Identifying the Electronic Properties Relevant to Improving the Performance of High Band-Gap Copper Based I-III-VI2 Chalcopyrite Thin Film Photovoltaic Devices: Final Subcontract Report, 27 April 2004-15 September 2007 Page: 58 of 76

                                
FIG. 39.   Effects of changing absorber        80
electronic properties on the cell fill-factor
and open circuit voltage calculated using                      :10-13/#10-14
SCAPS. Parameters varied to produce this       75 -
diagram included: the bandgap (BG) over a
range 1.1 to 1.3eV, the deep acceptor density                         10- >> 10-1
(NA) from  1.5 x 101 to 6 x 105 cm-3, the      70 -e: 10     3
deep acceptor electron capture cross section  L- -                     BG: 1.2 >>1.3
(6e) from 10"11 to 10" cm2, the hole carrier               'NA l'5>6 e: 102 10
density (p) from 1.5 x 1015 to 6 x 1015 cm 3, 65 BG: 1.2 >> 1.1 -. p: 1.5 >> 5
the density of a second deep, donor defect
(ND) from 1.5 x 101 to 5 x 101 cm-3, and the          N : 1.5  5
minority carrier mobility (pe) from  10 to         0.50  0.55  0.60  0.65  0.70  0.75
1000 cm2V~ s~1.                                               Voc (V)
cross section, the hole carrier density, the density of a second deep (donor-like) defect, and the
minority carrier mobility (variations in majority carrier mobility have nearly no effect). Thus,
for example, one can see that variation in the deep acceptor density alone has only a very small
effect, while an increase in an assumed deep donor defect decreases both the fill-factor and Voc.
We have now examined 4 or 5 possible microscopic scenarios to try to account for the
measured changes in the JVI cell performance together with the observed metastable increases
in deep acceptor and hole carrier density. Motivated by the recent Lany-Zunger model for
metastability in CIGS [33], we illustrate one particular fairly successful scenario in some detail.
Here we have considered the conversion of (positively charged) deep donors to (neutral) deep
acceptors with the accompanying increase of hole carriers to maintain charge neutrality. Table X
lists the parameters used for the SCAPS modeling in the annealed (i.e., initial) state of the
device. For the series of metastable states induced by light-soaking to obtain the JVJ data in
Fig. 37(a), we then increased the deep acceptor density in the manner determined from our
DLCP measurements, while decreasing the magnitude of a mid-gap donor level (located at
Ev+0.6 eV or above) by the same amount. The parameters used for this series of metastable
states are listed in Table XI. The resultant calculated JVIr curves are displayed in Fig. 37(c).
The agreement with the experimental data in Fig. 37(a) is now quite good. In particular, this
model naturally accounts for the fact that Jsc drops while Voc remains nearly constant.
However, there is one complication in this otherwise successful analysis: to accurately
account for the observed changes in fill-factor in the assumed defect conversion scenario, the
charge density in the CdS layer had to be adjusted slightly (from 9.5 x 10" to 8.7 x 10" cm-3)
This may indicate actual metastable changes in the junction region (such as the fact that our
DLCP profiles indicate an increased metastable deep acceptor level increase in the CIGS region
near the junction). We are hoping that we can account for this inferred change in the CdS charge
density by a more direct measurement in the future.

46