A new approach to develop the Raman carbonaceous material geothermometer for low-grade metamorphism using peak width



The Raman spectra of carbonaceous material (CM) from 19 metasediment samples collected from six widely separated areas of Southwest Japan and metamorphosed at temperatures from 165 to 655°C show systematic changes with metamorphic temperature that can be classified into four types: low-grade CM (c. 150–280°C), medium-grade CM (c. 280–400°C), high-grade CM (c. 400–650°C), and well-crystallized graphite (> c. 650°C). The Raman spectra of low-grade CM exhibit features typical of amorphous carbon, in which several disordered bands (D-band) appear in the first-order region. In the Raman spectra of medium-grade CM, the graphite band (G-band) can be recognized and several abrupt changes occur in the trends for several band parameters. The observed changes indicate that CM starts to transform from amorphous carbon to crystallized graphite at around 280°C, and this transformation continues until 400°C. The G-band becomes the most prominent peak at high-grade CM suggesting that the CM structure is close to that of well-crystallized graphite. In the highest temperature sample of 655°C, the Raman spectra of CM show a strong G-band with almost no recognizable D-band, implying the CM grain is well-crystallized graphite. In the Raman spectra of low- to medium-grade CM, comparisons of several band parameters with the known metamorphic temperature show inverse correlations between metamorphic temperature and the full width at half maximum (FWHM) of the D1- and D2-bands. These correlations are calibrated as new Raman CM geothermometers, applicable in the range of c. 150–400°C. Details of the methodology for peak decomposition of Raman spectra from the low to medium temperature range are also discussed with the aim of establishing a robust and user-friendly geothermometer.


As organic matter is heated, hydrogen, oxygen, and nitrogen are expelled, and the carbon atoms become progressively organized into stacked graphene sheets. This range of materials is referred to as carbonaceous material (CM). Numerous petrological studies have focused on the close relationship between the crystallinity of CM and metamorphic temperature. This crystallinity–temperature (T) correlation has been documented by various analytical techniques such as X-ray diffraction (XRD, e.g. French 1964; Landis 1971; Itaya 1981; Tagiri 1981), vitrinite reflectance (e.g. Kisch 1980; Mori & Taguchi 1988; George et al. 2001), and high-resolution transmission electron microscopy (HRTEM, e.g. Buseck & Huang 1985; Jehlicka & Rouzaud 1990; Beyssac et al. 2002b; Nakamura & Akai 2013). In recent years, laser Raman spectroscopy has also become a widely used approach to characterize CM structure. Important advantages of this method are: (i) it enables an in-situ, non-destructive analysis of target grains buried in the thin sections or rock chips; and (ii) it is sensitive to amorphous structures. Therefore, Raman spectroscopy can be used to characterize not only crystallized graphite but also amorphous carbon (e.g. coal, kerogen) without changing the structure of the sample (e.g. Jehlička et al. 2003).

Raman spectra of CM in metasediments exhibit a systematic change with metamorphic grade (e.g. Pasteris & Wopenka 1991; Wopenka & Pasteris 1993; Yui et al. 1996). From these Raman spectra, previous workers evaluated the metamorphic grade using spectral features such as intensity (i.e. height) ratio, area (i.e. integrated intensity) ratio, and width (i.e. full width at half maximum; FWHM) between several different peaks. These results indicate that the CM structure records the peak metamorphic temperatures and does not revert to its original state during the retrograde history. Beyssac et al. (2002a) first proposed an empirical geothermometer using the Raman spectrum of CM. They analyzed 54 samples from 10 regional metamorphic terrains over the world, and suggested that the geothermometer applicable is within the temperature range of 330–650°C. In addition to the regional metamorphic rocks, Aoya et al. (2010) showed that this geothermometer can also be applied to contact metamorphic rocks around a large igneous intrusion body. Rahl et al. (2005) proposed an alternative Raman CM geothermometer that is applicable to the lower range of metamorphic temperatures. This geothermometer combines the results of Beyssac et al. (2002a) with additional data from the Olympics Mountains, USA. The applicable range of this thermometer is given as 100–700°C. In the thermometry of Rahl et al. (2005), Raman spectra of CM in low-grade rocks were decomposed into four bands as proposed by Beyssac et al. (2002a). However, the Raman spectra of CM for the low-T range are more complex than those of the high-T range, and applying the same methodology of peak decomposition to low-grade rocks leads to the problem of non-unique solutions. As a way of overcoming this difficulty, Lahfid et al. (2010) proposed a new Raman CM geothermometer for low-grade rocks in the Glarus Alps incorporating a robust peak-fitting method for low-grade CM. The proposed geothermometer is applicable in a range of c. 200–320°C. However, this geothermometer was constructed using the samples from only one metamorphic region, and, as stated by the authors, it is unclear how far this method can be accurately applied to other areas. In addition, the estimated metamorphic temperatures of the samples used for the calibration of this geothermometer all lie within a restricted range, and it is unclear if the proposed peak-fitting method is applicable to temperatures < 200°C and >320°C.

As outlined above, there have been several studies that formulate the correlation between various parameters of Raman spectra of CM and metamorphic temperature. However, the structural changes that take place in CM as a result of metamorphic temperature increasing from the low- to high-T range have not been discussed in detail. In addition, none of the proposed Raman CM geothermometers are universally applicable especially in the low-T range. The main geological framework of convergent margins and island arcs, such as the Japan archipelago, commonly consist of widespread accretionary complexes that have experienced very low- to low-T metamorphism of <300°C. Developing methods to determine this thermal structure has an important role to play in documenting the structure and hence also the growth history of accretionary complexes and island arcs. However, few of the available methods are capable of accurately estimating metamorphic temperatures in the low-T range. It is also desirable to develop low-T geothermometers that are easy to use and don't involve lengthy sample preparation such as mineral separation or acid dissolution. The aims of the present study are: (i) to document the changes in Raman spectra of CM traversing from very low to high metamorphic temperature and discuss the associated structural changes of CM; and (ii) to construct a robust and user-friendly Raman CM geothermometer with particular focus on its applicability in the low-T range. To achieve these aims, we analyzed 19 samples metamorphosed at temperatures from 165 to 655°C collected from widely spaced locations in Southwest Japan. In the following, we discuss the most effective method of peak-fitting for measured Raman spectra of CM especially for low-T range, and then report the change of intensity ratio, area ratio, center position, and full width at half maximum (FWHM) of peaks with metamorphic temperature. Based on these data, we discuss the structural change of CM and propose a new Raman CM geothermometer applicable to very low- to medium-T metamorphic range using the FWHM of disordered bands that show linear correlation with metamorphic temperature.

Sample Selection

Eleven of the analyzed samples were collected for this study. These samples were selected from the Shirataki, Kitagawa, Kure, and Nobeoka areas in Southwest Japan (Fig. 1) and their metamorphic temperature range is from very low- to medium-T (c. 150–380°C). The Shirataki area is part of the subduction-type Sambagawa metamorphic belt. The Kitagawa, Kure, and Nobeoka areas are parts of accretionary complexes found in the Kurosegawa, Chichibu, and Shimanto belts. For our sample selection, we focused on massive homogenous rocks that are the least affected by deformation. We also incorporate the data sets reported in Aoya et al. (2010) for eight samples from the Kasuga and Daimonji contact metamorphic aureoles (Fig. 1). These samples cover a medium- to high-T range of 340–655°C. More detailed background information for each sample is summarized in Table 1.

Figure 1.

Locations of the Kasuga, Daimonji, Shirataki, Kitagawa, Kure, and Nobeoka areas and distributions of the Ryoke, Mino-Tamba, Sambagawa, Chichibu, Shimanto, and Kurosegawa belts in Southwest Japan. Abbreviation: MTL, Median Tectonic Line.

Table 1. Background information for each sample
SampleLocalitySettingProtolith ageMetamorphic ageRock typeT (°C)GeothermometerReference
Kr6KureShimanto beltLate CretaceousEoceneMudstone165 ± 35Vitrinite reflectanceMukoyoshi et al. (2006)
Kr4KureShimanto beltLate CretaceousEoceneMudstone200 ± 30Vitrinite reflectanceMukoyoshi et al. (2006)
Kr1KureShimanto beltLate CretaceousEoceneMudstone230 ± 30Vitrinite reflectanceMukoyoshi et al. (2006)
NB02NobeokaShimanto beltEarly CretaceousEoceneMudstone233 ± 30Vitrinite reflectanceMukoyoshi et al. (2007)
NB01NobeokaChichibu beltMiddle-Late JurassicEocene?Silty shale270 ± 30Vitrinite reflectanceMukoyoshi et al. (2007)
RS5KitagawaKurosegawa beltLate PermianEarly Jurassic?Shale277 ± 30Illite crystallinityHara et al. (2013)
RS4KitagawaKurosegawa beltLate PermianEarly Jurassic?Shale283 ± 30Illite crystallinityHara et al. (2013)
RS9KitagawaKurosegawa beltLate PermianEarly JurassicShale301 ± 30Illite crystallinityHara et al. (2013)
NB06NobeokaShimanto beltLate CretaceousEoceneShale305 ± 30Vitrinite reflectanceMukoyoshi et al. (2007)
TK22ShiratakiSambagawa beltCretaceousLate CretaceousPelite330 ± 30Mineral assemblageEnami et al. (1994)
K08KasugaMino terraneJurassicLate CretaceousPsammite340 ± 25Chl geothermometerAoya et al. (2010)
TK15ShiratakiSambagawa beltCretaceousLate CretaceousPelite380 ± 20Mineral assemblageEnami et al. (1994)
N33DaimonjiTamba beltJurassicLate CretaceousPelite387 ± 36Thermal modellingAoya et al. (2010)
N30DaimonjiTamba beltJurassicLate CretaceousPelite410 ± 30Grt-Chl geothermometerAoya et al. (2010)
N29DaimonjiTamba beltJurassicLate CretaceousPelite443 ± 29Thermal modellingAoya et al. (2010)
N27DaimonjiTamba beltJurassicLate CretaceousPelite475 ± 26Thermal modellingAoya et al. (2010)
N9DaimonjiTamba beltJurassicLate CretaceousPelite550 ± 19Thermal modellingAoya et al. (2010)
N3DaimonjiTamba beltJurassicLate CretaceousPelite604 ± 25Thermal modellingAoya et al. (2010)
K04KasugaMino terraneJurassicLate CretaceousPelite655 ± 25Cal-dol solvusWada and Suzuki (1983)

Samples from the Shirataki area (TK22 and TK15) were collected from the chlorite zone in the Sambagawa metamorphic belt. TK22 was sampled from a lower-grade pumpellyite-stable region (lower-grade chlorite zone) and TK15 was from a higher-grade pumpellyite-free region (higher-grade chlorite zone; Nakajima 1982). Estimated PT conditions in the lower- and higher-grade chlorite zones are 6 kbar at 330 ± 30°C and 380 ± 20°C, respectively (Enami et al. 1994). Both TK22 and TK15 are metapelite and were metamorphosed during the Cretaceous (e.g. Isozaki & Itaya 1990; Wallis et al. 2009).

Samples from the Kitagawa area of the Kurosegawa belt (RS5, RS4, and RS9) belong to the Late Permian accretionary complex (Ishida 1985; Yamakita 1986). RS5 and RS4 were sampled from a mélange unit and RS9 was from a phyllite unit. The metamorphic age of the sample RS9 was estimated at 194.9 ± 4.1 Ma using K–Ar isotopic dating of illite (Hara et al. 2013). The metamorphic temperatures for the samples RS5, RS4, and RS9 are estimated to be 277 ± 30, 283 ± 30, and 301 ± 30°C, respectively, using illite crystallinity data (Hara et al. 2013).

Samples from the Kure area (Kr1, Kr4, and Kr6) belong to the Cretaceous Shimanto accretionary complex (Mukoyoshi et al. 2006). Kr1 was sampled from the Kure mélange, and Kr4 and Kr6 were from the Nonokawa Formation. The Kure mélange and Nonokawa Formation are cut by a NW–SE en-echelon fault system with a large cumulative displacement (Mukoyoshi et al. 2006). The samples from the Kure area are best described as mudstone. Vitrinite reflectance data indicate metamorphic temperatures of Kr6, Kr4, and Kr1 are 165 ± 35, 200 ± 30, and 230 ± 30°C, respectively (Mukoyoshi et al. 2006).

Samples from the Nobeoka area (NB01, NB02, and NB06) were collected from the Chichibu and Shimanto accretionary complexes (Mukoyoshi et al. 2007). NB01 was sampled from the Jurassic Chichibu accretionary complex whereas NB02 and NB06 were from the Saiki subgroups of the Cretaceous Shimanto accretionary complex. Radiolarian fossils give the depositional age of the Saiki subgroup as Cretaceous (Teraoka et al. 1990). The samples from these areas are intermediate between shale and mudstone. The metamorphic temperatures of NB02, NB01, and NB06 are estimated to be 233 ± 30, 270 ± 30, and 305 ± 30°C, respectively, using vitrinite reflectance data (Mukoyoshi et al. 2007).

Sample Preparation and Spectra Acquisition

Raman spectra of CM were obtained with a quasi-backscattering geometry using a Nicolet Almega XR (Thermo Scientific, Yokohama, Japan) with a 532 nm Nd-YAG laser passed through a confocal microscope (Olympus, BX51: Olympus, Tokyo, Japan) with a 100× objective (Olympus Mplan-BD 100X, NA = 0.90). The laser power at the sample surface was set at 1–3 mW. The scattered light was collected by backscattered geometry with a 25 μm pinhole and a holographic notch filter, and finally dispersed using a 2400 lines/mm grating and analyzed by a Peltier cooled CCD detector of 256 × 1024 pixels (Andor Technology, Belfast, Northern Ireland). Spatial resolution is about 1 μm, and wavenumber resolution is about 1 cm−1. Frequencies of the Raman bands were calibrated by monitoring the position of plasma lines from a Ne lamp.

The Raman spectra of CM were obtained using standard petrographic thin sections. Several studies reported that the mechanical polishing dramatically influences the Raman spectra of CM (Beyssac et al. 2003a; Nasdala et al. 2004). To avoid the effect of this mechanical damage on the results, we carefully selected CM grains that are embedded within other transparent minerals and do not lie at the sample surface. Acquisition time was 30 s and at least 30 different CM grains were measured for each sample. Representative Raman spectra of CM for several samples are shown in Figure 2.

Figure 2.

Raman spectra of CM of the several representative samples used in the present study (left) and the methodology used in each cases for the spectrum decomposition (Fittings A to G: right). Asterisks in the band name in (e) to (g) mean that the center position of that band was fixed.

The orientation of CM with respect to the incident laser beam can have a significant effect on the Raman spectrum, especially in the case of well-crystallized graphite (Katagiri et al. 1988; Wang et al. 1989; Compagnini et al. 1997). On the other hand, Aoya et al. (2010) compared the measurement results of Raman spectra with the laser irradiation parallel and perpendicular to the c-axis, and they concluded that the direction of irradiation makes no significant difference to the R2 ratio [= D1/(G + D1+D2)area] in the metamorphic temperature range of 340–650°C. We also checked other parameters (intensity ratio, area ratio, center position, and FWHM) in c-axis parallel and perpendicular orientations for several samples in the Daimonji area and no significant differences were detected. In addition, the effect of polarization direction on the CM in the lower temperature range is considered to be negligible because such CM has a polycrystalline structure and the crystallite size is very small (<10 nm, e.g. Ferrari & Robertson 2000). For these reasons, the orientation of CM was not taken into account in the present study. Sedimentary clastic CM grains in the low-T samples whose Raman spectrum profile is distinctly different from the others were not measured.

Peak Decomposition of Raman Spectra

In this section, we focus on the methodology for peak decomposition of Raman spectra of CM. Although many studies have been carried out concerning Raman spectra of CM (e.g. Tuinstra & Koenig 1970; Rouzaud et al. 1983; Ferrari & Robertson 2000; Beyssac et al. 2003a,b), there is not yet a clear consensus on the best way to treat CM Raman bands especially for lower grade CM (i.e. amorphous carbon). This difficulty reflects the complexity of CM structure at low metamorphic grades. Establishing a reliable and reproducible method for the decomposition of Raman spectrum of CM is the largest single problem that needs to be overcome to develop the Raman CM geothermometer applicable over a wide range of metamorphic temperatures. Thus, we first review results from previous studies of Raman spectra in the section ‘Generalities’ and then examine the methodology for peak decomposition that provides a robust solution set over a wide temperature range in the section ‘Spectral analysis’. To ensure our proposed geothermometer is easy to use for a wide field of researchers, we provide a flowchart that gives a clear practical guide for peak decomposition (Fig. 3). For reference, a more detailed discussion of the peak decomposition of Raman spectra is given in Supporting Information Appendix S1.

Figure 3.

Flowchart for the peak fitting procedure used in this study.


The Raman spectrum of CM is composed of first-order (1000–1800 cm−1) and second-order (2500–3100 cm−1) regions (e.g. Nemanich & Solin 1979; Pasteris & Wopenka 1991; Beyssac et al. 2002a). This study focuses on the first-order region, which is associated with up to five discriminative bands for CM (G, D1, D2, D3, and D4; Fig. 2).

The crystal structure of well-crystallized graphite belongs to the P63/mmc (D6h4) space-group symmetry with vibration modes 2E2g + 2B2g + E1u + A2u. The E2g1 and E2g2 modes of the graphite structure are Raman active, but the E2g1 mode occurs close to the Rayleigh line and thus only one Raman band is observed at around 1580 cm−1 (e.g. Wang et al. 1990). This band is referred to as the G-band (‘G’ denotes graphite; Fig. 2). The G-band corresponds to doubly-degenerated (LO and iTO) E2g phonons at the Brillouin zone center and is assigned to in-plane C–C stretching vibration (e.g. Tuinstra & Koenig 1970; Song et al. 1976; Ferrari & Robertson 2000). With increasing degree of disorder, additional Raman bands for CM appear at around 1350 cm−1 (D1-band), 1620 cm−1 (D2-band), 1510 cm−1 (D3-band), and 1245 cm−1 (D4-band) as shown in Figure 2. The naming of these as ‘D-bands’ refers to their association with a disordered crystal structure or the presence of crystal defects. However, the origin and vibrational mode of these ‘D-bands’ are contentious and several plausible interpretations have been proposed (e.g. Tuinstra & Koenig 1970; Robertson 1986; Wang et al. 1990; Cuesta et al. 1994; Jawhari et al. 1995; Dippel et al. 1999; Escribano et al. 2001). The center position and relative intensity of these D-bands change with excitation wavelength (λ0) due to the resonance effect (e.g. Wang et al. 1990; Matthews et al. 1999; Sato et al. 2006).

For amorphous carbon (Fig. 2e–g), several different peak-fitting procedures have been proposed. In some studies, the broad band at around 1600 cm−1 (referred to as the O-band or G-band in a broad sense) is treated as a single peak (e.g. Tuinstra & Koenig 1970; Wopenka & Pasteris 1993; Rahl et al. 2005), whereas in other studies it is decomposed into two bands: a G-band in a strict sense and a D2-band (e.g. Sadezky et al. 2005; Lahfid et al. 2010). Whether the peak at around 1600 cm−1 (e.g. Fig. 2e–g) is treated as one peak or two causes significant differences in the best approach to obtain reasonable solutions to the peak decomposition. From a structural point of view, the G-band in the broad sense occurs at all sp2 sites in both rings and chains of amorphous carbons, and does not require the presence of six-fold aromatic rings (e.g. Ferrari & Robertson 2000). On the other hand, in the strict sense, G-band should be defined as representing the fundamental vibration mode of crystallized graphite as mentioned above (e.g. Fig. 2a). Hereafter we use the term ‘G-band’ in the strict sense in order to distinguish the G-band s.s. from the D2-band present at a similar wavenumber.

Spectral analysis: methodology for peak decomposition

The Raman spectra of CM were decomposed into several peaks using the computer program PeakFit 4.12 (SeaSolve Software Inc., Massachusetts, USA) with a pseudo-Voigt function (Gaussian-Lorentzian Sum). The spectra were corrected for the fluorescence background by subtracting a linear baseline in the spectral range of 1000–1750 cm−1. The area (integrated intensity), intensity (height), center position, and full width at half maximum (FWHM) were determined for all bands. We examined the effects of difference in the method of peak decomposition for samples metamorphosed at different temperatures (see Supporting Information Appendix S1 for details). In the following, we describe the most effective decomposition method for establishing a thermometer. The methodologies (fittings A, B, C, D, E, F, and G) are illustrated in Figure 2 and summarized in the flowchart of Figure 3, which presents the procedure for determining the appropriate method for samples whose metamorphic temperatures are unknown.

Fittings A, B, C, and D

In the metamorphic temperature range of 340–655°C, up to four bands can be recognized in the first-order region (1000–1800 cm−1): the G-, D1-, D2-, and D3-bands (fittings A, B, C, and D; Figs 2a–d and 3). In these D4-band-absent cases, a unique solution for peak decomposition can be obtained without fixing any parameters (e.g. Beyssac et al. 2002a).

Fitting E

In the metamorphic temperature range of < c. 340°C, the D4-band appears at around 1245 cm−1. The peak at around 1600 cm−1 can be divided into G- and D2-bands for metamorphic temperatures of c. 300–340°C and we can obtain a unique fit to the spectra using the five bands, G-, D1-, D2-, D3-, and D4-bands (method E; Figs 2e,3). The center position of the D4-band was fixed at 1245 cm−1. Analyses of samples RS9 (301 ± 30°C) and TK22 (330 ± 30°C) show that fixing the D4-band has no significant effect on other key band parameters such as center position and FWHM (see Supporting Information Appendix S1; Fig. A2).

Fittings F and G

In the metamorphic temperature range of < c. 300°C, the intensity of the D4-band increases (Fig. 2f,g). In addition, the peak at around 1600 cm−1 becomes broad and it cannot be decomposed into G- and D2-bands with a Voigt function (Wopenka & Pasteris 1993; Beyssac et al. 2003b). Hereafter, the peak at around 1600 cm−1 will be referred to as the GL peak (subscript ‘L’ represents ‘low’) regardless of if the peak can be decomposed into G- and D2-bands or not. To obtain insights into the condition of G- and D2-bands of CM in the low-T range, we examined the change in the profile of the GL peak with metamorphic temperature.

Figure 4 shows the relation between the FWHM of the GL peak and the intensity ratio of GL/D1 for the series of samples from Kr6 to K08, which have estimated metamorphic temperatures in the range from 165 to 340°C. Here, we treat the peak at around 1600 cm−1 as a single GL peak even for the samples whose Raman spectrum of CM can be decomposed into G- and D2-bands (i.e. samples RS9, NB06, TK22, and K08). The values of the GL/D1 intensity ratio of the samples from Kr6 to RS5 are almost constant around 1.7 to 2.2 and the FWHM values of the GL peak gradually decrease from 69 to 34 cm−1 with increasing metamorphic temperature. This simple decreasing trend of FWHM clearly changes in higher-temperature samples (RS4 or above). For the samples from RS4 to K08, the values of the GL/D1 intensity ratios decrease from 1.8 to 0.5 and, more importantly, the FWHM values of the GL peak ‘increase’ from 35 to 55 cm−1 with increasing temperature. The abrupt increase of FWHM of the GL peak at around 280°C can be interpreted to represent the transition from one dominant peak to two coexisting peaks. It is, therefore, necessary to use two different decomposition methods for the lower- and higher-temperature sides.

Figure 4.

Assessment of the peak at around 1600 cm−1 peak (GL peak) in the low temperature range. Correlation of FWHM of GL peak and the intensity ratio of GL/D1 for each CM grain in the samples metamorphosed at temperatures from 165 to 340°C are shown. Numbers in the parentheses next to the sample numbers in the legend are the mean values of independently estimated metamorphic temperatures. One band (D2-band) dominates the GL peak in the temperature range of c. 150 to 280°C, and two bands (G- and D2-bands) coexist at GL peak in the temperature range of around > 280°C. The value of 1.5 for the intensity ratio of GL/D1 is critical when determining whether fitting F or G (Fig. 2f,g & 3) should be applied for the peak decomposition.

In the present study, the GL peak of the Raman spectra for most samples < c. 280°C was treated as a single peak and the spectrum was decomposed into four bands (fitting G; Figs 2g,3). This allows a unique fit to the spectrum to be derived. The single GL peak for the samples < c. 280°C will, hereafter, be referred to as the D2-band. The validity of this treatment will be discussed further in this paper. In contrast, in the restricted temperature range of c. 280–300°C, the GL peak was treated as a combination of G- and D2-bands and, in this case, the center position of the G-band was fixed at 1593 cm−1 with a Lorentz function (fitting F; Figs 2f,3) in order to obtain a unique solution. The value, 1593 cm−1, is similar to that of the G-band positions obtained in unique solutions for samples c. 300°C (Fig. 5c). A suitable criterion to decide whether fitting F or G should be applied can be derived from the intensity ratio of GL /D1 (Fig. 4). If the GL /D1 intensity ratio is <1.5, the spectrum should be decomposed using fitting F (Fig. 2f), and if the GL /D1 intensity ratio is >1.5, the spectrum should be decomposed using fitting G (Fig. 2g). In addition to the above considerations, the center position of the D3-band was fixed at 1510 cm−1 in fitting G and that of the D4-band was fixed at 1245 cm−1 in fitting F and G (Fig. 3). Further details of the peak decomposition are given in Supporting Information Appendix S1.

Figure 5.

Parameters of decomposed spectra of the measured CM as a function of the independently estimated temperature. (a) Intensity ratio and area ratio of D2/D1. (b) Intensity ratio and area ratio of D1/G. (c) Center positions of D1-, D2-, and G-bands. (d) FWHM of D1-, D2-, and G-bands. Bold dashed lines divide the types of CM from low- to high-grade.


Spectral parameters were determined by peak fitting as discussed above and the corresponding data set are given in Figure 5 and Table 2. The data set of Table 2 includes mean values and standard deviations for center position, FWHM of the D1-, D2-, and G-bands, and values of the D2/D1 and D1/G intensity ratios. At least 30 CM grains were measured for each sample and the data deviating more than 2σ from the mean were omitted to eliminate the possibility of including clastic CM. The processed data sets were used to determine the final means and standard deviations that are listed in Table 2. Although the known metamorphic temperature of each sample also has errors (Table 1), they are not shown in Figure 5 to avoid complication. The histograms of FWHM for D1- and D2-bands are shown in Figure 6. In the following, the samples are designated by the expression ‘T##’, where ## indicates the known metamorphic temperature (°C) of the sample.

Table 2. Results of Raman spectral analysis. Mean values (mean) and standard deviations (1σ) are listed
SampleD1 centerD1 FWHMG centerG FWHMD2 centerD2 FWHMD2/D1 InetnsityD1/G intensity
Mean1 σMean1 σMean1 σMean1 σMean1 σMean1 σMean1 σMean1 σ
  1. *Not listed because there are few CMs containing the band. **Fixed at this position with Lorenz function.
Kr61355.17.1136.77.6  1600. 
Kr41352.85.8131.46.0  1600. 
Kr11349.33.8122.04.3  1602. 
NB021348.92.7112.24.1  1600. 
NB011338.  1602. 
RS51339.91.497.02.5  1603. 
RS41335.61.495.52.6* * 1603.21.537.21.21.660.06* 
K041353.31.631.813.51581.10.615.10.9* * * 0.030.03
Figure 6.

Histograms showing the frequency distribution of the FWHM values of (a) D1- and (b) D2-bands (FWHM-D1 and FWHM-D2). Frequency is indicated by the ratio against total number of measurements. Although the data more than 2σ away from the mean value were omitted in the text, all measured data are indicated in these histograms. Dashed line indicates the final mean value and gray shaded area indicates the 2σ area in all data.

Intensity and area ratio

The D2/D1 intensity and area ratios of the six lowest-T samples, T165 to T277, are almost constant at around 1.8–2.1 and 0.6–0.9, respectively (Fig. 5a). For these data the D2/D1 intensity ratios are >1.0, while the D2/D1 area ratios are <1.0, meaning that the D2-band is higher and narrower than the D1-band (Fig. 2g). At temperatures >c. 280°C, the values of the D2/D1 intensity and area ratios start to decrease, and those of the samples T330 or above become nearly constant at around 0.1–0.3 and 0.1, respectively (Fig. 5a). The D2/D1 intensity ratio of every CM grain in the sample T283 (RS4) shows D2/D1 > 1.0 and that in the sample T301 (RS9) shows D2/D1 < 1.0. This means that in the temperature range of > c. 300°C, the D1-band has greater height than the D2-band (Fig. 2c-e). For the sample T604, the D2-band could not be recognized in most of the measured spectra (Fig. 2b) and the values of the D2/D1 intensity and area ratios are nearly zero (Fig. 5a). For the sample T655, most spectra of CM grains show no discernable D1- and D2-bands (Fig. 2a), hence the values of D2/D1 ratios are not shown in Figure 5a.

The values of the D1/G intensity and area ratios show similar trends that decrease with increasing metamorphic temperature from 2.0 to 0.0 and 2.8 to 0.0, respectively (Fig. 5b). The values of the D1/G intensity ratios of the samples T301 to T387 are greater than 1.0, which means the D1-band is higher than the G-band (Fig. 2d,e). In the sample T410, the D1/G intensity ratios for different measurement spots are variable with values above and below 1.0. For samples T443 or more, the D1/G intensity ratio is less than 1.0, which means the G-band is higher than the D1-band (Fig. 2b,c). The absolute intensities of Raman spectra decrease with increasing metamorphic temperature.

Center position

For the seven lowest-T samples, T165 to T283, the wavenumber of the center position for the D1-band decreases with increasing metamorphic temperature from 1355 to 1336 cm−1 (Fig. 5c). Standard deviations of these values are large because the associated D1-band is broad (Fig. 2g). In contrast, the center positions of the D2-band are almost constant at around 1600 cm−1 in these samples (Fig. 5c). For samples T301 to T387, the values of the center positions of the D1-bands increase from 1338 to 1350 cm−1 and those of the D2-band also increase from 1606 to 1620 cm−1 with increasing metamorphic temperature (Fig. 5c). For samples T410 or more, the values of the center positions of the D1- and D2-bands are almost constant at around 1350 cm−1 and 1620 cm−1, respectively (Fig. 5c). For the sample T655, the error for the value of the D2-band is relatively large (Fig. 5c) because the intensity of the D2-band is weak and the center position is not clear.

The G-band was included in the peak decomposition at > c. 280°C. For the two samples T283 and T301, the values of the center positions of the G-bands were fixed at 1593 cm−1 with a Lorentz function. The center positions of the G-band for samples T305, T330, and T340 are almost constant at around 1593 cm−1 although the G-band for these three samples was not fixed (Fig. 5c). For samples T380 to T655, the center positions of G-band slightly shifts to lower wavenumbers and becomes almost constant at around 1580 cm−1 for samples T410 or more (Fig. 5c).

Full width at half maximum (FWHM)

The FWHM values of the D1-band decrease linearly from 137 to 43 cm−1 for samples T165 to T387, and become almost constant at around 40 cm−1 for samples T410 or more (Fig. 5d). The FWHM values of the D2-band decrease almost linearly from 56 to 7 cm−1 for samples T165 to T604 (Fig. 5d). Overall the FWHM values of the G-band monotonously decrease from 56 to 15 cm−1 for samples T301 to T655, although the gradient changes at around 400°C (Fig. 5d).

Structural Evolution of Carbonaceous Materials

The CM can be classified into four types based on the change in the features of the Raman spectrum: low-grade CM, medium-grade CM, high-grade CM, and well-crystallized graphite. Here, we describe the characteristics of each type and consider the structural evolution of CM with increasing metamorphic temperature.

Low-grade CM

The shapes of the Raman spectra of CM for samples T165 (Kr6) to T277 (RS5) (Fig. 2g) are similar to those of amorphous carbon such as coal or kerogen (e.g. Zerda et al. 1981; Wopenka & Pasteris 1993). The low crystallinity of samples in this temperature range suggests that the G-band s.s., which corresponds to the vibration mode of crystallized graphite, probably cannot be defined, or even if it does exist it is too small to be clearly distinguishable. In this study, therefore, we regard the Raman peak at around 1600 cm−1 (GL peak) of low-T samples as a single D2-band. If this peak is treated as a G-band as in the previous studies, a discontinuity for the trend of FWHM of G-band occurs at around 280°C. The obtained smooth trend for the FWHM of D2-band (Fig. 5c), therefore, at least partly justifies the treatment of the peak at around 1600 cm−1 (GL peak) as D2-band in the low temperature range. This result can be taken to suggest that the vibration mode of the D2-band attributes to the same vibration mode of the G-band (i.e. the E2g mode) but it may reflect the presence of ‘incomplete’ graphite structures with a defect in the 6-fold ring. We refer to this type of CM without a recognizable G-band as low-grade CM. The corresponding metamorphic temperature for the low-grade CM is around 150–280°C (Fig. 5).

Medium-grade CM

Figure 4 shows that the abrupt increase of FWHM of GL peak is due to appearance of the G-band s.s., which occurs at > c. 280°C. It may imply this is the temperature range over which the structure of CM starts to transform from amorphous carbon to crystallized graphite in metamorphic rocks. With increasing temperature from T283 (RS5) to T387 (N33), the center positions of D1- and D2-bands show clear shift to higher wavenumbers (Fig. 5c) and both D2/D1 intensities and area ratios show abrupt decreases (Fig. 5a). These results are in agreement with the suggestion above that the transition of CM occurs in this temperature range. We refer to this type of CM metamorphosed at around 280 to 400°C as medium-grade CM.

High-grade CM

The D1/G intensity ratio becomes less than 1.0 at > c. 400°C (Fig. 5b), which means the intensity of the G-band becomes stronger than that of the D1-band. At the same time, at > c. 400°C the center positions of the D1-, D2-, and G-bands and the FWHM of D1-band become nearly constant (Fig. 5c,d). These changes suggest that the transition of amorphous carbon to crystallized graphite is complete at around 400°C. Above this temperature, the CM grain can be termed ‘crystallized graphite’ and we refer to this type of material as high-grade CM.

Well-crystallized Graphite

Many CM grains in the highest temperature sample, T655 (K04), show spectra consisting of a single G-band without recognizable D1- or D2-bands. This result indicates that CM reaches well-crystallized graphite at metamorphic temperatures of 650°C or more. This result is consistent with the study of Beyssac et al. (2002a) that shows the R2 ratio [= D1/(G + D1+D2)area] approaches 0 at metamorphic temperatures of 650°C or more.

Correlation Between FWHM and Temperature

Calibration of Raman CM Geothermometer

In this study, smooth decreasing trends of the FWHM of G-, D1-, and D2-bands with increasing metamorphic temperature were obtained (Fig. 5d). The FWHM values of D1- and D2-bands (FWHM-D1 and FWHM-D2) show almost linear relations with temperature in the range of approximately 150–400°C. In this temperature range, therefore, these parameters can be used to estimate the metamorphic temperature. Correlations between the metamorphic temperature and average FWHM-D1 and FWHM-D2 using 13 samples are shown in Figure 7. These correlations can be described by the following equations:

display math(Equation 1)
Figure 7.

FWHM of D1- and D2-bands vs. peak metamorphic temperature for the samples metamorphosed at 165 to 655°C. The diamond and triangle symbols denote D1- and D2-bands, respectively. The data of the samples indicated by filled symbols were not used for the calibration. The lines were drawn using a least squares method to find the best-fit curves for the samples from T165 (Kr6) to T387 (N33). The gray shaded area indicates the applicable temperature range for Equations (Equation 1) and (Equation 2) and dashed line indicates the approximate boundary of peak intensity ratio of D1- and G-bands, which can be used as an indicator for the upper limitation of the geothermometers.


display math(Equation 2)

The number of analyses used for calculation of the average FWHM values (Table 2) is at least 28 for each sample. The effective temperature range covered by these two thermometers is around 150–400°C (Fig. 7).

Assessment of Error

Figure 8 shows the distribution of TFWHM-D1 − Tknown and TFWHM-D2 − Tknown, where TFWHM is the temperature determined by application of Equations (Equation 1) and (Equation 2), and Tknown is the known temperature of the samples T165 (Kr6) to T410 (N30). As shown in Figure 8, more of the TFWHM-D2 − Tknown values exceed ±30°C than the results for TFWHM-D1 − Tknown. This trend probably reflects the fact that the gradient of Equation (Equation 2) is about 3 times steeper than that of Equation (Equation 1) (Fig. 7). This means that a small difference in FWHM-D2 generates a relatively large change in the value of TFWHM-D2. It can, therefore, be said that the temperature estimation using FWHM-D1 is generally more accurate than that using FWHM-D2. However, there is an exception for this general trend in the lowest temperature range of < c. 200°C. The diagonal lines shown in Figure 8 indicate the difference of TFWHM-D1 and TFWHM-D2. Most CM grains plot within the ±30°C bracket, but some CM grains of samples T165 (Kr6) and T410 (N30) plot outside of this bracket. In particular, a lot of CM grains of T165 (Kr6) plot beyond the limit of +30°C when using Equation (Equation 1) (Fig. 8). This means that the calculated temperature using Equation (Equation 1) overestimates the temperature compared to Equation (Equation 2). In addition, the TFWHM-D2 − Tknown values for most CM grains of T165 (Kr6) fall within the range of ±30°C. This fact implies that Equation (Equation 2) is more reliable than Equation (Equation 1) in the temperature range of 150–200°C.

Figure 8.

The distribution of TFWHM-D1 − Tkonwn and TFWHM-D2 − Tkonwn, where TFWHM is the temperature determined by Equations (Equation 1) and (Equation 2), and Tknown is the known temperature of the measured CM grains of the samples Kr6 to N30. Numbers in the parentheses next to the sample numbers in the legend are the mean values of independently estimated metamorphic temperatures. The diagonal lines are those of constant difference between TFWHM-D1 and TFWHM-D2.

For the values of TFWHM-D1 − Tknown, almost all CM grains other than the samples T165 (Kr6), T305 (NB06), T387 (N33), and T410 (N30) have values < ±30°C. The histograms of the first three samples show broad and bimodal distributions (Fig. 6a). The estimated errors of the samples T165 (Kr6) and T387 (N33) are over ±30°C (Table 1) and these two samples originally have variation. For the sample T305 (NB06), most TFWHM-D1 values are higher than those of Tknown, which is near the limit of temperature estimation of vitrinite reflectance and it might be lower than the exact value. The sample T410 is beyond the effective temperature range of this thermometer. It can, therefore, be said that the errors associated with temperature estimates using Equation (Equation 1) are generally less than ±30°C, but it becomes more than ±30°C if the FWHM histogram of the sample shows non-unimodal distribution or the estimated temperature is close to the end of the applicable temperature range of 150°C or 400°C. On the other hand, most values of TFWHM-D2 − Tknown scatter over a wide range around +40 to −50°C (Fig. 8) even excluding the data of T410 (N33), which is out of the applicable temperature range, The values of TFWHM-D2 are more dispersed than those of TFWHM-D1 (Fig. 8) and this result suggests that the error of Equation (Equation 2) is about ±50°C. We, therefore, recommend using Equation (Equation 1) as the principle method of temperature estimation with Equation (Equation 2) as a subsidiary method of confirmation. In addition, the above discussions suggest that care is needed for the temperature estimation for the sample whose FWHM histogram shows broad or bimodal distribution.

All CM grains of the sample T410 (N30) plot in the negative region for both temperature estimations in Figure 8. This means that Equations (Equation 1) and (Equation 2) tend to underestimate temperature for the samples metamorphosed at around 400°C. This temperature corresponds to the transition of CM from medium- to high-grade and several parameters show major changes (Fig. 5). In particular, the FWHM-D1 becomes constant at around 40 cm−1, which correspond to TFWHM-D1 = 392°C (Fig. 5d). So care is needed when using Equations (Equation 1) or 2 to estimate the metamorphic temperatures in the range of around 400°C or more. We suggest that the D1/G intensity ratio can be used as a guide to judge whether or not it is suitable to use Equations (Equation 1) and (Equation 2). The average D1/G intensity ratios of the samples T380 (TK15) and T387 (N33) are greater than 1.0 and that of sample T410 (N30) smaller than 1.0 (Fig. 5b). We, therefore, suggest that if the value of D1/G is >1.0, the sample is suitable for temperature estimation using Equations (Equation 1) and (Equation 2), but if this value is less than or equal to 1.0, other temperature estimates should be used. We recommend doing a cross-check using alternative Raman CM geothermometers (e.g. Beyssac et al. 2002a; Rahl et al. 2005; Aoya et al. 2010) which can be used in this temperature range.

To summarize the above discussions: (i) Equation (Equation 1) is generally more accurate than Equation (Equation 2) in the range 200–400°C, while Equation (Equation 2) is more accurate than Equation (Equation 1) in the range 150–200°C; (ii) the error of Equation (Equation 1) is around ±30°C and that of Equation (Equation 2) is around ±50°C provided the associated FWHM histogram shows neither a broad nor bimodal distribution; and (iii) the applicable temperature ranges for using Equations (Equation 1) and (Equation 2) are around 150–400°C and D1/G intensity ratio can be used as a guide for application in higher temperature limit.

Consistency of FWHM

It is important to consider the general applicability of any proposed new geothermometer. In previous studies, the ratios of peak areas or intensities have been adopted as the parameters for Raman CM geothermometers (Beyssac et al. 2002a; Rahl et al. 2005; Aoya et al. 2010; Lahfid et al. 2010). An advantage of using ratios is that the device dependence of spectral parameters is reduced, meaning the measurements can be made without standards. However, our new Raman CM geothermometer is based on a single value of FWHM and is not based on ratios. We explored the possibility of using some ratios of several parameters, but we were unable to find a good correlation with temperature. Another possible approach is to employ standard material to calibrate the FWHM, but it is difficult to make a robust CM standard material resistant to repeated laser irradiation. We therefore examined the interlaboratory variation by using four different machines with six different measurement conditions. These results show that while the center position of D1-band and intensity ratio of D1/G are appreciably affected by the laser wavelength, the FWHM measurements are robust and not subject to significant interlaboratory variation (see Supporting Information Appendix S2 for details).

There is also the possibility that the structures of CM are affected not only by temperature, but also by pressure, time, deformation, catalytic species, and precursor material. Especially, samples from the Kitagawa and Kure areas (RS5, RS4, and Kr1) were collected from a mélange unit, and the samples have potentially been affected by the deformation. However, we carefully selected massive homogeneous samples to minimize the effect of deformation. An additional issue is the effect that precursor material may have on Raman spectra of CM. Such an effect has been reported for lower-temperature metamorphism (Quirico et al. 2009; Bower et al. 2013). The broader and more dispersed histograms of FWHM-D1 and FWHM-D2 for the low-T samples T165 (Kr6) and T200 (Kr4) might reflect the difference of precursor material. Although other factors may also be significant, it is difficult to evaluate the effects of each of these factors from natural samples. We considered it more significant that despite such complications the FWHM-D1 and FWHM-D2 of all samples metamorphosed at around 150 to 400°C and collected from five different areas where the metamorphic condition, age, and rock types are different (Fig. 1, Table 1), all lie on clear linear correlation trends with metamorphic temperature (Fig. 7). This fact suggests that the Equations (Equation 1) and (Equation 2) give acceptable estimates of metamorphic temperature, and are widely applicable to CM hosted by metasediments.

Comparison With Other Raman CM Geothermometers

Raman CM geothermometers have already been proposed by Beyssac et al. (2002a), Rahl et al. (2005), Lahfid et al. (2010), and Aoya et al. (2010). The accuracy and precision of our new Raman CM geothermometers using Equations (Equation 1) and (Equation 2) were compared with those of previous studies (Fig. 9).

Figure 9.

Tknown v.s. TCM diagram, where Tknown is the known temperature and TCM is the temperature determined by Raman CM geothermometers. The error bars represent 1σ errors. (a) Comparison of Tknown and TCM using the parameters of FWHM-D1 and FWHM-D2 proposed in this study. (b) Comparison of Tknown and TCM using the parameters of R2 and R1-R2 after Aoya et al. (2010) and Rahl et al. (2005), respectively. (c) Comparison of Tknown and TCM using the parameters of RA1 and RA2 after Lahfid et al. (2010). Dashes indicate the Tknown − TCM concordance line and gray areas indicate the applicable temperature range for each geothermometers.

Beyssac et al. (2002a) and Aoya et al. (2010) proposed Raman CM geothermometers using the R2 parameter, which is the area ratio of D1/(G + D1 + D2). Although both give similar results, the equation given in Beyssac et al. (2002a) for the Raman CM geothermometer is linear whereas that in Aoya et al. (2010) is quadratic. The applicable temperature range is 330–650°C. Rahl et al. (2005) proposed a Raman CM geothermometer using a combination of the R1 and R2 parameters. The R1 parameter is the intensity (height) ratios of D1/G. The suggested appropriate temperature range is 100–700°C and Raman spectra were decomposed with a Voigt function into the G-, D1-, D2-, and D3-bands. Lahfid et al. (2010) proposed a Raman CM geothermometer using the RA1 and RA2 parameters, which are the area ratios of (D1 + D4)/(D1 + D2 + D3 + D4 + G) and (D1 + D4)/(D2 +D3 + G), respectively. The suggested applicable temperature range is c. 200–320°C and Raman spectra were decomposed with a Lorentz function using the G-, D1-, D2-, D3-, and D4-bands. To compare the accuracy of these Raman CM geothermometers in terms of the difference from the known metamorphic temperatures of the samples used in this study, the Raman spectra were re-analyzed following the procedures given in Rahl et al. (2005) and Lahfid et al. (2010). Since the peak at around 1600 cm−1 (GL peak) in low- to medium-grade Raman spectra cannot be uniquely decomposed into G- and D2-bands when the peak fitting procedure of Rahl et al. (2005) is used, the geothermometer of Rahl et al. (2005) does not give unique temperature estimates for the samples T165 to T301. Therefore, we compared the three solutions (relative peak intensities: G > D2, G  D2, G < D2; see Supporting Information Appendix S1 Figure A1(a) combinations II, III, and IV) and adopted the solution from which estimated temperature obtained is the nearest to the known temperature.

Figure 9 shows the relationship between temperatures determined by several different Raman CM geothermometers (TCM) and known temperature (Tknown) used in this study. The geothermometric calibrations of this study (using parameters of FWHM-D1or FWHM-D2), Rahl et al. (2005; using R1 and R2), Aoya et al. (2010; using R2), and Lahfid et al. (2010; using RA1 or RA2) were compared. Since the estimated temperatures using Raman CM geothermometers proposed by Beyssac et al. (2002a) and Aoya et al. (2010) are similar, only the results using the calibration of Aoya et al. (2010) are shown for the R2 geothermometer. Figure 9 (a) shows that TCM values of this study using either FWHM-D1 or FWHM-D2 correspond to Tknown within 1σ errors in the temperature range of around 150 to 400°C, while they deviate significantly from the TCM − Tknown concordance line at temperatures >400°C. TCM values derived from a combination of R1 and R2 parameters (Rahl et al. 2005) generally lie within 1σ errors of the TCM − Tknown concordance line (Fig. 9b). However, the 1σ errors of TCM value are relatively large, especially for metamorphic temperatures of <300°C and >500°C (Fig. 9b). TCM values derived from R2 parameter using the calibration of Aoya et al. (2010) fall close to the TCM − Tknown concordance line at temperatures > 330°C, while they overestimate in the lower temperature range of <330°C (Fig. 9b). TCM values derived from RA1 or RA2 parameters (Lahfid et al. 2010) plot close to the TCM − Tknown concordance line in the temperature range from around 200 to 300°C (Fig. 9c). In contrast, these TCM values overestimate the temperature by > +50°C at temperatures of <200°C and >300°C (Fig. 9c). As shown by the above review, our new Raman CM geothermometer shows several advantages over previously proposed methods. In particular, our new thermometer is applicable over a wide temperature range, from very low- to medium-T, and the estimates of metamorphic temperature are associated with only relatively small errors.


We used Raman spectra of natural CM metamorphosed under a wide range of temperatures from 165 to 655°C to examine the thermally induced structural evolution. The Raman spectra of CM can be classified into four types (Fig. 5): (i) low-grade CM that shows features characteristic of amorphous carbon at around 150 to 280°C; (ii) medium-grade CM that shows a gradual transition from amorphous carbon to crystallized graphite at around 280 to 400°C; (iii) high-grade CM that completes the transition and becomes crystallized graphite at around 400°C to 650°C; and (iv) well-crystallized graphite at > c. 650°C.

Our results show that the FWHM of the D1- and D2-bands decreases linearly with increasing metamorphic temperature in the range of around 150 to 400°C (Fig. 5d). These correlations can be used as the basis for new Raman CM geothermometers (Eqns (Equation 1) and (Equation 2); Fig. 7). The error associated with using FWHM-D1 is around ±30°C and the corresponding value for FWHM-D2 is around ±50°C, if the histogram of FHWM shows neither a broad nor bimodal distribution. The estimated temperature tends to be uncertain when it is close to the end of temperature range and it is desirable to perform a cross-check using a combination of both Equations (Equation 1) and (Equation 2) or previously proposed geothermometers. Because the calibration range of this new geothermometer traverses the transition of CM from amorphous carbon to crystallized graphite, careful attention needs to be paid in choosing the appropriate method of peak decomposition to obtain a reliable temperature estimate: in the present study we suggest an appropriate procedure (Figs 2,3).


We are grateful to M. Enami and to all members of petrology laboratory at Nagoya University for their useful advice. Critical comments and suggestions by two anonymous referees, L. Nasdala and B. Wopenka greatly helped to improve this contribution. We are grateful to S. Arai (Kanazawa University), T. Hirajima (Kyoto University), H. Hidaka (Hiroshima University), K. Hoshino (Hiroshima University), T. Nishiyama (Kumamoto University), and all members of the laboratory for their constructive help with Raman spectroscopic measurements. We also thank H. Mukoyoshi for providing detailed information about the sample locality of Kure area.