This study used electron paramagnetic resonance (EPR) spectroscopy to investigate the effects of various processing techniques on defects in CdTe solar cell materials, including CdCl2 etching and the introduction of Cu. The EPR spectra of all samples were dominated by a signal from substitutional Mn impurities. However, subtle variations between samples indicated that the processing affected other defect centers. CdCl2 treatment introduced two new peaks and reduced the amplitude of a broad background signal. The addition of Cu doubled the amplitude of a narrow signal with g=2.002, likely increasing the concentration of a defect such as a Te vacancy. The most complex spectrum was seen in the sample treated with both CdCl2 and Cu, suggesting interaction between the

1. Electron Paramagnetic Resonance as a Probe of Technologically
Relevant Processing Effects on CdTe Solar Cell Materials
Michael Pigott, Liam Young, Patrick M. Lenahan
Pennsylvania State University, University Park, PA, 16802, USA
Adam Halverson, Kristian W. Andreini, Bas A. Korevaar
G.E. Global Research, Niskayuna, NY, 12309, USA
Abstract — The performance of polycrystalline thin film
CdTe solar cells is limited by as yet poorly understood deep level
defects which serve as recombination centers. Electron
paramagnetic resonance (EPR) has unrivaled analytical power
and sensitivity in the identification of deep level defects in
semiconductors; however, very little EPR literature exists with
regard to technologically relevant processing of present day
polycrystalline CdTe solar cells. In this study we explore the
effects of several features important in these present day
processing techniques: (1) effects of CdCl2 etching and (2) effects
of Cu on the CdTe. Cu is widely used with CdTe solar cells and is
thought to play a significant role in CdTe solar cells performance.
Although the EPR spectra in all the cases we explored are
dominated by a substitutional Mn site signal, our results are
clearly sensitive to multiple changes caused by the
aforementioned processing parameters.
Index Terms — CdTe, thin film, deep level defects,
recombination centers, EPR.
I. INTRODUCTION
The Shockley-Queisser Limit for CdTe based solar cells is
about 30%. The present day performance limit of these
devices is considerably lower; the current record efficiency is
18.3% [1]. A dominating factor in this reduced efficiency is
the presence of deep level defects which serve as
recombination centers. Unfortunately, the physical and
chemical nature of these centers is not yet well understood.
Electron paramagnetic resonance (EPR) has a sensitivity of
about 1010
total defect density within the sample and
analytical power to identify CdTe deep level defects. In fact, a
significant number of EPR studies have already been carried
out [2]-[7]. However, very little in the CdTe EPR literature
deals with the effects of technologically meaningful
processing parameters. In this study we utilize EPR to
investigate the effect of CdCl2 treatment and the introduction
of Cu into CdTe.
EPR spectroscopy, places a sample in a large, slowly
varying magnetic field while simultaneously exposing the
sample to microwave frequency electromagnetic radiation. In
the simplest of cases, EPR takes place when the resonance
condition (1) is satisfied:
Hghv β= (1).
Here h is Planck’s constant, ν is microwave frequency, β is the
Bohr magneton, and H is the field of resonance. The g is, for
a single defect orientation within the magnetic field, a number
typically close to 2. It depends upon the orientation of the
defect within the magnetic field and is often expressed as a
second rank tensor. The resonance condition for a perfectly
isolated electron would be satisfied for g = 2.00232. In
essentially all systems, deviations from this free electron value
are observed; the deviations are due to spin orbit coupling. In
many defect systems a second important contribution to the
resonance condition results from interactions between
electrons and the magnetic moments of nuclei. These so
called hyperfine interactions can also provide convincing and
often detailed information about the physical and chemical
nature of paramagnetic defect sites as well as provide
information about defect electron wave functions. Again, in
quite simple cases with only a single magnetic nucleus, the
hyperfine interactions alter the resonance condition to that of
expression (2):
AmHghv I+= β (2),
where h, ν, g, β, and H are defined as in expression (1) and mI
represents the nuclear spin quantum number and A is an
orientation dependent hyperfine parameter, also generally
expressed as a second rank tensor. A nuclear spin of I can
have nuclear spin quantum number, mI, of –I, -(I-1), -(I-2), …,
(I-2), (I-1), I. Thus, a nucleus with a spin of 1/2 can exhibit mI
= -1/2 and +1/2. A nucleus with spin 5/2 can exhibit mI = -5/2,
-3/2, -1/2, 1/2, 3/2, 5/2. Generally 2I+1 local fields and thus
2I+1 resonance lines are generated by the presence of a single
nucleus of spin I. Often more than one magnetic nucleus
and/or more than one defect orientation exists within the
sample under study. It is generally a straightforward process to
extend the expressions (1) and (2) to these more complex
cases [8]-[9].
EPR results are typically presented as the first derivative of
the absorption signal. The x-axis is the magnetic field (Gauss)
of the EPR spectrum and y-axis is in arbitrary units [8]-[9].
II. EXPERIMENTAL TECHNIQUES
Our measurements utilized X Band Bruker Biospin EPR
spectrometers and a Janis Helium Cryostat. All involved EPR
measurements were taken at 5 K. All samples were
polycrystalline CdTe powders.
978-1-4799-3299-3/13/$31.00 ©2013 IEEE 0635

2. III. EXPERIMENTAL RESULTS
Figure 1 shows the EPR results taken on four
at 5 K. The samples are labeled: CdTe,
CdTe+Cu, and CdTe+CdCl2+Cu. The labels
explanatory. The CdTe+CdCl2 sample was subjected to a
CdCl2 treatment step. The CdTe+Cu samples were exposed to
Cu. The CdTe+CdCl2+Cu samples were exposed to both
CdCl2 and Cu in consecutive steps.
Fig. 1. EPR spectra of four CdTe samples. All measurements were
made at 5 K.
Figure 1 shows that the EPR traces of all the samples are
dominated by six strong lines in the central part of the scans.
In all four samples, the six lines are about the same size and
have a g value near the free electron g (2.0023)
CdTe samples, the measured six line spectrum g=2.006. The
measured hyperfine splitting is approximately 61 Gauss.
six-line spectrum is consistent with a near
spin 5/2 nucleus. There are relatively few atoms
100% abundant spin 5/2 nucleus. Al, Mn, I
only four elements that meet these requirements
Mn the most plausible. An earlier study by Schwartz
dealing with the Mn in CdTe reports an isotropic
2.00597 and a hyperfine splitting of 61 Gauss
entirely consistent with results presented in this paper
Therefore, with a high degree of certainty, we associate the six
line spectrum with Mn. Although the spectra are dominated by
a Mn associated defect, of much greater interest are
ESULTS
four CdTe samples
d: CdTe, CdTe+CdCl2,
The labels are self-
sample was subjected to a
step. The CdTe+Cu samples were exposed to
+Cu samples were exposed to both
. All measurements were
Figure 1 shows that the EPR traces of all the samples are
dominated by six strong lines in the central part of the scans.
are about the same size and
the free electron g (2.0023). In all the
CdTe samples, the measured six line spectrum g=2.006. The
is approximately 61 Gauss. The
line spectrum is consistent with a near 100% abundant
There are relatively few atoms with a near
. Al, Mn, I, and Pr are the
only four elements that meet these requirements, with Al and
by Schwartz et al.
reports an isotropic g value of
of 61 Gauss [6], results
irely consistent with results presented in this paper.
Therefore, with a high degree of certainty, we associate the six
line spectrum with Mn. Although the spectra are dominated by
of much greater interest are the subtle
sample to sample variations resulting from the differences in
processing.
In Figures 2, 3, and 4 we compare the CdTe+CdCl
CdTe+Cu and CdTe+CdCl2+Cu spectra respectively to that of
CdTe. In these figures we have normalized the amplitudes
based on the dominating 6 line spectrum. We normalized the
amplitudes in order to compensate for slight differences in
sample volume. These comparisons clearly show that these
processing steps have effects on the EPR spectrum.
Fig. 2. EPR spectra of CdTe compared to CdTe+CdCl
Measurements were made at 5 K.
Figure 2 clearly shows that the CdCl
small reduction in the amplitude of the wide background
signal that encompasses most of the scan and interferes with
the 4th
Mn peak. The CdCl2 treatment also introduces two
additional peaks at 3173 G and 3633 G.
The comparison illustrated in Figure 3 does not
gross differences between the simple CdTe and CdTe+Cu
save for a change around g=2.002, indicating that the Cu by
itself does not introduce high densities of any new
paramagnetic defects.
Figure 4 illustrates very significant differences between the
simple CdTe and the CdTe+CdCl
also indicates a sharpening of the spectra at 3173 G and 3633
G. Several other large differences are also noticeable
resulting from the differences in
In Figures 2, 3, and 4 we compare the CdTe+CdCl2 and
+Cu spectra respectively to that of
CdTe. In these figures we have normalized the amplitudes
based on the dominating 6 line spectrum. We normalized the
o compensate for slight differences in
sample volume. These comparisons clearly show that these
processing steps have effects on the EPR spectrum.
compared to CdTe+CdCl2.
shows that the CdCl2 treatment causes a
small reduction in the amplitude of the wide background
signal that encompasses most of the scan and interferes with
treatment also introduces two
additional peaks at 3173 G and 3633 G.
e comparison illustrated in Figure 3 does not show any
een the simple CdTe and CdTe+Cu,
save for a change around g=2.002, indicating that the Cu by
itself does not introduce high densities of any new
ustrates very significant differences between the
simple CdTe and the CdTe+CdCl2+Cu sample. This figure
he spectra at 3173 G and 3633
are also noticeable.
978-1-4799-3299-3/13/$31.00 ©2013 IEEE 0636

3. Fig. 3. EPR spectra of CdTe compared to CdTe+Cu.
Measurements were made at 5 K.
Fig. 4. EPR spectra of CdTe compared to CdTe+C
Measurements were made at 5 K.
compared to CdTe+Cu.
compared to CdTe+CdCl2+Cu.
In order to more clearly identify the
sample-to-sample differences we utilized the EPR simulation
program Easy Spin [10] to
dominating Mn signal. By subtracting this simulated Mn
spectrum from the four traces, we largely, though still quite
imperfectly, eliminated the dominating effect
reveal more subtle effects of the processing variation
Another set of spectra, utilizing this spectral subtraction
approach, is plotted in figure 5. Although the approximation
we obtain in the spectral subtraction is not perfect, it
effectively reduces the amplitudes of all the Mn peaks
extent that additional EPR spectra are
Fig. 5. EPR spectra of the CdTe samples
dominating Mn signal (imperfectly)
indicate a secondary EPR signal with a g value near that of the free
electron. Note the presence of the strongest and most complex
underlying spectrum in the sample with both chlorine and copper.
One spectrum which is now clearly
subtraction traces of figure 5 is a narrow
electron g with a value close to 2.002
signal nearly doubles in both samples that contain Cu
amplitude is approximately 80 percent bigger in C
relative to CdTe. This near doubling in amplitude
indicate that the addition of Cu also causes the concentration
of this defect to double. However,
energy might also account for this
shift in Fermi level is very possible, as Cu acts as a dopant for
CdTe. The CdCl2 appears to slightly reduce the
to more clearly identify the processing induced
differences we utilized the EPR simulation
to closely approximate the
. By subtracting this simulated Mn
spectrum from the four traces, we largely, though still quite
eliminated the dominating effects of the Mn to
of the processing variations.
a, utilizing this spectral subtraction
. Although the approximation
we obtain in the spectral subtraction is not perfect, it
reduces the amplitudes of all the Mn peaks to the
are clearly visible.
CdTe samples taken at 5 K with the
removed. The red arrows
with a g value near that of the free
Note the presence of the strongest and most complex
underlying spectrum in the sample with both chlorine and copper.
clearly evident in the spectral
subtraction traces of figure 5 is a narrow line near the free
close to 2.002. The amplitude of this
signal nearly doubles in both samples that contain Cu. The
80 percent bigger in CdTe+Cu
CdTe. This near doubling in amplitude may
indicate that the addition of Cu also causes the concentration
However, changes in the Fermi
energy might also account for this doubled amplitude. The
el is very possible, as Cu acts as a dopant for
o slightly reduce the g=2.002
978-1-4799-3299-3/13/$31.00 ©2013 IEEE 0637

4. signal. Its amplitude in the CdTe+CdCl2 sample is reduced by
approximately 20 percent compared to CdTe.
A third feature of all four spectra is a very broad underlying
spectrum several hundred Gauss wide. A plausible source of
this very broad line is the Cd vacancy, which has a strongly
anisotropic g tensor (gǁ = 1.9555 and g┴ = 2.143) [7]. In a
polycrystalline sample, such a g tensor would yield a very
broad line reasonably consistent with our admittedly crude
observations [10]. With regard to the narrow line with a g
close to that of the free electron: color center like defects
typically produce an EPR signal with a nearly isotropic or
isotropic g value near the free electron g. Such defects in
CdTe translate to a Te vacancy which has a well-documented
g=2.000 [7]. Unfortunately, the EPR signal for the Te vacancy
is typically only 50 G wide, so such an assignment does not
explain the very broad background signal.
Another defect studied previously with a g value near that
of the free electron is Cd+
. Cd+
is formed by the oxidation of
CdTe during the annealing process when temperatures exceed
300 o
C. Cd+
is characterized by one intense, sharp, isotropic
signal with a g=2.0002 [3]. The Cd+ signal is only a few
Gauss wide and is not consistent with the wide background
signal associated with this unknown defect. It may however be
responsible for the narrow g=2.0025 and 2.0031 lines
observed in all of the samples. The modest discrepancy in g
could presumably be associated with the limits on our
admittedly crude spectral subtractions.
It is of some interest that by far the most complex EPR
pattern is that of the CdTe+CdCl2+Cu sample. These results
suggest complex interactions between CdCl2 and Cu. The
other interesting features are the signals that appear at 3173 G
and 3633 G when CdTe is treated with CdCl2. An
interpretation of these EPR lines is currently not clear.
IV. CONCLUSIONS
The results of this preliminary study indicate that EPR has
the analytical power and sensitivity to identify defect centers
created in processing steps, which are technologically
meaningful in CdTe photovoltaic manufacturing. Although
more work needs to be done to conclusively link the structure
and energy levels to the various EPR spectra, the results
strongly suggest that EPR could be a powerful tool for
technologists involved in optimizing CdTe solar cell
processing.
REFERENCES
[1] M.A. Green, K. Emery, Y. Hishikawa, W. Warta, E.D. Dunlop,
Progress in Photovoltaics, 21:1, 1-11 (2013)
[2] K. Saminadayar, J. M. Francou, J. L. Pautrat, "Electron-
paramagnetic resonance and photoluminescence studies of cdte-
cl crystals submitted to heat-treatment", Journal of Crystal
Growth, vol.72, no.1-2, pp. 236-241 1985.
[3] Padam GK, Gupta SK, “Electron-spin resonance and
transmission electron microscopy studies of solution-grown
CdTe thin films.” Applied Physics Letters 53(10). 1988.
[4] P. Christmann, B. K. Meyer, J. Kreissl, R. Schwarz, K. W.
Benz, "Vanadium in cdte: An electron-paramagnetic-resonance
study", Physical Review B, vol.53, no.7, pp. 3634-3637, Feb
1996.
[5] G. Brunthaler, W. Jantsch, U. Kaufmann, J. Schneider,
"Electron-spin-resonance analysis of the deep donors lead, tin,
and germanium in cdte", Physical Review B, vol.31, no.3, pp.
1239-1243 1985.
[6] Schwartz RN, Wang CC, Trivedi S, Jagannathan GV, Davidson
FM, Boyd PR, Lee U, “Spectroscopic and photorefractive
characterization of cadmium telluride crystals codoped with
vanadium and manganese.” Physical Review B 55(23):15378-
15381. 1997.
[7] Meyer BK, Hofmann DM, “Anion and Cation Vacancies in
CdTe.” Applied Physics a-Materials Science & Processing
61(2). 213-215. 1995.
[8] Poole CP, Electron Spin Resonance: A Comprehensive Treatise
on Experimental Techniques. New York, New York: John Wiley
and Sons, 1983.
[9] Weil JA, Bolton JR, Wertz JE, Electron Paramagnetic
Resonance. New York, New York: John Wiley and Sons, 1994.
[10] Easy Spin, a comprehensive software package for spectral
simulation and analysis in EPR. J. Magn. Reson. 178(1), 42-55.
2006
978-1-4799-3299-3/13/$31.00 ©2013 IEEE 0638