Oxygen during vapor CdCl2 VCC treatments significantly reduced resistive shunts observed in CdSCdTe

1
Polycrystalline Thin Film Device Degradation
Studies D. Albin1, S. Demtsu2, T. McMahon1, A.
Davies2, J. Pankow1 and R. Noufi1 1 National
Renewable Energy Laboratory, Golden, CO 2
Colorado State University, Fort Collins, CO
Abstract
General Processing Effects
Variations in small-area efficiency for CdTe
devices based upon differences in CdTe and CdS
thickness, and the presence or absence of both NP
etch and oxygen during the VCC process are shown
in Figure 1. Three general behaviors are
apparent. For thinner CdS devices (left column),
efficiency is observed to decrease in a
non-continuous, catastrophic behavior which can
also be reversible. The remaining two behaviors
are determined by their etch status. NP etched
devices degrade to a stabilized level. Non-etched
devices show some of this same characteristic
though the stabilized level may be much lower.
Oxygen during vapor CdCl2 (VCC) treatments
significantly reduced resistive shunts observed
in CdS/CdTe polycrystalline devices using thinner
CdS layers during 100 C, open circuit, 1-sun
accelerated stress testing. Cu oxidation
resulting from the reduction of various trace
oxides present in as-grown and VCC treated films
is the proposed mechanism by which Cu diffusion,
and subsequent shunts are controlled. Graphite
paste layers between metallization and CdTe
behave like diffusion barriers and similarly
benefit device stability. Ni-based contacts form
a protective Ni2Te3 intermetallic layer that
reduces metal diffusion but degrades performance
through increased series resistance.
These degradation behaviors are reflected by
changes in the J-V characteristics of devices
during stress. The degradation observed in
Figure 2(a) manifests primarily as a loss in Voc
and FF and is typically seen in thicker CdS/CdTe
devices using NP etch. This degradation mode is
due to increased recombination associated with Cu
Experimental Background
diffusion from the contact to the junction and
is also highlighted by an increase in 1st
quadrant roll-over. Changes shown in Figure 2(b)
correspond to thicker devices where NP is absent
and, in addition to recombination loss, resistive
losses are observed. Finally, Figure 2(c) shows
the type of catastrophic shunting observed when
thinner CdS devices are used.
Results from two separate studies are
presented. The first study focuses on more
general CdS/CdTe device processing CdS and CdTe
film thickness (60-100 nm and 8-11 ?m
respectively), the effect of O2 during VCC
treatments, and the use of nitric-phosphoric (NP)
pre-contact etches. The second study focuses on
back contact processing in which the use (or
absence) of graphite dag layers (used to
deliver Cu-dopant) is studied with two different
(Ag,Ni) metallic pastes. In both studies,
stability was determined by device analysis
performed during 1-sun, open-circuit bias, 100 C
stressing. Dark soak stabilization at 25 C for
12-24 hr preceded all measurements.
Oxygen during VCC clearly lessens the
probability of resistive shunts during stress
when thinner CdS layers are used. GIXRD analysis
of the CdTe surface after different processing
stages (Figure 3) show the presence of TeOx,
CdTeOx, and possibly CuTeOx phases near the CdTe
back surface.
CuxO phase formation, either by direct
oxidation, or reduction of existing oxides
(Figure 4) is proposed as the mechanism by which
Cu diffusion (and perhaps other contact metals)
is reduced, thus improving device resistance to
shunting.
The diffusion of Ag in CdTe, suspected at grain
boundaries, resulted in a wider distribution
(greater non-uniformity) in absolute quantum
efficiency measured using a 100-?m, 638 nm laser
(Figure 8) which developed after stress. The
apparent reduced diffusion of Ni was attributed
to the formation of a Ni3Te2 intermetallic layer
which was detected by GIXRD at the Ni/CdTe
interface after removing the Ni. No such
intermetallic was observed at the Ag/CdTe
interface. Resistive shunting, observed in one
Ag-only cell was not observed in any Ni-only
device. However, the increased resistance of the
Ni2Te3 layer contributed to a significant
increase in device series resistance.
Back contact Processing
The stability of CdTe devices made using 4
different back contact structures was studied
(Figure 5). In half of these devices, the Cu
doped graphite layer was left intact between the
metal and CdTe surface (graphite/Ag and
graphite/Ni). In the remaining devices, the
graphite was removed after the 280 C dopant
drive (Ag-only and Ni-only).
No significant difference in performance or
stability was observed among devices with the
graphite layer present. The graphite appears to
behave like a diffusion barrier, possibly due to
a cation exchange mechanism involving the
polyacrylic acid (PAA) constituent of this paste.
In the absence of graphite, devices showed much
faster degradation.
Compositional mapping using Auger electron
spectroscopy also identified a possible
degradation mechanism associated with the
Ag-paste used in our devices (Figure 10). Ag
appears to react with Cl from the PVC used in
paste binder to form AgCl, an insulator.
Acknowledgements
Capacitance-Voltage (CV) measurements before
and after stress for these devices (Figure 7)
appear to behave identically with the sole
exception of the Ag-only case. The decreased
carrier density and increased depletion width
after stress observed in the graphite/Ni,
graphite/Ag, and Ni-only devices is commonly
observed in cells where Cu is depleted from the
back contact. The absence of this behavior in the
Ag-only devices may be due to replenishment of Ag
from the Ag contact which behaves like an
infinite source. Ag is known to be a rapid
diffuser, as well as dopant in CdTe.
We acknowledge the support of the DOE under
contract DE-AC36-99-GO10337. We also thank Jim
Sites (CSU) and Sally Asher (NREL) for many
useful discussions.