Vous êtes sur la page 1sur 7

Thin Solid Films 520 (2011) 11671173

Contents lists available at ScienceDirect

Thin Solid Films


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t s f

The role of transparent conducting oxides in metal organic chemical vapour


deposition of CdTe/CdS Photovoltaic solar cells
S.J.C. Irvine , D.A. Lamb, V. Barrioz, A.J. Clayton, W.S.M. Brooks, S. Rugen-Hankey, G. Kartopu
Centre for Solar Energy Research (CSER), Glyndr University, OpTIC Glyndr, St Asaph Business Park, LL17 0JD, UK

a r t i c l e i n f o a b s t r a c t

Available online 4 May 2011 A systematic study is made between the relationship of Cd0.9Zn0.1S/CdTe photovoltaic (PV) device properties
for three different commercial transparent conducting oxide (TCO) materials and some experimental CdO to
Keywords: determine the role of the TCO in device performance. The resistance contribution from the TCO was measured
Transparent conducting oxides after depositing the gold contact architectures directly onto the TCOs. These were compared with the
Indium tin oxide Cd0.9Zn0.1S/CdTe device properties using the same contact arrangements. Series resistance for the commercial
Cadmium oxide
TCOs correlated with their sheet resistance and gave good agreement with the PV device series resistance for
Cadmium telluride
Cadmium zinc sulphide
the indium tin oxide (ITO) and uorine doped tin oxide (FTO) 15 /Sq. superstrates. The devices on the
Fluorine doped tin oxide thicker FTO 7 /sq superstrates were dominated by a low shunt resistance, which was attributed to the rough
Solar cells surface morphology causing micro-shorts. The device layers on the CdO substrate delaminated but devices
were successfully made for ultra-thin CdTe (0.8 m thick) and compared favourably with the comparable
device on ITO. From the measurements on these TCOs it was possible to deduce the back contact resistance
and gave an average value of 2 .cm 2. The correlation of ll factor with series resistance has been compared
with the predictions of a 1-D device model and shows excellent agreement. For high efciency devices the
combined series resistance from the TCO and back contact need to be less than 1 .cm 2.
2011 Elsevier B.V. All rights reserved.

1. Introduction TCO combination of zinc stannate and cadmium stannate for the TCO
[4]. Performance of CdTe PV cells using commercially available TCOs is
Transparent conducting oxides (TCO) play a vital part in the more in the 1415% range and for modules, where the limitations of
overall efciency of thin lm photovoltaic (PV) solar cells. However, the TCO conduction tend to be more apparent, the best is around 11%
the precise contribution to the overall performance of the PV cell is [5]. The issues in transferring from idealised laboratory PV cells to PV
often difcult to quantify. The TCO provides the so called front modules have been highlighted in the recent paper by Jingquan et al.
contact which has to allow the solar radiation to be transmitted to [6] where it was shown that the laser groove width in the TCO, and
the underlying cell and to conduct the generated electricity to the metallisation width, can signicantly affect the module efciency and
external circuit. Thus it is clear that for efcient cells the TCO even under the best conditions the ll factor (FF) did not exceed 58%.
transmission of solar radiation must be highly coupled with very low This indicates the difculty in translating performance from idealised
electrical resistance; however other properties of the TCO can also laboratory PV cells to module device architectures. For laboratory cells
inuence the cell performance in more subtle ways. A comprehensive the FF is expected to be over 70% and the decrease in modules is
review of the TCOs used for CdTe PV cells is given in the review by indicative of higher series resistance.
Bosio et al. [1]. The commonly used TCOs are uorine doped SnO2 This paper investigates the role that different TCOs play in the
(FTO) and indium tin oxide (ITO) with an increasing interest in performance of CdTe thin lm PV devices. The deposition method for
trivalent doped ZnO. The need for improved conductivity, achieved by the window and absorber layer in this study is metal organic chemical
improving TCO mobility rather than the carrier concentration, is vapour deposition (MOCVD) which enables excellent control over the
recognised as a factor in improving PV cell and module performance lm thickness and physical properties of the deposited layers [7]. For
[2]. High carrier concentrations in the region of 10 20 to 10 21 cm 3 can example, the conventionally used window layer CdS is alloyed with
reduce transmission, particularly in the infrared part of the solar Zn in the MOCVD process to allow more of the blue light to pass
spectrum [3]. The world record CdTe cell performance has remained through to the absorber layer than with CdS alone. Further, the CdTe
on the value set by Wu et al. in 2001 and this was achieved using a layer is doped, in situ, with arsenic (As) to increase the p-type carrier
concentration and this doping is increased towards the back contact
which enables non-rectifying contacts to be formed without any etch
Corresponding author. treatment [79]. These features were considered to be important in
E-mail address: Stuart.Irvine@optictechnium.com (S.J.C. Irvine). the present study where the purpose is to investigate the role of the

0040-6090/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.tsf.2011.04.194
1168 S.J.C. Irvine et al. / Thin Solid Films 520 (2011) 11671173

TCO in the nal performance of the device. A study of amorphous the bus-bars around the edge of the sample. For the PV device
silicon thin lm PV on uorine doped tin oxide (FTO) has shown the fabrication the CdTe layer was removed from the areas shown by the
importance of morphology of the TCO in addition to the usual bus-bars prior to evaporation of the gold. Thus the front and back
parameters of high transmission and low resistance [10]. Four contacts were made in a single evaporation step where the revealed
different TCOs are investigated in this paper: two commercial FTO areas of TCO provided the front contact. The difference between the
materials (NSG TEC C15 and NSG TEC A7), commercial indium tin two device architectures was in the network of front contact bus-bars.
oxide (ITO), and some experimental cadmium oxide (CdO). The The simple conguration in Fig. 1(a) relies on larger distances for
device layers of Cd0.9Zn0.1S and CdTe are deposited onto these lateral conduction in the TCO compared to the improved congura-
different TCO superstrates in order to investigate the relationship tion shown in Fig. 1(b) where the bus-bar extends further around the
between the properties of TCO and PV device performance on periphery of the sample and there is an addition of gold ngers
laboratory cells. extending between the rows of devices. The typical evaporated gold
thickness was 50 to 60 nm, as measured using a dektak150 from
2. Experimental details Veeco. For the PV devices it was possible to test both congurations on
the same sample to remove potential effect of small variance in the
The commercial NSG TEC C15 and NSG TEC A7 superstrates on CdTe and Cd0.9Zn0.1S between the MOCVD growth runs. The rst stage
soda lime glass were supplied by Pilkington Technology Centre. The was to deposit the contact architecture in Fig. 1(a) and following the
ITO on boro-alumino silicate glass was supplied by Delta Technologies IV device measurements using an Abet Technologies AM1.5
and the CdO was grown in our own laboratory onto Corning 1737 calibrated solar simulator, the contact architecture in Fig. 1(b) was
boro-alumino silicate glass substrates using MOCVD [12]. The overlaid and again IV measurements were repeated.
precursors for the CdO deposition were dimethylcadmium (DMCd)
and tertiarybutanol (tBuOH). The 500 nm thick lms were grown at
291 C which had been shown previously to give low resistance lms 3. Results
for the undoped CdO [12].
The Cd0.9Zn0.1S/CdTe device structure was grown using the MOCVD 3.1. Properties of TCOs
process described previously [7]. This entailed in situ doping of the 2 m
thick CdTe layer with As from trisdimethlyaminoarsenic (tDMAAs) with The results of the Hall measurement and 4-point probe measure-
a termination of high concentration of As (N1 1019 cm 3) to give a low ments for the 4 types of TCO used in this study are shown in Table 1.
back contact resistance. A CdCl2 layer was deposited and annealed in The thicknesses of the TCOs vary such that the thinnest sample is the
situ in the MOCVD reactor, as described by Barrioz et al. [12], and no Delta Tech ITO but still achieves a low sheet resistance of 11 /sq
etching of the surface was required prior to evaporating the gold back through the high carrier concentration of 1.26 1021 cm 3. The two
contacts. NSG TEC glass samples differ in their thickness with values of 340 nm
The electrical properties of the TCO layers were characterised and 630 nm for TEC C15 and TEC A7 respectively. This leads to a lower
using a Jandel RM3-AR 4-point probe for sheet resistance and an sheet resistance for TEC A7 and the higher carrier mobility of
Ecopia HMS 3000 Hall measurement system to determine carrier 36.9 cm2/Vs compared with 27.8 cm2/Vs also contributes to this low
concentration and carrier mobility. The Hall measurements were sheet resistance. The CdO sample had the highest mobility and the
made, at room temperature, on cleaved 1 cm 2 samples with In-Sn lowest carrier concentration with values of 44.9 cm 2/Vs and
eutectic contacts soldered onto the extreme corners of the sample. 2.20 1020 cm 3 respectively. This led to a modest sheet resistance of
The optical transmittance was measured over the spectral range from 19.0 /sq. This set of TCO samples gives a signicant range of electrical
300 to 2000 nm using a Varian Cary 5000 spectrometer. properties and lm thickness. The latter also relates to surface
Laboratory cell device architectures were tested on 5 5 cm 2 TCO morphology and to microstructure where Oswald ripening mechanisms
superstrates by evaporating gold contact congurations directly onto will lead to larger grain size for thicker layers and is reected in the RMS
the TCO. This enabled measurement of lateral conduction through the surface roughness measured by A Veeco Dimension 2400 Atomic Force
different TCOs between the front contact bus-bars and the individual Microscope (AFM) shown in Table 1. All these properties are relevant
devices. These are shown in Fig. 1 where the devices are dened by when considering the nucleation, morphology and grain structure of the
the 0.5 0.5 cm 2 gold contacts and the front contacts are dened by MOCVD grown Cd0.9Zn0.1S/CdTe structure.
The specular transmittance spectra are shown in Fig. 2 for each of
the four types of TCO. The lower optical transmission on the NSG TEC
samples is not indicative of higher absorption in these lms but is
related to the rougher surfaces on these compared to the ITO and CdO
samples. The decrease in maximum transmission follows the trend of
increasing RMS roughness in Table 1. The short wavelength cut-offs
on these TCOs are in line with their optical bandgaps with the longest
wavelength cut-off for CdO around 500 nm corresponding to the
bandgap of 2.5 eV. This will cause some loss of photocurrent when

Table 1
Measured properties of the TCO lms using Hall effect and 4-point probe. All
measurements were made at 294 K.

Property CdO ITO NSG TEC NSG TEC


C15 A7

Thickness (nm) 20 501 140 340 630


Carrier concentration (cm 3) 2.201020 1.261021 5.561020 4.061020
Carrier mobility (cm2/Vs) 44.9 33.7 27.8 36.9
Fig. 1. A schematic of the two variations of evaporated gold contacts used in this study,
Resistivity ( cm) 6.3310 4 1.47104 4.04104 4.1710 4
showing (a) the standard simple layout and (b) an improved layout to reduce the
Sheet resistance (/sq) 19.0 11 15 6
effect of lateral resistance in the TCO layers. All samples measured with air as reference
RMS roughness from AFM (nm) 7.1 3.3 13.8 28.3
NSG TEC are on 3 mm glass, ITO and CdO are on 0.7 mm glass.
S.J.C. Irvine et al. / Thin Solid Films 520 (2011) 11671173 1169

100
NSG TEC A7 a Resistance cm2
NSG TEC C15
Transmittance (%)

80 ITO

60 CdO
3.0
40 2.6

cm2
2.2
20 1.8
1.4 D
1
0 C
200 400 600 800 1000 1200 1400 1600 1800 2000 2
Wavelength (nm) 3 B
A
Fig. 2. Spectrophotometer measurements for the four TCO samples over the 4
wavelength range from 200 to 3300 nm. contact

using the wider bandgap window layer Cd0.9Zn0.1S but not with CdS Resistance cm2
b
which has a similar bandgap. The most striking feature of CdO is the
high transmission to longer wavelengths where free carrier absorp-
tion starts to become signicant for wavelengths longer than 1200 nm
in the ITO and FTO superstrates. The excellent long wavelength
transmission for CdO has been discussed in a previous publication
2.4
[11]. For the purposes of a single junction CdTe device this does not
offer any advantage but would be attractive for multi-junction thin 2.0

cm2
lm devices. 1.6
1.2
3.2. Device architecture 1 D

2 C
The two different contact layouts shown in Fig. 1 enable a
3 B
comparison to be made of the effect of lateral conduction in the TCO
on series resistance in a PV device. By testing the contact architectures 4 A
on the TCOs, without the PV layers, it will help to interpret the series
contact
resistance and other properties of the Cd0.9Zn0.1S/CdTe devices grown
onto these TCO superstrates. In addition to the lateral conduction of
Fig. 3. Maps of resistance across NSG TEC C15 superstrate using (a) the simple contact
the TCO, the contact resistance of the gold to the TCO and probe
architecture in Fig. 1(a), (b) the improved contact architecture from Fig. 1(b).
contact resistance must also be considered. The front contact bus-bars
were connected with crocodile clips (both sides) to a 2400 Keithley
source meter, with the other terminal being connected via gold
probes to each of the back contact pads in turn. The contact resistance The pattern in Fig. 3(b) is different with the lowest resistance values
of the crocodile clips was checked by contacting the same bus bar in columns 2 and 3, increasing to each end. This pattern was more
different distances apart and extrapolating to zero separation. This usual than that shown in Fig. 3(a) across most of the samples
gave a contact resistance for the crocodile clips to the gold bus bar of measured and is consistent with the 2-fold symmetry of the contact
less than 0.25 .cm 2. The resistance was determined from the slope of layout. However, this valley pattern of resistance indicates that the
the IV plot where in all cases they were Ohmic. Tests were carried lateral conduction in the gold bus-bars is adding signicant series
out with and without the AM1.5 solar simulator and no signicant resistance to the measurements for columns 1 and 4. From the centre
difference was observed. In order to make a direct comparison of to edge measurement it is possible to estimate the maximum
resistance with the device series resistance, the resistance values here contribution of the gold bus bar resistance to the series resistance
are multiplied by the back contact area (0.25 cm 2) but this must not and this is 0.42 .cm 2. The average valley resistance of 1.64 .cm 2 is
be confused with resistivity of the TCO as this would only be correct attributed to the lateral conduction in the NSG TEC C15 for the
for vertical conduction through the back contact pad and not lateral improved contact architecture. Carrying out a similar calculation for
conduction being considered here. An example of the resistance maps the Delta-Tech ITO (which displayed similar valley behaviour) gave
is shown in Fig. 3 for the NSG TEC C15 superstrate, where a a bus-bar series resistance of 0.27 .cm 2 and an average valley
comparison is made between the two contact architectures. The resistance of 1.45 .cm 2. Some variance is expected on the series
simple contact architecture was used for the resistance map in Fig. 3 resistance depending on the variance in gold thickness. For the Delta-
(a) and shows a systematic variation across the sample from average Tech ITO sample the average column 1 thickness was 59 nm and
of 2.5 .cm 2 in column 1 to 1.8 .cm 2 in column 4. The absence of any average column 4 thickness was 47 nm. For the NSG TEC A7 sample
random variation in resistance across the area indicates that the probe with the same, improved, contact architecture, the estimated bus-bar
contact resistance to each pad is very small but the systematic resistance was 0.20 .cm 2 and the valley resistance was 1.23 cm 2.
variation does not reect the 2-fold symmetry of the contact pattern. It was not possible to obtain sufciently uniform CdO over the
The thickness of the gold contacts was measured using a stylus same 5 5 cm 2 area so a 5 2.5 cm 2 sample was used instead.
prolometer and showed an increase in thickness of the gold going Although this is not a direct comparison with the 5 5 cm 2 samples
from column 1 to column 4, mirroring the reduction in resistance. The it provided an average series resistance to compare with the PV device
average thickness in row 1 was 50 nm and in column 4 was 64 nm. It results that were on similar small format architecture. The average
is also noted that the column 3 and column 4 resistance values are resistance was 3.1 0.3 .cm 2, which is signicantly higher than for
similar and correlates with similar thicknesses for columns 3 and 4. the other TCOs but in agreement with the higher sheet resistance.
1170 S.J.C. Irvine et al. / Thin Solid Films 520 (2011) 11671173

In order to establish the primary contribution to the series Table 2


resistance a correlation graph was plotted from the series resistance Correlation of series resistance between the inner and outer rows of contacts for each of
the TCO superstrates.
for each superstrate, taken from the valley region of contacts, with
the corresponding sheet resistance values (Fig. 4). Comparison is Average of rows B and C Average of rows A and D
made for both contact architectures and in both cases they show good (.cm2) (.cm2)

agreement between series resistance and TCO sheet resistance, Delta-Tech ITO 1.73 0.13 1.64 0.20
conrming that these series resistance measurements are attributable NSG TEC A7 1.40 0.11 1.36 0.15
NSG TEC C15 2.07 0.26 2.02 0.20
to the TCO. It is clear that the additional gold bus-bars reduce the
series resistance, giving a shallower slope for the improved contact
architecture, which was designed to be less dependent on the lateral
conductivity of the TCO. However, the overall uniformity is good on the Delta-Tech ITO and
The remaining question for the lateral conduction in the ITO and provides a suitable template for testing the properties of this TCO. The
FTO superstrates is whether the inner rows of contacts (away from the AM1.5 conversion efciency (Fig. 5(a)) ranged from 9.5 to 11.9%. The
bus-bars) in the simple contact arrangement (Fig. 1(a)) have a higher series resistance was measured from the forward bias section of the
series resistance than the rows adjacent to the bus-bars. For this IV curve. These were all Ohmic with no roll-over that would be
comparison an average was taken for rows B and C and compared indicative of a rectifying contact. The series resistance values ranged
with the average for A and D. These results are shown in Table 2. It is from 2.7 .cm 2 to 4.4 .cm 2, with an average of 3.4 .cm 2 (Fig. 5(e)),
clear that there is no signicant difference between the inner and with no particular pattern across the surface. The ll factor (FF) shown
outer sets of contacts. This might seem surprising considering the in Fig. 5(d) is a parameter that depends on a low series resistance as
dependence of the series resistance on the TCO sheet resistance. well as a high shunt resistance. The shunt resistance is reliant on good
However, the integrated conduction paths with these contact surface coverage of the window layer and CdTe absorber layers and
architectures are complex and contacting from bus-bars on both for this device were mostly high (N1000 ) but with more variance
sides of the 5 5 cm 2 sample will provide overlap of spreading than for the other parameters. Similarly, the FF is variable but with all
resistance and an averaging between the bus-bars. This was values over 60% and six contacts over 70%.
conrmed by contacting from just one bus bar on the ITO superstrate The same device sample was also processed with the simple
with the simple contact architecture. The average series resistance of contact pattern (Fig. 1(a)) with average values given in Table 3, and
the row adjacent to the contacted bus-bar was 1.73 .cm 2 and showed similar uniformity of device parameters with efciency
increased for each row to 3.13 .cm 2 for the row furthest away. These ranging from 9.5 to 11.6%. The series resistance was in the range 2.3
results also provide further insight to the interplay between TCO sheet to 3.5 .cm 2, slightly lower than for the results shown in Fig. 5(e) but
resistance and the device contact architecture. considering the variance in this measurement no conclusions can be
drawn from this. The difference between the two device architectures
3.3. Results on CdZnTe/CdTe devices on different TCOs is very small and in line with expectation from the relatively small
differences in series resistance from Section 3.2. The more signicant
Having established the contributions to series resistance from the difference is with the device series resistance and the TCO series
chosen contact architecture and from the sheet resistance of the TCO, resistance measurements from Section 3.2. Average series resistance
the next objective was to see if this will translate into the PV devices. over the same area of the device, as for the TCO valley
The full contact array was made onto the Cd0.9Zn0.1S/CdTe deposited measurements gives a value of 3.4 0.5 .cm 2, using the improved
(and CdCl2 treated) layers, except for the CdO, which had to be made contact architecture. When compared with the series resistance for
on a smaller superstrate size. A map of IV characteristics under the ITO using the same (improved) contact architecture, which was
AM1.5 illumination could be created by probing each of the contacts. 1.45 .cm 2, there is an additional series resistance of approximately
The full set of maps is shown in Fig. 5 for the PV structure deposited 2 .cm 2 arising from the PV devices that can't be attributed to the TCO
onto the Delta-Tech ITO superstrate, using the improved contact resistance. If we use the lower value for the simple architecture,
architecture. A summary of average values, for all the commercial considering the variance in series resistance, this non-TCO resistance
substrate material, is given in Table 3. The ow direction in the could be as low as 1 .cm 2.
MOCVD reactor causes some non-uniformity due to small tempera- Turning now to the NSG TEC C15 superstrate, where average
ture variations and depletion of precursors from the gas stream. device measurements are given in Table 3, the range of AM1.5
efciency for the simple contact architecture was somewhat lower
2
than that for the devices on ITO with a range of 8.310.1% over an
Improved array of 2 3 contacts within the larger matrix. The FF was also poorer
Series resistance .cm2

contact over this contact array with an average value of 56%. However, the
1.8
shunt resistance was over 1000 cm 2 for all these contacts but the
simple series resistance had signicantly increased, compared with the ITO,
1.6 contact with an average of 6.7 2.2 .cm 2. For the improved contact
architecture on this sample the AM1.5 efciency is similar to the
1.4 results from the simple architecture, ranging from 8.2 to 10.4%, with
an average FF of 63% and is consistent with a lower average series
1.2 resistance of 3.9 0.3 .cm 2 compared with the simple architecture.
Comparing this value with the estimated TCO series resistance for NSG
TEC C15 superstrates of 1.64 gives a difference of 2.3 cm 2, similar
1
0 5 10 15 20 to the difference for the results using the ITO superstrate. The error on
Sheet resistance /Sq. the series resistance on the simple contact architecture was too large
for any meaningful calculation of the residual series resistance. The
signicant decrease in series resistance, when using the improved
Fig. 4. Correlation graph of series resistance (averaged from columns 2 and 3 in both
contact architectures shown in Fig. 1 where simple just has the straight bus bar front
contact architecture, was expected for this relatively high impedance
contacts and improved has the additional gold bus bars dividing the rows of gold superstrate where the difference in contact architecture can make a
contacts) versus the measured sheet resistance values for each superstrate. larger difference in the device series resistance (as seen in Fig. 4).
S.J.C. Irvine et al. / Thin Solid Films 520 (2011) 11671173 1171

a b
% JscmA cm-2

12.0
23.5
11.5 23.0

mA cm-2
11.0 22.5
10.5 22.0
%

10.0 21.5
9.5 21.0
9.0 D 20.5 D
3 3
4 C 4 C

5 5
B B
6 6
A A

contact contact

c d
VocmV FF%

750
75
71
700
67
mV

63
650
59
600 D 55 D
3 3
C C
4 4
B 5 B
5
6 A
6 A
contact contact

e
Rs cm2

4.5
4.1
cm-2

3.7
3.3
2.9
2.5 D
3
C
4
B
5

6 A

contact

Fig. 5. Maps of PV device characteristics under AM1.5 illumination for Cd0.9Zn0.1S/CdTe on Delta-Tech ITO superstrate, showing (a) the cell efciency, (b) the Jsc, (c) the Voc, (d) the ll
factor (FF) and (e) the series resistance Rs. The contact architecture was the improved architecture in Fig. 1(b). The ow direction of the gases in the horizontal reactor is marked with
the arrow.

The device results for the lowest resistance TCO in this series, the the simple and improved contacts respectively. This is the main cause
NSG TEC A7, were disappointingly low and with larger variance of the poor FF with average values of 61% and 54% respectively. The
between the devices. The AM1.5 efciency was in the range of 6.9% to series resistance for the devices on the NSG TEC A7 was much lower
9.3% for the simple contact architecture and 5.2% to 8.1% for the (as expected) than for the NSG TEC C15 with average values of 2.9
improved architecture. The overlaying of the additional contacts has 0.3 .cm 2 and 3.4 0.5 .cm 2 for simple and improved contacts
clearly caused some deterioration but the main cause of the large respectively. The difference between these series resistance values is
variance and poor performance can be seen in the shunt resistance not signicant considering the observed variance and would be
with average values shown in Table 3 of 270 .cm 2 and 327 .cm 2 for difcult to distinguish for a low sheet resistance TCO. The difference in
1172 S.J.C. Irvine et al. / Thin Solid Films 520 (2011) 11671173

Table 3 devices on the other substrates and is consistent with this interpre-
Comparison of Cd0.9Zn0.1S/CdTe PV devices on three different commercial superstrates, tation of the poor FF and shunt resistance.
Deltatec ITO, NSG TEC A7 and NSG TEC C15. The device parameters are AM1.5 efciency
(), short circuit current density (Jsc), open circuit voltage (Voc), series resistance (Rs)
An attempt to repeat the PV device structure growth onto the CdO
and shunt resistance (Rsh). For each superstrate two different contact architectures superstrate resulted in delamination of the CdO lm from the
were used (shown in Fig. 1) where simple just has the straight bus bar front contacts substrate. The adhesion of the as grown lm was sufcient to pass
and improved has the additional gold bus bars dividing the rows of gold contacts. the Scotch tape test so it appeared that interfacial strain caused by the
Contacts Jsc Voc FF Rs Rsh thicker CdTe layer had caused the delamination. A further PV device
(%) (mA cm 2) (mV) (%) (.cm2) (.cm2) was prepared where the MOCVD grown CdTe was only 800 nm thick.
Delta-Tech ITO Simple 10.9 23.3 665 70 2.6 1573 Previous reports on ultra-thin CdTe devices have shown that good
Improved 10.9 22.5 711 68 3.4 1500 device quality can be obtained but with some loss of Jsc due to the
NSG TEC A7 Simple 8.5 20.8 655 61 2.9 270 reduction in solar absorption [13,14]. An 800 nm CdTe absorber layer
Improved 7.2 20.1 654 54 3.4 327 was grown onto 240 nm thick CdS onto CdO superstrate with the
NSG TEC C15 Simple 9.0 21.6 670 56 6.7 2105
characteristics given in Table 1. This was compared with growth onto
Improved 9.3 22.0 680 63 3.9 1894
an ITO substrate in the same MOCVD batch run. The adhesion of this
CdS/CdTe lm was good on both substrates and the results are given
series resistance between the devices and the TCO series resistance in in Table 4 along with the series resistance. Although the TCO had a
Fig. 4 is 1.7 and 2.1 cm 2 respectively for the simple and improved lower series resistance, with corresponding improvement in the FF
architectures. These values are in agreement with the non-TCO series over the CdO superstrate device, the Voc is signicantly higher for the
resistance derived from the other superstrate devices. The explana- latter and indicates factors relating to the TCO that might favour
tion for the poor FF and low shunt resistance lies with the incomplete improved CdS nucleation and a reduction in micro-shorts. This feature
coverage of the Cd0.9Zn0.1S window layer creating micro-shorts. A of the MOCVD growth where nucleation can be carefully controlled is
separate growth of the Cd0.9Zn0.1S under identical conditions to the now being investigated in more detail. Comparing the series
device layer showed a patchy nucleation onto the NSG TEC A7 surface. resistance from Table 4 for the CdO superstrate device with the CdO
This was conrmed in the devices by taking a blue laser beam induced series resistance from Section 3.2, gives a difference of 1.9 .cm 2
current (LBIC) image of the device as shown in Fig. 6. The laser which is in agreement with the non-TCO series resistance for the
wavelength of 405 nm is highly absorbing in the Cd0.9Zn0.1S, leading other TCO superstrate devices.
to a very low photo-current over most of the contact area. The spikes In this results section a careful analysis of the series resistance of the
are regions of very thin Cd0.9Zn0.1S where the blue laser light PV device has been made and contributions associated with lateral
penetrates into the CdTe absorber layer. The optimal red laser conduction in the TCO, the contact architecture and the non-TCO
image doesn't show these spikes, where penetration through the contribution attributed to the back contact. The observed range of device
window layer is good. This conrms the non-uniformity seen in the series resistance for the 5 5 cm2 format is 2.6 to 6.7 .cm2 and has been
Cd0.9Zn0.1S coatings and helps to illustrate where micro-shorts are attributed to the TCO sheet resistance. The SCAPS model [15] was used to
likely to occur where the CdTe layer might contact the underlying interpret impact on PV device performance. This is a 1-D device model
TCO. For the rougher morphology on the NSG TEC A7 superstrate the taking into account properties of the window layer, absorber layer, back
RMS roughness was 28 nm from Table 1 and peak to valley height of contact and series resistance. The cell parameters and materials
over 100 nm was observed in the AFM image. For the very thin regions properties used in this modelling were the same as those reported by
of the window layer it is likely that TCO spikes uncoated with the Jones et al. [13]. Fig. 7 shows the predicted correlation between FF and
window layer would punch through into the CdTe layer, causing series resistance from the SCAPS model and the from the device results
micro-shorts. The reason for the regions of poor Cd0.9Zn0.1S nucleation for the Delta Tech ITO and NSG TEC C15 superstrates. The NSG TEC A7
is not clear but uniform nucleation by the MOCVD process does was not included in this comparison as the FF was dominated by the poor
depend on a high surface cleanliness and this was more difcult to shunt resistance and therefore not limited by series resistance. The
achieve on the rougher surfaces. The open circuit voltage Voc for the results in Fig. 7 show excellent agreement between the experimental
devices on NSG TEC A7 superstrates is also signicantly lower than for results and theoretical prediction, conrming that for these devices,
where the shunt resistance is N1000 .cm2, the FF is limited by the series
resistance. The SCAPS model predicts that to achieve a FF of 75% the
series resistance will have to be less than 1 .cm 2. This means that the
target series resistance contributions from both the TCO and back
contact will both have to be signicantly below 1 .cm2.

4. Conclusions

A detailed study has been made of the role of TCOs in the


performance of Cd0.9Zn0.1S/CdTe PV solar cells where the solar cell
structure has been deposited by MOCVD. It was found that the factors
inuencing device performance were:
a) The device architecture where front contacts can mitigate against
poor lateral conduction in the TCO.

Table 4
Comparison of ultra-thin CdTe devices on a Deltatec ITO superstrate with a CdO
superstrate.

(%) Jsc (mA cm 2) Voc (mV) FF (%) Rs (.cm2)

Delta-Tech ITO 6.7 16.1 596 69 1.5


Fig. 6. LBIC image over one of the device contacts on the NSG TEC A7 superstrate devices
MOCVD CdO 7.2 16.8 626 68 5.0
taken at 405 nm showing higher response from areas of thin Cd0.9Zn0.1S window layer.
S.J.C. Irvine et al. / Thin Solid Films 520 (2011) 11671173 1173

100
achieve greater control over the nucleation and hence improve the
80
quality of the window layer. This appears to strongly inuence the Voc
and further work is being carried out to see what effect different
60 nucleation treatments have on the Voc. It proved to be difcult to
FF% make a direct comparison with CdO as we were only able to obtain
40 adherent devices on ultra-thin CdO superstrates. Further work will be
SCAPS FF%
carried out on doping the CdO to reduce sheet resistance and work
20 Experimental FF%
with thinner CdO to see if this avoids delamination from thin lm
0 strain. A nal comment is that the sheet resistance of the TCO is not
0 5 10 15 necessarily the limitation on device performance provided suitable
Series resistance cm2 contact architecture is used to increase lateral conduction. Therefore,
further work will be carried out using the NSG TEC C15 superstrates
Fig. 7. A comparison of the theoretical ll factor (FF) versus the series resistance from with improved nucleation.
SCAPS model with the results for devices on the Delta-Tech and NSG TEC C15
superstrates.
Acknowledgements

b) The gold contact bus-bars can contribute to series resistance The authors would like to acknowledge EPSRC for their nancial
(estimated to be 0.4 .cm 2 for Au bus-bars less than 60 nm support through the PV SUPERGEN project (PV Materials for the 21st
thickness) Century). We would also like to thank Dr Paul Warren and Dr Neil
c) Series resistance measured using the two different contact McSporran of the Pilkington Group Ltd. for providing the NSG TEC
architectures, directly onto the different TCOs, correlated with glass. Technical support from Fraser Hogg in measuring large numbers
the measured sheet resistance and the improved contact archi- of devices is also gratefully acknowledged.
tecture falling on a line below the simple architecture.
d) The PV devices on the Delta-Tech. ITO and on the NSG TEC C15
superstrates were limited by series resistance arising from lateral References
conduction in the TCO and a residual series resistance associated [1] A. Bosio, N. Romeo, S. Mazzamuto, V. Canevari, Prog. Cryst. Growth Charact. Mater.
with the back contact that fell in the range from 1 to 2.3 .cm 2 52 (2006) 247.
with most of the results in the range 1.7 to 2.3 .cm 2. Therefore, a [2] M.I. Baraton, Mat. Res. Soc. Symp. Proc. 1209 (2010) 03.
[3] P.K. Biswas, A. De, N.C. Pramanik, P.K. Chakraborty, K. Ortner, V. Hock, S. Korder,
reasonable estimate for the back contact resistance for these Mater. Lett. 57 (2003) 2326.
devices is 2 .cm 2. [4] X. Wu, J.C. Keane, R.G. Dhere, C. Dehert, D.S. Albin, A. Dude, T.A. Gessert, S. Asher,
e) Surface morphology of the TCO and consequent nucleation of the D.H. Levi, P. Sheldon, Proceedings of the 17th European Photovolt. Sol. Energy
Conf., Munich, Germany, vol. II, 2001, p. 995.
Cd0.9Zn0.1S window layer was shown to be the limiting factor with [5] D. Cunningham, M. Rubcich, D. Skinner, Prog. Photovolt. Res. Appl. 10 (2002) 159.
the rougher NSG TEC A7 superstrate causing shunt resistance [6] Z. Jingquan, F. Lianghuan, L. Zhi, C. Yaping, L. Wei, W. Lili, L. Bing, C. Wei, Z. Jiagui,
b1000 .cm 2 and poor FF. Sol. Energy Mater. Sol. Cells 93 (2009) 966.
[7] S.J.C. Irvine, V. Barrioz, D. Lamb, E.W. Jones, R.L. Rowlands-Jones, J. Cryst. Growth
f) A good correlation was obtained between the device series 310 (2008) 5198.
resistance and the FF which was also in good agreement with [8] V. Barrioz, Y.Y. Proskuryakov, E.W. Jones, J.D. Major, S.J.C. Irvine, K. Durose, D.A.
the prediction of SCAPS modelling. Lamb, Mater. Res. Soc. Symp. Proc. 1012 (2007) 367.
[9] Y.Y. Proskuyakov, K. Durose, J.D. Major, M.K. Al Turkestani, V. Barrioz, S.J.C. Irvine,
Although the sheet resistance of the TCO was shown to contribute E.W. Jones, Sol. Energy Mater. Sol. Cells 93 (2009) 1572.
[10] D.W. Sheel, H.M. Yates, P. Evans, U. Dagkaldiran, A. Gordijin, F. Finger, Z. Remes, M.
to the device series resistance it is clear from this study that other
Vanecek, Thin Solid Films 517 (2009) 3061.
factors are also important and no one factor dominates. Other factors [11] D. Lamb, S.J.C. Irvine, Thin Solid Films 518 (2009) 1222.
that have been shown to be important are the back contact resistance, [12] V. Barrioz, S.J.C. Irvine, E.W. Jones, R.L. Rowlands, D. Lamb, Thin Solid Films 515
the surface morphology of the TCO and the nucleating surface. The (15) (2007) 5808.
[13] E.W. Jones, V. Barrioz, S.J.C. Irvine, D. Lamb, Thin Solid Films 517 (2009) 2226.
latter may be particularly important for MOCVD where surface [14] A. Gupta, V. Parikh, A. Campaan, Sol. Energy Mater. Sol. Cells 90 (2006) 2263.
chemistry is important. However, this provides an opportunity to [15] M. Burgelman, P. Nollet, S. Degrave, Thin Solid Films 361362 (2000) 527.

Vous aimerez peut-être aussi