Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Received: 5 June 2020 Revised: 11 July 2020 Accepted: 18 August 2020

DOI: 10.1002/jrs.6003

RESEARCH ARTICLE

Reliable estimation of Raman shift and its uncertainty for a


non-doped Si substrate (NMIJ CRM 5606-a)

Nobuyasu Itoh | Katsuhiro Shirono

National Metrology Institute of Japan


(NMIJ), National Institute of Advanced
Abstract
Industrial Science and Technology (AIST), Raman shift is one of the most important parameters in Raman spectroscopy
Japan
and is theoretically calculated from the difference in the wavenumbers of
Correspondence excitation and Raman scattering. At present, there are no standard procedures
Nobuyasu Itoh, National Metrology available to reliably estimate the Raman shift and evaluate its uncertainty from
Institute of Japan (NMIJ), National
experimental results. In this study, we reliably estimated the Raman shift and
Institute of Advanced Industrial Science
and Technology (AIST), Central 3, 1-1-1 evaluated its uncertainty for a non-doped crystalline Si substrate (NMIJ
Umezono, Tsukuba, Ibaraki 305-8563. CRM5606-a) using a He-Ne laser and its plasma lines (Ne emission lines). The
Japan.
Email: nobuyasu-itoh@aist.go.jp
Raman scattering wavenumber of NMIJ CRM 5606-a was estimated by the
bracket calibration method using the wavenumbers of the closest Ne emission
lines. The Raman shift was determined from the differences in the theoretical
wavenumbers of a He-Ne laser and the observed Raman scattering. To
evaluate the combined standard uncertainty, we considered the uncertainty
originating from the He-Ne laser and the Ne emission lines, the repeatability
of experimental results, inhomogeneity of measuring points and spectral
dispersion. The estimated Raman shift was 520.45 cm−1 ± 0.28 cm−1 (coverage
factor k = 2). The Raman shift was validated by the other estimated Raman
shifts obtained with other two Raman spectrometers.

KEYWORDS
He-Ne laser, Ne emission lines, Raman shift, Si substrate, uncertainty

1 | INTRODUCTION using a common material to calibrate all the Raman


spectrometers.[5]
Raman shift originates from the difference between Wavelengths/wavenumbers of elemental emission
the energies of excitation and Raman scattering.[1–4] lines such as Ne and Raman shift values of
Although Raman shifts are theoretically independent of well-characterized materials are commonly used for
the Raman spectrometer used, in practice, the observed calibration.[1–3] Use of elemental emission lines are suit-
shifts are strongly dependent on the Raman spectrome- able for accurate calibration and well-characterized mate-
ter. Furthermore, users must calibrate their Raman rials are convenient for daily calibration. Materials
spectrometers according to the references and protocols provided in ASTM E1840 were obtained from 8 laborato-
provided by the manufactures because there are no ries and are widely used to calibrate Raman spectrome-
standardized protocols available for calibration. Under ters.[1,3,4,6] Some of these materials (polystyrene,
these circumstances, Raman shifts obtained from differ- cyclohexane, and paracetamol) are also described in
ent Raman spectrometers are not so consistent; however, European Pharmacopoeia (Ph. Eur.) along with their tol-
their consistency can be significantly improved when erance standards.[7] Although the Raman shift of some

2496 © 2020 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/jrs J Raman Spectrosc. 2020;51:2496–2504.
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ITOH AND SHIRONO 2497

materials described in ASTM E1840 (polystyrene, cyclo- 2 | EXPERIMENTAL


hexane, and benzonitrile) were in good agreement with
the most of independent interlaboratory comparison 2.1 | Materials
data,[5] Raman shifts of materials provided in ASTM
E1840 are not associated with uncertainty information, The non-doped Si substrate for positron defect measure-
which is essential for obtaining reliable values. Thus, it is ments (NMIJ CRM 5606-a, No. 003) was obtained from
necessary to reliably estimate the Raman shifts and eval- National Metrology Institute of Japan (NMIJ). This CRM
uate their uncertainty for materials; however, requisite constitutes two pieces, each of which is 15 mm × 15 mm
procedures have not been provided. with 1 mm thickness, with a crystal orientation of
To evaluate the uncertainty in Raman shifts, the {100} ± 1 .[17] One side of the substrate is mirror-finished,
uncertainty in the wavenumbers of excitation and Raman while the other side has a rough surface. The measure-
scattering should be known. International Committee for ments were performed on the mirror-finished side.
Weights and Measures (CIPM) provides the reliable
frequency (f = 473.6127 THz) and wavelength in vacuum
(λ = 632.9908 nm), along with their relative uncertainty 2.2 | Raman spectrometers
(1.5 × 10−6), for unstabilised He-Ne lasers.[8,9] Thus, the
use of an unstabilised He-Ne laser (He-Ne laser com- Three types of independent Raman spectrometers were
monly used for Raman spectrometers) is important for employed in this study. One of them was for the estima-
evaluating the uncertainty in the Raman shift. Ne emis- tion of the Raman shift of NMIJ CRM 5606-a and uncer-
sion lines are also widely used for the calibration of tainty evaluation (System A), while the other two were
Raman spectrometers, and their reliable wavenumbers in for the validation of that estimation (Systems B and C).
vacuum, along with their uncertainties, are also avail- System A is a LabRAM HR Evolution Raman spectrome-
able.[10] The bracket calibration method allows the esti- ter, and Systems B and C are an NRS-7100 Raman spec-
mation of the wavenumbers and the evaluation trometer and an NRS-5500 Raman spectrometer,
uncertainties of unknown peaks from those of the Ne respectively. Unstabilised He-Ne lasers were used for all
emission lines in vacuum.[11,12] Thus, reliable Raman the systems to apply its reliable wavenumber and uncer-
shifts of unknown peaks may be able to be estimated by tainty provided from CIPM. Ne emission lines used in
using the wavenumbers and the uncertainties of He-Ne this study were originated from the He-Ne laser itself
laser and Ne emission lines. (plasma lines) in all the analytical systems. These lines
Crystalline Si substrates are widely used for the daily are more reliable than Ne emission lines using another
calibration of Raman microscopes because they give a Ne lamp because their optical path is the same as that of
single strong peak at 520 cm−1.[4,5] On the other hand, the He-Ne laser and Raman scattered light.
the Raman shift of Si substrates is not provided in ASTM A LabRAM HR Evolution Raman spectrometer (focal
E1840 partly because of the differences in the Raman length: 800 mm, objective: 100×, laser power at sample:
shifts that arise from not only the residual stresses and 1 mW; HORIBA, Kyoto, Japan) with a 1800 gr/mm
preparation processes[13–16] but also the inapplicability in (System A) was used to estimate the Raman shift of NMIJ
calibration for a wide spectral range.[5] Thus, reliable esti- CRM 5606-a. The measurement range was
mation of Raman shift for Si substrate is important for 15123.29–15445.66 cm−1 (wavelength: 647.43–661.23 nm)
users of Raman microscopes. in air.
In this study, we reliably estimated the Raman shift To validate the Raman shift of NMIJ CRM 5606-a, an
of Si substrate (NMIJ CRM 5606-a) using an unstabilised NRS-7100 Raman spectrometer (focal length: 500 mm,
He-Ne laser and its plasma lines (Ne emission lines). This objective: 100×, laser power at sample: 1 mW; JASCO,
CRM can be reasonably used as a typical Si substrate Tokyo, Japan) equipped with a 1800 gr/mm grating
because NMIJ CRM 5606-a (single-crystal silicon for posi- (Systems B) and an NRS-5500 Raman spectrometer
tron defect measurements) is fabricated from a non- (focal length: 300 mm, objective: 100×, laser power at
doped Si substrate.[17] The combined standard uncer- sample: 1 mW; JASCO) equipped with a 1800 gr/mm
tainty was evaluated from the uncertainties in the grating (System C) were also used. The measurement
wavenumbers of the He-Ne laser and the Ne emission ranges in air were 15021.94–15534.75 cm−1
lines, the repeatability of experimental results, inhomoge- (wavelength: 643.72–665.69 nm) for System B and
neity of measuring points and spectral dispersion. The 14953.90–15648.14 cm−1 (wavelength: 639.05–668.72 nm)
obtained Raman shift of NMIJ CRM 5606-a was validated for System C.
by the other estimated Raman shifts obtained with other Five points on a piece of Si substrate (four points
Raman spectrometers. close to the corners and a point around the centre) were
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2498 ITOH AND SHIRONO

selected for the measurement. 10 times repetitions were 3 | R E S U L T S AN D D I S C U S S I O N


implemented on each point. Peak top wavenumbers of
Ne emission lines and Si substrate obtained as the output 3.1 | Estimation of Raman scattering
data were used for the analysis without any curve fittings wavenumber of the Si substrate
in this paper.
It should be noted that the uncertainty evaluation
depends on not only the measurement procedure but also
2.3 | Estimation of Raman shift the definition of the measurand. The definition should
be, therefore, given in an explicit way. The measurand in
The wavelength in vacuum (λ = 632.9908 nm), frequency this study is defined as the Raman shift on an arbitrary
(f = 473.6127 THz), and relative standard uncertainty point on a Si-substrate piece. The uncertainty evaluation
(1.5 × 10−6) of the unstabilised He-Ne laser are of the Raman shift on a specific point given by the posi-
available.[8,9] Wavelengths of the Ne emission lines in air tional information, and the general reference value for
and their wavenumber in vacuum are also available.[10] the Raman shift of a bulk of NMIJ CRM 5606-a with the
In this study, we used wavenumbers in vacuum and their uncertainty are different from what is evaluated in this
corresponding uncertainties; the wavenumber of the study. More information is given in 3.3 Uncertainty in the
He-Ne laser in vacuum (15798.018 cm−1) was calculated inhomogeneity of the measurement points subsection.
from the reciprocal of its wavelength in vacuum Five points on a piece of Si substrate were selected for
(λ = 632.9908 nm). the measurement. Specifically, four points close to the
Theoretical Raman scattering wavenumber of NMIJ corners and a point around the centre. 10 times
CRM 5606-a was estimated with the two closest Ne repetitions were implemented on each point. We use the
emission lines using the bracket calibration method. subscript j and k to mean the identifications for the
Then, the Raman shift was estimated from the difference positions (j = 1, …, 5) and the repetitions (k = 1, …, 10),
in the wavenumber of the He-Ne laser and that of the respectively.
Raman scattering of NMIJ CRM 5606-a. For this purpose, we applied the bracket calibration
method (Equation 1) to obtain the theoretical
wavenumber of the Si substrate[11,12] as there is only one
2.4 | Evaluation of uncertainty target peak at 520 cm−1 in this study. We used the two
closest Ne emission lines (lines 2 and 3, Table 1 and
The uncertainty of the Raman shift was evaluated based Figure 2) to obtain the Raman shift of the Si substrate for
on ISO/IEC Guide98–3 “Evaluation of measurement the k-th repetition at the j-th position, ΔνSi, j, k, as follows:
data -- Part 3: Guide to the expression of uncertainty in
measurement (GUM)”.[18] In Figure 1, we show the
2   Ob  3
νTh
Ne2 −ν Th
Ne3 × ν Si,j,k −ν Ob
Ne3,j,k
ΔνSi,j,k = νTh 4 5
summary of the uncertainty evaluation in this study. HeNe − + νTh
Ne3 ,
The standard uncertainty of the Raman shift u (ΔνSi) νOb
Ne2,j,k −νNe3,j,k
Ob

was evaluated step by step according to this figure. The ð1Þ


uncertainty sources taken into consideration are
follows:
−1 −1
where νThNe2 (cm ), νNe3 (cm ) and νHeNe are the theoreti-
Th Th

i. Uncertainty due to the wavenumbers given in the cal wavenumbers of Ne emission line-2, Ne emission
references line-3 and He-Ne laser, respectively. Table 1 gives the
−1 −1
ii. Uncertainty due to the inhomogeneity (difference Si (cm ) and νSi,j,k (cm ) are
specific values of them. νTh Ob

between measurement points) the theoretical and observed wavenumbers, respectively,


−1
Si,j,k , νNe2,j,k (cm )
of Raman scattering for Si substrate. νOb Ob
−1
iii. Uncertainty due to the line selection and νNe3,j,k (cm ) are Raman scattering of the Si
Ob

substrate, the observed wavenumbers of Ne emission


iv. Uncertainty due to the spectral dispersion line-2 and Ne emission line-3, respectively.
(wavenumber resolution) The average of the 50 values are computed as the
v. Uncertainty due to the repetitions in the measurement value as follows:
measurement
1X 5 X 10
Finally, the expanded uncertainty was obtained by the ΔνSi = ΔνSi, j, k : ð2Þ
50 j = 1 k = 1
coverage factor (k = 2) for the 95% level of confidence.
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ITOH AND SHIRONO 2499

F I G U R E 1 Summary of the
uncertainty evaluation in this
study [Colour figure can be
viewed at wileyonlinelibrary.
com]

TABLE 1 Wavenumbers used in this study and their


uncertainty

Wavenumber in Standard
vacuum (cm−1) uncertainty (cm−1)
He-Ne laser*1 15798.018 0.024
Ne line-1 15364.9354 0.0012
Ne line-2 15302.9512 0.0012
Ne line-3 15149.7353 0.0011
Ne line-4 15140.677 0.009

Wavenumbers and the uncertainties of Ne emission lines were


obtained from literature (Reference 10). F I G U R E 2 Raman spectrum of NMIJ 5606-a with Ne
*1 Calculated from the reciprocal of wavelength in vacuum and its emission lines obtained with System A. Labels correspond to those
relative uncertainty (References 8 and 9). in Table 1

Table 2 shows the averaged Raman shift on each mea- table. We applied the three-decimal-point expression,
surement point together with the total averaged values. and the total averaged value of ΔνSi = 520.446 cm−1 is
For later discussion, the standard deviations for the repe- employed as the reported value for the measurand in
titions on individual measurement points are given in the this study.
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2500 ITOH AND SHIRONO

T A B L E 2 Averaged Raman shifts and the standard deviations ∂ΔνSi 1


cTh
Ne3 = ≈ Ave Ave −1, ð8Þ
of a measured Raman shifts obtained on individual measurement ∂νNe3 νNe2 − νNe3
Th
points together with the total averaged Raman shift using System A

Averaged Raman Standard ∂ΔνSi


cTh
HeNe = ≈1: ð9Þ
shift (cm−1) deviation (cm−1) ∂νTh
HeNe
Top-right 520.4627 0.1583
Top-left 520.3944 0.0758
Thus, the uncertainty due to the wavenumbers can be
Centre 520.5076 0.1740
computed as follows:
Bottom-right 520.3703 0.0310
h  Th 2  Th  Th 2  Th  Th 2 i1=2
Bottom-left 520.4924 0.1761
uwave = cTh
Ne2 u νNe2 + cNe3 u νNe3 + cHeNe u νHeNe
Total average 520.4455
= 0:0237 cm − 1 :
Indices of top-right to bottom-left are given in convenience and no ð10Þ
absolute direction is defined for NMIJ CRM 5606-a.

3.2 | Uncertainty of the wavenumbers 3.3 | Uncertainty in the inhomogeneity of


the measurement points
As mentioned, we evaluated the uncertainties of the
wavenumbers by converting those of the wavelengths The Raman shift of a Si substrate is affected by the
available from references.[8,9] Not only the standard changes in residual stresses[13–16]; thus, the estimated
uncertainties of individual sources but also the sensitivity Raman shifts can vary with the measuring points,
coefficients, the conversion factors of them to the depending on the cutting processes. Therefore, we
reporting Raman shift,[18] must be clarified in the compu- obtained data at five different measuring points.
tation. To compute the sensitivity coefficients, we need to The inhomogeneity of the measuring points was
give an approximate measurement model to specify the assessed by measuring the wavenumbers at five different
influence of individual wavenumbers. The following measuring points (four corners and the centre). The one-
approximate measurement model suggested in this study: way ANOVA was performed, and the mean squares for a
measurement point (Vrep) and among measurement
   Ave Ave  
Ne2 − νNe3 × νSi − νNe3
νTh Th points (Vpoint) were calculated as follows:
ΔνSi ≈νTh
HeNe − + νNe3 ,
Th
ð3Þ
Ne2 −νNe3
νAve Ave

1X 5 X 10  2
V rep = ΔνSi, j −ΔνSi , ð11Þ
4 j=1 k=1
Si , νNe2 and νNe3 are given as follows:
where νAve Ave Ave

1X 5 X 10 1X 5 X 10  2
νAve = νOb , ð4Þ V point = ΔνSi, j, k −ΔνSi, j , ð12Þ
Ne2
50 j = 1 k = 1 Ne2,j,k 45 j = 1 k = 1

1X 5 X 10
where
νAve
Ne3 = νOb , ð5Þ
50 j = 1 k = 1 Ne3,j,k
1X 10
ΔνSi, j = ΔνSi, j, k : ð13Þ
1X 5 X 10 10 k = 1
νAve
Si = νOb : ð6Þ
50 j = 1 k = 1 Si, j, k
Vrep and Vpoint can be computed using the data in
Based on equation 3, the sensitivity coefficients for νTh
Ne2 , Table 2. Usually, the standard deviations between the
νTh
Ne3 and ν Th
, c , c
HeNe Ne2 Ne3 and c HeNe , are given as follows: points (spoint) were calculated as follows[19]:

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
∂ΔνSi 1 V point − V rep
cTh = Th ≈− Ave Ave , ð7Þ spoint = = 0:0423 cm − 1 : ð14Þ
Ne2
∂νNe2 νNe2 −νNe3 10
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ITOH AND SHIRONO 2501

It should be noted that the uncertainty due to the because the standard uncertainty
pffiffiffi for the rectangular dis-
inhomogeneity is not spoint. We must take the two tribution is given as H/ 3 , where H means the half
matters into consideration: (i) The representative value width of the rectangular distribution.[18]
ΔνSi varies because of the inhomogeneity, and (ii) the
arbitral points on the target piece of Si substrate has the
Raman shift fluctuate with the standard deviation spoint. 3.5 | Uncertainty originated from spectral
For the first, we can evaluate the standard
pffiffiffi deviation in dispersion
the determination of ΔνSi as spoint/ 5 , because there
were the five measurement points. Thus, we can compute We also considered spectral dispersion as a source of
the uncertainty due to the inhomogeneity in the uncertainty when estimating the Raman scattering
measurement points as follows: wavenumbers, because it depends on the grating sys-
tems under the same focal length and strongly affects
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s2point the accuracy of the experimental results. Here, we used
upoint = + s2point = 0:0464 cm − 1 : ð15Þ peak-top numbers of Ne emission lines and Si
5
substrate without any curve fitting to reasonably
evaluate the standard uncertainties originated from
In this study, the measurand is the Raman shift at an spectral dispersion (udisp) using the standard deviation
arbitrary point on the target piece. However, if we specify of rectangular distributions with the half-width
the measurement point with the positional information, corresponding to the resolution of the employed plots.
the uncertainty due to the inhomogeneity is not basically Table 4 shows the most frequently observed peak-tops
required because the measured value is associated with for Ne lines 2 and 3 and Si substrate together with the
the specific point. two closest wavenumbers to the peak-tops in larger and
smaller sides. For each line, we computed the larger
value between the two differences (of the closest
3.4 | Uncertainty in the selection among wavenumbers and the peak-top), which is shown in
the Ne emission lines Table 4 as “larger interval”.
Since the interval of the observed wavelength vary
The Raman shifts computed with the different pairs of in the larger and the smaller sides of the peak-top, the
the Ne emission lines varies from each other. Specific rectangular distribution may not be applied in a strict
values can be computed using the same data to obtain sense. However, the differences of the intervals are
ΔνSi with Ne lines 2 and 3. In Table 3, we show the three small enough to approximate the distribution to be
values of the Raman shift with using three different pairs rectangular. In this approximation, we employ the
of Ne emission lines (1, 3), (1, 4), and (2, 4). We can find larger interval as the half-width of the rectangular dis-
the difference of 0.0828 cm−1 at the maximum. tribution. Thus, we evaluate the standard uncertainties
The uncertainty in variance between the Ne emission Ne2 , νNe3 , and νSi as due to the spectral dispersion as
of νAve Ave Ave

lines (uline) was evaluated in this study. For this purpose, follows:
we assume the uniform distribution with the half width
of the maximum difference of 0.0828 cm−1. Thus, the   0:3280 −1
Ne2 = pffiffiffi cm
u νAve = 0:0947 cm − 1 , ð17Þ
uncertainty uline is computed as follows: 3

0:0828   0:3215
uline = pffiffiffi cm − 1 = 0:0478 cm − 1 , ð16Þ Ne3 = pffiffiffi cm
u νAve −1
= 0:0928 cm − 1 , ð18Þ
3 3

  0:3269
u νAve
Si = pffiffiffi cm − 1 = 0:0944 cm − 1 : ð19Þ
3
T A B L E 3 Averaged Raman shifts computed using different Ne
lines for System A The sensitivity coefficients for these values are
Averaged Raman Difference from evaluated using the approximation with equation 3 as
shift (cm−1) ΔνSi (cm−1) follows:
Ne lines 1 and 3 520.5283 0.0828
   Ave Ave 
Ne lines 1 and 4 520.4940 0.0485 ∂ΔνSi νThNe2 −νNe3 × νSi −νNe3
Th
cAve = Ave ≈  Ave Ave 2 , ð20Þ
Ne lines 2 and 4 520.4308 – 0.0147
Ne2
∂νNe2 νNe2 −νNe3
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2502 ITOH AND SHIRONO

T A B L E 4 Most frequently observed peaktops for Ne lines 2 and 3 and Si substrate together with the two closest wavenumbers to the
peak-tops in larger and smaller sides, respectively, for System A. see text for the computation of the larger difference to the peak-top

Most frequently Closest wavenumber Closest wavenumber


observed the peaktop to the peaktop to the peaktop Larger interval to
(cm−1) (larger side) (cm−1) (smaller side) (cm−1) the peaktop (cm−1)
Ne line 2 15306.6926 15307.0206 15306.3645 0.3280
Ne line 3 15153.4667 15153.7882 15153.1682 0.3215
Si substrate 15281.4071 15281.7340 15281.1035 0.3269

   Ave Ave  srep


∂ΔνSi νThNe2 −νNe3 × νNe2 − νSi
Th
urep = pffiffiffiffiffi = 0:0193 cm − 1 , ð25Þ
cAve = Ave ≈  Ave Ave 2 , ð21Þ 50
Ne3
∂νNe3 νNe2 −νNe3

because 50 measurements were performed in total for


∂ΔνSi νTh −νTh
cAve = Ave ≈− Ne2 Ne3
: ð22Þ five points in 10 repetitions.
Si
∂νSi νAve
Ne2 −νNe3
Ave

Thus, the uncertainty due to the wavenumbers can be 3.7 | Combined standard uncertainty and
computed as follows: the expanded uncertainty
h  Ave 2  Ave  Ave 2  Ave  Ave 2 i1=2 The results of the uncertainty evaluation are summarized
udisp = Ne2 u νNe2
cAve + cNe3 u νNe3 + cSi u νSi
in Table 5. The combined standard uncertainty u (ΔνSi) is
= 0:1214 cm −1 : computed the square root of the sum of the square of the
ð23Þ five standard uncertainties, i.e. uwave, upoint, uline, udisp,
and urep, as follows:
Strictly, this spectral dispersion uncertainty is not per-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fectly systematic but partially randomly shown in the uðΔνSi Þ = u2wave + u2point + u2line + u2disp + u2rep = 0:1418 cm −1 :
repetition. However, the analysis on the separation of the
fluctuation into the systematic and the random effects ð26Þ
are complicated.[20] We hence regard the spectral disper-
sion uncertainty to be perfectly systematic and indepen- The expanded uncertainty (U) in the Raman shift is equal
dent of the variance in the repetitions evaluated in 3.6 to k × uCRM, where k = 2 is the coverage factor, for the
Uncertainty originated from repetition subsection, and 95% level of confidence.
allow some over-evaluation of the uncertainty caused by
this measure. U = kuðΔνSi Þ = 0:28 cm − 1 : ð27Þ

Thus, the Raman shift of the Si substrate was estimated


3.6 | Uncertainty originated from to be 520.45 ± 0.28 cm−1.
repetition

As described in 3.5 Uncertainty originated from spectral 3.8 | Validation of the obtained Raman
dispersion subsection, uncertainty originated from repeti- shift with other systems
tion was separately evaluated from uncertainty originated
from spectral dispersion (udisp). The standard deviation in The obtained Raman shift with System A (focal length:
the repetition (srep) can be computed from the data 800 mm and grating: 1800 gr/mm) was validated with
shown in 3.3 Uncertainty in the inhomogeneity of the mea- other two different systems (System B, focal length:
surement points subsection as follows: 500 mm, and groove density: 1800 gr/mm; System C,
focal length: 300 mm, and groove density: 1800 gr/mm).
pffiffiffiffiffiffiffiffiffi
srep = V rep = 0:1364 cm − 1 : ð24Þ As the same with System A, five points on a piece of Si
substrate (top-right, top-left, centre, bottom-right,
bottom-left) were selected and 10 times repetitions were
Thus, we can compute the uncertainty due to the repeti- implemented on each point in Systems B and
tion as follows: C. Obtained averaged Raman shift at each point was from
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ITOH AND SHIRONO 2503

TABLE 5 Uncertainty budget for the Raman shift of a Si substrate for System A

Sub-source

Standard Converted standard Standard and


uncertainty of uncertainty with expanded
Source sub-source Sensitivity the sensitivity uncertainty for
(symbol) Symbol (cm−1) coefficient coefficient (cm−1) ΔνSi (cm−1) Note
Wavenumber 0.0237
(uwave)
 
u νTh
Ne2 0.0012 0.0065 0.0000
 
u νTh
Ne3 0.0011 – 0.9935 0.0011
 Th 
u νNeHe 0.0237 1 0.0237
Inhomogeneity 0.0061
(upoint)
Selection in the 0.0478
lines (uline)
Spectrum 0.1214
dispersion
(udisp)
 
u νAve
Ne2 0.0944 0.8089 0.0763
 
u νAve
Ne3 0.0947 0.1904 0.0180
 Ave 
u νSi 0.0928 – 0.9993 0.0927
Repetition 0.0193
(upoint)
u(ΔνSi) 0.1418 Relative standard
uncertainty; 0.027%.
U 0.28 k = 2. Relative expanded
uncertainty; 0.054%.

520.4161 cm−1 to 520.5284 cm−1 with System B and from (0.1418 cm−1) < System B (0.1884 cm−1) < System C
520.1723 cm−1 to 520.3132 cm−1 with System C. The (0.2562 cm−1).
detailed information is given in the Electric Supplemen- The significance of the differences in the obtained
tary Information (Tables S1–S3 for System B and results was evaluated using inequation (28)[19] for the
Tables S5–S7 for System C). The obtained Raman shifts comparison between Systems A and B:
with Systems B and C were 520.52 ± 0.38 cm−1 and
520.25 ± 0.51 cm−1, respectively (k = 2). The evaluated pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jΔνSi − ΔνSi, B j ≤ k × u2 ðΔνSi Þ + u2 ðΔνSi, B Þ : ð28Þ
Raman shifts with Systems B and C are referred to as
ΔνSi,B and ΔνSi,C, respectively, for the simplicity in the
description. When evaluated using this method, the difference is not
The uncertainty budgets of Systems B and C are sum- significant if this equation is satisfied. For the compari-
marized in Tables S4 and S8, respectively. uwave son between Systems A and C, the evaluation can be
(0.0237 cm−1) is a common source for all systems because implemented with the similar inequation. The both
all systems commonly used He-Ne laser and the same Ne results obtained with Systems B and C satisfied these
emission lines. upoint of Systems B (0.0133 cm−1) and C inequations, respectively, meaning the obtained Raman
(0.0461 cm−1) are larger than System A (0.0061 cm−1) shift with System A (520.45 ± 0.28 cm−1) is reasonable.
whereas uline of Systems B (0.0127 cm−1) and C
(0.0311 cm−1) are better than System A (0.0478 cm−1).
Commonly, udisp is originating from the combination of 4 | CONCLUSIONS
focal length and groove density, and the largest uncer-
tainty source for all the systems. Thus, the combined Raman shift and its expanded uncertainty (coverage fac-
uncertainty increased following order: System A tor k = 2) of a non-doped Si substrate (NMIJ CRM5606-a)
10974555, 2020, 12, Downloaded from https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/jrs.6003 by Sabanci University, Wiley Online Library on [10/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2504 ITOH AND SHIRONO

was reliably evaluated to be 520.45 ± 0.28 cm−1 with an [9] International Committee for weights and Measures (CIPM),
unstabilized He-Ne laser and its plasma lines Recommended values of standard frequencies for applications
(Ne emission lines) using a bracket calibration method. including the practical realization of the metre and secondary
representation of the second -Helium neon laser (unstabilized)
To the best of our knowledge, this is the first estimation
(f ≈ 474 THz), 2007. Restrieved Jun. 05, 2020 from https://
of reliable Raman shift for a Si substrate. Although this www.bipm.org/utils/common/pdf/mep/M-e-P_unstab-HeNe_
approach is also suitable for reliable estimation of Raman 633.pdf
shifts for specific peaks of other materials close to Ne [10] E. B. Saloman, C. J. Sansonetti, J. Phys. Chem. Ref. Data 2004,
emission lines using fine dispersive resolution system, 33, 1113.
reflective materials should be better to use Ne emission [11] L. Cuadros-Rodríguez, M. G. Bagur-González,
lines originated from He-Ne lasers. For Raman shifts of M. Sánchez-Viñas, A. González-Casado, A. M. Gómez-Sáez,
J. Chromatogr. A 2007, 1158, 33.
peaks without close Ne emission lines, using low disper-
[12] N. Itoh, A. Sato, T. Yamazaki, M. Numata, A. Takatsu, Anal.
sive resolution system, or the both, other approaches
Sci. 2013, 29, 1209.
should also be developed. [13] T. Englert, G. Abstreiter, J. Pontcharra, Solid-State Electron.
1980, 23, 31.
A C K N O WL E D G E M E N T [14] I. D. Wolf, H. E. Maes, J. Appl. Phys. 1996, 79, 7148.
This work was partly performed by the VAMAS TWA42 [15] Y. Kang, Y. Qiu, Z. Lei, M. Hu, Opt. Lasers Eng. 2005, 43, 847.
corresponding committee in Japan. We are grateful to [16] T. Beechem, S. Graham, Rev. Sci. Instrum. 2007, 78, 061301.
Mr. Shinsuke Kashiwagi (HORIBA) and Mr. Kohei [17] National Metrology Insitute of Japan, Reference Material
Certificate of NMIJ CRM 5606-a (Single-crystal silicone for
Tamura (JASCO) for the measurements and valuable dis-
positron defect measurements), National Metrology Insitute of
cussions. We thank Tottori Institute of Industrial Tech- Japan, Tokyo, 2015.
nology for the usage of the Raman spectrometer. [18] International Organization for Standardization (ISO), ISO/IEC
98–3: 2008 Guide to the expression of uncertainty in measure-
ORCID ment, ISO, Geneva, 2008.
Nobuyasu Itoh https://orcid.org/0000-0003-0163-4454 [19] International Organization for Standardization (ISO),
Katsuhiro Shirono https://orcid.org/0000-0002-4930- ISO Guide 35: Reference materials -- Guidance for
799X characterization and assessment of homogeneity and stability,
ISO, Geneva, 2017.
[20] R. B. Frenkel, Statistical Background to the ISO “Guide to
R EF E RE N C E S the Expression of Uncertainty in Measurements”, CSIRO,
[1] R. L. McCreery, Raman spectroscopy for chemical analysis, Sydney 2002.
Wiley Interscience, New York 2000.
[2] I. R. Lewis, H. Edwards, Handbook of Raman Spectroscopy:
From the Research Laboratory to the Process Line (Practical SU PP O R TI N G I N F O RMA TI O N
Spectroscopy), CRC Press, New York 2001. Additional supporting information may be found online
[3] J. R. Ferraro, K. Nakamoto, C. W. Brown, Introductory Raman in the Supporting Information section at the end of this
Spectroscopy, 2nd ed., Academic press, Amsterdam 2003.
article.
[4] E. Smith, G. Dent, Modern Raman spectroscopy - A practical
approach, Wiley, Chichester 2005.
[5] N. Itoh, K. Shirono, T. Fujimoto, Anal. Sci. 2019, 35, 571. How to cite this article: Itoh N, Shirono K.
[6] ASTM International, ASTM E1840–96 (Reapproved 2014):
Reliable estimation of Raman shift and its
Standard guide for Raman shift standards for spectrometer,
uncertainty for a non-doped Si substrate
PA, 2014.
[7] The European Pharmacopoeia Commission, European (NMIJ CRM 5606-a). J Raman Spectrosc. 2020;51:
Pharmacopoeia, 9th ed., Strasbourg 2016. 2496–2504. https://doi.org/10.1002/jrs.6003
[8] J. A. Stone, J. E. Decker, P. Gill, P. Juncar, A. Lewis,
G. D. Rovera, M. Viliesid, Metrologia 2009, 46, 11.

You might also like