Micro-Raman Spectroscopy Assessment of Chemical Compounds of Mantle Clinopyroxenes

The composition of clinopyroxenes is indicative for chemical and physical properties of mantle substrates. In this study, we present the results of Raman spectroscopy examination of clinopyroxene inclusions in natural diamonds (n = 51) and clinopyroxenes from mantle xenoliths of peridotites and eclogites from kimberlites (n = 28). The chemical composition of studied clinopyroxenes shows wide variations indicating their origin in different mantle lithologies. All clinopyroxenes have intense Raman modes corresponding to metal-oxygen translation (~300–500 cm−1), stretching vibrations of bridging O-Si-Obr (ν11~670 cm−1), and nonbridging atoms O-Si-Onbr (ν16~1000 cm−1). The peak position of the stretching vibration mode (ν11) for the studied clinopyroxenes varies in a wide range (23 cm−1) and generally correlates with their chemical composition and reflects the diopside-jadeite heterovalent isomorphism. These correlations may be used for rough estimation of these compounds using the non-destructive Raman spectroscopy technique.


Introduction
Mineral inclusions in natural diamonds carry important information about the chemical physical conditions of the Earth's mantle. The sources of mineral inclusions in diamonds can be divided into three common parageneses: peridotite, eclogite, and rare websterite types. The major inclusions of the peridotite associations are olivine (forsterite), orthopyroxene (enstatite), garnet (Cr-rich pyrope), Cr-spinel (chromite), clinopyroxene (Cr-diopside), and Fe-Ni-sulfides; eclogite associations are usually represented by garnet (pyrope-almandine-grossular), clinopyroxene (omphacite), quartz (coesite), and Fe-sulfides [1]. Websteritic inclusions are generally presented by clinopyroxene and orthopyroxene, which associate with garnet and olivine in some rare cases.
The analysis of the chemical composition of inclusions traditionally involves the partial destruction of the samples, e.g., study of the chemical composition of inclusions in diamonds by EPMA (electron probe microanalysis) requires specific sample preparation, such as polishing or crushing of the host. Therefore, the methods of study of mineral inclusions using non-destructive methods are becoming more demanded [4][5][6][7][8][9]. Raman spectroscopy is an alternative high-resolution non-invasive non-destructive technique, which provides information about specific vibrations in solid materials reflecting the chemical and structural features of minerals, including those presented as inclusions in natural diamonds. In this study, we present new data on the Raman spectroscopic characteristics of clinopyroxenes inclusions in natural diamonds and clinopyroxenes from the series of peridotite and eclogite xenoliths from kimberlite. These data are used to define the correlations between Raman features of mantle clinopyroxenes and their chemical composition.

Materials and Methods
In total, 51 natural diamonds were selected from different occurrences of Yakutian diamondiferous province (Internatsional'naya (n = 9), Yubileynaya (n = 4), Udachnaya (n = 4), Mir (n = 1), Komsomolskaya (n = 1), Nyurbinskaya (n = 2) kimberlite pipes, Ebelyakh river deposits (n = 30)). The studied diamonds were represented by both octahedral and rounded dodecahedral crystals with the inclusions of peridotite or eclogite parageneses ( Figure 2). The sample preparation includes polishing of a diamond parallel to the (110) plane in order to recover the inclusions to the surface for further analysis of chemical composition by EMPA. Some inclusions were examined by Raman spectroscopy before and after the polishing and several fully extracted inclusions from diamond were studied to avoid the ambiguity as clinopyroxene inclusions may be affected by the residual strain of a diamond host [10,11]. In total, 28 clinopyroxene grains were also selected from peridotite (n = 8) and eclogite (n = 20) xenoliths from Udachnaya kimberlite pipe. The characterization of the clinopyroxenes from these samples is provided in [12,13]. Cr# of clinopyroxenes-lherzolitic, websteritic, and eclogitic compositional fields and the transition from peridotite to eclogite clinopyroxenes at Cr# 0.07-0.10 (gray vertical stripe) are taken from [2]; (b) Na 2 O vs. MgO contents in clinopyroxenes (A, B, C are the groups of eclogites are shown according to [3]). 1-peridotite clinopyroxene inclusions in diamonds, 2-clinopyroxenes from peridotites, 3-eclogite clinopyroxene inclusions in diamonds, 4-clinopyroxenes from eclogites; 5-single clinopyroxene inclusion in diamond with transitional composition [2].
The analysis of the chemical composition of inclusions traditionally involves the partial destruction of the samples, e.g., study of the chemical composition of inclusions in diamonds by EPMA (electron probe microanalysis) requires specific sample preparation, such as polishing or crushing of the host. Therefore, the methods of study of mineral inclusions using non-destructive methods are becoming more demanded [4][5][6][7][8][9]. Raman spectroscopy is an alternative high-resolution non-invasive non-destructive technique, which provides information about specific vibrations in solid materials reflecting the chemical and structural features of minerals, including those presented as inclusions in natural diamonds. In this study, we present new data on the Raman spectroscopic characteristics of clinopyroxenes inclusions in natural diamonds and clinopyroxenes from the series of peridotite and eclogite xenoliths from kimberlite. These data are used to define the correlations between Raman features of mantle clinopyroxenes and their chemical composition.

Materials and Methods
In total, 51 natural diamonds were selected from different occurrences of Yakutian diamondiferous province (Internatsional'naya (n = 9), Yubileynaya (n = 4), Udachnaya (n = 4), Mir (n = 1), Komsomolskaya (n = 1), Nyurbinskaya (n = 2) kimberlite pipes, Ebelyakh river deposits (n = 30)). The studied diamonds were represented by both octahedral and rounded dodecahedral crystals with the inclusions of peridotite or eclogite parageneses ( Figure 2). The sample preparation includes polishing of a diamond parallel to the (110) plane in order to recover the inclusions to the surface for further analysis of chemical composition by EMPA. Some inclusions were examined by Raman spectroscopy before and after the polishing and several fully extracted inclusions from diamond were studied to avoid the ambiguity as clinopyroxene inclusions may be affected by the residual strain of a diamond host [10,11]. In total, 28 clinopyroxene grains were also selected from peridotite (n = 8) and eclogite (n = 20) xenoliths from Udachnaya kimberlite pipe. The characterization of the clinopyroxenes from these samples is provided in [12,13].  [14]. The chemical composition of some clinopyroxene inclusion was taken from the published data (MT, HI- [15]; NBI- [16]; Ud10- [17]).
Raman spectra were collected using a Horiba Jobin Yvon LabRAM HR800 Raman microspectrometer with a multichannel air-cooled (−70 °C) CCD detection system and a 532-nm Nd:YAG laser, equipped with an Olympus BX41 microscope (×50). Spectra were recorded in backscattering geometry in the range 100-1200 cm −1 , through 7-10-s exposure and 10-15 acquisitions. The width of slit was 100 μm and the power of the laser was 10 mW. The spectra were calibrated using known emission lines of the Ne-lamp, also the Si-band at 520.6 cm −1 was used for monitoring of the frequencies of the Raman spectra. Additionally, the spectra were obtained at different crystallographic orientations via rotation (for each 15°) because some modes can change their intensity and disappear or are overlapped [18]. The spectroscopic software OPUS 8.2. (Bruker Optik GmbH, Ettlingen, Germany) was used for manipulation with spectra: the peak position was obtained by a regression least-squares curve fitting calculation using a Gaussian-Lorentzian mixed peak shape and baseline correction. This procedure yields a precision for peak of ±0.5 cm -1 .
Raman spectra were collected using a Horiba Jobin Yvon LabRAM HR800 Raman microspectrometer with a multichannel air-cooled (−70 • C) CCD detection system and a 532-nm Nd:YAG laser, equipped with an Olympus BX41 microscope (×50). Spectra were recorded in back-scattering geometry in the range 100-1200 cm −1 , through 7-10-s exposure and 10-15 acquisitions. The width of slit was 100 µm and the power of the laser was 10 mW. The spectra were calibrated using known emission lines of the Ne-lamp, also the Si-band at 520.6 cm −1 was used for monitoring of the frequencies of the Raman spectra. Additionally, the spectra were obtained at different crystallographic orientations via rotation (for each 15 • ) because some modes can change their intensity and disappear or are overlapped [18]. The spectroscopic software OPUS 8.2. (Bruker Optik GmbH, Ettlingen, Germany) was used for manipulation with spectra: the peak position was obtained by a regression least-squares curve fitting calculation using a Gaussian-Lorentzian mixed peak shape and baseline correction. This procedure yields a precision for peak of ±0.5 cm -1 .

Results
Most of the studied inclusions are presented by clinopyroxenes, which correspond to either peridotite (n = 16) or eclogite (n = 34) parageneses and single inclusion (Cr# 0.08 and Mg# 0.89) falls into the field of compositions transitional between these two principal parageneses (Figure 1a (Table 1). The structure of clinopyroxene is known to consist of SiO 4 -tetrahedra chains parallel to [001] linked by cation layers of M1O 6 -octahedron and M2O 8 -coordinated polyhedral. The factor group analysis of the clinopyroxene structure (C2/c) defines 30 Raman active modes (14A g + 16B g ) [19][20][21][22][23]. The Raman spectra of the studied clinopyroxenes generally exhibit several intense modes near~300-500, 670, and 1000 cm −1 assigned to metal-oxygen translation, stretching vibrations of bridging O-Si-O br (ν 11 Figure A1). Thus, the stretching O-Si-O br mode may be precisely fitted and seems to be most suitable for estimation of the possible dependencies with composition.

Discussion
The elements can continuously substitute each other in the structure of clinopyroxene (e.g., diopside CaMg[Si 2 O 6 ]-omphacite Na 0.5 Ca 0.5 Mg 0.5 Al 0. 5 [24]. Isostructural substitutions of the cations influence the lengths and angles of bonds, which changes the energy of Si-O vibrations and, consequently, the wavenumbers of Raman peak positions. In clinopyroxenes, heterovalent isomorphism Na + Al 3+ -Ca 2+ Mg 2+ affects Raman shifts close to linear [18]. These Raman shifts may be calibrated with the chemical composition of clinopyroxenes obtained by other techniques. It makes it possible to use the Raman technique for quantitative estimation of the contents of the major chemical compounds of this mineral. We compared the peak position frequencies of the stretching O-Si-O br mode (ν 11 ) with the chemical composition of the studied clinopyroxenes. Large variations in the composition, accompanied by wide Raman shifts of the ν 11 mode, allowed determination of adequate linear dependences and the regression parameters of these correlations. The correlation coefficients for all observed linear dependences for α (the level of significance or the probability of a type I error) = 10 −4 are summarized in Table 1  The data calculated from the regression equations and EPMA data yield sufficiently good convergence ( Figure 5). Hereby, the equations of regression lines (1-4) can be used for a quantitative assessment of the chemical composition of mantle clinopyroxenes using only Raman spectroscopy technique. The discrepancies between Raman calculated data and EPMA data are <5.0 wt.% for CaO (mean 1.1 wt.%), <4.2 wt.% for MgO (mean 0.6 wt.%), <2.5 wt.% for Na 2 O (mean 0.4 wt.%), and <2.5 wt.% for Al 2 O 3 (mean 0.5 wt.%). Some high discrepancies may occur because Cr 2 O 3 and FeO were not considered in the calculations of the regression lines. It is known that strain pressure can also affect the Raman shifts of peak positions for anisotropic crystals [27]. However, the completely extracted and in situ inclusions of clinopyroxenes and clinopyroxenes from eclogites and peridotites give similar regression lines. It may suggest that the Raman shift of v 11 stretching mode for clinopyroxene inclusions in diamonds depends on the composition rather than residual internal pressure.
Minerals 2020, 10, x FOR PEER REVIEW 7 of 11 The data calculated from the regression equations and EPMA data yield sufficiently good convergence ( Figure 5). Hereby, the equations of regression lines (1-4) can be used for a quantitative assessment of the chemical composition of mantle clinopyroxenes using only Raman spectroscopy technique. The discrepancies between Raman calculated data and EPMA data are <5.0 wt.% for CaO (mean 1.1 wt.%), <4.2 wt.% for MgO (mean 0.6 wt.%), <2.5 wt.% for Na2O (mean 0.4 wt.%), and <2.5 wt.% for Al2O3 (mean 0.5 wt.%). Some high discrepancies may occur because Cr2O3 and FeO were not considered in the calculations of the regression lines. It is known that strain pressure can also affect the Raman shifts of peak positions for anisotropic crystals [27]. However, the completely extracted and in situ inclusions of clinopyroxenes and clinopyroxenes from eclogites and peridotites give similar regression lines. It may suggest that the Raman shift of v11 stretching mode for clinopyroxene inclusions in diamonds depends on the composition rather than residual internal pressure. The previously published data reveal that peridotite and eclogite clinopyroxenes can be discriminated by the Cr# [1,2]. The lack of sufficient correlations of the stretching mode ν11 with the Cr and Fe contents do not allow assessment of these components using solely Raman data. It makes it difficult to discriminate the clinopyroxenes of different parageneses. Despite this, the The previously published data reveal that peridotite and eclogite clinopyroxenes can be discriminated by the Cr# [1,2]. The lack of sufficient correlations of the stretching mode ν 11 with the Cr and Fe contents do not allow assessment of these components using solely Raman data. It makes it difficult to discriminate the clinopyroxenes of different parageneses. Despite this, the clinopyroxenes of peridotite paragenesis and eclogite parageneses can roughly be separated by the peak position of the stretching ν 11 mode. We calculated two separate distributions by the peak position of ν 11 mode for peridotite and eclogite clinopyroxenes ( Figure 6). These distributions are characterized by different parameters of the mean value and standard deviation (µ = 670.2, σ = 2.3 for peridotite clinopyroxenes Minerals 2020, 10, 1084 8 of 10 and µ = 678.9, σ = 4.0 for eclogite clinopyroxenes), assuming that these distributions are normal. The Raman spectra of peridotite clinopyroxene has the stretching ν 11 mode with a peak position frequency lower than 672 cm −1 with a probability of 85% (α = 0.15). Similarly, the clinopyroxene is likely to belong to eclogite (with probability of 85%, α = 0.15) if the peak position of the stretching ν 11 mode is higher than 675 cm −1 . The uncertainty range takes place between 672 and 675 cm −1 . This extended uncertainty range is due to the existence of Na 2 O rich with peridotite clinopyroxenes (4.19-5.09 wt.%), which significantly overlap the range of eclogite clinopyroxenes and have the upshifted peak position of ν 11 Raman mode. Some eclogite clinopyroxenes with relatively low Na 2 O~2.25 wt.% and elevated Cr 2 O 3~0 .27 wt.% also fall into the uncertainty range.
Minerals 2020, 10, x FOR PEER REVIEW 8 of 11 clinopyroxenes of peridotite paragenesis and eclogite parageneses can roughly be separated by the peak position of the stretching ν11 mode. We calculated two separate distributions by the peak position of ν11 mode for peridotite and eclogite clinopyroxenes ( Figure 6). These distributions are characterized by different parameters of the mean value and standard deviation (μ = 670.2, σ = 2.3 for peridotite clinopyroxenes and μ = 678.9, σ = 4.0 for eclogite clinopyroxenes), assuming that these distributions are normal. The Raman spectra of peridotite clinopyroxene has the stretching ν11 mode with a peak position frequency lower than 672 cm −1 with a probability of 85% (α = 0.15). Similarly, the clinopyroxene is likely to belong to eclogite (with probability of 85%, α = 0.15) if the peak position of the stretching ν11 mode is higher than 675 cm −1 . The uncertainty range takes place between 672 and 675 cm −1 . This extended uncertainty range is due to the existence of Na2O rich with peridotite clinopyroxenes (4.19-5.09 wt.%), which significantly overlap the range of eclogite clinopyroxenes and have the upshifted peak position of ν11 Raman mode. Some eclogite clinopyroxenes with relatively low Na2O ~2.25 wt.% and elevated Cr2O3 ~0.27 wt.% also fall into the uncertainty range. Figure 6. Raman shift of stretching mode ν11 vs. Cr# of clinopyroxenes: 1-peridotite clinopyroxene inclusions in diamonds, 2-clinopyroxenes from peridotites, 3-eclogite clinopyroxene inclusions in diamonds, 4-clinopyroxenes from eclogites; 5-clinopyroxene inclusion in diamond with transitional composition [2]. The transition from peridotite to eclogite clinopyroxenes occurs at the range 672-675 cm −1 of the peak position of ν11 mode; PDF-probability density function of normal distribution with parameters μ (mean) and σ (standard deviation) for clinopyroxenes of peridotite (Pcpx) and eclogite (E-cpx) parageneses, α-the level of significance for boundary lines of uncertainty area. The transition from peridotite to eclogite clinopyroxenes at Cr# 0.07-0.10 is taken from [2].

Conclusions
Raman spectroscopy is a non-destructive technique for the identification of structural and chemical properties of minerals, and in particular is appropriate to study inclusions in natural diamonds. The new dataset of the Raman spectroscopic features of clinopyroxene inclusions in natural diamonds of different paragenesis (peridotite and eclogite) and clinopyroxenes from xenoliths peridotites and eclogites in kimberlite were obtained in this study. The wide variations of the peak position in a wide range for the ν11 mode of stretching vibrations of Si-O-Sibr were found. These variations correlate well with the chemical composition of the studied clinopyroxenes. The correlations, described as linear regression lines, generally reflect heterovalent jadeite-diopside isomorphism and can potentially be applied to evaluate the chemical composition of mantle Figure 6. Raman shift of stretching mode ν 11 vs. Cr# of clinopyroxenes: 1-peridotite clinopyroxene inclusions in diamonds, 2-clinopyroxenes from peridotites, 3-eclogite clinopyroxene inclusions in diamonds, 4-clinopyroxenes from eclogites; 5-clinopyroxene inclusion in diamond with transitional composition [2]. The transition from peridotite to eclogite clinopyroxenes occurs at the range 672-675 cm −1 of the peak position of ν 11 mode; PDF-probability density function of normal distribution with parameters µ (mean) and σ (standard deviation) for clinopyroxenes of peridotite (P-cpx) and eclogite (E-cpx) parageneses, α-the level of significance for boundary lines of uncertainty area. The transition from peridotite to eclogite clinopyroxenes at Cr# 0.07-0.10 is taken from [2].

Conclusions
Raman spectroscopy is a non-destructive technique for the identification of structural and chemical properties of minerals, and in particular is appropriate to study inclusions in natural diamonds. The new dataset of the Raman spectroscopic features of clinopyroxene inclusions in natural diamonds of different paragenesis (peridotite and eclogite) and clinopyroxenes from xenoliths peridotites and eclogites in kimberlite were obtained in this study. The wide variations of the peak position in a wide range for the ν 11 mode of stretching vibrations of Si-O-Si br were found. These variations correlate well with the chemical composition of the studied clinopyroxenes. The correlations, described as linear regression lines, generally reflect heterovalent jadeite-diopside isomorphism and can potentially be applied to evaluate the chemical composition of mantle clinopyroxenes using only the Raman spectroscopy technique. The peak position of the stretching Si-O-Si br Raman mode (v 11 ) can be satisfactorily used to separate clinopyroxenes of different parageneses. The clinopyroxenes with the peak position of v 11 mode lower than 672 cm −1 are likely to belong to peridotites; and the clinopyroxenes with the peak position of v 11 mode higher than 675 cm −1 are likely to belong to eclogites.