High Precision in Raman Frequency Achieved Using Real-Time
Calibration with a Neon Emission Line: Application to ThreeDimensional Stress Mapping Observations
SHOKO ODAKE, SATOSHI FUKURA, and HIROYUKI KAGI*
Geochemical Laboratory, Graduate School of Science, The University of Tokyo, 7-3-1 Hongo, Tokyo 113-0033, Japan
A three-dimensional (3D) Raman mapping system with a real-time
calibration function was developed for detecting stress distributions in
solid materials from subtle frequency shifts in Raman spectra. An atomic
emission line of neon at 918.3 cm1 when excited at 514.5 nm was used as
a wavenumber standard. An emission spectrum of neon and a Raman
spectrum from a sample were introduced into a single polychromator
using a bifurcated optical fiber. These two spectra were recorded
simultaneously on a charge-coupled device (CCD) detector using
double-track mode. Energy deviation induced by the fluctuation of
laboratory temperature, etc., was removed effectively using the neon
emission line. High stability during long measurements was achieved. By
applying curve fitting, positions of the Raman line were determined with
precision of about 0.05 cm1. The present system was applied to
measurements of residual pressure around mineral inclusions in a natural
diamond: 3D stress mapping was achieved.
Index Headings: Raman spectroscopy; Three-dimensional mapping; 3D
mapping; Two-dimensional mapping; 2D mapping; High precision.
INTRODUCTION
Raman spectroscopy measures molecular and lattice vibrations by detecting the frequency of inelastically scattered light.
It is sensitive to the detection of internal stress in a material. In
particular, micro-Raman spectroscopy has a high spatial
resolution on the order of 1 lm; furthermore, the precise
determination of frequency of Raman band enables nondestructive mapping observations of stress in solid-state materials. Local mechanical stress in semiconductor materials has
been evaluated using micro-Raman spectroscopy. Uniaxial
stress in the [100] direction of silicon causes a peak shift of 1
cm1 for the first-order Raman peak at 520 cm1.1 Using a very
sensitive and stable instrument, it is possible to detect a stress
of 25 MPa and a strain of 104 from the Raman peak shift of
0.05 cm1. Nevertheless, it is extremely difficult to achieve
such an ideal measurement, as described in this paper.
Micro-Raman spectroscopy has also been applied to
mineralogical samples originating from deep in the Earth in
order to measure the distribution of residual stress around an
inclusion in a host mineral and decipher the original depth at
which the inclusion was trapped in the host mineral. Diamond
is a good sample that can retain a measurable stress distribution
around the trapped inclusions. Diamond is chemically and
physically stable; it can encapsulate mantle-originated minerals
during transport to the surface of the Earth. The residual
pressure retained in inclusions might be the same as the
confining pressure acting upon the host diamond when the
inclusions and the host diamond were formed at some initial
temperature and pressure. Later, the diamonds rose to the
Received 4 April 2008; accepted 28 July 2008.
* Author to whom correspondence should be sent. E-mail: kagi@eqchem.
s.u-tokyo.ac.jp.
1084
Volume 62, Number 10, 2008
Earth’s surface. The pressure inside the inclusion is expected to
differ from that in the host diamond because compressibilities
and thermal expansivities of the inclusions and diamond are
different. For this reason, residual stresses on the order of 1
GPa around various types of inclusions in natural diamonds
have been reported.2–6 Recently, two-dimensional (2D) Raman
mapping of diamond samples has disclosed the stress
distribution around mineral inclusions by mapping the peak
shift of Raman spectra of diamond.7
High resolution and precision are necessary for determining
peak positions of Raman spectra to measure the residual stress
in materials from Raman spectra. In general, the spectral
resolution of Raman spectra is constrained by the groove
number of a grating, the length of a spectrometer, the pixel size
of a charge-coupled device (CCD) detector, etc. On the other
hand, the reproducibility of the frequency of a Raman band is
considerably biased by the laboratory environment. Fukura et
al.8 reported that the Raman shift drifts in synchronization with
a deviation of laboratory temperature and that the drifts are not
negligible for studying subtle frequency shifts. According to
their study, a periodical temperature change in a laboratory
with an amplitude of 0.8 8C caused oscillation of the Raman
frequency of 0.1 cm1. This oscillation is considerably large
when observing the stress distribution from a subtle frequency
shift of Raman spectra. This deviation in Raman shift
corresponds to 35 MPa for diamond and 32 MPa for
olivine.9,10 The fluctuation of laboratory temperature can
impart considerable effects on the evaluation of stress
distribution obtained from Raman spectra.
In general, residual stress in a solid is three-dimensionally
heterogeneous because of the anisotropic elastic properties of
crystals. Consequently, three-dimensional (3D) Raman mapping is necessary for detailed and precise estimation of stress
distributions. Raman imaging techniques are separable into
three groups: point-by-point illumination, line illumination,
and plane illumination.11 Point-by-point illumination, which
obtains mapping data through sequential translation of the
sample stage, provides the best spatial resolution, but much
time is necessary to scan a sample. The other two methods can
save considerable amounts of time used for mapping
measurements. Nevertheless, we adopt the point-by-point
illumination for achieving the best spatial resolution. A 3D
Raman measurement using point-by-point illumination requires
a long time (e.g., tens of hours or more depending on the
sample size and increment of measurements); the peak
oscillation induced by changes in room temperature, as
described above, can substantially influence the measurement.
Additional stability and precision in acquiring Raman spectra
are necessary for 3D Raman mapping measurements. The
present study is intended to develop a new system that can
remove spectral fluctuation attributable to the laboratory
0003-7028/08/6210-1084$2.00/0
Ó 2008 Society for Applied Spectroscopy
APPLIED SPECTROSCOPY
FIG. 1. Schematic diagram of the experimental setup for sequential
wavenumber calibration. See text for details.
environment and reveal the 3D stress distribution using-micro
Raman spectroscopy.
EXPERIMENTAL PROCEDURES
Real-time calibration was applied to a spectrometer to reduce
the deviation of Raman frequency induced by the laboratory
environment and achieve stability in determining Raman
frequency. Neon emission spectra were chosen as an absolute
wavelength standard during measurements because wavelengths of these emission lines are practically independent of
the laboratory environment. A Raman spectrum from a sample
and an emission spectrum from a neon lamp were collected
simultaneously using a CCD detector, and real-time calibration
was performed during each exposure of a spectrum to cancel
out the frequency deviation induced by changes in the
laboratory environment.
Micro-Raman Measurement System. The micro-Raman
system comprises an optical microscope (BX60; Olympus
Corp.), a 30 cm single polychromator (250is; Chromex), an Ar þ
laser (5500A; Ion Laser Technology, Inc.), and a silicon-based
charge-coupled device (CCD) camera capable of capturing
images of 1024 3 128 pixels (DU-401-BR-DD; Andor
Technology). The excitation laser beam was focused on a
spheroidal spot of approximately 2 3 2 3 10 lm in volume using
a 503 objective lens (NA ¼ 0.80; Olympus Optical Co. Ltd.).
The slit width of the spectrometer was fixed at 130 lm.
Scattered light was dispersed using a grating with 1800 grooves
per millimeter. The spectral resolution was approximately 1.5
cm1 per pixel. Samples were excited using 514.5 nm emission
of an Ar þ ion laser. The laser power was about 5 mW at the
sample surface, which is sufficiently below the threshold for the
heat-induced spectral change. The Rayleigh line was removed
using a holographic high-pass filter (HIPF-514.5-1.0; Kaiser
Optical Systems, Inc.). The CCD camera was cooled electronically to 70 8C using a Peltier device for thermal noise
reduction.
Before measurements, the Raman shift was initially
calibrated using five Raman bands of naphthalene (C10H8) at
513.8, 763.8, 1021.6, 1382.2, and 1464.5 cm1. The peak
position of the Raman band of the diamond sample was
determined by curve fitting using a Gaussian function. The
average full width at half-maximum (FWHM) value of the
fitted Gaussian peak was 2.9 pixels. Although the spectral
resolution of the spectrometer is 1.5 cm1 per pixel, the center
FIG. 2. Emission spectrum of a neon lamp. Three measurable emission lines
were observed at 676.0, 712.2, and 918.3 cm1 in Raman shift.
of the fitted Gaussian curve can be determined with much
higher precision, typically 0.05 cm1, unless Raman frequency
is perturbed by the laboratory temperature.8,12,13
Neon Emission Spectra for the Absolute Wavelength
Standard. Neon emission spectra as the wavelength calibrator
and Raman scattering signal were introduced simultaneously
into the spectrometer using a bifurcated optical fiber made
entirely of silica. Emission from a compact neon lamp was
introduced into one branch of the bifurcated optical fiber. The
Raman signal from a sample was introduced into another
branch of the bifurcated optical fiber. The other end of the fiber
was connected to a polychromator; signals from the neon lamp
and Raman scattering were directed simultaneously into the
polychromator and imaged as two independent rows of bands
on a CCD detector. These two rows of spectra were read out
independently and do not interfere with each other (see Fig. 1).
Three-Dimensional Measurement System. A computercontrolled, motorized XY sample stage (FC-101G; Sigma Koki
Co. Ltd.) was installed on the optical microscope; the stage can
be manipulated by 0.1 lm steps with accuracy. Additionally,
the stage height (Z-axis) was controlled using an autofocus
controller (AF/r; Flovel Co. Ltd.). The minimum spatial
resolution of the Z-stage is 0.05 lm. Focusing on the sample
surface was adjusted automatically by maximizing the
brilliance of the reflected laser spot on the sample surface.
The following describes a cycle of the data acquisition process
for 3D Raman mapping measurements in the present study.
First, the initial point of measurement was located on the focal
line of the laser beam by translating the XY stage. Second, the
focus was adjusted on the sample surface using the auto focus
function. After obtaining a Raman spectrum on the surface, the
stage was moved up for a given height step and Raman spectra
of the deep part of the sample were measured.
RESULTS AND DISCUSSION
A representative emission spectrum of the neon lamp is
shown in Fig. 2; three emission lines were observed in the
APPLIED SPECTROSCOPY
1085
FIG. 3. Time changes in the peak positions of neon spectra. The upper line
shows the calibrated spectra.
studied energy region. The emission line at 540.06 nm (918.3
cm1 in Raman shift with excitation by the 514.5 nm line of the
Arþ laser) was chosen as the wavelength standard because this
peak is sufficiently intense for precise determination of the
wavelength and the position of this line is close to the diamond
Raman peak. Fukura et al. (2006) showed that all Raman peaks
in the region from 400 cm1 to 1500 cm1 deviated
synchronously in their direction and amplitude. Thus, the
deviation in the Raman bands can be compensated using a
single neon line at 918.3 cm1 in the Raman shift. Before
applying real-time calibration to measurements of Raman
spectra, the availability of the real-time calibration system in
the present instrument was evaluated. Another neon lamp was
set on the sample stage and the emission was introduced into
the spectrometer in addition to the neon lamp for the standard
using the bifurcated optical fiber. Consequently, two neon
emission spectra were recorded on the CCD detector
simultaneously.
Figures 3a and 3b show the time change of the observed
wavelength of a neon emission line at 540.06 nm for the two
obtained spectra. Both profiles exhibit considerable fluctuation
of the frequency by as much as 0.37 cm1 in Raman shift,
resulting from change in the laboratory temperature. The
corresponding oscillation of laboratory temperature was 1.0 8C.
Applying the neon emission spectra as the standard, the time
change in the neon line was calibrated (see Fig. 3c). The
calibration process eliminated the fluctuation. The final
fluctuation was reduced to 0.05 cm1 in two sigma for 4 h.
Results confirmed that the use of the neon emission line as the
wavelength standard reduced the fluctuation of the Raman shift
efficiently. This calibration potentially cancels the unstable
behavior in the initial measurement.
Two-Dimensional Mapping of a Diamond Containing
Inclusions and Application of Real-Time Calibration. Realtime calibration was applied to a 2D Raman mapping
measurement. The diamond specimen, containing olivine and
chromite as inclusion minerals, was from the Internationalnaya
Mine in Russia. The (110) surface of the diamond was
polished. Using a 350 objective lens and confocal arrangement, Raman spectra from the 40 lm subsurface of the sample
were collected in the region of 200 lm 3 200 lm using a step
size of 4 lm. Raman spectra were obtained with an exposure
time of 1 s for each measured point; the total time for acquiring
the 2D map presented in Fig. 4 was approximately 35 h. Figure
4a shows a 2D frequency map of the F2g Raman band of
diamond of the sample after the curve-fitting procedure without
real-time calibration. The relative shifts of the Raman spectra
of diamond from an unstressed region were countered with
gray-scale gradation. The brighter regions around the inclusions correspond to positive shifts of Raman spectra of
diamond and the stressed region. In Fig. 4a, streak noise
resulting from room-temperature fluctuation was observed.
Because the sample stage was scanned in a horizontal
direction, the streaks induced by the oscillation of the
laboratory temperature were observed in the horizontal
direction. The streak is comparable to the stress distribution.
It is necessary to remove this noise by application of real-time
calibration. Figure 4b shows the Raman map after application
of the real-time calibration using neon emission spectra. The
streak was removed by application of the calibration using the
neon standard line. Figure 4b shows clearly that the diamond
around the inclusions was more highly stressed than other
regions.
Application to Three-Dimensional Mapping Measurements. The real-time calibration system was applied to 3D
FIG. 4. A 2D Raman map of the sample, (a) without real-time calibration; a streak attributable to temporal change in room temperature was observed; and (b) with
real-time calibration; the streak is absent.
1086
Volume 62, Number 10, 2008
around the olivine and chromite inclusions were respectively
estimated as 0.69 GPa and 0.75 GPa. Because of the
anisotropic elastic properties of minerals, the stress field can
be highly anisotropic: 3D stress mapping is necessary to
support detailed discussion. The obtained Raman maps in
layers can be reconstructed into tomographic 3D stress
distribution. The determination of the stress values around
two different mineral species can be used to estimate the
pressure and temperature at which these inclusions were
trapped in the host diamond. X-ray tomography determining
the shape of inclusions will complement the 3D Raman
mapping measurement for precise discussion of the anisotropic
stress field in diamond around the inclusions. It is noteworthy
that the present calibration system can facilitate the detection of
subtle frequency shifts in Raman spectra for various purposes.
CONCLUSION
A micro-Raman spectrometer system was developed with a
real-time calibration function using neon emission spectra. The
developed system removed the deviation in Raman frequency
derived from the temporal change in laboratory temperature.
By applying curve fitting to the diamond Raman line, the
precision in determining Raman frequency was reduced to less
than 0.05 cm1. High stability and precision established by the
real-time calibration enabled 3D Raman mapping, which
necessitates a long time for measurements. For a natural
diamond sample containing two mineral inclusions, 3D Raman
mapping was achieved. The stress distribution was observed
clearly by applying real-time calibration.
ACKNOWLEDGMENTS
FIG. 5. A 3D Raman map of the sample. Raman maps of diamond stacking
from the sample surface to the interior.
mapping observation on the same diamond sample containing a
couple of mineral inclusions using the same measurement
system. Figure 5 shows a Raman map of the sample stacking
from the surface to the interior. The sample was scanned three
dimensionally, as described in the Experimental Procedure
section. The scanned region was 400 3 400 3 60 lm. The
scanning steps were 4 lm in the horizontal directions and 10
lm in the perpendicular direction. A Raman spectrum was
acquired at each sampling point with exposure time of 1 s. In
all, 71 407 pairs of Raman spectra and neon emission spectra
were obtained during the approximately 35 h spent for
measurement.
Figure 5 shows the obtained Raman frequency maps for 10
lm depth steps. The layered Raman maps clearly show the 3D
stress distribution around the inclusions. With increasing
pressure, the frequency of the lattice vibration of the diamond
increases with a gradient of 2.2 cm1/GPa along the ,111.
direction.14 Figure 5 shows that the maximum stress values
We are grateful to Dr. Dimitry Zedgenizov for diamond samples and
scientific discussion. This study was financially supported by a Grant-in-aid for
Exploratory Research (18654098), a Grant-in-aid for Creative Scientific
Research (19GS0205) from the Japan Society for the Promotion of Science,
and the Global COE Program for Chemistry Innovation.
1. De Wolf, Semicond. Sci. Technol. 11, 139 (1996).
2. N. V. Sobolev, B. A. Fursenko, S. V. Goryainov, J. F. Shu, R. J. Hemley,
H. K. Mao, and F. R. Boyd, Proc. Natl. Acad. Sci. U.S.A. 97, 11875
(2000).
3. M. Schrauder and O. Navon, Nature (London) 365, 42 (1993).
4. E. S. Izraeli, J. W. Harris, and O. Navon, Earth Planet. Sci. Lett. 173, 351
(1999).
5. H. Kagi, R. Lu, J. R. Hemley, and H. Mao, Mineral. Mag. 64, 1089 (2000).
6. D. A. Zedgenizov, H. Kagi, V. S. Shatsky, and N. V. Sobolev, Mineral.
Mag. 68, 61 (2004).
7. L. Nasdala, J. W. Harris, and J. Glinnemann, Am. Mineral. 90, 745 (2005).
8. S. Fukura, T. Mizukami, S. Odake, and H. Kagi, Appl. Spectrosc. 60, 946
(2006).
9. H. Boppart, J. van Straaten, and I. F. Silvera, Phys. Rev., B. 32, 1423
(1985).
10. S. Y. Wang, S. K. Sharma, and T. F. Cooney, Am. Mineral. 78, 469
(1993).
11. S. Nakashima, J. Phys.: Condens. Matter 16, S25 (2004).
12. Y. Kawakami, J. Yamamoto, and H. Kagi, Appl. Spectrosc. 57, 1333
(2003).
13. J. Yamamoto, H. Kagi, Y. Kawakami, N. Hirano, and M. Nakamura, Earth
Planet. Sci. Lett. 253, 369 (2007).
14. M. H. Grimsditch, E. Anastassakis, and M. Cardona, Phys. Rev., B. 19,
901 (1978).
APPLIED SPECTROSCOPY
1087