Você está na página 1de 3

20.

1% EFFICIENT LARGE AREA CELL ON 140 MICRON THIN SILICON WAFER




Nigel Mason
1
, Oliver Schultz
2
, Richard Russell
3
, Stefan Glunz
2
and Wilhelm Warta
2

1. BP Solar, Chertsey Road, Sunbury TW16 7LN, UK.
Tel: +44 1932 739634, E-mail: nigel.mason@bp.com
2. Fraunhofer ISE, Heidenhofstrasse 2, 97110 Freiburg, Germany
3. BP Solar Espaa, Pol. Ind. Tres Cantos, 28760 Tres Cantos (Madrid), Spain



ABSTRACT: The laser-grooved buried contact (LGBC) crystalline silicon solar cell has a selective emitter with
excellent response to the solar spectrum in the wavelength range 300-900 nm. However, the cell efficiency is limited
by a reduced response in the 900-1200 nm region of the spectrum due to the high recombination characteristics of the
rear aluminium back surface contact and reduced internal reflection at this interface. The LGBC cell has now been
fabricated with an aluminium laser-fired contact (LFC) through a passivating rear silicon oxide dielectric. The cells
were processed using FZ wafers of resistivity 0.5 ohm.cm to produce a 140 m thin cell of 20% efficiency. The
performance gain produced with the LFC process is clearly seen in the spectral response and internal reflection in the
long wavelength region of the spectrum (900-1200 nm). This process offers the potential for premium efficiency,
cost-competitive photovoltaic cells with a silicon feedstock consumption of less than 6 g/Wp.
Keywords: c-Si, High-Efficiency, Laser Processing


1 INTRODUCTION

The laser grooved buried contact (LGBC) solar cell
has been used in premium crystalline silicon PV products
for more than a decade. Originally invented by Stuart
Wenham and Martin Green [1] in 1983, it has been
developed and manufactured by BP Solar in Spain since
1992 [2] with a cumulative total production of 120
MWp. These cells are used in modules marketed by BP
Solar under the Saturn product name with a premium 72-
cell module having a minimum output of 190 Wp under
standard test conditions [3].
In todays silicon constrained environment, the need
to optimise the PV output from the available feedstock is
driving the use of increasingly thinner wafers. The
performance of the present commercial LGBC cell is
limited to around 18% efficiency on 250 m thick
Czochralski grown (Cz) wafers. The cell performance is
constrained by the high surface recombination
characteristics of its full-area rear aluminium contact and
poor internal reflection (light trapping) from the rear
surface. Furthermore, this high rear surface
recombination velocity results in a declining cell
performance as the wafer thickness is reduced.
The laser fired rear contact process developed at the
Fraunhofer ISE [4] has very low surface recombination
characteristics and offers the prospects of a cost-
competitive fabrication process for the cell rear contact
that is compatible with the LGBC cell front surface. The
two technologies have been combined into a single cell
and applied to thin silicon wafers to demonstrate the
performance potential of the structure.
The motivation for the present study was, therefore,
to demonstrate the stretch-potential of the LGBC cell to
realise over 20% cell efficiency on silicon wafers of less
than 150 m thickness. Assuming current commercial
manufacturing yields, cells of this performance could be
produced with a silicon feedstock consumption of only 6
g/Wp half the current industry requirement. Cells were
fabricated on silicon wafers produced by the Floating
Zone (FZ) process and incorporated a silicon oxide front
surface film. Whilst neither of these features are suitable
for commercial production they serve as well
characterised materials on which to demonstrate proof-
of-concept. A parallel programme is underway to
develop a cost-competitive commercial product based on
the LGBC/LFC cell combination [5].


2 EXPERIMENTAL

2.1 Cell structure and fabrication
A schematic illustration of the cell structure is shown
in figure 1. Boron-doped FZ wafers of (100) surface
crystal orientation, resistivity 0.5 ohm.cm and thickness
350 m were lapped (ground) to a thickness of 150 m
prior to cell fabrication. After etching the surface crystal
damage (generated by the grinding process) the wafer
surface was oxidized with the front surface oxide
subsequently removed to facilitate random pyramid
texturing in KOH solution. This resulted in a final cell
thickness of 140 m.


Figure 1: Schematic x-section illustrating the structure
of the LGBC/LFC cell

A lightly doped n
+
emitter (>100 ohm/square) was
formed by phosphorus doping using a POCl
3
/O
2
vapour
source, followed by the formation of a 200 nm thick
silicon oxide passivating/dielectric layer. Laser grooves
of dimensions 20 m wide by 30 m deep were micro-
machined into the front silicon surface and etched to
remove residual silicon and surface crystal damage. The
grooves were then doped to a sheet resistivity of less than
10 ohm/square again using a POCl
3
/O
2
vapour source.
The resulting phosphor-silicate glass (PSG) in the groove
was removed using buffered HF solution and resulted in
a final front oxide layer thickness of 105 nm. An
aluminium film of thickness 2 m was evaporated over
the oxide on the rear surface. A thin (~100 nm) nickel
film was electroless plated in the heavily doped (n
++
-
type) silicon grooves and sintered to form a mechanically
robust nickel silicide interface. A neodymium-YAG laser
was used to thermally fire small (100 m diameter)
areas of the aluminium layer through the oxide film to
form a local back contact. Silver metal was electroplated
into the grooves in the front of the cell.

2.2 Comments on yield and commercial feasibility
The process sequence described in section 2.1 to
make the present LGBC/LFC cells is not regarded as
being commercially viable. The silicon oxide surface has
poor antireflection properties once incorporated in a glass
laminate and the use of FZ wafers is considered
uneconomic. The large number of process steps in the
present sequence would significantly add to both the
capital cost and operational cost of such a production
line. Recognising these issues, work is underway to
identify a cost-effective process sequence using lower
cost Czochralski-grown (Cz) wafers and incorporating a
silicon nitride front surface dielectric/antireflection layer
with a substantially reduced process sequence [5].
By commercial standards, the wafer-to-cell yield
experienced with the present study using 140 m thin
wafers was poor. A total of 57 discrete wafer handling
steps were identified in the current sequence. The yield
was further compromised by the fact that the cells were
part-processed at both Fraunhofer ISE (Germany) and BP
Solar (Spain) and the cells made four separate journeys
between the two sites to complete the cell processing. An
analysis of the wafer breakage as a function of each
operation showed no inherent problem with the process.
Breakage was largely considered to be a result of
mechanical stress on the wafer edge as a consequence of
the extensive handling. The contribution of the 30 m
deep laser grooves in the silicon surface were
investigated but not considered a significant source of
breakage at least for the present 140 m thick wafers in
this study.


3 RESULTS AND DISCUSSION

3.1 Cell Performance
Table I gives the electrical characteristics of a
selection of LGBC/LFC cells fabricated at 140 m
thickness representing cells with the highest open circuit
voltage, highest short circuit current and highest
efficiency. The best cell, measured under standard test
conditions, recorded an efficiency of 20.1%. This is
almost 5% higher than our previous best LGBC cell
using an all aluminium rear contact and fabricated on a
boron-doped FZ wafer of 300 m thickness [6].
Cell E in table I was fabricated on a Czochralski
wafer grown using magnetic confinement (MCz) that
results in significantly improved minority carrier lifetime
as compared to conventional Cz wafers. The resistivity
of the wafer was 1.0 ohm.cm and thus resulted in a lower
open circuit voltage (Voc) than the 0.5 ohm.cm FZ
wafers in this study. The short circuit current density
(Jsc) of cell E is comparable to the cells on FZ wafers but
the cell Fill Factor (FF) was low. This resulted in a cell
efficiency of only 18.2%.

Table I: Selection of LGBC/LFC cells illustrating a
range of cell characteristics at irradiance 1 kW/m
2
and
temperature 25C with cell area 149.4 cm
2
and thickness
140 m.

Cell Voc
(mV)
Jsc
(mAcm
-2
)
FF
(%)
Efficiency
(%)
A(FZ) 674 37.9 78.7 20.1
B(FZ) 678 36.5 78.8 19.5
C(FZ) 674 38.1 74.7 19.2
D(FZ) 674 37.6 78.5 19.9
E(MCz) 652 37.1 75.0 18.2


3.2 Cell Characterisation
The spectral response and reflectance of the cells as a
function of wavelength were measured. Figure 2
compares the spectral reflectance and derived internal
quantum efficiency (IQE) of the 20.1% efficient 140 m
thin LGBC/LFC cell (cell A) with that of a 300 m thick
LGBG cell of 18.3% efficiency incorporating a
conventional aluminium back contact. Note that the
LGBC/LFC cell (open squares) has a silicon oxide front
surface whilst the LGBC cell has a silicon nitride front
surface.

0.0
0.2
0.4
0.6
0.8
1.0
300 400 500 600 700 800 900 1000 1100 1200
photon wavelength (nm)
R
e
f
l
e
c
t
a
n
c
e

&

I
n
t
e
r
n
a
l

Q
E
140 um LGBC/LFC cell (20.1%)
300 um LGBC cell (18.3%)
140 um LGBC/LFC cell (20.1%)
300 um LGBC cell (18.3%)
Reflectance
IQE


Figure 2: Internal quantum efficiency (upper curves)
and reflectance (lower curves) for the 20.1% efficient
LGBC/LFC cell and an 18.3% efficient LGBC cell with
conventional aluminium rear contact.

The two cells in figure 2 show similar excellent
response characteristics in the 300-800 nm wavelength of
the solar spectrum. Above 800 nm the cell with the LFC
rear shows superior internal quantum efficiency due to
the passivating properties of the rear oxide dielectric
interrupted only by rear metal point-contacts. Similarly,
above a wavelength of 1000 nm the LGBG/LFC cell
shows increased reflectance that can be attributed to
internal reflection from the rear surface.
Figure 3 compares the spectral reflectance and
internal quantum efficiency of cell A ( 20.1%) made
using an FZ wafer with cell E ( 18.2%) made with a
MCz wafer. The internal quantum efficiency (upper
curves) are almost identical for the two cells
demonstrating that the MCz wafer has the potential to
realize the same cell performance as the FZ wafer. The
slightly lower reflectance observed for cell E in the
wavelength range 1000-1200 nm is attributed to reduced
internal reflection from the rear aluminium interface due
to the presence of a non-planar rear surface on this cell.
[To simplify cell processing with the MCz wafers, cell E
was texture etched on both surfaces prior to surface
oxidation].

0.0
0.2
0.4
0.6
0.8
1.0
300 400 500 600 700 800 900 1000 1100 1200
photon wavelength (nm)
R
e
f
l
e
c
t
a
n
c
e

&

I
n
t
e
r
n
a
l

Q
E
140 um LGBC/LFC cell (20.1%), FZ 0.5 ohm.cm
140 um LGBC/LFC cell (18.2%), MCZ 1.0 ohm.cm
140 um LGBC/LFC cell (20.1%), FZ 0.5 ohm.cm
300 um LGBC cell (18.3%), MCZ 1.0 ohm.cm
Reflectance
IQE


Figure 3: Internal quantum efficiency (upper curves) and
reflectance (lower curves) for LGBC/LFC cells made on
FZ and MCz wafers.

3.3 Silicon feedstock consumption (g/Wp)
In a constant drive to improve costs, the PV industry
has been steadily reducing the thickness of the silicon
wafers used to manufacture solar cells from a typical
thickness of around 450 m in the mid-1980s to a
thickness of 200-250 m today. With the increased
demand for PV in the past few years the need to
accelerate the progress to thinner wafers is desirable to
minimize the silicon feedstock constraint on industry
growth. A number of roadmaps [7, 8] have recently been
published charting projected reduction in average wafer
thickness and increase in average cell efficiency that is
necessary to grow the MW output of the industry at a
higher rate than possible simply from increased silicon
feedstock production. The single metric that combines
these two output growth enablers is the silicon feedstock
consumption per unit peak rated power output (g/Wp). It
is generally recognized that the industry today is
operating at a typical consumption of 10-12 g/Wp and the
expectation of published roadmaps is that this figure will
fall to below 10 g/Wp by the year 2010 and be around 5-
8 g/Wp by 2020.
If the cell result of 20% efficiency on 140 m thin
wafers reported here can be realized in a commercially
viable process, whilst maintaining current production
yields, it would result in a feedstock consumption of less
than 6 g/Wp - comfortably within the roadmap targets.
With the anticipated advances in wire sawing technology
over the coming years, silicon kerf loss is expected to be
reduced from around 200 m to 160 m. Factoring this
gain would reduce feedstock consumption for these cells
to around 5 g/Wp.

3.4 Future outlook
As discussed in section 2.2, the process and materials
used to demonstrate the present cells are not considered
appropriate for commercially viable production of flat
plate (1-sun) solar modules. However, our modeling and
experimental studies have shown that cell efficiencies in
excess of 19.5% should be realized on 150 m thin CZ
wafers produced by magnetic confinement (100 s
lifetime). The short wavelength blue response of the
commercial LGBC cell (section 3.2) incorporating a
silicon nitride front surface dielectric/antireflection
coating is compatible with 20% cell performance. The
remaining challenge is to incorporate the rear
dielectric/LFC process using a simple manufacturing
sequence. The early results on this work are reported
elsewhere [5].


4 SUMMARY

As part of a programme to raise the cell efficiency
above 20% and facilitate device production on thin
wafers, LGBC cells have been fabricated using the rear
surface laser-fired contact (LFC) developed by
Fraunhofer ISE. This combination offers high
performance, efficient use of silicon feedstock and cost-
effective industrial fabrication.
LGBC/LFC cells fabricated with a wafer thickness of
140 m from 0.5 ohm.cm FZ wafers have realised large
area cell efficiencies of 20.1% (area 149.4 cm
2
). Internal
rear surface reflectance and quantum efficiency
measurements show the performance gain is realised by a
significantly improved collection from photons in the
wavelength range 800-1200 nm.
The silicon feedstock requirement for this cell is
expected to be less than 6 g/Wp which is almost half of
the current consumption level and will reduce further to 5
g/Wp with the anticipated developments in wafer sawing
technology. This result is an important development for
the silicon PV industry in the current climate where
growth is constrained by silicon feedstock supply.
This development offers the prospects for a cost-
effective process for high efficiency cells that can drive
PV costs towards grid-parity with domestic electricity
and substantially reduce the silicon feedstock
consumption per Watt-peak of PV power


5 REFERENCES

[1] SR Wenham and MA Green, Australian Patent
570309 (1993).
[2] NB Mason et al., Proc. 10
th
European Photovoltaic
Solar Energy Conference, Lisbon 1991, 280.
[3] BP7190 product, www.bpsolar.de
[4] E Schneiderlchner et al., Progress in Photovoltaics:
Research Applications, 10 (2002), 29-34
[5] O Schultz et al., paper 2CV.3.17 this conference
[6] NB Mason et al., Proc. 19
th
European Photovoltaic
Solar Energy Conference, Paris 2004, 620
[7] www.epia.org/documents/Roadmap_EPIA.pdf (May
2004)
[8] European PV Technology Platform, Draft Strategic
Research Agenda, June 2006. www.eupvplatform.org

Você também pode gostar