Full text of DOI:10.1557/JMR.2000.0209

Micro-Raman spectral analysis of the subsurface damage
layer in machined silicon wafers
Long-Qing Chen, Xin Zhang, and Tong-Yi Zhang
Department of Mechanical Engineering, Hong Kong University of Science and Technology,
Clear Water Bay, Kowloon, Hong Kong
H.Y. Lin and Sanboh Lee
Department of Materials Science, National Tsing Hua University,
Hsinchu, Taiwan, Republic of China
(Received 25 October 1999; accepted 25 April 2000)
In the present work we studied the depth of damage layer in machined silicon wafers
that was incorporated with chemical etching using micro-Raman spectroscopy.
Subsurface damage causes changes in the shape and intensity for the shoulder
(450–570 cm−1) of the most intense band (519 cm−1) and the second band (300 cm−1)
regions of the Raman spectrum. Etching reduces the thickness of the damage layer
and, hence, the intensities at the shoulder and the second band. The intensities at the
shoulder and the second band become stable when the damage layer is completely
etched out. The shoulder consists of two Gaussian profiles: the major and the minor.
The band for the major profile is independent of etching depth, but the band for the
minor profile shifts toward the longer wave numbers with increasing etching period
until the damage layer is completely etched out. The depth of the damage layer is
determined by the profiles of the shoulder and the second band and confirmed by the
band shift of the minor profile. Transmission electron microscopy (TEM) further
verified the results with respect to the depth of the damage layer. TEM observation
showed that dislocations and stacking faults are responsible for the subsurface damage.
Silicon is the basic material widely used in the semi-
conductor industry. More than 90% of semiconductor
devices are made of silicon. Because of the trend toward
using larger and larger silicon wafers in fabrication of
microelectronic devices, new requirements in wafer ma-
chining and quality control have become increasingly
important.1,2 Silicon wafers are manufactured from sili-
con ingots through the machining process including inner
diameter cut-off grinding, lapping, normal grinding,
chemical etching, and polishing. The machining proc-
essing may induce surface and subsurface damage in the
wafers.3,4 The depth of damage layer is defined as the
distance from the specimen surface to the layer below
which the lattice remains perfect.5 The damage layer
consists of various defects such as dislocations, voids,
precipitates, or/and microcracks, etc.6 These defects de-
teriorate the physical and chemical properties of silicon
wafers and may locally break the lattice symmetry.7–10
Various techniques are employed to characterize the
surface and subsurface damage.10–13 Raman spectros-
copy (or inelastic scattering) is an important tool for the
analysis of surface layer of silicon wafers. For example,
Raman spectroscopy was used to identify the phase and
composition and to determine the short-range and long-
range order and impurity configurations.14–16 The effect
of crystal size leads to shift the excitation maximum to
lower energies and to additional band broadening.16 Re-
sidual stresses were also studied on the basis of the band
shift of Raman spectra.5,7,17,18 Subsurface damage may
induce residual stresses and, thus, can be assessed by
Raman spectroscopy on the basis of the band shift.5,9,12
FIG. 1. Etching depth of silicon as a function of etching time at 60 °C.
J. Mater. Res., Vol. 15, No. 7, Jul 2000
© 2000 Materials Research Society
1441
http://journals.cambridge.org
Downloaded: 12 Apr 2015
IP address: 129.93.16.3

Rapid Communications
In this paper, we use micro-Raman spectroscopy to esti-
mate the depth of the damage layer in the machined
silicon wafers.
after etching was measured, and their difference yields
the etching depth. The Raman spectroscopy was con-
ducted with a Renishaw model 3000 Micro-Raman/
Photoluminescence System made by Renishaw,
Gloucestershire, England, at room temperature. The ar-
gon ion laser beam of 514.5-nm wavelength with a power
of 0.5 mW was focused onto a 2-␮m spot on the speci-
men using an optical microscopy with a 50× objective
lens. The wavelength was scanned in the range from 200
to 800 cm−1.
Transmission electron microscopy (TEM) was con-
ducted on a model JEM-2010 TEM system made by
JEOL, Tokyo, Japan, with a working voltage 200 kV.
Specimens were prepared by mechanical polishing to
about 30 ␮m and then thinned by ion milling.
Two kinds of polished silicon wafers (with 20- and
10-cm diameters) were provided by Mitsubishi Com-
pany, Tokyo, Japan, and Sino-American Silicon Products
Inc., Hsinchu, Taiwan, respectively. All wafers were first
annealed at 900 °C for 2 h and then machined with dif-
ferent conditions. Let S1, S2, S3, S4, and S5 denote
wafers with the five different final processes of grinding
(1200 mesh, 4000 mesh, 8000 mesh, grinding and pol-
ishing, and lapping, respectively). The wafer diameter
is 20 cm for S2 and S3 and 10 cm for S1, S4, and S5. All
specimens cut from the machined wafers have a size of
1 × 1 cm.
The chemical etching was carried out in a solution
containing 40 wt% potassium hydroxide (KOH) inside a
sealed container at 60 °C. A specimen was etched in
KOH solution for 1, 3, 5, 10, 30, 60, 180, or 300 min. A
shelter spot of low-temperature oxide (LTO) was de-
signed and fabricated as a reference for measurement of
the etching depth of the specimen. The step height be-
tween the LTO spot and the wafer surface before and
Figure 1 shows the etching depth as a function of etch-
ing time. The experimental data are well fitted with a
straight line, and its slope yields the etching rate of
0.36 ␮m/min. Figures 2(a)–2(e) illustrate the Raman
spectra for specimens S1, S2, S3, S4, and S5, respec-
tively, with different etching times. It can be seen from
Fig. 2 that intensities at the shoulder (450–570 cm−1) for
the most intense band (519 cm−1) and the second band
FIG. 2. Raman spectra for different specimens (a) S1, (b) S2, (c) S3, (d) S4, and (e) S5.
1442
J. Mater. Res., Vol. 15, No. 7, Jul 2000
Downloaded: 12 Apr 2015
http://journals.cambridge.org
IP address: 129.93.16.3

Rapid Communications
2
2
regions (450–570 cm−1) decrease with increasing etching
I ס A1e−[(k−k )/␣ ] + A2e−[(k−k )/␣ ] + B
,
(1)
1
1
2
2
period until the profile becomes stable. The spectra be-
come stable after 10, 3, 5, 1, and 10 min of etching for
S1, S2, S3, S4, and S5, respectively. Thus, the depth of
subsurface damage for S1, S2, S3, S4, and S5 are esti-
mated to be about 3.6, 1.08, 1.80, 0.36, and 3.60 ␮m. The
Raman spectrum at the shoulder for the most intense
band consists of two overlapped features between 450
and 570 cm−1. The Raman spectra in this range can be
fitted by the sum of two Gaussian profiles as
where A, ␣, and k are the amplitude, Gaussian radius, and
band position, respectively. B is a constant. Subscripts 1
and 2 denote respectively the major profile and minor
profile. Figure 3 shows an example of shoulder consist-
ing of two Gaussian profiles for specimen S1 before
etching. The values of A1, A2, ␣1, ␣2, k1, k2, and B for the
five specimens with different etching times are calcu-
lated and listed in Table I. It is found that the band for the
most intense band is independent of etching time and
located at 517.1, 523.1, 520.6, 519.4, and 521.6 cm−1,
respectively. The band for the minor profile shifts toward
the long wave number when the etching time is increased
for all machining processes. The band position of minor
profile becomes stable at 514.9, 519.5, 520.3, 518.7, and
520.9 cm−1 after the etching time is greater than 10, 3, 5,
1, and 10 min, for specimens S1, S2, S3, S4, and S5,
respectively. Thus, the depth of subsurface damage for
S1, S2, S3, S4, and S5 samples are calculated to be 3.6,
1.08, 1.8, 0.36, and 3.6 ␮m. This means that the grinding
and polishing is the least effective damage process in
this study.
According to the experimental results of Raman spec-
tra, the damage induced by lapping is comparable to the
damage ground by 1200 mesh. The microcracks were
examined using an optical microscope. There are many
microcracks in the subsurface layer of the lapped speci-
men. In the ground specimen, however, no microcracks
were found. Raman spectroscopy is very sensitive to re-
FIG. 3. Solid line indicated by the shoulder for the most intense band
for specimen S1 before etching decomposed into two dashed lines of
Gaussian profiles. The parameters of Gaussian profiles are listed in
Table I.
TABLE I. The values of A1, A2, ␣1, ␣2, k1, k2, and B for specimens S1, S2, S3, S4, and S5 with different etching times, t.
Specimen
t (min)
A1
A2
␣
1
(cm−1
)
␣
(cm−1
2
)
k1 (cm−1
)
k2 (cm−1
)
B
0
1
3
5
10
4108
3368
3376
4193
4159
2072
445
410
447
440
11.46
7.95
8.39
6.33
6.38
55.90
54.23
27.02
22.30
22.47
515.7
517.2
516.9
516.4
517.1
473.8
489.5
514.9
514.5
514.9
537.1
107.8
105.2
69.0
S1
74.0
0
1
3
5
2079
2294
2605
2599
531
406
483
488
12.14
12.70
11.64
11.56
51.99
51.99
29.36
29.11
522.8
521.0
523.3
523.1
495.0
495.6
519.5
519.5
220.4
227.5
175.3
172.6
S2
0
1
3
5
2663
1490
1109
1636
566
409
427
823
10.90
13.93
11.49
11.06
57.73
55.05
25.48
25.73
520.8
521.0
521.0
520.6
491.1
491.7
519.4
520.3
209.5
191.7
192.1
168.4
S3
S4
S5
0
1
3
4937
1811
2483
649
327
231
8.93
9.62
9.17
29.11
28.63
30.29
519.7
519.1
519.4
518.5
518.7
518.7
89.6
324.4
80.9
0
5
10
30
2063
2058
1862
1989
303
543
208
195
13.04
10.23
13.36
12.91
33.33
28.63
33.52
39.53
522.0
521.7
522.0
521.6
523.1
519.8
518.8
520.9
192.1
101.7
98.3
91.8
J. Mater. Res., Vol. 15, No. 7, Jul 2000
1443
http://journals.cambridge.org
Downloaded: 12 Apr 2015
IP address: 129.93.16.3

Rapid Communications
sidual stresses rather than to microcracks. The damage in
the lapped specimen is more severe than that in the
ground specimen.
520.7 cm−1 for specimens S1, S2, S3, S4, and S5, respec-
tively. The band position for the minor profile shifts to-
ward the long wave number as the etching time is
increased until it reaches a stable value. The above re-
sults show that the depth of damage layer is 3.6, 1.08,
1.8, 0.36, and 3.6 ␮m for specimens S1, S2, S3, S4, and
S5, respectively.
The micro-Raman spectrum can be applied to measure
the depth of damage layer. Direct evidence of damage
layer is provided by the microstructure. According to
TEM results, dislocations and stacking faults are re-
sponsible to this damage layer. However, many micro-
cracks appear in lapped specimens and not in the ground
specimens.
TEM examination was conducted to find defects in the
damaged layers. Figure 4 shows the cross-sectional TEM
micrograph for specimen S1, indicating dislocations and
stacking faults. However, dislocations were not found in
the cross-sectional TEM micrograph. The quantitative
relation between defects and Raman spectroscopy will
be investigated in the future. The distance from defect
to surface is estimated to be about 3.5 ␮m in Fig. 3
and 0.30 ␮m in Fig. 4. The result is consistent with the
micro-Raman analysis.
This work studies subsurface damage in machined sili-
con wafers using micro-Raman spectroscopy that was
incorporated with wetting etching. The silicon wafers
under different machining processes of grinding (1200,
4000, and 8000 mesh, grinding and polishing, and lap-
ping) are assigned as S1, S2, S3, S4, and S5, respectively.
The machined specimens were etched in 40% KOH so-
lution for different periods. Both the shape and intensity
for the shoulder (450–570 cm−1) of the most intense band
and the second band (300 cm−1) regions are changed for
the machined wafers. Both shoulder and second band
intensities decrease with increasing etching depth until
the damage layer was etched out. The shoulder for the
most intense band can be fitted using the sum of two
Gaussian profiles. The band for the major profile is al-
ways located at 514.7, 519.5, 519.9, 519.2, and
ACKNOWLEDGMENTS
The work of L-Q.C., X.Z., and T-Y.Z. was supported
by the Hong Kong Research Grants Council, and the
work of S.L. was supported by the National Science
Council of Taiwan.
REFERENCES
1. J.F. Scott and B. Pouligny, J. Appl. Phys. 64, 1547 (1988).
2. C.A.P. Dearaujo, L.D. McMillan, B.M. Melnik, J.D. Cuchiaro,
and J.F. Scott, Ferroelectrics 104, 241 (1990).
3. P.A. McKeown, Ann. CIRP 36, 495 (1987).
4. N. Taniguchi, Ann. CIRP 32, 573 (1983).
5. U. Bismayer, E. Brinksmerier, B. Gu¨ttler, H. Seibt, and C. Menz,
Precis. Eng. 16, 139 (1994).
6. H. Hashimoto, H. Itoh, H. Eda, and K. Kishi, Proc. Int. Conf.
Prod. Eng. 5, 603 (1984).
7. M. Cardona, Proc. SPIE 822, 2 (1987).
8. R.G. Sparks and M.A. Paesler, J. Appl. Phys. 71, 891 (1992).
9. J. Verhey, U. Bismayer, B. Gu¨ttler, and H. Lundt, Semicond. Sci.
Technol. 9, 404 (1994).
10. E. Brinksmeier, Precis. Eng. 11, 211 (1989).
11. S. Johansson and J.A. Schweitz, J. Am. Ceram. Soc. 71, 617
(1988).
12. R.G. Sparks and M.A. Paesler, Precis. Eng. 10, 191 (1988).
13. H.K. Tonshoff, W.V. Schmieden, I. Inasaki, W. Konig, and
G. Spur, Ann. CIRP 39, 621 (1990).
14. R.J. Nemanich, in Materials Characterization, edited by N.W.
Cheung and M-A. Nicolet (Mater. Res. Soc. Symp. Proc. 69,
Pittsburgh, PA, 1986), p. 23.
15. G. Contreras, A.K. Sood, M. Cardona, and A. Campaan, Solid
State Commun. 49, 303 (1984).
16. P.M. Fauchet and I.H. Campbell, CRC Crit. Rev. Solid State
Mater. Sci. 14, S79 (1988).
17. E. Anastrassakis, A. Pinezuk, E. Burstein, F.H. Pollak, and
M. Cardona, Solid State Commun. 8, 133 (1970).
18. F. Cerdeira, C. Buchenauer, F. Pollak, and M. Cardona, Phys.
Rev. B 5, 580 (1972).
FIG. 4. Cross-sectional TEM micrograph for specimen S1 (13,600×).
There are many dislocations and (111) stacking faults at the subsurface
damage layer.
1444
J. Mater. Res., Vol. 15, No. 7, Jul 2000
Downloaded: 12 Apr 2015
http://journals.cambridge.org
IP address: 129.93.16.3