Determination of the Resolution of a Multichannel Raman

Spectrometer Using Fourier Transform Raman Spectra

BRYAN T. BOW IE* and PETER R. GRIFFITHS†

Department of Chemistry, University of Idaho, Moscow, Idaho 83844-2343

The resolution of a grating polychromator for Raman spectro scopy beam passing through the interferometer is sufŽ ciently
has been simulated by measuring spectra on a Fourier transform collimated. Similarly, under most circumstances the full-
(FT) Raman spectrometer and selecting the FT of the apodization width at half-height (FW HH) of the spectral slit function
function so that the instrument line shape function mimics the tri-
of a CCD-Raman spectrometer is the product of the re-
angular spectral slit function of the polychromator. To this end, FT-
Raman spectra measured with a nominal resolution of 0.5 cm 2 1
ciprocal linear dispersion of the polychromator, R d, the
were modiŽ ed through the application of sinc 2 apodization functions width of the entrance slit, and the m agniŽ cation of the
of various widths to simulate spectra m easured on a polychrom ator beam at the CCD relative to the entrance slit.1 For ex-
at lower resolution. The success of this approach was validated us- ample, for the instrum ent being used for this project (a
ing the 1085 cm 2 1 band of calcite. When the modiŽ ed FT-Raman Renishaw System 2000 with 782 nm excitation), it is
spectra were compared with spectra measured on a grating poly- known that R d 5 45 cm 2 1 /mm at 1000 cm 2 1 Raman shift
chrom ator equipped with slits of widths 100 and 150 mm, the res- and the magniŽ cation factor is 1.5. 2 Thus, for an entrance
olution of the polychromator was estimated to be 6.3 and 7.8 cm 2 1, slit width of 100 mm, the resolution should be equal to
respectively. This conclusion was veriŽ ed experimentally by mea-
45 3 100 3 10 2 3 3 1.5, or 6.75 cm 2 1. Similarly, for an
suring the separation of two bands in the Raman spectrum of
BaSO 4 at ;460 cm 2 1 .
entrance slit width of 150 mm, the resolution should be
equal to 10.1 cm 2 1 . According to the manufacturer, how-
Index Head ings: Raman spectroscopy; Charge-coupled device-Ra-
man; CCD-Raman; Fourier transform Raman; Resolution; Instru-
ever, if the slit width is smaller than 25 mm, the resolution
ment line shape function; Apodization. is limited by the number of lines on the grating, while if
the slit width is greater than 50 mm , the resolution is
determined by the width of the entrance slit. For slit
INT RODUCTIO N widths between 25 and 50 mm , the resolution cannot be
readily calculated from either the reciprocal linear dis-
As Raman spectrometers equipped with charge-cou- persion or the number of lines on the grating. Further-
pled device (CCD) array detectors are applied to an in- more, the reciprocal linear dispersion in cm 2 1 /mm is not
creasingly diverse number of applications, certain issues constant across the spectrum for several reasons, the most
regarding the instrument performance become of great important of which is that the resolution of grating mono-
importance. For example, the resolution of CCD-Raman chromators is approximately constant in wavelength rath-
spectrometers should be known if these instruments are er than wavenum ber. An R d of 45 cm 2 1 /m m at a Raman
to be used for the construction or application of spectral shift of 1000 cm 2 1 with 782 nm excitation should, there-
databases or when calibration data measured on one in- fore, be converted to R d 5 3.25 nm/mm if the resolution
strument are to be transferred to another. W hen the res- at any other wavelength is to be calculated. Even then,
olution is varied, both the peak heights and the widths of the angle of diffraction varies with wavelength and the
bands across the spectrum are altered. Such variations in illuminated area of the grating change across the spec-
spectra from instrument to instrument can degrade the trum when using a Czerny–Turner type m onochromator.
results of library searching, spectral subtraction, and Thus, converting the reciprocal linear dispersion from
quantitative analysis when the reference or calibration cm 2 1 /mm to the corresponding value in nm/m m still does
data are not measured under conditions identical to the not allow the precise resolution at any other wavenumber
sample spectra. Now that inexpensive CCD-Raman spec- in the Raman spectrum to be readily calculated. Although
trometers with very low resolution (Dñ . 20 cm 2 1 ) have the resolution in limited regions of the spectrum can be
been introduced commercially, it is even possible that one calculated in this way, R d in nm/mm still varies across
group may report that a certain compound gives rise to the spectrum when the same grating is used for collecting
a single band in a certain spectral region while another the Raman spectrum with different excitation wave-
group using a higher-resolution spectrometer observes a lengths, e.g., 532 and 782 nm.
multiplet.
Because the spectral slit function of a scanning grating
One approach is to calculate the resolution from known
monochromator operating far from diffraction-limited
instrum ent parameters. This is easily done for a Fourier
resolution is the result of con volvin g rectan gu larly
transform (FT) Raman spectrometer since resolution is
shaped entrance and exit slits of equal widths, the spectral
determined only by the maximum optical path difference
slit function is triangular, with a full-width at half-height
(OPD) and the apodization function, provided that the
(FW HH) of S g . The baseline resolution of spectra mea-
sured under such a condition has been deŽ ned as 2S g.1
Received 3 March 2000; accepted 23 September 2002.
* Present address: Naval Air Warfare Center, Code 4T4230D, China
The resolution of a spectrometer depends on the exact
Lake, CA 93555. deŽ nition used (baseline, Rayleigh, Sparrow, etc.), but
† Author to whom correspondence should be sent. two lines m easured with a triangular slit function will

0003-7028 / 03 / 5702-0190$2.00 / 0
190 Volume 57, Number 2, 2003 q 2003 Society for Applied Spectroscop y
APPLIED SPECTROSCOPY
 F IG . 1. Spectrum of a neon lamp taken off-axis (solid line) compared to the spectrum of the lam p when correctly aligned (broken line).

usually be visibly resolved, albeit not to baseline, if they mated by m easuring the relative height or width of one
are separated by a wavelength between S g and 2S g. or more bands in the spectrum.
The spectral slit function of a CCD-Raman spectrom- Because the height of bands can usually be measured
eter or the instrum ent line shape (ILS) function of an FT- more accurately than their width, it is possible in prin-
Raman spectrometer is often m easured experimentally by ciple to estimate the resolution from a spectrum contain-
observing the spectrum of one or m ore narrow atomic ing two neighboring bands, one of which is broad
emission lines from, for example, a neon lamp. To ac- (FW HH 5 gb, peak height 5 A b ) and the other narrow
curately simulate the measurem ent of a Raman spectrum, (FW HH 5 g n, peak height 5 A n ). In a classic paper dis-
however, the neon lamp must be located at precisely the cussing the effect of a triangular slit function on the peak
same position as the sample. Improperly positioning the absorbance of Lorentzian bands in infrared spectra, Ram-
neon lamp can skew the band and hence m ake the de- say 5 deŽ ned a resolution param eter, r, as:
termination of the wavenumber of the band center and its
FW HH inaccurate,3 as shown in Fig. 1. It has also been spectrometer resolution
r5 (1)
reported 4 that lines in the emission spectrum of a neon FW HH of the band
lamp can become distorted and their relative intensity
may change as the detector reaches saturation, as shown W hen the resolution is increased (i.e., Dñ is sufŽ ciently
for an extreme case in Fig. 2. small) so that r for the narrowest band becom es less than
In light of the above discussion, therefore, it is appar- ;0.2, any further increase in band height with decreasing
ent that an independent m easurement of resolution should r becomes negligibly small. At this point, the measured
be made using some type of standard sample, which we heights (A b 9 and A n 9) and widths (gb 9 and gn 9) of the
will call a ‘‘resolution standard’’. The ideal sample bands will be equal to their true values and the ratio of
should be a stable solid that can withstand high laser their heights will be approximately equal to A b /A n. As
power without photo-bleaching to ensure its usefulness Dñ is gradually increased, the m easured height of the nar-
over a long period of time. Ideally, a resolution standard row band, A n 9, starts to decrease before that of the broad
should also have no absorption bands in the region where band A b 9. Thus, over the range g n # Dñ # gb, the ratio
the Raman spectrum is measured to avoid the effects of A b 9/A n 9 is strongly dependent on Dñ . When Dñ ¾ gb, A b 9/
resonance enhancement or self-absorption. The FW HH of A n 9 is approximately equal to A b gb /A n gn and will stay con-
bands in a resolution standard should vary from very stant as Dñ is increased. Provided that gb ± g n , the ratio
wide (up to about 20 cm 2 1 ) to quite narrow (;1 or 2 of the m easured values of the band heights varies signif-
cm 2 1 ) since Raman spectrometers are comm only operated icantly with Dñ over at least the range gn , Dñ , gb .
from resolutions between 1 and 20 cm 2 1. In practice, Thus, in principle at least, the ratio of the m easured peak
these bands should be well separated since instrumental heights of the two bands, A b 9/A n 9 can be used to estimate
resolution is rarely estimated by detecting when two the resolution at which the measurement was made. This
neighboring bands are just resolved and is better esti- technique was used to determine the resolution of mid-

APPLIED SPECTROSCOPY 191

 F IG . 2. Spectra of neon lamp m easured with a low power setting (broken line) and when the detector was overloaded (solid line).

infrared (IR ) grating spectrometers for many years by to be the best approximation for the slit function for Ra-
measuring the heights of neighboring broad and narrow man spectra m easured at typical resolution settings.
bands in the spectrum of indene.6 Fourier Transform-Raman Spectrometers. The ILS
W hen we started working on this project, we were in- function of Fourier transform spectrometers is very well
tending to apply an analogous approach to estimate the understood.7 The change in the m easured FW HH of any
resolution of Raman spectrometers (after correction for band measured on an FT-Raman spectrometer with r .
the instrument response). It was found, however, that this 0.2 as the resolution is varied is only dependent on the
technique is not as readily applicable to Raman spectrom- maximum optical path difference and apodization func-
etry as it is for IR spectrometry because the instrument tion, provided that the beam passing through the inter-
response function (IR F) cannot be measured accurately ferometer is sufŽ ciently collimated. If parameters for the
enough. The feasibility of using other techniques will, computation of a spectrum m easured on an FT-R aman
therefore, be described in this report. spectrometer could be found such that the spectrum ex-
actly matched the spectrum of the same sample measured
TH EORY on a CCD-Raman spectrometer, the resolution of the lat-
Charge-Coupled Device-Raman Spectrometers. For ter instrument could be estimated with a high level of
grating spectrometers operating well above the diffraction accuracy. For spectra measured on an FT-R aman spec-
limit, the resolution is equal to the FW HH of the spectral trometer to reproduce spectra measured on a grating spec-
slit function, t(l), which is a triangle of the form: trometer, the apodization function must be selected such
that the ILS of the FT-Raman spectrom eter mimics the

5 )
t (l) 5 1 2
(l 2 lo )
Sg )6 for lo 2 S g # l # lo 1 S g
triangular spectral slit function of the grating spectrom-
eter. This m ay be achieved by convolving the FT-Raman
spectra with a triangular function, i.e., multiplying the
and
interferogram by a sinc 2 function of the appropriate
t (l) 5 0 elsewhere (2) width.
where l is the incident wave length, lo is the setting of In the above discussion, we have assumed that the slit
the m onochromator, and S g is the geometric spectral band function is triangular. This assumption is not accurate for
pass.1 In practice, the effect of aberrations is to cause the very low-resolution Raman spectrometry using a CCD-
measured shape of t(l) to be m ore Gaussian than trian- Raman spectrometer, as the exit slit width is equal to the
gular, as can be seen from Fig. 3. Presumably this is pixel width and not the width of the entrance slit (see
caused in part by the effect of optical aberrations and in Fig. 1). Convolving a large entrance slit with the rela-
part because the spectrum is measured by a CCD in the tively narrow pixel should yield a trapezoidal spectral slit
focal plane of the exit slit and not by a single detector function for which the corresponding apodization func-
located after the exit slit. Nonetheless, a triangle seems tion has the form:

192 Volume 57, Number 2, 2003

 F IG . 3. Spectra of a single neon line measured with slit widths of 10, 25, 50, 150, and 300 mm .

f (x) 5 A sinc 2x 1 B sinc 2x (3) form Raman spectra of the sample were collected on a
PerkinElmer Spectrum 2000 FT spectrometer equipped
We did not attempt to m imic a trapezoidal instrument line
with a 500-m W Nd 31 : YAG laser emitting at 1064 nm,
shape function through the use of an appropriate apodi-
an InGaAs detector, and IRDM software. All spectra were
zation function in this project because more than one var-
measured using 1808 back-scattering at several nom inal
iable must be adjusted, making the result susceptible to
resolutions (0.5, 1, 2, 3, 4, 5, 6, 7, and 8 cm 2 1 ), where
operator bias.
the nominal resolution is the reciprocal of the maximum
EXPERIMENTAL optical path difference. Each spectrum was acquired by
coadding 40 double-sided, bi-directional interferograms
Samples. A ;1-cm cube of cr ystalline calcite was ob- while m aintaining a Jacquinot-stop setting appropriate to
tained from Professor Mickey Gunter of the Department a resolution of 0.25 cm 2 1.
of Geology at the University of Idaho. BaSO 4 powder All FT-Raman spectra were corrected for the IRF with
(Aldrich Chemical Co., M ilwaukee, WI) was ground and a white-light source using a procedure reported by Fryl-
used with no further treatment. ing et al.8 This procedure assumes the m easured white-
Fourier Transform Raman Spectra. Fourier trans- light source closely approximates the size, geometry, and
position of the sample.9,10 The white-light spectra used to
TABLE I. The effect of sample orientation on peak height, FWHH,
correct for the IRF were collected using a source and
and area on the three most intense bands of calcite. The spectra power supply supplied by Perkin-Elmer. The same data
were acquired with a 203 objective and a 100-mm-wide entrance acquisition param eters and sample position were used for
slit. the collection of the interferograms of the white-light
% Dif- % Dif- source and the sample.
ference ference The collected interferogram s of the sample and white-
in in light source at each resolution were then imported into
Height Width Area height width
Matlab 5.3 (The M athWorks, Inc.) using a program writ-
Band 1 (280 cm 2 1 ) ten in our laboratory. Each interferogram was then mul-
Orientation 1 1877 13.18 33 900 tiplied by the FT of Eq. 2 with S g equal to the nominal
Orientation 2 1813 13.25 32 700 3.7% 0.5%
resolution of the respectively m easured sample or white-
Band 2 (710 cm ) 21
light-source interferogram. The phase-corrected spectra
Orientation 1 697 7.62 6000 were then calculated by taking the FT of each interfero-
Orientation 2 1252 7.70 10 900 80.7% 1.1%
gram.
Band 3 (1085 cm 2 1 ) A 10 cm 2 1 resolution spectrum of the known response
Orientation 1 620 7.08 48 645 of the white-light source, supplied by PerkinElmer, was
Orientation 2 1252 7.70 120 600 148% 1.6%
imported into Matlab 5.3. CoefŽ cients from a sixth-order

APPLIED SPECTROSCOPY 193

F IG . 4. Effect of orientation on the spectrum of a single cr ystal of calcite. Spectrum obtained by rotating the sample until the band at ;1085 cm 2 1
reached a m aximum (broken line); spectrum acquired by rotating the cr ystal 908 from its previous position (solid line).

polynom ial Ž t were used to reconstruct the spectrum at model a mixed Gaussian and Lorentzian function. The
each of the resolution settings. Each sample spectrum was calculated FW HH of each band was then transferred into
then corrected for the IRF. The corrected spectra were Microsoft Excel where the FW HH of the bands could be
saved as ASCII data and imported into Gram s/32 ver. plotted vs. the nominal resolution (i.e., the reciprocal of
4.01 (Galactic Industries Corp., Salem, NH). the maximum optical path difference).
The FW HH of each band of interest was calculated Charge-Coupled Device-Raman Spectra. The test
using the Grams/32 ‘‘curve Ž t’’ routine program med to spectra were collected on a Renishaw System 2000 CCD-

F IG . 5. Plot of the m easured full-width at half-height of the 1085 cm 2 1 band of calcite measured on an FT-Raman spectrometer vs. the nominal
resolution (reciprocal of the maximum optical path difference).

194 Volume 57, Number 2, 2003

F IG . 6. FT-Raman spectra of BaSO 4 powder calculated with a sinc 2 apodization function at nominal resolutions of 6, 7, and 8 cm 2 1 compared
with the CCD-Raman spectra collected with 100-mm and 150-mm slit widths.

Raman micro-spectrometer equipped with a ;20-mW di- spectrum of calcite to give a measure of resolution. Re-
ode laser (;5 m W at the sample focus) emitting at 782 grettably, when investigating the feasibility of this ap-
nm and a back-illuminated silicon CCD array detector proach, we found that two factors caused it to yield un-
cooled to 270 8C. The sample was placed under a 203 acceptable results. First, even though the IRF of our spec-
microscope objective for 1808 back-scatter collection and trometer was known to 615%, this accuracy was appar-
brought into focus on the surface by using the visual ently not high enough. Secondly, the relative intensity of
image of the aperture stop of the microscope light source. bands in the Raman spectra of single crystals varies dra-
Renishaw ‘‘Wire’’ software (ver. 1.2.76) was set to collect matically with the polarization of the incident laser beam
2 coadded scans in the ‘‘continuous’’ m ode from 150 to and the orientation of the sample, as shown by the spectra
1200 cm 2 1 (Raman shift) with an exposure time of 30 s of calcite mounted in two mutually perpendicular orien-
per point for each scan. tations (see Fig. 4). The ratios of the peak heights and
Type 2412 Kopp glass obtained from Gray Glass Co. widths of the three strong bands in the spectrum of calcite
(Jamaica, NY) was used as a standard to correct the in- for the two orientations are summarized in Table I. W hile
strument response of the Renishaw spectrometer. The the widths were not changed by sample orientation to
 uorescence spectrum of the glass was imported into within experimental error, the relative peak heights were
MatLab 5.3 using a program written in our laboratory. dramatically affected. The use of powdered samples ame-
To obtain the IRF, this spectrum was then ratioed against liorated this effect somewhat, but when samples are
a synthetic background constructed from a sixth-order
mounted in a Raman microspectrometer, as was the case
po lyn om ial, w h ose coefŽ cients w ere o btain ed from
for these measurements, the effect of sample orientation
McCreery’s web site.11 To conŽ rm the accuracy of the
published coefŽ cients, a spectrum of cyclohexane was is not eliminated entirely because only a limited num ber
acquired and corrected using the calculated IRF. The ex- of particles are observed.
perimentally determined relative Raman cross-sections of Calcite appears to be a good choice for a reference
cyclohexane from the IRF-corrected spectrum were com- material because it is nontoxic, inexpensive, has a
pared to those reported by Frost and McCreery 12 and the smooth, hard surface, and its spectrum has three com-
results agreed within 15%. All Raman spectra measured pletely resolved bands of differing widths over a range
on the CCD-Raman spectrometer were corrected by this of 1000 cm 2 1. These bands are denoted on Fig. 4 as bands
IRF. The corrected spectra were then saved as ASCII data 1 (280 cm 2 1), 2 (710 cm 2 1), and 3 (1085 cm 2 1 ). Calcite
and imported into Grams/32 for calculation of the peak is a birefringent material and the bands in its Raman
heights, widths, and areas. spectrum are strongly affected by the orientation of the
crystal relative to the plane of polarization of the laser
RESULTS AND DISCUSSION (as shown in Fig. 4). Nonetheless, there does not appear
We Ž rst investigated the feasibility of using the ratio to be a problem in the m easurement of the FW HH of the
of the heights of the three strong bands in the Raman bands in its spectrum.

APPLIED SPECTROSCOPY 195

 Construction of the Calibration Curve. Spectra of FT-R aman spectrometer, which is consistent with the val-
calcite were measured at a variety of resolutions on an ues predicted above.
FT-R aman spectrometer using sinc 2 apodization. From
measurem ents m ade at high resolution on an FT-Raman CONCLUSION
spectrometer, the true FW HH of bands 1, 2, and 3 in the In this paper, we have laid the foundation for deter-
spectrum of calcite are ;9, 2, and 1 cm 2 1 , respectively. mining the resolution of a m ultichannel Raman grating
The exact value of the FW HH of the narrower bands spectrometer by using a calibration curve constructed
cannot be determined because the FW HH of the Nd 31 : from m easurements m ade on an FT-Raman spectrometer.
YAG laser line is about 0.5 cm 2 1. Thus, the value of r As both FT-Raman and CCD-Raman spectrometers be-
for bands 2 and 3 can never be made low enough for an come more prevalent in the workplace, som e common
accurate determination of their FW HH. To illustrate this ground needs to be set. There are several deŽ nitions for
point, the FW HH of band 3 was plotted vs. the resolution the determination of the resolution of a grating spectrom-
of the FT-Raman spectrometer (see Fig. 5). It can be seen eter, whereas there is only one simple deŽ nition for the
that the measured FW HH is still decreasing when the nominal resolution of an FT-Raman spectrometer, which
resolution is reduced from 2 cm 2 1 to 1 cm 2 1. is deŽ ned as the reciprocal of the optical path difference
Two spectra of calcite were taken on the CCD-Raman of the interferometer, assuming a point source or sufŽ -
spectrometer, one with a slit width of 100 mm and the ciently small Jacquinot stop. Thus, the use of an FT-Ra-
other with a slit width of 150 mm. For these two slits, man spectrometer to deŽ ne the resolution of any other
the FW HH of band 3 was determined to be 7.1 and 8.6 type of Raman spectrometer is a logical choice, provided
cm 2 1 , respectively. When these two values were used as that the FT-R aman spectrometer can be m ade to mimic
input to the calibration curve shown in Fig. 5, the reso- the response of a dispersive Raman spectrometer by the
lution was determined to be ;6.3 and ;7.8 cm 2 1 for the application of a suitable apodization function. This study
slit widths 100 mm and 150 mm, respectively. This m ay suggests that the proposed method will produce an an-
be compared to the values calculated above from the re- swer close to the true resolution of a CCD-Raman spec-
ciprocal linear dispersion of 6.7 and 10.1 cm 2 1 , respec- trometer. However, a closer investigation of the spectral
tively. Although the Ž rst of these values agrees well with slit function of a modern grating spectrometer using a
the value obtained by the method we have described CCD detector is required before an apodization function
above, the second value is in error by about 20%, indi- can be developed that accurately mimics this slit function.
cating that when Dñ is large, the resolution of a CCD-
Raman spectrom eter is not easily calculated even if R d is ACK NOW LEDGM ENTS
known accurately.
We would like to express our appreciation to Drs. Steven Choquette,
To further validate our proposed approach, an unre- Edgar Etz, and Wilbur Hurst of NIST and the other attendees of the
solved doublet in the spectrum of BaSO 4 at ;455 cm 2 1 meetings of the ASTM sub-committee on Raman spectroscopy for their
was closely inspected. Bands with a Raman shift less than useful dialogue and genuine enthusiasm in this project. We would also
1085 cm 2 1 will be slightly less resolved than predicted like to express our appreciation to Dr. Zhenhuan Chi of Renishaw Inc.
for useful information on the reciprocal linear dispersion of the Reni-
by the calibration curve because resolution, in cm 2 1 , is shaw Spectrum 2000. B.T.B. thanks Thermo Nicolet for allowing him
not constant across spectra measured on a grating spec- to continue his work on this project.
trometer. However, assuming the reciprocal linear disper-
sion has a constant value of 3.25 nm/m m across the spec- 1. J. D. Ingle, Jr., and S. R. Crouch, Spectrochemical Analysis (Pren-
trum, the resolution at a Raman shift of 455 cm 2 1 should tice-Hall Inc., New Jersey, 1988).
be 7.38 cm 2 1 when using a 100-mm entrance slit. The 2. Z. Chi, Renishaw, Inc., personal comm unication to P.R.G. (2002).
two bands of BaSO 4 at 455 cm 2 1 become resolved when 3. C. Tseng, J. F. Ford, C. K. Mann, and T. J. Vickers, Appl. Spectrosc.
the spectrum is measured on an FT-Raman spectrom eter 47, 1808 (1993).
4. A. W. Fountain III, T. J. Vickers, and C. K. Mann, Appl. Spectrosc.
with a nominal resolution of better than 8 cm 2 1 and the 52, 462 (1988).
spectra are computed with a sinc 2 apodization function. 5. D. A. Ramsay, J. Am. Chem. Soc. 74, 72 (1952).
The FT-Raman spectra of this sample m easured at a nom- 6. Coblentz Soc. Board of Management, Anal. Chem. 38, 27A (1966).
inal resolution of 8, 7, and 6 cm 2 1 are shown as the lower 7. P. R. GrifŽ ths and J. A. de Haseth, Fourier Transform Infrared
Spectrometry (Wiley Interscience, New York, 1986), Chap. 1.
three traces of Fig. 6. CCD-Raman spectra of the same 8. M . Fryling, C. J. Frank, and R. L. McCreer y, Appl. Spectrosc. 47,
sample m easured with 100- and 150-mm slits are shown 1965 (1993).
above the FT-Raman spectra. It is clear that the CCD- 9. H. Hamaguchi, Appl. Spectrosc. Rev. 24, 137 (1988).
Raman spectrum m easured with the 100-mm slit has a 10. K. Iwata, H. Hamag uchi, and M. Tasumi, Appl. Spectrosc. 42, 12
(1988).
resolution between 7 and 8 cm 2 1 while the resolution of 11. K. J. Frost and R. L. McC reer y, Appl. Spectrosc. 52, 1614 (1998).
the spectrum measured with the 150-mm slit is numeri- 12. R. L. McCreery, ‘‘http://chemistry.ohio-state.edu/;rmccreer/intensty/
cally greater than the 8 cm 2 1 spectrum measured with the table1.html’’.

196 Volume 57, Number 2, 2003