Making Solar Cells
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Making Solar Cells
We will look at some the differences that stand between
solar cells that are research and commercial in particular
the following features
- materials that are used (semiconductor, quality of
semiconductor)
- processing used (complexity, number of steps)
- designs
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Research
• Generally have more freedom in design since cost
effectiveness is not main concern
• However, the idea is to develop a high efficiency design and
then try to implement as a commercially viable process
sometimes
• Because of the lack of restrictions in materials and
processes the efficiencies reported for research devices are
generally significantly higher than for commercial solar cells
• Sizes tend to be small also, not always striaght forward to
scale up
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Commercial
• Want material that is generally inexpensive though can be
mitigated by high efficiency
• Processing has to be relatively straightforward (in general
people with PhDs will not be doing this)
• Want as few processing steps as possible (more steps =
more expensive and more chance for something to go wrong)
• Processes must be robust i.e. it works all the time if you have
set up correctly
• Must be easily scaleable
• Ideally the disruption to current production is minimal – we
don’t want to have to build a new factory
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Materials
• Materials for solar cells are usually one or more of the
following:
• Group IV (Si and Ge)
• III-V (GaAs, InP and
variants)
• II-VI
– CuInSe2 (CIS)
– CuInGaSe2 (CIGS)
– CdTe
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Silicon
• First need to decide on type of silicon to be used
– Float Zone: gives best diffusion lengths etc.. As impurities are
reduced particularly oxygen. Popular for high efficiency devices in
research
– Czochralski: good quality material with low impurities. Mainstay for
single crystalline devices both research and commercial. Not as
expensive as Float Zone
– Multi(Poly)crystalline: not as good as CZ or FZ, but much cheaper,
efficiency will be highly variable from cell due to grain boundaries
etc.. Very popular commercially.
– Amorphous: very cheap, not crystalline so low efficiencies. Easy to
deposit with relatively high quality. Some research done but mostly
used in ‘disposable’ consumer applications
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Refining Silicon
•
•
Start with silica or sand (want low impurity concentrations)
Fire with carbon to get metallurgical grade silicon
SiO2 + C → Si + CO2
•
To get semiconductor grade need
to purify – react with anhydrous
HCl at 300 C
Si + 3HCl → SiHCl3 + H2
resulting solution distilled to
reduce impurities
Reacted with Hydrogen (1100 C for 200-300 hours!) to give pure silicon
SiHCl3 + H2 → Si + 3HCl
End up with polysilicon rods – broken up for feedstock
•
•
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Silicon - Multicrystalline
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Silicon - Czochralski
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Silicon – Float Zone
• Molten region contains the
impurities move high temperature rf
coil over ingot
• End up with high impurity stub –
throw away (actually recycle)
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Silicon - Amorphous
• Low temperature deposition typically using Plasma
Enhanced Chemical Vapour Deposition
Can deposit on glass
– thin film solar cells
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Ribbon Silicon
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
High η Silicon - PERL
• Passivated Emitter Rear Locally diffused (PERL(d))
• Current champion – 25% efficiency
• Photolithography used for inverted pyramids on front
surface
• Emitter and base passivated with high quality oxide
• Local diffusion to minimize
metal-sc contact area
• Metal-sc contact area is
also minimized at front
• Float zone
• Purely research
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
PERL cont’d
Inverted pyramids
Metal-semiconductor contact area minimized
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Back Contact
• Completely avoid shading losses since no contact on the
‘front’
• Also interconnection is easier good for concentration
• Processing is necessarily more complicated
• Problems with the increased area of the pn junction
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Buried Contact
• Reduce shadowing losses whilst also minimizing resistance
losses
• Some commercial uptake, still heavily researched
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Screen Printed
•
•
•
•
Texturing using pyramids - easy to do
Contact is screen printed – no buried contact feature
Processing is straightforward
Reasonable efficiencies
~ 16%
• Processes can be easily
transferred
• Easily scaleable
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Emitter formation
• Semiconductor material (“wafer”) comes doped (typically
p type for silicon solar cells)
• Need to form a highly n type doped emitter layer for our
solar cell
• Can grow with dopant incorporated – methods such as
MBE, MOCVD this is trivial
• To keep costs down, however, better to diffuse the n
type dopant into the substrate
• For silicon, this means diffusing Phosphorus
• Since diffusing the dopant in, the doping density will vary
with depth and junction is not sharp
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Emitter Diffusion
• Must ‘drive’ Phosphorus into the substrate so that
N D >> N A
• High temperature process is required – take too long
E
otherwise
− a
D = Ae kT
• We already know that a diffusion flux is determined by the
concentration gradient and its diffusivity, this is actually
known as Fick’s First Law
∂φ
J = −D
∂x
• However, this applies when we have a steady state i.e. the
concentration distribution isn’t changing with time
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Emitter Diffusion
• When concentration is changing we need to use Fick’s
second law
∂ 2φ
∂φ
=D 2
∂x
∂t
• Analogous to the heat equation
• If we have a concentration n(0) at the surface (we assume
we have an unending supply) then a solution to this
equation is
⎛ x ⎞
n( x, t ) = n(0)erfc⎜
⎟
⎝ 4 Dt ⎠
• Junction is formed where n(x,t) > NA
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Emitter Diffusion
• When have a limited source, solution is a little different
⎡ − x2 ⎤
S
n ( x, t ) =
exp ⎢
⎥
4
Dt
πDt
⎣
⎦
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Emitter Diffusion
• Typically the diffusion is two stage
– First, a shallow pre-deposition using a semi-infinite
source
– Second, a higher temperature drive-in diffusion using the
pre-deposition as the source
• Remember dopant concentration mustn’t be too high – end
up with solubility problem leading to a dead layer
• Since dopant concentration isn’t constant the sheet
resistivity is not so simple
1
ρ† = t
∫ µ ( x) N D ( x)dx
0
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Emitter Diffusion
• Solid source Phosphorus source can be used with
diffusion furnace
• Spin on dopants also used
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Ion Implantation
• Accelerate ions and fire them at semiconductor
• Best for shallow junction
• Good control of dose
• Creates damage to the
crystal lattice
• Can also create some
amorphization
- it can melt!
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Solar Cell Production
• Screen printed is dominant technology
• Start with 10 cm x 10 cm ~ 0.5 mm thick wafer of Silicon
• Saw damage means surfaces are rough – give a
chemical polish using a strong alkaline solution
– Cleans wafer of any crud
– Removes damaged Silicon
• Now ready to begin making solar cell
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Solar Cell Production
• First create junction by n type diffusion
• Wafer is heated to 800-1000 C in a phosphorus atmosphere
giving thin layer of n type
• The n type layer at the edges means the top and bottom are
connected
– Need to remove to make our solar cell
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Solar Cell Production
• Stack together so only edges are exposed
• Etch the edges away using a highly reactive plasma to
remove the junction
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Solar Cell Production
• Screen Print the rear contact
• Process the same as screen printing for a tee shirt
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Solar Cell Production
• First heated gently to remove organic binders in the paste
• Second firing at much higher temperature to fire the metal
into the Silicon
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Solar Cell Production
• Finish up with screen printing of front contact grid
• Heated to fire the metal into the Silicon
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Solar Cell Production
• Get something that looks like this
• What about AR coatings?
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Other Bits
• Back contact firing will give a BSF improving performance but
Al is relatively expensive and when fired require second metal
layer to contact
• Dead layer effect reduced by shallower emitters, important
since we need high doping under contacts
• AR coatings and/or surface texturing incorporated into process
– Texturing often done at the start, difficult to manage with multicrystalline
– AR coating typically applied at the end, popular materials are TiO2 and
Si1-xNx
• Remember, decisions are based on economics not just what
gives best performance
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Testing
• Not all solar cells are created equal, especially when multicrystalline
• Testing of finished solar cell done by I-V, end up with bins of
similar performance solar cells or grades
• Also want to test during the process
– Don’t want to continue processing if there is no chance of a good
solar cell
– Identify any problems in the process flow easily and quickly
• Movement towards techniques such as photoluminescence
to check quality of feedstock
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Imaging
• Inline
BT Imaging, Sydney Australia
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Imaging
• Find Shunts
Thermography
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner
Imaging
• Find cracks
• Series resistance
ELEG620: Solar Electric Systems University of Delaware, ECE Spring 2009 S. Bremner