cahen, S. Cohen,
K. Gartsman, G. Karuppiah, J. Kavalakkatt, W. Li, O. Millo, Th. Rissom,
Y. Rosenwaks, H.-W. Schock, A. Schwarzman, T. Unold, Nanometer-scale
electronic and microstructural properties of grain boundaries in Cu(In,Ga)Se2,
Thin Solid Films, in preparation.
[36] T.E. Everhart, P.A. Hoff, Determination of kilovolt electron energy dissipation
vs. penetration distance in solid materials, J. Appl. Phys. 42 (13) (1971)
58375846.
[37] J. Rechid, A. Kampmann, R. Reinecke-Koch, Characterising superstrate CIS solar
cells with electron beam induced current, Thin Solid Films 361362 (2000)
198202.
[38] C. Donolato, An alternative proof of the generalized reciprocity theorem for
charge collection, J. Appl. Phys. 66 (9) (1989) 45244526.
[39] R. Kniese, M. Powalla, U. Rau, Characterization of the CdS/Cu(In,Ga)Se2
interface by electron beam induced currents, Thin Solid Films 515 (2007)
61636167.
[40] L. Jastrzebski, J. Lagowski, H.C. Gatos, Application of scanning electron
microscopy to determination of surface recombination velocity: GaAs, Appl.
Phys. Lett. 27 (10) (1975) 537539.
[41] Donolato, Evaluation of diffusion lengths and surface recombination velocities
from electron beam induced current scans, Appl. Phys. Lett. 43 (1) (1983)
120122.
[42] J. Kavalakkatt, D. Abou-Ras, R. Caballero, M. Nichterwitz, T. Rissom, T. Unold,
H.-W. Schock, unpublished.
[43] R.E. Dunin-Burkowski, The development of Fresnel contrast analysis, and the
interpretation of mean inner potential proles at interfaces, Ultramicroscopy
83 (2000) 193216.
[44] C.T. Koch, A. Lubk, Off-axis and inline electron holography: A quantitative
comparison, Ultramicroscopy 110 (2010) 460471.
[45] L. Reimer, Transmission Electron Microscopy, 4th ed., Springer-Verlag, 1997.
[46] R.F. Egerton, Electron Energy-Loss Spectroscopy in the Electron Microscope,
Plenum Press, New York, 1996.

You might also like