View
1
Download
0
Category
Preview:
Citation preview
Iron in crystalline silicon solar cells: fundamental properties, detection techniques,
and gettering
Daniel Macdonald, AnYao Liu, and Sieu Pheng Phang
Research School of Engineering The Australian National University, Canberra
2
• Origins of Fe in multicrystalline Si ingots
• Chemical states and recombination activity of Fe in silicon
• Measuring [Fei] by QSSPC and PL imaging
• Gettering of Fe during ingot growth and cell fabrication:
– Internal gettering – at GBs, dislocations and within grains
– External gettering by P, Al and B diffusions
Outline
3
• One of the most common metal contaminants
• Total Fe concentration, measured by NAA before and after P gettering
• Comes from the crucible, not the feedstock
• Typically between 1012 -1015 cm-3
Origins of Fe in multicrystalline Si ingots
Macdonald et al. 29th IEEE PVSC New Orleans (2002)
4
• Concentration increases towards top - segregation
• Also increases at bottom – solid-state diffusion from crucible
• Only a small fraction is interstitial - around 1%
• Remainder is precipitated (or substitutional)
• The dissolved fraction has a much larger impact on lifetime
Origins of Fe in multicrystalline Si ingots
0.1 1
1011
1012
1013
1014
1015
1016
1017
keff = 0.65B
keff < 0.05
keff < 0.05
interstitial Fe
Impu
rity
conc
entra
tion
(cm
-3)
Fraction from top of ingot
total Fe
Macdonald et al. J. Appl. Phys. 97 033523 (2005)
5
Recombination activity of interstitial Fe in silicon
conduction band
valence band
FeB0/+ (EV+0.1 eV)
FeB0/- (EC-0.26 eV)
Fei0/+ (EV+0.38 eV)
• Interstitial Fe (Fei) introduces a deep donor level
• Positively charged in p-type Si - mobile at RT - forms pairs with ionised acceptors.
• Two FeB levels – acceptor and donor
• Pairs break under illumination – only Fei present in working cells Courtesy of J. Schmidt, ISFH
6
Recombination activity of Fe in p-type silicon • Different energy levels and capture cross sections - different lifetime
curves • SRH modelling on left, QSSPC data on right (trapping restricts range)
1012 1013 1014 1015 1016 1017
1
10
100
crossoverpoint
1 Ω.cm p-type Si
FeB lifetime
Carri
er li
fetim
e (µs
)
Excess carrier density ∆n (cm-3)
Fei lifetime
Macdonald et al. J. Appl. Phys. 95 1021 (2004)
7
Detecting Fei using FeB pairing • Manipulating the SRH equations
shows that:
• Zoth and Bergholz developed a famous method based on SPV
• Extended to other methods (uW-PCD and QSSPC)
• Very sensitive (similar to DLTS) • Only works in p-Si • Must avoid the crossover point
1011 1012 1013 1014 1015 1016 1017
10
100
1000
crossover pointQSSPC
SPV
Fei lifetime FeB lifetime Auger lifetime
µW-PCD
Carri
er li
fetim
e (µs
)
Excess carrier concentration (cm-3)
p-Si, NA=1016cm-3, [Fei]=1012cm-3
−×=
FeBFei
i
CFeττ
11][
8
Detecting Fei using FeB pairing
• Pre-factor C is a function of doping and excess carrier density.
• In true low injection, C becomes constant – ideal measurement region.
• Not accessible to QSSPC or uW-PCD (trapping effects).
109 1011 1013 1015 1017-1x10-13
0
1x10-13
2x10-13
3x10-13
4x10-13
5x10-13
6x10-13
NA= 3x1016cm-3
1x1016
3x1015
1x1015
3x1014
1x1014
1/C
(µs
cm-3
)
Excess carrier density ∆n (cm-3)
−×=
FeBFei
i
CFeττ
11][
Macdonald et al. J. Appl. Phys. 95 1021 (2004)
9
Iron imaging with photoluminescence (PL)
• Band-to-band PL imaging - rapid and highly-resolved method for low-injection lifetime imaging – no trapping effects.
• Allows low-injection Fe imaging - similar to original SPV technique • Two PL images required, before and after breaking FeB pairs.
1011 1012 1013 1014 1015 1016 10171
10
100
1000 100% Fei : 0% FeB 75% : 25% 50% : 50% 25% : 75% 0% : 100%
Augerlifetime
crossover point
QSSPC
SPV PL imaging
µW-PCD
Carri
er lif
etim
e (µ
s)
Excess carrier concentration ∆n (cm-3)
p-Si, NA=1016cm-3, [Fei]+[FeB]=1012cm-3
Macdonald et al. J. Appl. Phys. 103, 073710 (2008)
10
Recombination activity of Fe in n-type silicon • Neutral charge state in n-type – less attractive for minority carriers
compared to p-type. • Higher lifetime in n-type in low- to mid-injection. • Possible incentive for using n-type substrates…
1012 1013 1014 1015 1016 10170.1
1
10
100
1000
10000 n-type FZ SiND=2.4x1015cm-3
[Fei]=3.8x1012cm-3
p-type FZ SiNA=2.8x1015cm-3
[Fei]=3.4x1012cm-3
Reco
mbi
natio
n life
time
(µs)
Excess carrier concentration (cm-3) 1010 1011 1012 1013 10140.01
0.1
1
10
100
1000
10000
n-type
p-type
Effe
ctive
lifet
ime τ Fe
+int
rinsic
(µs)
Interstitial Fe concentration [Fei ] (cm -3)
NA/D = 1016 cm -3, 0.1 suns illumination
Macdonald and Geerligs, Appl. Phys. Lett. 85 4061 (2004)
11
Fe images on mc-Si
• Wafer 20% from bottom of ingot
• High [Fei] (1013 cm-3)
• Internal gettering of Fe during ingot cooling at GBs, dislocation clusters
Liu et al. Progress in PV, 19 649 (2011)
12
Fe images on mc-Si • Wafer from near very
bottom of ingot
• Moderate [Fei] (1012 cm-3)
• Small grains
• Fewer dislocation clusters
• Lower [Fei] within grains – presence of precipitation sites
Liu et al. Progress in PV, 19 649 (2011)
13
Fe images on mc-Si • Wafer from middle of
ingot
• Low [Fei] (1011 cm-3)
• Reduced gettering at GBs (due to precipitation starting at lower T)
Liu et al. Progress in PV, 19 649 (2011)
14
Internal gettering of Fe at GBs during ingot cooling • Line-scans of PL images with resolution of 25 microns • 1D diffusion/capture model – 2 free parameters – diffusion length of Fei LD(Fei) and
precipitation velocity P of the GB • Have to take care of PL artifacts!
– Image smearing in the CCD camera - use point-spread function de-convolution – Carrier spreading in the sample – use low-lifetime – i.e. high [Fei] samples)
Liu and Macdonald, IEEE JPV 2, 479, (2012)
15
Internal gettering of Fe at GBs during ingot cooling • Same GB on different wafers reveals diffusion length of Fei LD(Fei) depends on initial [Fei] • Lower [Fei] means that precipitation begins later during cooling • Modelling ingot cooling time reveals data can only be explained if precipitation at GBs
commences after super-saturation ratio of about 50 is reached!
Liu and Macdonald, IEEE JPV 2, 479, (2012)
16
Internal gettering of Fe at GBs during annealing • Annealing at low temps can drive further precipitation at GBs, and within
grains. • Higher temperatures tend to homogenise the dissolved Fe
17
• At 600 ºC, [Fei] is far above solubility limit: • Strong super-saturation • Drives precipitation at GBs,
and within grains
• At 800 ºC, [Fei] is approx equal to the solubility limit: • No precipitation
• At 900 ºC and above, [Fei] is
below solubility limit: • No precipitation • Homogenization by diffusion • Dissolution of precipitates
also possible
Internal gettering of Fe at GBs during annealing
18
Internal gettering of Fe at GBs during annealing
• 500 C annealing for various times
• 1D diffusion/capture model: • Widening denuded zone • Reduced intra-grain [Fei]
19
• At fixed temp annealing, diffusion length of Fe can be calculated from literature values.
• Very good agreement with fitted values (500 ºC)
• A method to measure diffusivity?
Internal gettering of Fe at GBs during annealing
20
• After homogenization on left, after 14 hour anneal at 500 °C on right. • Precipitation rate varies from grain to grain.
Internal gettering of Fe within the grains
21
• Precipitation rate varies despite initial Fe concentrations being similar
Internal gettering of Fe within the grains
22
Final [Fei] with respect to intra grain dislocation density, after annealing at 500oC for 14.5 hours
Microscope image of a defect etched mc-Si wafer
Internal gettering of Fe within the grains
• Less Fe precipitation in grains of low dislocation density • Average distance between dislocations is less than Fe diffusion length • Dislocations act as nucleation sites for Fe precipitation within grains
23
External gettering of Fei by P, Al and B diffusions • Fe-implanted, annealed, mono FZ p-Si, detected by QSSPC.
implant Fe anneal P, B or Al diffusion
24
P, 8
50 C
P, 8
00 C
P, 7
50 C
P, 8
50 +
650
C -- -- -- -- -- --
0.01
0.1
1
10
100
Per
cent
age
of F
e re
mai
ning
(%)
Phosphorus gettering of Fei • P gettering removes between 90-99% of Fe – better at lower temp • Adding a post-getter anneal improves gettering further – segregation ratio
improves
Phang and Macdonald, JAP 109, 073521 2011
25
Phosphorus gettering of Fei • Driving in P diffusion reduces gettering effectiveness
• Very heavily doped region (>1020 cm-3) required for best gettering
Phang and Macdonald, IEEE JPV accepted
26
Aluminium gettering of Fei • Al gettering removes more than 99.9%! • However, typical BSF firing is too short for Al gettering to rear side…
P, 8
50 C
P, 8
00 C
P, 7
50 C
P, 8
50 +
650
CAl
, 850
C, 5
5 m
inAl
, 750
C, 5
5 m
inAl
, 750
C, 1
5 se
c -- -- --
0.01
0.1
1
10
100
Per
cent
age
of F
e re
mai
ning
(%)
Phang and Macdonald, JAP 109, 073521 2011
27
Boron gettering of Fei • B gettering very effective if Boron Rich Layer (BRL) is present. • However, BRL is oxidised in-situ to allow low Joe - re-injects Fe into base. • Even a low temp anneal does not help much…
P, 8
50 C
P, 8
00 C
P, 7
50 C
P, 8
50 +
650
CAl
, 850
C, 5
5 m
inAl
, 750
C, 5
5 m
inAl
, 750
C, 1
5 se
cB,
950
C +
BRL
B, 9
50 C
no
BRL
B, 9
50+6
50 n
o BR
L
0.01
0.1
1
10
100
Per
cent
age
of F
e re
mai
ning
(%)
Phang et al., IEEE JPV 3, 261, 2013
28
Boron gettering of Fei • One solution is to remove the BRL at low temp by etch-back
• Preserves gettering • Allows lower Joe
Phang et al., IEEE JPV 3, 261, 2013
29
Boron gettering of Fei • Another idea is to overlay a light phosphorus diffusion – ‘buried emitter’. • Gives very good gettering
Phang and Macdonald, IEEE JPV accepted
30
Conclusions • Fe is common in mc-Si wafers, both dissolved and precipitated • Dissolved iron more active in p-type than n-type due to charge state • QSSPC allows very sensitive [Fei] measurements • PL allows spatially resolved Fe imaging • Gettering of dissolved Fe is critical for mc-Si cells
• Internal gettering at GBs, dislocations and in intra-grain regions • Strong super-saturation required
• External gettering via P, B or Al diffusions can be very effective • Al and B more effective than P • However, B diffusions require the presence of a BRL
Acknowledgements: this work has been supported by the Australian Research Council (ARC).
Recommended