Raman Spectroscopy from Laboratory and Proximal to Remote Sensing: A Tool for the Volcanological Sciences

: Here we explore and review some of the latest ideas and applications of Raman spectroscopy to the volcanological sciences. Firstly, we provide a brief overview of how Raman spectral analysis works and how spectra from silicate glasses are interpreted. We then look at specific applications of Raman spectral analysis to the volcanological sciences based on measurements on and studies of natural materials in the laboratory. We conclude by examining the potential for Raman spectral analysis to be used as a field based aid to volcano monitoring via in situ studies of proximal deposits and; perhaps; in remote sensing campaigns


Introduction
One of the central goals of modern volcanology is to understand the physics and chemistry of processes that operate during transport, storage and eruption of magmas. These endeavors include both forensic reconstructions of pre-historic volcanic eruptions and real-time analysis of active volcanic systems. From the microscopic to the macroscopic scale, magmatic and volcanic processes are primarily governed by the physical and chemical properties of the melt. These properties of the melt are, themselves, highly transient because of cooling, differentiation, assimilation, mixing, and volatile exsolution that commonly attend transport and eruption.
Volcanic glasses are particularly informative in that they represent the quenched melt fractions of magmas at the time of eruption. These natural glasses can be investigated in the laboratory to track the transient evolution of volcanic systems. Lab-based tools and techniques, for example, can be used to estimate the compositions and thermochemical properties of volcanic glasses and, thus, can inform on pre-and syn-eruptive magma conditions. Such data can contribute to long and short-term hazard planning at active volcanic centers.

Raman Spectral Analysis of Natural Glasses
The versatile, non-destructive character of Raman spectroscopy allows for effective and rapid (from a few seconds to few minutes) characterization of a wide variety of geologic materials, including organic materials, minerals, silicate melts and glasses [13][14][15][16]. Raman spectroscopy can be performed on samples with little to no preparation, and small volumes of material. Confocal Raman spectroscopy imaging can work with a horizontal resolution down to 1-2 µm 2 which allows for detailed (micro-scale) mapping of compositional or structural variations in glassy samples [10,17,18]. Relative to infrared (IR) spectroscopy, Raman spectroscopy allows for a better selection of peaks that are narrower thereby allowing discrimination of different species within a single spectrum (i.e., less or no overlapping bands). Raman can also detect IR-inactive vibrational modes as well as FTIR can detect Raman-inactive vibrations. Compared with electron microscopy (EM) and X-ray diffraction (XRD), sample preparation and measuring environment do not need particular care and do not need vacuum, heating/cooling/relative humidity controlled stages. A Raman workbench is much more compact than those of XRD or EM techniques. Compared to EM, Raman allows to measure multilayer materials and compared to XRD, Raman permits the measurement of both crystalline and amorphous materials.
For anhydrous glasses and melts, Raman spectra are correlated to their structural properties, themselves a reflection of chemical composition (Figure 1). Differences in the Raman spectra collected in the range of 180-1300 cm −1 on silicate glasses depend on both structural and compositional aspects of the glasses. These differences reflect the variable contribution of network forming cations and network modifying cations, which eventually determine the arrangement of bridging oxygens (BO) and non-bridging oxygens (NBO) of the silicate network. Above the so-called boson region (10-250 cm −1 ) [11], iron-bearing alumino-silicate glasses typically exhibit two main broad and asymmetric bands (referred to as Low-Wavenumber (LW) and High-Wavenumber (HW), respectively), the first between 200 and 700 cm −1 , centered at about 490 cm −1 , and the second between 800 and 1300 cm −1 with center at about 960 cm −1 . The LW band is correlated mainly to bending vibrations of Si-O-Si bonds, and in consequence, to the tetrahedral chains, while the HW bands depends on the stretching vibrations of Si-O units. A third, weaker band is observed, at intermediate wavenumbers (650-850 cm −1 ; here following reported as medium wavenumber [MW]), with a center, that, as a function of composition varies from ca 680 to about 810 cm −1 ( Figure 1 [2], where the lowest peak wavenumber values are associated to the most depolymerized melts. Several others secondary peaks are observed as a function of the investigated glass compositions and the presence of mineral phases. A short summary of the main characteristics of the LW, MW and HW vibrational bands, their position and the relationships with the main kinds of vibration modes associated to them for silicate melts is presented in the Appendix A2 (Table A1). Importantly to the Earth Sciences, Raman spectroscopic analysis can also be used to estimate the species and abundances of volatile components dissolved and preserved in natural and experimental glasses or in exsolved fluids preserved as fluid inclusions within glasses and crystals. Raman detection limits for volatile components are in the order of a few ppm allowing for very low volatile contents to be measured even in in nominally anhydrous minerals. For volatile bearing glasses and melts, stretching (translational) vibrations are observed at higher frequencies such as 2331 for N2 [19]; 4125 for H2 [20], 990 cm −1 for SO4 2− [21,22]), 1062-1092 cm −1 for CO2 and CO3 2− [23,24] vibrations of water in both the molecular and hydroxyl form can be observed between 3000 and 4000 cm −1 approximately centered at 3550 cm −1 for hydrous glasses [8,14,16,[25][26][27][28][29].

Volcanological Applications
Over the last decade, Raman spectroscopy has had a rapidly expanding number of applications in the Earth Sciences [30,31]. The ability of Raman spectroscopy to differentiate phases (e.g., glass versus microcrystalline material), to estimate compositions of silicate glasses and to characterize structural properties of glasses has tremendous import for the volcanological sciences.
Here, we introduce three state-of-the-art applications of Raman spectroscopic analysis to the volcanological sciences. These include: (a) estimation of melt composition, including the Fe-redox state, directly from Raman spectra of glasses, (b) estimation of volatile contents in experimental and natural (e.g., volcanic products; glass inclusions) silicate glasses, and (c) prediction of chemical proxies of melt structure (e.g., Non-Bridging Oxygens over Tetrahedra, NBO/T; and Structural Modifiers parameter, SM) from Raman spectra as a means to connect to melt transport properties (e.g., viscosity). Raman spectroscopy is also an excellent tool for detection of crystalline phases at the nano-scale, which are invisible with standard techniques (EPMA, SEM). Below we explain each of these applications and explore their implications and consequences for volcanological research.

Estimating Composition and Redox State of Silicate Glasses and Melts
Raman spectra of glasses show few characteristic large bands whose shape depends on several contributors. Separating the effects of each factor can be a complicated process requiring the use of different models (e.g., baseline corrections) and techniques. This is in sharp contrast to crystalline materials, whose spectra show many distinctive peaks making for a relatively straight interpretation. Raman spectra for glasses, on the other hand, are generally devoid of well-defined and wellseparated spectral bands, except in the case of those glass compositions with dissolved volatiles (e.g., SO2 or H2O).
Methods for estimating qualitatively or quantitatively the composition of silicate glasses and melt commonly rely on calibration strategies based on: (a) the estimation of the intensity of the heights of peaks and bands; (b) the integrated area of specific band envelopes or (c) the position of peaks bands or their normalized counterparts [2,14,15,[32][33][34]. The various models adopt different protocols for spectral correction (e.g., [35]), baseline subtraction, or analysis of vibrational features. These models provide a rapid and inexpensive means of reliably estimating the elemental or oxide composition of natural glasses in the laboratory, in the field, and, potentially on other terrestrial planets and moons.
Previous authors [16,29,33,[36][37][38], investigating the effect of iron on the vibrational response of glasses, demonstrated that both iron content and its redox state strongly influence the Raman spectra. These authors showed that the ratio Fe 3+ /Fetot could be calculated by using different treatments of the spectra involving baseline fitting and/or deconvolution in Gaussian components associated to Q nspecies (Q n distinguishes silicon atoms according to the number n of coordinated bridging oxygens (BO), i.e., Q 4 stands for silicon coordinated by four BO, Q 3 corresponds to three BO and one nonbridging oxygen (NBO), etc.) [9]. Di Muro and coauthors [33] demonstrated that several calibrations can be obtained for the determination of the Fe 3+ /Fetot ratio in glasses by exploiting the sensitivity of the HW envelope (at ca 1000 cm −1 ) to iron redox state. The spectral analysis of Di Muro and coauthors [33] was performed on glasses of peralkaline rhyolite (i.e., SiO2-rich, iron-rich rhyolite) and basaltic (SiO2-poor, iron-rich) compositions. Their analyses were performed by measuring the ratios of the band heights; the integrated areas of specific lines as well as the position of HW bands (e.g., band IV for rhyolites and band III for basalts, Figure A2) obtained on fully characterized reference glasses. They observed that the Fe 3+ /Fetot ratio is highly dependent on composition and composition has shown to control the direction of redox change [39]. Peralkaline glasses are excellent systems to accurately characterize the redox state of iron, as consequence of the high sensitivity of Raman spectra to network former cations (Si, Al and Fe 3+ ). The sensitivity decreases significantly for basaltic compositions, such as Stromboli and Etna trachybasalts but continued work may overcome this challenge.
More recently, Di Genova and coauthors [39,40] carried out similar investigations on peralkaline rhyolites from Pantelleria Island (i.e., Pantellerites) and basaltic glasses having compositions resembling some Martian rocks. They provided a refined quantitative method for estimating Fe 3+ /Fetot ratios based on the main peaks composing the HW band (at 970-1040 cm −1 for pantellerites; 950-970 cm −1 for Martian basalts). The method uses an empirical formula [40] to parametrize the spectra as function of a Raman parameter (Rp). A Raman spectrum of a glass of any unknown composition can be approximated by combining spectra from two end-members, one for the most oxidized (IOX) material and the other, the most reduced (IRED) sample, according to the equation: Their model allows calculation of Fe 3+ /Fetot ratios from Raman spectra based on the following equation: where a, b and c are fit parameters that are particular to each family of glass compositions (e.g., basalt, rhyolite) [39]. The same method proposed by [39] can be used to approximate, on a weight percent basis (wt%), the composition (y) of each oxide concentration in wt. %. After calculation of Rp for a given spectrum by combination of two known endmembers (i.e., using an ideal mixing equation), oxide concentrations can be calculated through a simple 2nd or 4th order polynomial equations such as: where a, b and c are adjustable parameters. The broad shape of the main bands of the Raman spectra of glasses presents a challenge for using Raman spectroscopy to estimate the composition of multicomponent glasses. These broad bands are more or less insensitive to chemical variability and, thus, it is a complicated process to attribute individual vibrational modes to specific cations or molecules within the silicate glass/melt network. At present, there are no models for predicting the glass composition of natural multi-components silicates from the corresponding Raman spectra more accurately than those proposed by Di Genova et al. [34,35]. Furthermore, their approach was optimized to predict compositions falling between two end-members which, themselves were aligned along a linear mixing trend. Recent work by González-García and co-authors [41], however, represents a new approach to constructing a general model for retrieval of bulk chemical composition of glasses from their Raman spectra.

Predicting Dissolved Volatile Contents in Silicate Glasses and Melts
Volatile contents in natural melts and glasses can reach levels of several wt.%. Establishing the concentration of the main dissolved and exsolved volatile phases in a magmatic system and in the erupted volcanic products is of critical importance to define the pre-eruptive conditions and the eruption dynamics and validate existing models used to simulate eruptive events. Raman, commonly used in association with FTIR (Fourier Transform Infrared), is an ideal tool for measuring residual water contents in glassy volcanic materials, including crystal-hosted melt and fluid inclusions. The Raman spectra of hydrous glasses contain two spectral regions of interest between 200 and 1700 cm −1 corresponding to Si-O stretching vibration and H-O-H bending vibrations. The peak between 2800 and 3800 cm −1 , centered near 3500-3650 cm −1 , correspond to OH vibrations [42] (Table A1). The band 2800-3800 cm −1 is complex to interpret due to the individual contribution of different H2O vibrational modes.
The recent models of Mercier et al. [14,15], Le Losq et al. [44]) and Schiavi et al. [29] are amongst the most promising methods for predicting the H2O content dissolved in glasses. The model of Di Genova et al. [16] is adequate but is based only on a small dataset of hydrous glass compositions. Nonetheless, defining which model is more promising is complex as, so far, there is still a lack of a standardization in the procedures which would allow to compare the results obtained by different studies. In most cases, in fact, comparison with other studies may not be possible, however, as the protocols to estimate H2O content are substantially different and commonly the data are not easily available. Given the considerable amount of new data so far available, we believe that standardization procedures could be easily attainable by the construction of database sharing facilities together with the establishment of the best specific protocols and the comparison of the results obtained. In addition, samples produced by different laboratories should be shared in order to optimize interlaboratory calibration as done by Giordano et al. [2] and create networks of laboratories devoted to face and solve specific scientific issues.

Toward a Structural Viscosity Model for Silicate Melts
More recently, Le Losq and Neuville [48], Giordano and Russell [34]) and Giordano et al. [42] showed that the viscosity of simple and multicomponent anhydrous silicate melts over a temperature interval of ~ 700 to 1600 °C, can be predicted from the Raman spectra obtained from the corresponding glasses (i.e., fast quenched melts) in the range between 100 and 1500 cm −1 . These methods proved to be very promising for in situ rheological investigations and may find substantial application in planetary sciences studies [34,48]. In particular, Le Losq and Neuville [48] developed a model to predict viscosity of the simple SiO2-Na2O-K2O system through the comparison of viscosity data and the Q n -species abundance retrieved from analysis of Raman spectra of the glasses. Giordano and Russell [34] and Giordano et al. [42], following the intuition of Mercier et al. [14,15]), presented a first order model predicting the viscosity of multicomponent natural melts on the basis of the dependence of the characteristic viscosity descriptive parameters (BVFT, CVFT) on the Raman spectral data collected from the corresponding glasses ( Figure 2). Model VFT parameters BVFT (R) and CVFT (R) as defined by [28] and relationships between pseudostructural parameters According to [28]: BVFT (R) = b1R b2 and CVFT (R) = c1R c2 +c3 where b1, b2, c1, c2, c3 are adjustable parameters.
The authors defined a Raman ratio (R) as the ratio between the low frequency band (LW) and the high frequency band (HW) which they normalized (Rn) to the Raman ratio obtained on a reference glass, one amongst those with high R values. They showed that strong relationships exist between constitutive parameters used to describe the viscosity and the normalized Raman ratios which allows the viscosity of anhydrous multicomponent natural melts to be predicted with a great accuracy. Although the model requires expansion to use of the structural information of volatile-bearing melts, it allows accurate description of the viscosity of anhydrous melts by the employment of a simple equation with 6 adjustable parameters and the measured R (details in [2,34]). Also, the SM and NBO/T parameters, calculated from compositions, are shown to be strongly correlated with R.

Raman in the Field: Proximal and Remote Sensing Campaigns
At present most laboratories can acquire spectra from a wide interval of locations within its close environment, but the possibility to perform Raman spectroscopy measurements at certain distances will enable to collect data of great relevance in the field and for planetary explorations. For terrestrial applications stand-off Raman instruments are commonly used in field geology and mineralogy and recently several efforts have been adopted to adapt and miniaturize the instrumentation for the more ambitious task of planetary exploration [49][50][51][52][53][54][55].
The use of an optimized optical system made it possible to acquire Raman spectra over larger target distances. More than 20 years ago, Angel et al. [8] were able to identify organic and inorganic compounds on targets up to 20 m away. The coupling of the Raman system with a telescope in Cassegrain configuration is at present the most sensible configuration to reach big distances, up to hundreds of meters [56]. The coupling can be made by means of fiber optics or directly. In the case of direct coupling, the excitation laser can be focused using an external path (oblique configuration) or in a coaxial configuration. A coaxial configuration is preferred as it provides greater stability, but with a cost of lesser brightness.
The main improvement has been the use of a pulsed laser as the excitation source, coupled to a time-gated charge coupled device (CCD) or, better, intensified CCD (iCCD), as detector [56]. The time resolution allows to acquire the signals only during the short time where the excitation is concentrated, reducing the interference of environmental light (allowing measurements in full daylight). A second, very important, advantage is the removal of the fluorescence due its slower answer respect to the Raman scattering. The combination of the high power peaks during the laser pulses, the enhanced signal to noise ratio due to time gating, and the use of telescopic systems allowed to obtain spectra at very large distances: the group operating at the University of Hawai'i realized different compact and large systems operating between 1.5 m and 430 m ( [57] and references therein) including a scanning system able to realize Raman maps with 2 mm of space resolution at a distance of 100 m [58]; Wu et al. [59] realized a mobile standoff UV Raman system able to obtain measurements up to 500 m and Johansson et al. [60] tested their standoff visible Raman in rainy conditions obtaining spectra from 50 m of distance.

•
There are two main frontiers and challenges to expanding the use of Raman spectroscopy in the Earth Sciences. These are to collect and analyze data from complex dynamic systems in real time and to collect data remotely. Volcanic materials are complex to analyze in a real-time sense because they are highly transient. The volcanic products evolve during cooling, degassing and flow (i.e., deformation) as they control the onset, sequence, and amount of crystallization. It is here that Raman spectroscopy will require the development of new protocols capable of accounting for the real-time variation of the mineralogical assemblage. This will require extensive new information and experiments to provide a template for interpreting Raman spectroscopy in terms of mineral phases and glasses pertinent to magmatic and volcanic materials. Currently such data are compiled and being continuously updated in at least a few well-known open access downloadable databases including these web-based examples: • The RRUFF Project website (http://rruff.info/about/about_general.php).
However, these future field-based applications lead to the second great challenge, namely, the ability to collect data remotely. Clearly, Collecting Raman spectra at distances larger than the typical working distance of a microscope objective is in principle relatively simple. The only fundamental requirement is to focus the laser beam at the desired distance and to collect the light from the same point with a suitable optical system. In order to collect enough Raman signal, the most important parameter is the solid angle at which the focused spot is seen by the optical entrance of our system. That means that a large lens or mirror should be used as objective to maximize the signal at large distances. For example, it would be possible to setup a crude but functional instrument for collecting Raman spectra at a distance of few meters in the field using a handled Raman spectrometer and a focussing lens. The most important parameter is the width of the lens, because it determines the solid angle of collection of the scattered light  Here, we performed our measurements using an inexpensive EnSpectr RaPort portable Raman, equipped with a 532 nm green laser coupled with a lens with 6 cm diameter (and 25 cm focal distance) (see Figure 3). Using that lens it was possible to collect Raman spectra from a set of target minerals, including quarts, sulfur, aragonite and a piece of shell (pink coloured) at a distance of 7 m. The pink shell is included because the strong resonant Raman signal of carotenoids is one of the motivations of the use of Raman instruments in the search for life on Mars. Figure 3b shows the spectra of some minerals obtained at different distances in full environmental light: in front of a sunny windows or in presence of fluorescent bulbs using the standard background removal procedure of the instrument. The analyzed spot is in the order of a millimeter of diameter; the spectra of a sub-millimetric crystal of anatase on gneiss matrix is included in Figure 3b.
A functional Scanning Standoff Raman Spectrometers (SSTRS) system has already been developed for mapping the spatial distribution of some minerals (e.g., feldspar, quartz, gypsum) and glasses at long distances [6]. The pan-tilt scanner and laser beam pointing onto the distant samples are computer controlled, and spectra are obtained in an x-y grid on the area of interest. Then, spectra are gathered at each point in the grid and processed to determine the distribution of minerals from their respective Raman fingerprints. The new SSTRS system has been successfully tested, offering the capability to produce spatial distribution maps of mineral species from their remote Raman fingerprints. Remote Raman spectral analysis will contribute greatly to planetary exploration programs [55]. Remote data collection would have immediate applications to volcano monitoring activities. As smaller, more robust Raman spectrometers that collect data from greater distances are developed further, their use in volcanology will intensify. Miniaturized Raman spectrometers used in combination with drones are a means to collect new types of real-time data on volcanoes before, during and after eruptions without putting scientists at risk. For example, the volcanological sciences could benefit immensely by using Raman to monitor lava flows wherein Raman spectra can be analysed for temperature, melt composition, and mineralogy. Raman spectra could be collected with time at the vent to track the evolution of the erupting magma or could be collected from lava as it moves downstream from the vent to its distal edge marking the rheological death of the lava [43,61]. In a similar manner, Raman-based monitoring could be applied to the products of explosive eruptions (i.e., tephra). Technically, the employment in the field under sun light exposure as well as the exposure to toxic gases and the high temperature of lava flows, for instance, would also require to consider the temperature-frequency dependence of the relative Raman scattering efficiencies due to the temperature-dependent population density of the vibrational energy levels [6,49]. The blackbody radiation generated by high temperature bodies provide an increased noise-signal background which have to be reduced. Such complication can be reduced through the employment of high energy lasers and the employment of fast-gated pulsed spectroscopy and CCD (Charge-Coupled Device) detection [49,62]. Acknowledgments: The micro-Raman equipment at UniTO belongs the Interdepartmental Center "G. Scansetti" for Studies on Asbestos and Other Toxic Particulates and was acquired with support from the Compagnia di San Paolo, Torino. D. Giordano is grateful to the Center for Advanced Studies (CAS) at the Ludwig-Maximilians-Universität (LMU) München (https://www.en.cas.uni-muenchen.de/index.html) for the support of this research through the LMU CAS Research Group "Magma to tephra: ash in the earth system" (DBD).

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A1. Main Vibrational Modes
The interatomic forces between bonded atoms control the exact frequency of the vibrations, in much the same way as a spring connecting two vibrating masses. Atoms, molecules or more complex structures (e.g., rings of tetrahedral or polymers) commonly vibrate through pure stretching (translational) and bending (rotational) modes or composition of them (roto-translational) (Fig. A1).
For silicate melts, above the LW region the vibrations are mostly of stretching type. Bending modes are instead the main feature of LW region. More details are provided in the following paragraph. Figure A1. Main vibrational modes of atoms in molecules or in complex structures (e.g., rings of tetrahedra or polymers -skeletal type of vibrations), shown here for polyethylene. Figure modified after [63].

Appendix A2. LW, MW and HW Bands in Silicate Melts and Glasses
According to previous authors [2,14,15,[64][65][66][67], the LW domain are dominated by a broad band resulting from bending vibrations of T-O-T units and, to some extent, deformation vibrations of T-O-T bridges (Figure 1). The LW envelope is usually interpreted as the combination of delocalized vibrational modes mainly related to symmetric stretching vibrations of BO in TO4 rings; rocking motion in fully polymerized units; vibration of T against its tetrahedral frame accompanied by slight displacement of the oxygen; T-O-T bending of NBO bonds and the so-called "defect bands", likely related to breathing modes of 3-4-membered rings of TO4 tetrahedra. In this region, the characteristics generally allocated to vibrations of 3-to 6-membered rings of tetrahedra in the aluminosilicate network [68,69]. In particular, a strong band near 500 cm −1 (D1) and sometimes a shoulder near 600 cm −1 (D2) are present in this domain, and can be respectively assigned to breathing vibrations of 4and 3-membered rings of tetrahedra; while the long tail at the LW region develops from stretching (translational movements) of Si-O units in 5-, 6-or higher-membered rings [68,69] forming the aluminosilicate framework of the glass [14,15,70]. The presence of alkalis or alkaline Earths these bands develop a peak centered at 580 cm −1 and a shoulder at 600 cm −1 . The peak centered at 580 cm −1  [12] and references therein). In general, the shape and positions of bands will evolve with varying glass composition and structure. Above the LW region the vibrations are mostly of stretching type. The weak band in the MW domain (600-850 cm −1 ) (e.g., [2,33]) is commonly ascribed to Si-O stretching (translational) mode [71] involving oxygen motions in the Si-O-Si plane [64,72] with intensity correlated to silica content [73] or to the motion of the Si atom in its oxygen cage [74]. The main characteristic in the envelope centred at ~800 cm −1 arises from two bands resulting from deformation oscillations of T-O-T bonds within precise domains, with a structure resembling crystalline SiO2 [67,75]. This band is more prominent in more polymerized melts and glasses (i.e., rhyolitic).
Bands in the HW region of the spectra (~800-1300 cm −1 ) contains information related to the Si-O, Al 3+ -O and Fe 3+ -O stretching (translations) in the glass structure. These are caused by oscillations of terminal covalently bonded units in TO4 tetrahedra with a variable number of bridging oxygens (BO), commonly referred to as Q n -units or Q n -species*. Here Q represents a TO4 tetrahedron and n is the number of bridging oxygens (BO), varying between 0 and 4. The HW envelope is the most sensitive to changes of redox conditions and is commonly deconvoluted in order to quantify the effect of iron oxidation state [16,33]. The investigation of vibration in the HW region has also the advantage of avoiding the contributions potentially provided by the presence of nanolites of oxide crystals (e.g., 16,77) and for the same reason is frequently used for studies devoted to characterize the effect of oxidation state of iron in basaltic to pantelleritic compositions [14,16,33,38,76,77]. Table A1 reports a summary of the main kind of vibration frequencies for some of the most well-known alumino-silicate and iron-bearing alumino-silicate melts.  [14] 565-595 Si-O-Si bending in Q 2 [9] 605 Oxygen breathing in SiO4 in Q 4 [9] 450, 500, 600 Antisymmetric Si-O stretching vibrations of SiO4 tetrahedra (Q 4 units) [9] 1070 C-O symmetric stretch in CO3 groups [9] 1083 SiO4 asymmetric stretching vibration [9] 1100 Antisymmetric Si-O-(NBO) in Q 3 [9] Volatile bearing compositions 900−970 Si-OH stretching mode [7,9] 1280 Molecular CO2 [9] 1600−1650 H-O-H bending of molecular H2O [9] 2350 Si-OH groups involved in intratetrahedral hydrogen bonding across an edge of the SiO4 tetrahedron 2800−4000 O-H+H2O stretch (wide band) [9] According to previous investigations seven to eight bands were used to fit the HW domain [10,33,78] (Figure A2) due to the entry of several types of T cations in a given Q n unit. [33], for instance, have subdivided the band in Band I (770-800 cm −1 ); II (~890 cm −1 ); III (~935 cm −1 ); IV (~ 980 cm −1 ); V (~1035 cm −1 ); VI (~1090 cm −1 ); VII (1150 cm −1 ), VIII (1000 cm −1 ). Figure A2. Curve-fitted HF envelope of Raman spectra (Long-corrected), figure from [33]. Band IV (dark colored curve) results from stretching of coupled Si-Fe tetrahedra with iron in alkali-balanced Q 3 units. The abundance of Q n species is controlled mainly by the following disproportionation reactions (5-7): 2Q n = Q n+1 +Q n−1 (n = 3, 2, 1)  (7) where Q n is the abundance of a particular unit.
According to the same authors bands II -VII result from to vibrations of Q n unit with the number of bridging oxygen (n) increasing from 0 to 4 ([33] for more details). These authors have shown that both peak intensities and areas and their ratios for the HW bands are significantly affected, other than by composition, also by both the Fe 3+ /Fetot ratio [14,16,33] and by the presence of water [14,33] (Figure  2).