scispace - formally typeset

Journal ArticleDOI

Ring defects in n-type Czochralski-grown silicon: A high spatial resolution study using Fourier-transform infrared spectroscopy, micro-photoluminescence, and micro-Raman

26 Dec 2018-Journal of Applied Physics (AIP Publishing LLCAIP Publishing)-Vol. 124, Iss: 24, pp 243101

Abstract: We investigated ring defects induced by a two-step anneal in n-type Czochralski-grown silicon wafers using a combination of high spatial resolution Fourier Transform Infrared Spectroscopy (FTIR), micro-photoluminescence (PL) mapping, and micro-Raman mapping. Through FTIR measurements, we show the inhomogeneous loss in interstitial oxygen with a positive correlation with the inverse lifetime. Using high-resolution micro-PL mapping, we are able to distinguish individual recombination-active oxygen precipitates within the rings with a decreasing density from the center to the edge of the sample. The radial inhomogeneity of the oxygen precipitates is likely to be related to variations in the distribution of grown-in defects. We also demonstrate that micro-Raman mapping reveals the oxygen precipitates without the smearing effects of carrier diffusion that are present in micro-PL mapping.

Summary (3 min read)

Introduction

  • J. Appl. Phys. 124, 243101 (2018); https://doi.org/10.1063/1.5057724 124, 243101 © 2018 Author(s).
  • The authors also demonstrate that micro-Raman mapping reveals the oxygen precipitates without the smearing effects of carrier diffusion that are present in micro-PL mapping.
  • During certain high-temperature treatments, Cz-Si wafers sometimes develop ring defects which influence the uniformity of the electrical properties of the wafer.
  • 1,2 Ring defects are usually associated with the inhomogeneous distribution of oxygen precipitates and their associated extended defects.
  • PL imaging yields lifetime images with a spatial resolution in the 10-100 μm range, and the relatively low injection level achieved during PL imaging results in significant image smearing due to the lateral diffusion of carriers, which typically extends many tens of micrometers.

A. Sample preparation

  • The samples used in this work were quartered sections of 4-in.
  • Diameter, 275 μm thick, phosphorus doped,a)E-mail: rabin.basnet@anu.edu.au electronic-grade Cz-Si wafers with a resistivity of 1Ω cm. Samples were divided into two different groups; Group I—for the FTIR measurements and Group II—for the micro-PL and micro-Raman mappings.
  • Oxygen precipitates were grown intentionally by subjecting the samples from both groups to a low-high, two-step anneal process.
  • After annealing, the Group II samples were chemically (HNO3:HF 10:1) etched to remove several micrometers of silicon and any denuded zone,21 before characterization.

B. FTIR scanning

  • Radial distributions of [Oi] were measured by line scans before and after annealing, with a step size of 160 μm using a Bruker Vertex 80 FTIR microscope equipped with the liquid-nitrogen-cooled InGaAs detector.
  • The FTIR line scans were conducted after removing oxide layers of annealed samples using a laser spot size of 10 μm and a spectral resolution of 4 cm−1.
  • 22 During scanning, room temperature transmission measurements were performed in the range from 500 to 2500 cm−1.
  • In order to extract the relative [Oi] from the corrected transmission spectra, a baseline value was constructed, being an expected transmission spectrum in the absence of interstitial oxygen.
  • The uncertainties in the [Oi] measurements arise mainly from variations in the wafer thickness and overlapping of the interstitial oxygen peak and the oxygen precipitate band and are estimated to be approximately 10%.

C. PL imaging

  • The pixel sizes of the PL images were approximately 162 μm and 23 μm for the standard and the magnifying lenses, respectively.

D. Micro-PL and micro-Raman mapping

  • Both the micro-PL and the micro-Raman maps were obtained at room temperature using a Horiba LabRam HR Evolution system equipped with a confocal microscope.
  • The samples were excited with a 532 nm continuous-wave laser focused onto a spot size of 1.2 μm in diameter with a 50× objective lens.
  • At room temperature, the absorption depth of the 532 nm laser light is around 1.2 μm.24.
  • The apparent precipitate density calculated from the micro-PL maps was expressed as an average areal density, to avoid the uncertainty of estimating the detection depth that would be required to estimate a volume density.
  • Micro-Raman maps were obtained using a silicon charge coupled device (CCD) detector to detect the backscattered Raman light.

A. The loss in [Oi]

  • Figure 1 presents the radial distributions of [Oi] in both the as-grown and the Group I (11 h annealed) samples, measured from the center to the edge of the samples.
  • It shows that in the as-grown sample, the [Oi] distribution is very uniform along the radial direction.
  • Note that the solubility limit of interstitial oxygen in silicon at 1000 °C in an oxygen ambient is around 9 × 1016 cm−3,11 significantly lower than the concentrations measured here, creating a strong driving force for precipitation.
  • 26 Based on Voronkov’s theory,27 the type of intrinsic point defect (vacancies and silicon interstitial) in a crystal depends on the so-called V/G criterion (where V is the crystal growth rate and G is the radial temperature gradient).
  • 1. Further, the authors observed mm-scale local fluctuations in the loss in [Oi] in the annealed sample.

B. The loss in [Oi] and carrier lifetime

  • Figure 2 shows PL-based lifetime images of both the as-grown and the Group I samples, expressed as effective carrier lifetimes via calibration with a photoconductance coil.
  • The as-grown sample shows a uniform and high average lifetime (τas-grown) of 1880 μs, as shown in Fig. 2(a).
  • Note that the dark spots in this image are surface artifacts.
  • Figure 3 shows the radial profiles of the loss in [Oi] and the inverse carrier lifetime for the Group I (11 h annealed) sample.
  • 3. This result further supports the conclusion that the lifetime striations are due to the local fluctuations of oxygen precipitates originating from grown-in precipitate nuclei as explained in Sec. III A.

C. PL images, micro-PL maps, and apparent oxygen precipitate density

  • Figure 4 shows the PL-based lifetime images of the as-grown sample with a standard lens and the Group II sample with both standard and magnifying lenses expressed as effective carrier lifetimes via calibration with a photoconductance coil.
  • The mapped areas in the samples are indicated by red points in Fig. 4, although they are not to scale.
  • 5. The recombination sites appear to be sub-micrometer sized and are closely spaced as a result; the conventional PL images were not able to resolve them.
  • This observation of precipitate growth accentuates the potential value of the micro-PL technique for studying oxygen precipitate evolution.
  • 20,37 Therefore, some of the nano-scale precipitates, which were undetected during the mapping of the 11 h annealed sample, might have grown to a detectable size after the additional 4 h of annealing, leading to the observed increase in the detected precipitate density.

D. The loss in [Oi] and apparent oxygen precipitate density

  • Figure 7 presents the apparent areal densities of the oxygen precipitates obtained from the micro-PL maps at several locations in a radial direction and the loss in [Oi] calculated from the FTIR after 11 h of precipitate growth anneal.
  • It shows no clear relationship between loss in [Oi] and apparent oxygen precipitate density.
  • This could be due to detection limits of the micro-PL, with some of the nanoprecipitates remaining undetected during the mapping.
  • Further, the micro-PL mapping is not able to detect the low recombination active oxygen precipitates due to the very high injection level employed here.
  • Thus, there could be some smaller sized and unstrained/less-active oxygen precipitates which remained undetected in the micro-PL maps, causing the weak correlation observed in Fig.

E. Micro-Raman map and oxygen precipitates

  • Figure 8 shows the micro-PL and the micro-Raman maps at the same location “B,” as shown in the PL image in Fig. 4(b).
  • From these two maps, the authors observe two different phenomena—first, they observed several new oxygen precipitates as indicated by “X” in the micro-Raman map and second, oxygen precipitates were much better resolved in the micro-Raman map.
  • The concentration of metallic impurities decorating particle “X” could be low or it might be less strained.
  • Figure 8(c) presents the line scans of the Raman intensity at 520 cm−1 and the PL intensity at 1130 nm, at the same precipitates as indicated by dotted arrows in Figs. 8(a) and 8(b).
  • To obtain a high signal-to-noise ratio and maintain a high spatial resolution (0.5 μm in this work), relatively long acquisition times in the order of 1-5 s per pixel are required.

IV. CONCLUSIONS

  • The authors have presented a high spatial resolution investigation of ring defects in two-step annealed Cz-Si wafers using a combination of FTIR scanning microscopy, micro-PL mapping at high injection conditions, and micro-Raman mapping.
  • The authors observed a direct local correlation between losses in [Oi] obtained from scanning FTIR measurements and the inverse lifetime extracted from PL images in a two-step annealed sample.
  • Further, this paper demonstrated the use of micro-PL mapping to investigate ring defects.
  • The authors results demonstrate that the ring defects or lifetime striations are indeed due to the cumulative effect of individual oxygen precipitates.
  • Furthermore, the micro-Raman maps confirm that besides the precipitates detected in the micro-PL maps, there are in fact smaller and/or less-active precipitates not revealed by the micro-PL technique due to the carrier smearing.

Did you find this useful? Give us your feedback

Content maybe subject to copyright    Report

J. Appl. Phys. 124, 243101 (2018); https://doi.org/10.1063/1.5057724 124, 243101
© 2018 Author(s).
Ring defects in n-type Czochralski-grown
silicon: A high spatial resolution study using
Fourier-transform infrared spectroscopy,
micro-photoluminescence, and micro-Raman
Cite as: J. Appl. Phys. 124, 243101 (2018); https://doi.org/10.1063/1.5057724
Submitted: 15 September 2018 . Accepted: 06 December 2018 . Published Online: 26 December 2018
Rabin Basnet , Chang Sun , Huiting Wu , Hieu T. Nguyen , Fiacre Emile Rougieux, and Daniel
Macdonald
ARTICLES YOU MAY BE INTERESTED IN
Tutorial: Hysteresis during the reactive magnetron sputtering process
Journal of Applied Physics 124, 241101 (2018); https://doi.org/10.1063/1.5042084
Perspective: Magnetic skyrmions—Overview of recent progress in an active research field
Journal of Applied Physics 124, 240901 (2018); https://doi.org/10.1063/1.5048972
Epitaxial growth of InN thin films by plasma-enhanced atomic layer deposition
Journal of Applied Physics 124, 243104 (2018); https://doi.org/10.1063/1.5054155

Ring defects in n-type Czochralski-grown silicon: A high spatial
resolution study using Fourier-transform infrared spectroscopy,
micro-photoluminescence, and micro-Raman
Rabin Basnet,
1,a)
Chang Sun,
1
Huiting Wu,
1
Hieu T. Nguyen,
1
Fiacre Emile Rougieux,
2
and Daniel Macdonald
1
1
Research School of Engineering, The Australian National University, Canberra, Australian Capital Territory
2601, Australia
2
School of Photovoltaic and Renewable Energy Engineering, The University of New South Wales, Sydney,
New South Wales, Australia
(Received 15 September 2018; accepted 6 December 2018; published online 26 December 2018)
We investigated ring defects induced by a two-step anneal in n-type Czochralski-grown silicon
wafers using a combination of high spatial resolution Fourier Transform Infrared Spectroscopy
(FTIR), micro-photoluminescence (PL) mapping, and micro-Raman mapping. Through FTIR mea-
surements, we show the inhomogeneous loss in interstitial oxygen with a positive correlation with
the inverse lifetime. Using high-resolution micro-PL mapping, we are able to distinguish individual
recombination-active oxygen precipitates within th e rings with a decreasing density from the center
to the edge of the sample. The radial inhomogeneity of the oxygen precipitates is likely to be related
to variations in the distribution of grown-in defects. We also demonstrate that micro-Raman
mapping reveals the oxygen precipitates without the smearing effects of carrier diffusion that are
present in micro-PL mapping. Published by AIP Publishing. https://doi.org/10.1063/1.5057724
I. INTRODUCTION
N-type Czochralski-grown silicon (Cz-Si) is the most
common material used for high-efciency crystalline silicon
solar cells. During certain high-temperature treatments, Cz-Si
wafers sometimes develop ring defects which inuence the
uniformity of the electrical properties of the wafer.
1,2
Ring
defects are usually associated with the inhomogeneous distri-
bution of oxygen precipitates and their associated extended
defects.
13
Ring defects can contribute up to 4% (absolute)
loss in solar cell efciency
4
owing to the strong recombina-
tion activity of these oxygen precipitates.
57
However, a
detailed understanding of the origin and properties of ring
defects remains incomplete. Improved techniques to observe
and quantify the recombination active precipitates within the
ring defects will help to further improve this understanding.
It is known tha t the interstitial oxygen concentration [O
i
]
distribution during ingot growth affects the formation of micro-
defects.
8,9
Nauka et al.
10
observed a quantitativ e rela tionship
between the [O
i
] distribution and the excess carrier lifetime, in
as-grown Cz-Si along the crystal growth direction. Similarly,
Varker et al.
11
observed the inhomogeneous distribution of
minority carrier lifetime along the radial direction after two-step
annealing, in Cz-Si wa fers. More recently , Gaspar et al.
12
dem-
onstra ted that the initial radial distribution of [O
i
] has an inu-
ence on the radial distribution of the minority carrier lifetime at
the millimeter (mm) scale. Previous work has used Fourier
Transform Infrared Spectroscopy (FTIR) to measure the [O
i
]
concentration with mm scale spatial resolution.
12
In this work,
we investigate the distribution of the [O
i
] concentration at the
sub-mm scale using an FTIR scanning microscope. We show
that the high resolution [O
i
] scans are well corr ela ted with pho-
toluminescence (PL) based images of the carrier lifetime on
n-type Cz wafers subjected to a two-step anneal.
PL imaging yields lifetime images with a spatial resolution
in the 10-100 μm range, and the rela tively low injection lev e l
achieved during PL imaging results in signicant image smear-
ing due to the lateral diffusion of carriers, which typically
extends many tens of micrometers.
13
As a result, ring defects
appear as continuous bands of a reduced lifetime in PL images
under constant and homogenous laser excitation, as has been
reported by numerous authors.
4,14
To address these problems,
we have utilized a micro-PL system with confocal optics with
a higher spatial resolution, in principle down to 1 μmbasedon
the optical conguration. In addition, a much higher injection
level can be achieved, which makes the carrier lifetime largely
Auger limited
1517
and reduces carrier smearing effects during
measurements. However, in pra ctice, carrier smearing effects
are still present and limit the achievable spatial resolution to
several micrometers. Furthermore, micro-PL only detects those
particles that are strongly recombination active. To further
improve the spatial resolution, we also applied micro-Raman
mapping. The Raman signal from crystalline silicon is very
sensitive to defects and impurities in the material such as dislo-
cations, metal precipita tes, and doping densities.
18,19
We will
utilize the spatial resolution below 1 μm to observe the oxygen
precipitates and demonstrate that the two mapping techniques
are complementary to each other.
II. EXPERIMEN TAL DETAILS
A. Sample preparation
The samples used in this work were quartered sect-
ions of 4-in. diameter, 275 μm thick, phosphorus doped,
a)
E-mail: rabin.basnet@anu.edu.au
JOURNAL OF APPLIED PHYSICS 124, 243101 (2018)
0021-8979/2018/124(24)/243101/7/$30.00 124, 243101-1 Published by AIP Publishing.

electronic-grade Cz-Si wafers with a resistivity of 1 Ω cm.
Samples were divided into two different groups; Group
Ifor the FTIR measurements and Group IIfor the
micro-PL and micro-Raman mappings. The samples from
Group I were saw-damage etched with tetramethylammonium
hydroxide (TMAH) solution and samples from Group II were
chemically (HNO
3
:HF 10:1) etched to remove 10-12 μmfrom
each side, before standard chemical cleaning steps prior to
further processing. Oxygen precipitates were grown intention-
ally by subjecting the samples from both groups to a low-high,
two-step anneal process. First, the samples were annealed at
650 °C for 5 h in nitrogen for the nucleation of the oxygen
precipitates.
20
Subsequently, in the second step, the samples
were annealed at 1000 °C for 11 h in oxygen for the growth
of the oxygen precipitates.
20
Further, some samples from
Group II were annealed at 1000 °C for an additional 4 h
(total 15 h). After annealing, the Group II samples were
chemically (HNO
3
:HF 10:1) etched to remove several micro-
meters of silicon and any denuded zone,
21
before characteri-
zation. We note that these annealing conditions are indeed
not typical for solar cell processing, but they were chosen in
this work to ensure a high precipitate density to enable a
clear demonstration of the methods.
To allow the measurement of PL images and micro-PL
maps, all samples were passivated with thermal aluminum
oxygen (Al
2
O
3
) layers deposited on both sides using a Beneq
TFS-200 atomic layer deposition (ALD) system.
B. FTIR scanning
In this study, radial distributions of [O
i
] were measured
by line scans before and after annealing, with a step size of
160 μm using a Bruker Vertex 80 FTIR microscope equipped
with the liquid-nitrogen-cooled InGaAs detector. The FTIR
line scans were conducted after removing oxide layers of
annealed samples using a laser spot size of 10 μm and a spec-
tral resolution of 4 cm
1
. In general, the broad infrared
absorption band near 1107 cm
1
at room temperature has
been ascribed to [O
i
].
22
During scanning, room temperature
transmission measurements were performed in the range
from 500 to 2500 cm
1
. The transmission spectra obtained
from the FTIR measurements were corrected by ltering out
high-frequency oscillations via wavelet transformation. In
order to extract the relative [O
i
] from the corrected transmis-
sion spectra, a baseline value was constructed, being an
expected transmission spectrum in the absence of interstitial
oxygen. The uncertainties in the [O
i
] measurements arise
mainly from variations in the wafer thickness and overlap-
ping of the interstitial oxygen peak and the oxygen precipi-
tate band and are estimated to be approximately 10%. The
relative [O
i
] values extracted from the FTIR measurements
were converted to absolute [O
i
] concentrations by using
SEMI MF standard 1188-1107.
C. PL imaging
PL images were obtained using an LIS-R1 PL imaging
tool from BT imaging.
23
The pixel sizes of the PL images
were approximately 162 μm and 23 μm for the standard and
the magnifying lenses, respectively. The injection level was
estimated to be in the range of 10
15
-10
16
cm
3
.
D. Micro-PL and micro-Raman mapping
In this work, both the micro-PL and the micro-Raman
maps were obtained at room temperature using a Horiba
LabRam HR Evolution system equipped with a confocal
microscope. The samples were excited with a 532 nm
continuous-wave laser focused onto a spot size of 1.2 μmin
diameter with a 50× objective lens. The laser power on the
sample surface was 19 mW.
Micro-PL maps were obtained using a liquid-nitrogen-
cooled InGaAs array detector and a step size (pixel size) of
2 μm in both x and y directions. The band-to-band transition
of free carriers in silicon is represented as a peak at around
1130 nm. The intensity maps of the samples were obtained by
integrating the luminescence intensity across the wavelength
range from 1050 to 1200 nm. At room temperature, the
absorption depth of the 532 nm laser light is around 1.2 μm.
24
The apparent precipitate density calculated from the micro-PL
maps was expressed as an average areal density, to avoid the
uncertainty of estimating the detection depth that would be
required to estimate a volume density. The detection depth
will be somewhat deeper than the absorption depth due to
the diffusion of excess carriers, which is estimated to be
several micrometers in magnitude. The injection level during
micro-PL mapping was estimated to be around 10
18
cm
3
.
Micro-Raman maps were obtained using a silicon charge
coupled device (CCD) detector to detect the backscattered
Raman light. In the absence of either internal or external
stress, crystalline silicon has a Raman peak at 520 cm
1
.We
mapped the intensity of this peak with a step-size of 0.5 μm
in both x and y directions.
III. RESULTS AND DISCUSSION
A. The loss in [O
i
]
Figure 1 presents the radial distributions of [O
i
] in both
the as-grown and the Group I (11 h annealed) samples,
FIG. 1. FTIR measured radial distribution of [O
i
] for the as-grown sample
and the Group I sample (650 °C-5 h; 1000 °C-11 h).
243101-2 Basnet et al. J. Appl. Phys. 124, 243101 (2018)

measured from the center to the edge of the samples. It
shows that in the as-grown sample, the [O
i
] distribution is
very uniform along the radial direction. However, in the
annealed sample, we observed a strongly non-uniform distri-
bution of [O
i
]lower toward the center in comparison to the
edge of the sample, and with mm-scale local variations. The
loss in [O
i
] can be determined by Δ[O
i
]=[O
i
]
as-grown
[O
i
]
annealed
, and indicates the magnitude of oxygen precipita-
tion in the samples. Note that the solubility limit of intersti-
tial oxygen in silicon at 1000 °C in an oxygen ambient is
around 9 × 10
16
cm
3
,
11
signicantly lower than the concen-
trations measured here, creating a strong driving force for
precipitation. In general, the loss in [O
i
] is expected to
depend on both the initial [O
i
] content and the thermal treat-
ment.
25
Further, oxygen precipitation is enhanced by the
presence of vacancies in the crystal.
26
Based on Voronkovs
theory,
27
the type of intrinsic point defect (vacancies and
silicon interstitial) in a crystal depends on the so-called V/G
criterion (where V is the crystal growth rate and G is the radial
temperature gradient). In principle, the relatively fast growth
rates used for commercial silicon ingots result in a relatively
vacancy-rich crystal.
26
This usually results in a radial vacancy
gradient which decreases toward the edge,
28
leading to less
oxygen precipitation and a lower loss of [Oi] toward the edge,
as we observed in the annealed sample shown in Fig. 1.
Further, we observed mm- scale local uctuations in
the loss in [O
i
] in the annealed sample. This is also likely to
be related to the crystal growth conditions, where small tem-
perature deviations can occur in the melt and at the crystal
growth interface due to variations in thermal convection,
crystal rotation, and crystal pull rate.
29,30
These temperature
variations cause microscopic growth rate uctuations, as a
result of which impurities such as dopants, oxygen, and
carbon are inhomogeneously incorporated into the growing
crystal.
3134
The inhomogeneously distributed impurities in
the crystal can trigger clustering of intrinsic point defects
(vacancies and silicon interstitials) and induced agglomeration
of supersaturated oxygen during cooling down.
35
This leads to
an inhomogeneous incorporation of grown-in oxygen precipi-
tate nuclei and subsequently induces local uctuations in the
loss of [O
i
] during annealing.
B. The loss in [O
i
] and carrier lifetime
Figure 2 shows PL-based lifetime images of both the
as-grown and the Group I samples, expressed as effecti ve
carrier lifetimes via calibration with a photoconductance
coil.
23
The as-grown sample shows a uniform and high
average lifetime (τ
as-grown
) of 1880 μ s, as shown in Fig. 2(a).
Note that the dark spots in this image are surface artifacts.
However, after the two-step annea l, ring defects were formed
which caused lifetime striations, and the average minority
carrier lifetime (τ
annealed
) had degraded to approximately
5 μs, as shown in Fig. 2(b). All PL images are shown at
an average injection level of around Δp ¼ 5 10
13
cm
3
by
varying the illumination intensity.
Figure 3 shows the radial proles of the loss in [O
i
] and
the inverse carrier lifetime for the Group I (11 h annealed)
sample. Note that for an approximately equal excess carrier
density, the inverse lifetime is approximately proportional to
the effective defect concentration, and represents the recom-
bination strength of the defects at a given location. Both the
loss in [O
i
] and the inverse lifetime proles were measured
with the same spatial resolution of 160 μm. We observed
clear correlations, both globally across the sample and at the
local scale, between the amount of interstitial oxygen loss
and the inverse lifetime, as shown in Fig. 3. This result
further supports the conclusion that the lifetime striations
are due to the local uctuations of oxygen precipitates origi-
nating from grown-in precipitate nuclei as explained in
Sec. III A. Thus, dissolving the grown-in oxygen precipitate
nuclei by a high-temperature pre-annealing step known as a
tabula rasa can be expected to mitigate the formation of the
FIG. 2. PL based carr ier lifetime
images of the (a) as-grown sample
andtheGroupIsample(650°C-5h;
1000 °C-11 h). The PL images were
captured at an average injection level
of Δp ¼ 5 10
13
cm
3
. The brig ht
circle in the images is an artifact due
to the conductance coil in the PL
imaging tool.
FIG. 3. Microscale radial distribution of the inverse lifetime extracted from
PL image and loss in [O
i
] obtained from FTIR measurements in a Group I
sample (650 °C-5 h; 1000 °C-11 h).
243101-3 Basnet et al. J. Appl. Phys. 124, 243101 (2018)

ring defects in subsequent thermal treatments, as has been
demonstrated by previous research work.
36
C. PL images, micro-PL maps, and apparent oxygen
precipitate density
Figure 4 shows the PL-based lifetime images of the
as-grown sample with a standard lens and the Group I I
sample with both standard and magnifying lenses expressed
as effective carrier lifetimes via calibration with a photocon-
ductance coil.
23
Figure 5 shows the micro-PL maps of the as-grown and
the Group II (650 °C-5 h; 1000 °C-15 h) samples. The mapped
areas in the samples are indicated by red points in Fig. 4,
although they are not to scale. In the as-grown sample, no
individual recombination sites were observed, whereas, in
the annealed sample, heterogeneously distributed individual
recombination sites were detected, as shown in Fig. 5.The
recombination sites appear to be sub-micrometer sized and
are closely spaced as a result; the conventional PL images
were not able to resolve them. However, the micro-PL maps
successfully resolved the individual recombination sites due
to the use of a pixel size of 2 μm and a much higher injection
level. Therefore, we were able to demonstrate that ring
defects are in fact a cumulative effect of the individual recom-
bination sites rather than continuous bands, as observed in the
PL images in Figs. 4(b) and 4(c). In this work, we found ran-
domly distributed oxygen precipitates with a higher apparent
areal density of 9 10
9
cm
2
at position 1, as shown in the
map in Fig. 5(b). Similarly, at position 2, we found more
FIG. 4. PL based carrier lifetime images of the (a) as-grown sample and the Group II sample (650 °C-5 h; 1000 °C-15 h) captured with (b) standard lens (pixel
size of 162 μm) and (c) magnifying lens (pixel size of 23 μm). The PL images were captured at an average injection level of Δp ¼ 5 10
13
cm
3
. The red
points indicate the locations (not the size) of the micro-PL maps in this work. The black box in (b) indicates the area of the PL image captured with the magni-
fying lens shown in (c). The bright circle in the images is an artifact due to the conductance coil in the PL imaging tool. The straight lines in (b) and (c) are
laser marks for locating specic regions.
FIG. 5. Micro-PL maps of 300 μm × 300 μm areas on (a) the as-grown sample and [(b)(d)] on the Group II sample (650 °C-5 h; 1000 °C-15 h). The locations
of the micro-PL maps are indicated in the PL images in Fig. 4. (e) Inverse lifetime extracted from the PL image as a function of apparent oxygen precipitate
areal density obtained from micro-PL maps on the Group II sample.
243101-4 Basnet et al. J. Appl. Phys. 124, 243101 (2018)

Figures (8)
Citations
More filters

Journal ArticleDOI
Chang Sun1, AnYao Liu1, Aref Samadi2, Catherine Chan2  +2 moreInstitutions (2)
Abstract: The concentrations of Cr, Fe, Ni, and Cu in a cast-monocrystalline silicon ingot grown for solar cell applications are reported. Wafers taken from along the ingot are coated with silicon nitride films and annealed, causing mobile impurities to be gettered to the films. Secondary ion mass spectrometry is applied to measure the metal content in the silicon nitride films. The bulk concentrations of the gettered metals in samples along the ingot are found to be: Cr (3.3 10–3.3 10 cm ), Fe (3.2 10–2.5 10 cm ), Ni (1.5 10–1.3 10 cm ), and Cu (7.1 10–3.2 10 cm ). For each metal, the lower limit is measured on the wafer from the middle of the ingot, and the higher limit is measured on wafers from the bottom or the top. The results are compared with similar data recently measured on a highperformance multicrystalline silicon ingot. The results provide insights into the total bulk concentrations of the metals in cast-grown ingots.

6 citations


Journal ArticleDOI
Rabin Basnet1, Sieu Pheng Phang1, Chang Sun1, Fiacre Rougieux2  +1 moreInstitutions (2)
Abstract: This paper presents experimental studies on the formation of ring defects during high-temperature annealing in both electronic-grade and upgraded metallurgical-grade (UMG) Czochralski-grown silicon wafers. Generally, a faster onset of ring defects, or shorter incubation time, was observed in the UMG samples ([Oi] = 6.3 × 1017 cm−3) in comparison to the electronic-grade samples ([Oi] = 3.9 × 1017 cm−3) used in this work. By applying a tabula rasa (TR) treatment prior to annealing, the incubation time can be increased for both types of wafers. We show that TR temperatures above 1000 °C are necessary to effectively dissolve grown-in oxygen precipitate nuclei and limit the subsequent formation of ring defects. A 30 min TR treatment at 1000 °C resulted in the longest incubation time for both types of samples used in this work, as it achieved the best balance between precipitate nuclei dissolution and precipitate re-growth/ripening. Finally, a nitrogen ambient TR step showed a short incubation time for the formation of ring defects in comparison to an oxygen ambient TR step.

3 citations



Journal ArticleDOI
Jonas Schön1, Tim Niewelt1, Di Mu1, Stephan Maus1  +3 moreInstitutions (2)
Abstract: Commercial silicon is prone to form silicon oxide precipitates during high-temperature treatments typical for solar cell production. Oxide precipitates can cause severe efficiency degradation in solar cells. We have developed a model describing the nucleation and growth of oxide precipitates that considers silicon self-interstitial defects and surface effects influencing the precipitate growth in ∼150 μm thick wafers during the solar cell processing. This kinetic model is calibrated with experiments that cause a well-defined and strong precipitate growth to give a prediction of the carrier lifetime limitation because of the oxide precipitates. We test the oxide precipitate model with scanning Fourier-transform infrared spectroscopy, selective etching, and lifetime measurements on typical Cz solar cell wafers before and after solar cell processes. Despite the relatively rough saw damaged etched surfaces and the thin wafers, we observe recurring ring patterns in the measurements of interstitial oxygen reductions, oxide precipitate etch pit density, and recombination activity by photoluminescence imaging. The concentration of precipitated oxygen correlates with the recombination activity and with the initial interstitial oxygen concentration. However, we found lifetime measurements to be a more sensitive technique to study oxide precipitates and using these we find smaller precipitates not detected by selective etching are very recombination active too. The measured concentrations of precipitated oxygen and lifetime agree fairly well with the predictions of the model.

1 citations


Cites background or methods from "Ring defects in n-type Czochralski-..."

  • ...The methods featuring a higher spatial resolution, such as micro-Raman, might be capable of detecting an additional share of small precipitates [12] and, thus, could help to understand the evolution of small precipitates....

    [...]

  • ...[11] and [12], and OP density (EPD) on these wafers....

    [...]



References
More filters

Journal ArticleDOI
Abstract: The incorporation of solute elements into single crystals of germanium grown from the melt is examined in terms of a simple model. The theory takes account of the contribution of solute transport in the melt, owing to diffusion and fluid motion, to the over‐all process of impurity incorporation during steady‐state crystallization. The analysis is extended to treat the transient inclusion of solute which results when the composition of the melt is abruptly changed.

1,198 citations


Journal ArticleDOI
Abstract: Photoluminescence imaging is demonstrated to be an extremely fast spatially resolved characterization technique for large silicon wafers. The spatial variation of the effective minority carrier lifetime is measured without being affected by minority carrier trapping or by excess carriers in space charge regions, effects that lead to experimental artifacts in other techniques. Photoluminescence imaging is contactless and can therefore be used for process monitoring before and after individual processing stages, for example, in photovoltaics research. Photoluminescence imaging is also demonstrated to be fast enough to be used as an in-line tool for spatially resolved characterization in an industrial environment.

608 citations


Journal ArticleDOI
Abstract: The observed influence of growth rate V and temperature gradient G 0 on swirl defects leads to the conclusion: the first stage of defect formation is recombination and diffusion of vacancies and self-interstitials in the vicinity of the crystallization front. The equilibrium concentrations C v0 and C i0 and the diffusion coefficients D v and D i of these points defects are determined from the experimental data; C v0 is slightly higher than C i0 but D v C v0 is lower than D i C i0 . The type and concent ration of point defects that remain in the crystal after the recombination, depends on the ratio V / G 0 (vacancies if V / G 0 >ξ t , interstitials if V / G 0 t where ξ t is a certain constant). Typical growth conditions correspond to the interstitial case V / G 0 t . The subsequent process consists of several successive stages: diffusion of interstitials to the crystal surface, nucleation of primary interstitial clusters, cluster growth, conversion of clusters into other forms (particularly dislocation loops). The quantitative results of the theory are in a fairly good agreement with the growth-stop and growth-quench experiments and the data on concentration and size of microdefects.

474 citations


Journal ArticleDOI
Abstract: An optical absorption band at 9\ensuremath{\mu} has been correlated with the oxygen content in silicon. Pulled silicon crystals were found to contain up to ${10}^{18}$ oxygen atoms per ${\mathrm{cm}}^{3}$ which seem to originate from the quartz crucible. The oxygen concentration in silicon crystals prepared by the floating zone technique in vacuum was found to be less than ${10}^{16}$ oxygen atoms per ${\mathrm{cm}}^{3}$. The 9\ensuremath{\mu} absorption due to silicon-oxygen bond stretching vibrations provides a possibility for a quantitative oxygen analysis of high sensitivity. A corresponding absorption in germanium at 11.6\ensuremath{\mu} is believed to be due to a germanium-oxygen vibration.

316 citations


Journal ArticleDOI
Vladimir V. Voronkov1, R. Falster1Institutions (1)
Abstract: A model of multi-step vacancy aggregation in dislocation-free silicon crystals is analyzed. In this model, voids are first nucleated (normally just below 1100°C). The vacancy loss to voids is retarded below some characteristic temperature (about 1020°C) as the vacancies become bound by oxygen into O2V complexes. These remaining vacancies control nucleation of oxide particles on further cooling. Some vacancies survive even this stage and control nucleation of oxygen clusters at still lower temperature (around 700°C). The oxygen clusters are major precipitation nuclei during subsequent heat treatments. It is through residual vacancies that the oxygen precipitation behavior is closely related to the grown-in microdefects (voids and particles). The microdefect properties and the residual vacancy concentration ( C res ) are computed in dependence of the starting vacancy concentration ( C 0 ). The C res (C 0 ) function is of a twin-peak type which results in a banded precipitation pattern if C 0 decreases gradually either in radial or axial direction. The model accounts for complicated (strongly banded) precipitation patterns, particularly those observed in halted and quenched crystals.

151 citations