Available via license: CC BY-NC-ND 3.0
Content may be subject to copyright.
doi:10.1016/j.proeng.2011.04.239
A
vailable online at www.sciencedirect.com
ICM11
Effect of mechanical surface damage on
Silicon wafer strength
Daisuke Echizenya
ab
*, Hiroo Sakamoto
a
, Katsuhiko Sasaki
b
a
Mitsubishi Electric Corporation, 8-1-1, Tsukaguchi-Honmachi, Amagasaki, Hyogo 661-8661, Japan
b
Hokkaido University, 13-8, kita, Sapporo,Hokkaido 060-8628, Japan
Abstract
Solar power generation using polycrystalline silicon wafers has been rapidly growing in recent years. As a result, it is
required to understand the strength characteristics of polycrystalline silicon wafers in order to enhance their quality.
Scratches and material defects should be taken into consideration when strength characteristics of polycrystalline
silicon are evaluated, since it is a brittle material. In this paper, bending strength of polycrystalline silicon wafers for
solar cells were measured, and evaluation regarding the cause of different strength values, which depend on
manufacturing conditions of the wafer, was conducted based on fracture mechanics. Residual stress measurements
using Raman spectroscopic and observation with TEM (Transmission Electron Microscope) were also conducted.
The results clarified the existence of numerous cracks on the wafer surface that are assumed to be generated during
slicing process. Thus, it was confirmed that wafer strength depends on the level of machining damage in slicing
process. We can establish high reliability for PV modules as a result of modifying the slicing conditions to minimize
the mechanical surface damage on wafers and increase the wafer strength.
Keywords: Crystalline silicon; Fracture strength; Slicing process; Raman spectroscopic; Crack; Residual stress
1. Introduction
In recent years, photovoltaic power generation has been extending all over the world in consideration
to environmental problems. Polycrystalline silicon is widely used since it is cost effective, and this
tendency is expected to continue [1].It is necessary to understand the strength characteristics of silicon
* Corresponding author. Tel.: +81-6-6497-7556
E-mail address: Echizenya.Daisuke@dx.MitsubishiElectric.co.jp.
Procedia Engineering 10 (2011) 1440–1445
1877-7058 © 2011 Published by Elsevier Ltd.
Selection and peer-review under responsibility of ICM11
Open access under CC BY-NC-ND license.
© 2011 Published by Elsevier Ltd.
Selection and peer-review under responsibility of ICM11
Open access under CC BY-NC-ND license.
Daisuke Echizenya et al. / Procedia Engineering 10 (2011) 1440–1445 1441
wafers in order to enhance the quality (ex. the thinner the wafer, reliability for a longer lifetime) of PV
modules.
Crystalline silicon is a brittle material and has high notch sensitivity. Therefore, crack evaluation is
indispensable in its quantitative strength assessment, and especially attention should be paid to
manufacturing damage in the slicing process from ingot to wafer production. In the past, there were some
reports of strength evaluation of silicon wafer for integrated circuits [2]. Because the manufacturing
process of wafers for solar cells is not the same as for integrated circuits it needed a single purpose
evaluation for polycrystalline silicon. Especially polycrystalline silicon has many crystal grain boundaries
and deficiencies in the crystal structure. It means that the clarification of a strength factor is important for
the polycrystalline silicon.
This paper presents the effects of surface damage on the silicon wafer strength. The polycrystalline
silicon wafers were manufactured by a wire saw. It is known that the affected layer of the surface occurs
when the ingot is sliced (Fig 1). First we measured strength and fracture toughness, so that the strength of
damaged polycrystalline silicon wafer was evaluated based on fracture mechanics assuming the damaged
layer as potential cracks on the wafer surface. For evaluation of surface damage, microscopic observation
by using both a transmission electron microscope (TEM) and Raman spectroscopic were conducted for
different strength polycrystalline silicon wafers.
2. Evaluation of
5
trength Characteristics
2.1. Test Samples
A polycrystalline silicon wafer for this study is shown in Fig 2. The sample was manufactured by
slicing an ingot into 150mm x 150mm square wafers of 200ȝm to 300ȝm thickness using a wire saw, and
then cutting 10mm widths using diamond dicing. The cut surface by diamond dicing is shown in Fig 3.
The cut surface is smoother compared with the wafer surface and it was confirmed that fracture did not
initiate from the cut surface. Three types of test pieces were prepared: samples are manufactured by
different slicing conditions. Sample A and B were made from the same ingot, but sample C was made
from the different ingot.
2.2. Bending Strength Test
A four point bending test was conducted to measure the tensile strength by using a tensile testing
machine (Fig 4). A four point bending test can get more accurate results than a three point bending test
and ring-on-ring test. Because it can apply high stress over a large area, there is little displacement from
the high stress point to the break point. Bending strength ı was calculated using the following equation,
2
)(3 21
2
LL
bt
F
V
(1)
Where, F is the applied load, b is the width of the sample, t is the thickness, L1 is the larger span, L2 is
the smaller span.
1442 Daisuke Echizenya et al. / Procedia Engineering 10 (2011) 1440–1445
Fig 5 shows measured results of bending strength. Samples A, B, and C have different bending
strengths. Especially, the bending strength of samples A and C are different value although samples A
and C are made from the same ingot. It suggests that the difference in bending strength is not caused by
the difference in ingots but caused by slicing damage on the surface.
2.3. Fracture Toughness Test
In order to evaluate the effect of surface cracks due to machining damage, fracture toughness (KIC) of
polycrystalline silicon wafers was measured by controlled surface flaws (CSF) method [3]. Same test
piece for bending strength test was used. Knoop indenter was pressed on the center of the test piece with
its longitudinal axis set to be perpendicular to the longitudinal direction of the test piece. Indentation load
was set at 5N and 10N. Sample A and sample B from different ingots were used. Then, the four point
bending test was conducted by loading a tensile stress on the surface where the Knoop indentation had
been applied. KIC was calculated by following equation for stress intensity factor,
KIC
˙
F
V
Ba
S
(2)
where ıB is the fracture stress obtained from the bending strength test, a is the crack depth due to Knoop
indentation, and F is a correction factor, respectively. Semi-elliptic crack due to Knoop indentation can be
seen and F is determined from the following equation [3].
F=1.1359-0.3929ȝ-0.3440ȝ2-0.2613ȝ3+Ȝ(-1.5184+0.4178ȝ+0.7846ȝ2-0.6329ȝ3)
+Ȝ2(4.3721-13.9152ȝ+16.2550ȝ2-6.4894ȝ3)+Ȝ3(-3.9502+12.5334ȝ-14.6137ȝ2+5.8110ȝ3)
and
ȝ=d/w, Ȝ=d/t (3)
Slurry
Affected
layer
Wire
Wire
150mm
Fig. 1. Image of affected layer when wire slicing Fig. 2. silicon wafer Fig. 3. Test piece
L1=30mm
L2=10mm
Fig. 4. Four-point bending test
Daisuke Echizenya et al. / Procedia Engineering 10 (2011) 1440–1445 1443
Where, d is crack depth, w is a half crack width, and t is thickness of the test piece.
Although silicon is an anisotropic material, measured KIC for different crystal orientations are
distributed from 1.11±0.07 to 1.18±0.03 MPa 㨯m0.5 according to the previous investigation [3], which is
little affected by crystal orientation. Therefore, crystal orientation was not taken into consideration in this
investigation.
Fig 6 shows measured results of fracture toughness (KIC). Fig 7 shows the fracture surface. Sample A
and B have almost same KIC values. Though the bending strength of sample A and B is different they
have almost same the KIC value. Therefore, it can be considered that the difference in bending strength is
caused by slicing damage on the surface as well.
40 60 80100 300
.01
.1
1
5
10
20
30
50
70
80
90
95
99
99.9
99.99
0.2 0.4 0.6 0.8 1
.01
.1
1
5
10
20
30
50
70
80
90
95
99
99.9
99.99
Fig. 5. Results of strength measurement Fig. 6. Results of fracture toughness measurement Fig. 7. Controlled flaw
2.4. Strength Test for the Wafer Eliminated Damage
We manufactured a sample whose surface damage was eliminated then we conducted a bending
strength test described in chapter 2.2. Sample D was manufactured from sample A by surface polishing
and reducing 20% of its thickness. Fig 8 shows the polished sample compared with an unpolished sample.
The results are shown in Fig 9. The bending strength of the sample D is about three times greater than
that of sample A in the initial condition, confirming the big contribution of surface cracks to the strength
deterioration of polycrystalline silicon wafers
Therefore this result shows that the crystal grain boundary and deficiency of the crystal structure of
polycrystalline silicon does not affect the strength of the wafer at least under 300MPa.
10 100 1000
.01
.1
1
5
10
20
30
50
70
80
90
95
99
99.9
99.99
Fig. 8. Polished sample Fig. 9. Results of strength measurement
Depth˖a
Width˖2b
10Pm
Initial crack by indentation
Cumulative probability(%)
Fracture toughness(MPa m0.5)
Cumulative probability(%)
Strength(MPa)
Sample A
Sample B
Sample C
Sample D
Cumulative probability(%)
Strength(MPa)
sample A(113MPa)
sample B(122MPa)
sample C(89.2MPa)
sample A(0.56MPa m0.5)
sam
p
le B
(
0.58MPa
m
0.5
)
$
%
Bare
Polished
1444 Daisuke Echizenya et al. / Procedia Engineering 10 (2011) 1440–1445
3. Evaluation of Surface Damage
3.1. Surface Observation by TEM
Surface observation by TEM was conducted for sample B and C in order to investigate the level of
surface damage. The sample for the observation was cut by Focused Ion Beam (FIB) to observe the
thickness direction of the wafers. Average sample thickness was 0.25 ȝm, and length was 20 ȝm.
Results of TEM observation is shown in Fig 10 and Fig 11 at the same scale. In spite of a narrow
observation surface of a couple of micron square meters, some interference fringes that indicated crack
existence were observed in both samples. Furthermore, the crack in sample C is apparently larger than
that of sample B, which indicates a correlation with bending strength.
Fig. 10. TEM image of cross section (sample B) Fig. 11. TEM image of cross section (sample C)
3.2. Residual Stress Measurement using a Raman Spectroscopic
Observed area by TEM is 10-8 times smaller compared to test piece area of strength sample. Therefore,
evaluation of residual stress distribution using Raman spectroscopic was tried in order to cover a wide
surface range. Raman spectroscopic analysis can measure residual stress as mechanical damage, and
evaluate the deterioration of the crystal structure [5]. Total value of three axis principal stress is measured
as a positive value of compressive stress in the Raman spectroscopic. Residual stress from the surface to
the thickness direction was measured in the wide cross section.
Measured residual stress for sample A and D are shown in Fig 12. In Fig. 12 axis of ordinate shows
residual stress and abscissa is a distance from the surface normalized by sample A thickness. The
compressive residual stress near the surface of sample A is high, while that of sample D is low due to
mirror polishing. Therefore, it was confirmed that damage by machining causes the residual stress.
Bending strength of sample D is higher while its compressive residual stress near the surface is low. It
can be considered that the singular stress field caused by the crack chapter has a greater effect on bending
strength than the enhanced strength caused by the compressive residual stress.
The residual stress of samples A, B, and C are shown in Fig 13 as a log-log plot. Residual stress
converges to some extent as the distance from the surface becomes greater. It was confirmed from
samples A, B, and C that surface machining damage has an effect on the wafer bending strength because
the bending strength reduces as the residual stress increases.
Focusing on the residual stress near the surface of samples A, B, and C described in Fig 13, it can be
assumed that transitions or cracks exist due to machining damage and that crystal structure is deteriorated
because the residual stress has a large variation. The maximum residual stress had a tendency to saturate
near 500 MPa for all samples, clarifying the existence of a residual stress threshold. This phenomenon is
probably due to the nonlinearity of crystal silicon such as the occurrence of transitions caused by shear
Wafer surface
Crack
Crack
d=0.01
Wafer surface
Thickness
direction
Crack
d=0.01
Daisuke Echizenya et al. / Procedia Engineering 10 (2011) 1440–1445 1445
stresses or the occurrence of cracks caused by the destruction of crystals. These transitions or cracks are
supposed to work as potential cracks [6] but work as strength parameters instead.
-100
0
100
200
300
400
500
00.02 0.0 4 0.06 0.08
0.1
1
10
100
1000
0.001 0.01 0.1
Fig. 12. Results of Raman spectroscopic (Linear scale) Fig. 13. Results of Raman spectroscopic (Log scale)
4. Conclusions
Investigations on surface damage and strength measurement of photovoltaic polycrystalline silicon
wafers were conducted and the following results were obtained:
(1) According to the bending strength test and the fracture toughness test, surface damage due to
machining was found to affect bending strength. The bending strength of the samples whose machining
damage was modified by mirror polishing was about three times higher than that of samples in the initial
condition. Also it was found that the crystal grain boundary and deficiency of the crystal structure of
polycrystalline silicon does not have an effect on the strength of the wafer at least under 300MPa.
(2) Surface observation by TEM clarified numerous cracks on wafer surface. It also showed that cracks
on a small strength sample have a strength that is apparently larger than that of a large strength sample,
which indicates correlation with bending strength.
(3) From surface observation by Raman spectroscopic it was confirmed that damage by machining
caused the residual stress.
The results clarified the existence of numerous cracks on the wafer surface that are assumed to be
generated during the slicing process. Thus, it was confirmed that wafer strength depends on the level of
machining damage in the slicing process. We can predict higher reliability for PV modules as a result of
modifying the slicing conditions so as to reduce the mechanical surface damage on wafers and thereby
increase the wafer strength.
References
[1] Photovoltaic Power Generation Market; Present Trend of Enterprises, Journal of Architectural Products, 2004, 283,p.21-24.
[2] S. Takyu, et al., Novel Wafer Dicing and Chip Thinning Technologies Realizing High Chip Strength, Electronic
Components and Technology Conference, 2006; p1623-1627.
[3] K. Hayashi, et al., Fracture toughness of single crystal silicon, Journal of the Society of Materials Science, Japan, vol. 40,
No.451,1991, p. 39-44
[4] The Society of Materials Science, Japan, STRESS INTENSITY FACTORS HONDBOOK Volume 2, 1987,p. 698-699
[5] E.Anastassakis, Effect of Static Uniaxial Stress on the Raman Spectrum of Silicon, Solid State Com., vol.8, 1970, p.133-138
[6] K. Tanaka, Y. Nakai, Fatigue growth threshold of small cracks, Int. Journ. of Fractureˈvol.17, No.5, 1981,p.519-533 .
Ccmpressive residual stress(Mpa)
Ccmpressive residual stress(Mpa)
Normalized distance from surface
sample A(113MPa)
sample B(122MPa)
sample C(89.2MPa)
Normalized distance from surface
sample A(Bare)
sample D(Polished)