Content uploaded by Timothy Gessert
Author content
All content in this area was uploaded by Timothy Gessert on Dec 26, 2017
Content may be subject to copyright.
Presented at the 2003 Spring Meeting of the MRS, April 21-25, 2003 San Francisco, MRS Proceedings Vol. 763, pp. 133-138 (2003)
Analysis of the ZnTe:Cu Contact on CdS/CdTe Solar Cells
T.A. Gessert, M.J. Romero, R.G. Dhere, and S.E. Asher
National Renewable Energy Laboratory (NREL), Golden, CO 80401, U.S.A.
ABSTRACT
We report on the recent use of cathodoluminescence (CL) to probe the depth-dependent
changes in radiative recombination that occur in CdTe devices during ZnTe:Cu contacting
procedures. These types of CL measurements may be useful to assist in linking impurity
diffusion (e.g., Cu) from the contact with depth-dependent variation in electrical activation
within the CdTe layer. Variable-energy CL suggests that diffusion from the ZnTe:Cu contact
interface may assist in reducing effects of shallow donors in the CdTe bulk, and yield higher
acceptor levels in the region near the contact. CL analysis near abrupt metal discontinuities
provides estimates of diffusion lengths for carriers associated with both excitonic and donor-to-
acceptor pair recombination. Finally, CL measurements at increasing excitation levels (i.e.,
increasing electron-beam current) provides estimates of the defect state density, as well as
providing evidence that discrete multiple defect bands may exist in CdTe prior to contacting.
INTRODUCTION
The demonstration of a manufacturable, stable, low-resistance, ohmic contact for p-CdTe
polycrystalline photovoltaic devices remains an important goal of the CdTe research community.
Devices with fill factors approaching 77% have been demonstrated by incorporating a Cu-doped
ZnTe contact interface layer between the CdTe absorber and a Ti metallization [1]. This
contacting process uses vacuum processing, including ion-beam milling instead of wet-chemical
etching, to fabricate the contact. In addition to potential manufacturing advantages, the high
degree of control afforded by the vacuum processes used for this contact can enable systematic
variation of critical aspects of the contact design. These types of studies suggest that the
contacting processes used in typical CdS/CdTe have a much greater role in device performance
than simply providing a low-resistance pathway for current transport. Previous work has shown
that various impurities enter the CdTe absorber layer during both the CdCl2 and the contacting
processes. Of these impurities, Cu is believed to be a very important component of ultimate
device performance, and significant effort has been directed toward understanding effects of Cu
in the device following various process steps and/or accelerated life testing procedures [2,3,4].
Understanding how contact processes lead to particular depth-dependent Cu concentrations is
critical to designing processes that produce reproducible device performance. However,
understanding how this baseline device operation in achieved, and identifying methods to
significantly improve performance requires that the depth-dependence of the electrical activation
of Cu in the CdS/CdTe device also be understood. Appropriate techniques in this type of study
would not only correlate Cu concentration with electrical properties, but provide insight into the
particular defect(s) that produce these properties. Because the formation of Cu-related defects
would be expected to affect recombination processes, our recent efforts toward analyzing the
depth-dependence of Cu activation have involved examination of the depth-dependent radiative
recombination using CL. This analysis indicates that significant modification of the radiative
recombination spectra occurs during contacting and suggests that understanding both the
Presented at the 2003 Spring Meeting of the MRS, April 21-25, 2003 San Francisco, MRS Proceedings Vol. 763, pp. 133-138 (2003)
diffusion profiles and depth-dependant recombination may yield insight into device evolution
and stability.
EXPERIMENTAL DETAILS
CdS/CdTe materials used in this investigation were produced by both First Solar and NREL.
The CdS and CdTe layers of the First Solar material were deposited by close-spaced sublimation
(CSS) at ~580°C onto 5-mm-thick soda-lime glass to nominal thicknesses of ~300 nm and ~4.5
µm, respectively. This was followed by a wet CdCl2 treatment at First Solar. Previous
development of the NREL ZnTe:Cu/Ti contact on this type material has produced devices with
open-circuit voltage (Voc) of ~820 mV, fill factors of 77%, and device efficiency of ~10%-11%
(performance of these devices is limited primarily by low current density due to thick CdS) [1].
The NREL CdS/CdTe device material used 1-mm-thick 7059 glass substrates, a 450-nm
SnO2:F/SnO2 bilayer formed by chemical-vapor deposition (CVD) of tetramethyltin and oxygen,
~100-nm-thick CdS formed by chemical-bath deposition, an 8-µm-thick CdTe layer formed by
CSS, and a vapor CdCl2 treatment [5]. Representative efficiencies for these particular NREL
devices (processed with a Cu-doped graphite-paste contact) were 11%-12%.
The ZnTe:Cu/Ti back contact was produced using a sequential process involving 2-h
temperature equilibration at ~350°C, ion-beam milling the CdTe surface to a depth of ~100 nm,
r.f.-sputter deposition of ~0.5 µm of ZnTe:Cu (~6 at.% Cu), and 0.5 µm of d.c.-sputter-deposited
Ti (as the sample cooled from ~300° to 100°C).
CL measurements were performed in a JEOL 5800 scanning electron microscope at various
electron-beam energies and beam currents. CL spectra were acquired with a Princeton LN/CCD-
1340/400 cryogenic charge-coupled device. Wavelength-dispersive images were reconstructed
from the CL spectra by synchronizing spectra acquisition with the electron-beam positioning
system. Secondary-ion mass spectroscopy (SIMS) was performed from the contacted side of the
devices using a Cameca IMS-5F instrument tuned for a mass resolution (M/∆M) of ~4000 to
allow for separation of Cu from the Te and S species.
DISCUSSION
Figure 1 shows representative SIMS
analysis of the NREL ZnTe:Cu/Ti contact, as
well as comparisons with contacts produced
on the same CdS/CdTe material using Cu-
doped graphite-paste produced by NREL and
the University of South Florida (USF). These
quantified depth profiles show that the
various contacting processes produce similar
Cu concentration profiles. The figure also
shows that a significant amount of the Cu
enters the CdTe layer during the CdCl2
treatment. It is also noteworthy that the
ZnTe:Cu contact produces a lower
concentration Cu in the CdS layer, even
though it experiences a higher processing
Figure 1. SIMS analysis of Cu levels in First Solar vapor-
transport deposited material that was wet-CdCl2 treated and
contacted with the USF C/Cu paste (USF), the NREL C/Cu
paste (NREL), and the NREL ZnTe:Cu/Ti (NREL ZnTe).
Presented at the 2003 Spring Meeting of the MRS, April 21-25, 2003 San Francisco, MRS Proceedings Vol. 763, pp. 133-138 (2003)
temperature than the paste contacts (~350°C vs. ~200°C). Potential effects of these differences
on device performance and/or stability have not been correlated. However, because both the
performance and Cu concentration profiles of these devices are similar, it seems likely that
improved understanding of the role of Cu in any of these contact may provide significant insight
into the function of other typical contacts.
Figure 2 shows the configuration used for
CL analysis. One of the methods available to
analyze the near-contact region is to vary the
energy of the electron beam. As indicated in
Table 1, this method can be used to probe
radiative recombination from ~0.1 to ~3 µm
from the contacted surface. In the example
shown in Fig. 3, the beam is incident through
either a ~50 nm or ~200 nm ZnTe:Cu contact
interface that has been deposited onto a
CdS/CdTe device following ion-beam milling.
The CL spectra show two primary luminescence
peaks: a high-energy peak at ~1.5-1.6 eV, and a
broader peak centered at ~1.4 eV. The higher-
energy peak is ascribed to excitonic and
shallow-donor transitions, whereas the broad
peak at ~1.4 eV is believed to be related to
donor-to-acceptor (DAP) transitions and related
phonon replicas. The acceptor states of the
DAP transitions are believed to correspond to
cadmium vacancy (VCd) complexes, such as the
A center [6,7]. Analysis of the exciton
transition in Fig. 3 shows that it is attenuated by the presence of the ZnTe:Cu layer. Deeper
penetration through the thinner ZnTe:Cu layers (i.e., 20-keV beam, 50-nm ZnTe:Cu) produces a
more intense CL signal and a substantial shift in the exciton peak. It has been suggested that the
shift may be due to the probed CdTe region containing more shallow donor states than the region
nearer the contact interface. If this is true, it may suggest that the uncontacted CdTe is less p-
type than often assumed, and that sufficient Cu diffusion from the contact is required to effect
not only higher acceptor level near
the contact, but also produce CdTe
bulk material with higher acceptor
levels [3]. In contrast to the exciton
peak, the position of the DAP
emission remains relatively
unchanged by increasing ZnTe:Cu
thickness and its intensity increases
with increasing penetration depth.
Similar CL analysis at low energy
(10 keV) suggests that the Cu
diffusion from the ZnTe:Cu (as
observed by SIMS analysis above)
Table 1. Maximum range of penetration for primary-
beam electrons and density of states associated with
DAP radiative recombination.
Eb (keV) 5 10 20 30
Re (µm) 0.12 0.41 1.39 2.83
*DAP
Density (cm-3)2x1011 2x1012 3x1013 8x1013
DAP
Density (cm-3)- 6x1010 5x1011 8x1012
*Before ZnTe:Cu contact
Re
e-beam
CL
Mirror
Glass Substrate
SnO2
CdS
CdTe
ZnTe:Cu
x
Metal
Figure 2. Schematic diagram illustrating the analysis
configuration used for CL analysis.
120x10 3
80
40
0
CL Intensity (Counts)
1.61.51.41.3 Ener
gy
(eV)
ZnTeCu500_10keV
ZnTeCu500_20keV
ZnTeCu2000_10keV
ZnTeCu2000_20keV
DAP
Exciton
Figure 3. Variable-energy CL spectra (10 eV or 20 eV) of CdTe
back surface following ZnTe:Cu/Ti deposition (500 Å or 2000 Å)
with Ti layer removed. Temperature =77 K.
Presented at the 2003 Spring Meeting of the MRS, April 21-25, 2003 San Francisco, MRS Proceedings Vol. 763, pp. 133-138 (2003)
produces emission consistent with the formation of CuCd defects [8] whereas higher energy
analysis (30 keV) shows that the DAP complex remains active deeper in the CdTe [9]. These
DAP results also suggest that electrical activation of Cu, and thus the electrical properties of the
CdTe, change significantly as a function of distance from the ZnTe:Cu/CdTe interface.
Another CL analysis technique useful for contact studies is based on the fact that, although
an electron beam can penetrate a thin metal pad, this same metallization will reflect optical
luminescence generated directly beneath it. However, if the lifetime of the minority carriers is
sufficient long, some recombination may occur outside of the metal pad, and a CL signal can be
detected. Figure 2 indicates the configuration used for this type of analysis. In this figure, as the
electron beam is scanned from right to left, the beam will encounter the edge of the metal pad.
Once this occurs, only the radiative recombination of carriers with diffusion length longer than
the distance to the metal edge (x) will be observed. Analysis of the CL intensity as a function of
x can provide separate estimates of the diffusion length for exciton and DAP-related diffusion
(Lx and LDAP, respectively) [7]. Successful application of this technique requires that an abrupt
metal edge can be formed, and that any surface layers (i.e., contact interface layers) remaining
outside of the metallization are transparent to the luminescence.
Because the bandgap of the ZnTe:Cu layer is ~2.25 eV, and because photolithographic
techniques and selective etches are used to pattern the contacts, we plan to use this technique to
probe this contact. In contrast, the Cu-doped paste contact used for NREL “baseline devices”
incorporates a chemical etch (nitric acid + phosphoric acid) to form a thick Te layer prior to Cu-
doped paste application. In this case,
the Te layer is optically dense to the
luminescence, and thus, a CL signal
cannot be acquired. Nevertheless, the
technique has been used to provide
baseline information on the NREL
CdTe device material prior to
contacting. In this case (data shown
in Fig. 4), a 2x2 pattern of 500-µm x
500-µm pads was formed using a 25-
50-nm thick Al layer deposited
through a shadow mask at room
temperature. Although we have not
yet explored if possible Al diffusion
affects the analysis, (because of the
low-temperature of the deposition
process) we assume diffusion is
minimal and the Al pad functions
only as a CL reflector.
Figure 4 shows that the exciton
peak of the CL spectrum attenuates
quickly as the beam proceeds into the
metallization. In contrast, the relative
contribution of the 1.4-eV peak
decreases much more slowly. Taken
together, these results suggest a
DAP
X
X
DAP
LX
LDAP
10 5 0
10
-3
10
-2
10
-1
10
0
CL Intensity (norm)
Distance
(µµ
µµ
m
)
Figure 4. Secondary-electron image of a Al metallized NREL-
deposited CdTe film (left side) and spectra acquired for the
different locations. The relative contribution of the DAP
transitions to the exciton emission (X) increases drastically over
the distance from the edge of the pattern, suggesting a longer
diffusion length for the DAP recombination (LDAP) compared with
the exciton diffusion length (LX).
Presented at the 2003 Spring Meeting of the MRS, April 21-25, 2003 San Francisco, MRS Proceedings Vol. 763, pp. 133-138 (2003)
longer diffusion length for carriers that recombine through the DAP-related emission relative to
diffusion length of carriers recombine via excitons. Indeed, straight-line fits of the appropriate
portion of these data in Fig. 4 indicated LX and LDAP of 1.5 µm and 5.5 µm, respectively [7].
Another method used to explore recombination processes near the contact region is to
monitor the CL spectra as a function of electron-beam current (i.e., excitation intensity). As the
number of beam-generated carriers increases, the relative intensities of the exciton and DAP
evolve. Although the number of excitonic states available is unlimited, the DAP-related
recombination is limited by the availability of these defects. Saturation of the DAP transition
occurs when the density of carriers generated by the electron beam is comparable to the density
of defect complexes involved in a particular transition [10].
Figure 5a shows CL emission spectra of a CdCl2-treated CdTe film prior to contacting for
four selected beam intensities (as defined relative to the S1 and S2 markers in Fig. 5b). These
spectra have been normalized to illustrate how the DAP emission shifts to higher energies as the
excitation level increases. An alternative method to illustrate this effect is shown by comparing
Figs. 5b and 5c. Figure 5b shows the same analysis as in Fig. 5a, but with the luminescent
intensity mapped against the log of the beam current. Figure 5c shows a sample that is
nominally identical to that used in Figs. 5a and 5b, but after contacting with a 0.5-µm-thick
ZnTe:Cu layer. Figure 5c reveals that the DAP shift to higher energy is not observed in the
contacted sample. Estimates of the DAP state density from these CL spectra (combined with
diffusion length on these samples performed similar to Fig. 4) are provided in Table 1 [7,11].
The spectral shift in the DAP luminescence shown in Figs. 5a and 5b may be a manifestation of
the existence of distinct bands of defect levels in the CdTe film prior to contacting. A schematic
illustration of this possibility is shown in Fig. 6. (It is also noted that the existence of multiple
discrete bands is not observed for similar analysis performed on polycrystalline CIGS [10]).
Because this shift is not observed for the ZnTe:Cu contacted sample, we considered that it may
1.2 1.3 1.4 1.5 1.6
0.0
0.2
0.4
0.6
0.8
1.0
> S2
S2
S1
<< S1
DAP X
CL Intensity (norm)
Photon energy (eV)
10
-1
10
0
XDAP
I
b
(nA)
10
-1
10
0
1.3 1.4 1.5 1.6
Photon energy (eV)
I
b
(nA)
S1
S2
Figure 5. (a) Normalized CL spectra of non-contacted
CdTe illustrating shift of DAP emission with increasing
beam current at various levels shown in (b). (b-c) Effect
of electron-beam current on emission spectrum (b) prior
to and (c) after contacting.
(a)
(
b
)
(c)
Ib
(nA)
Ib
(nA)
Presented at the 2003 Spring Meeting of the MRS, April 21-25, 2003 San Francisco, MRS Proceedings Vol. 763, pp. 133-138 (2003)
be possible to probe depth-
dependent Cu activation by using
CL at high beam current along
CdTe cross-sections.
Unfortunately, as has been
observed in other surface analysis
techniques, CdTe samples are
highly resistive unless Cu has
been allowed to diffuse into the
sample. Therefore, non-contacted
CdTe samples experience
significant charging at high beam currents, and this leads to significant problems in CL analysis.
CONCLUSIONS
Expanding our level of understanding of contact processes on the CdS/CdTe device will
require measurements that probe the extent of electrical activation of Cu in the various regions of
CdTe. We have begun to use CL to assess the depth-dependant radiative recombination in the
region near the back contact. These analyses have indicated that recombination processes differ
significantly with depth into the CdTe. Variable-energy spectroscopic CL analysis provided
baseline information related to how various steps of the contact process affect the radiative
recombination as a function of depth. Variations on these techniques have also provided insight
into the minority-carrier lifetime associated with particular recombination mechanisms. Finally,
by varying the electron-beam current (i.e., excitation level), the density of DAP defect levels was
estimated, and the possibility of multiple discrete bands of defect levels established.
ACKNOWLEDGEMENTS
The authors acknowledge First Solar LLC for samples and valuable technical discussion, Chitra
Narayanswamy of NREL for providing the NREL Cu-paste contact, and Chris Ferekides of the
University of South Florida for providing samples with the USF Cu-paste contact. This research
was supported under DOE Contract No. DE-AC36-99GO10337.
REFERENCES
1. T.A. Gessert, A. Duda, S.E. Asher, C. Narayanswamy, and D. Rose, Proc. 28th IEEE Photovoltaic Specialists Conf.,
Anchorage, Alaska, Sept. 15-22, 2000 (IEEE, Piscataway, NJ, USA) pp. 654-657.
2. S.E. Asher, et. al., Proc. 28th IEEE Photovoltaic Specialists Conf., Anchorage, Alaska, Sept. 15-22, 2000 (IEEE,
Piscataway, NJ, USA) pp. 479-482.
3. T.A. Gessert, M.J. Romero, S. Johnston, B. Keyes, and P. Dippo, Proc. 29th IEEE PV Specialists Conf., New Orleans,
LA, May 20-25, 2002 (IEEE, Piscataway, NJ, USA) pp. 535-538.
4. K.D. Dobson, I. Visoly-Fisher, G.Hodes, D. Cahen, Solar Energy Materials & Solar Cells 62 295-325 (2000)
5. D. Rose, F. Hasoon, R. Dhere, D. Albin, R. Ribelin, X. Li, Y Mahathongdy, T. Gessert, and P. Sheldon, Prog.
Photovolt: Res. Appl. 7, 3312-340 (1999).
6. D.M. Hofmann, P. Omling, and H.G. Grimmeiss, Phys. Rev. B, 45 (11) 1992 6247-6250.
7. M.J. Romero, T.A. Gessert, and M.M. Al-Jassim, Appl. Phys. Lett., 81, (17) 3161-3162 (2002).
8. D. Grecu, A.D. Compaan, D. Young, U. Jayamaha, and D.H. Rose, J. Appl. Phys. 88 (5) (2000) 2490-2496.
9. T.A. Gessert, M.J. Romero, and S. E. Asher, Proc. NREL NCPV Review Meeting, Lakewood, CO, Oct. 14-17, 2001,
pp. 189-190.
10. M.J. Romero, W. Metzger, T.A. Gessert, D.A. Albin, and M.M. Al-Jassim, Proc. NREL NCPV Review Meeting,
Denver, CO, March 24-26 (2003).
11. M.J. Romero, T.A. Gessert, M.M. Al-Jassim, R.G. Dhere, D.S. Albin, and H. R. Moutinho, Symposium F: Defect- and
Impurity-Engineered Semiconductors and Devices III, Spring 2002 MRS Meeting.
Figure 6. Schematic illustration of the type of defect bands that
may exist in CdCl2-treated CdTe, based on the CL analysis at
increasing beam energy. Figures left to right indicate evolution of
luminescent transitions with increasing beam current
C
V
C
V
C
V