Medium-range structure modifications induced by Fe³⁺/Fe_tot. in volcanic glasses: a low-frequency Raman spectroscopy study

Abstract

Here is carried out a Raman scattering study on the boson peak evolution of an iron-rich peralkaline rhyolite in function of both the iron oxidation state and the glass transition temperature. It is reported here that the distribution of low-frequency modes in the boson peak range is only slightly affected for an iron ratio (Fe³⁺/Fe_tot.) from 0.83 down to 0.24. Their distribution does not change in the boson peak range as a function of Fe³⁺/Fe_tot., until the reduction process starts to modify the glass network from a dominantly fourfold coordinated Fe³⁺ structures into a structure mostly governed by fivefold coordinated Fe²⁺. This trend is also related to a decreasing glass transition temperature peak, mirroring an increasing proportion of weakest bonds with respect to the stronger ones.

1 Introduction

Natural glass has always existed on Earth and in the solar system [1]. They are the result of high-energy events, ranging from volcanic eruptions to meteoritic impact revealing also their potential to encapsulate biosignature on Earth [2]. Volcanic glasses, in particular, represent the last melt fraction quenched in eruptive products (i.e. ash, lapilli, bombs, etc.), and thus, their structure is key in understanding volcanic eruptions [3].

One of the major elements of these glassy fractions is iron, which may act as network former (NF) and network modifier (NM) in the glass structure. Therefore, its role in tailoring the chemical and physical properties of glasses and melts is strongly dependent by the red-ox state, that in turn are affected by melt composition, temperature and oxygen fugacity that deeply influences thermal properties, elasticity and rheology. The oxidized iron Fe³⁺ is typically associated to a tetrahedral geometry, thus mostly fourfold coordinated [4, 5], while the reduced Fe²⁺ lies above fourfold [6,7,8] in particular between five- and sixfold [9, 10]. Disregarding the oxidation state, iron can be simultaneously found between four- and sixfold coordination, but experimental evidence uncovered that Fe²⁺ is mostly fivefold coordinated, suggesting that these systems, veer toward fivefold dominated structure [5, 11]. Additionally, the diagnostic on Fe³⁺ as NF (tetrahedral, ^IVFe³⁺) and NM (octahedral ^VIFe³⁺) is a pretty disputable and hotly endorsed subject, adding further degrees of complexity while analyzing its contribution in the glass structure [12,13,14]. Thus, the parametrization of the Fe³⁺/Fe_tot ratio is crucial in determining the chemical dependence of the silicate melt properties (i.e., viscosity and glass transition temperature T_g) obtaining attention in both material science and geosciences [15]. For this purpose, geoscientists employed Raman spectroscopy as a crucial tool for fast and thorough determination of the iron oxidation state in volcanic glasses [16,17,18,19,20,21].

The high-frequency region of Raman spectra (800–1300 cm⁻¹) is particular informative regarding the environment of the tetrahedrally coordinate cations (Si, Al or Fe) in characterizing the chemical structure of the glass network [22, 23] and its short-range order (SRO). However, below 150 cm⁻¹, the spectrum shows a peculiar low-frequency band in the reduced vibrational density of states (VDoS) $g(\omega )/\omega^{2}$ named boson peak, BP. Its origin, although highly disputed, seems to be related to a pile-up of the transverse acoustic modes at the edge of the first pseudo-Brillouin zone providing crucial information regarding the thermal conductivity and specific heat of glasses [24]. Additionally, it has been demonstrated that the analysis of the BP of silicate glasses can be used to estimate the melt properties since it embeds the detachment of the non-Arrhenian behavior of the liquid at the glass transition temperature (T_g) in the T_g/T plot [25], i.e., melt fragility (m) [26]. In particular, by determining the BP maximum position (${\omega }_{{{\text{BP}}}}$) and T_g of a glass, it is possible to build up its parental liquid’s viscosity in function of temperature [27, 28]. However, it represents the structural disorder of glass at the medium-range order (MRO) [29].

In this work is provided an extension of the analysis in [17, 18], showing new results from low-frequency Raman (in the BP spectral region) on a set of peralkaline iron-rich rhyolites named pantellerites. These volcanic glasses are characterized by a quite high iron content (~ 9 wt. %) having a known ratio between the Fe³⁺ and total iron content (Fe³⁺/Fe_tot.). By increasing Fe³⁺/Fe_tot., the BP moves toward higher intensities and lower frequencies suggesting a softening of the elastic constants distribution in the medium [30] linked to a strong liquid [27] expressed by the parameter m as defined in [26]. Conversely, the heat flow peak (T_g^peak) shifts toward higher temperatures, while the BP maximum becomes more and more suppressed. Moreover, here it is shown the existence of the scaling law for the BP, at least for Fe³⁺/Fe_tot. ratio ~ 0.40, below this value the scaling does not hold any more and the structure seems undergoes deep modifications.

2 Materials and methods

The starting material consists of anhydrous iron-rich rhyolitic glasses named pantellerites, previously synthesized and characterized in the studies [17, 18]. Chemistry, Fe³⁺/Fe_tot. and T_g^peak are summarized in Table 1. The samples consist on 7 glasses with different Fe³⁺/Fe_tot. ratios ranging between 0.24 and 0.74, plus 3 further samples for which T_g^peak is unknown having Fe³⁺/Fe_tot. from 0.83 to 0.26. The SiO₂ content varies between ~ 72–75 wt%, ~ 7–9 wt% FeO_tot and the alkalis (Na₂O + K₂O) between ~ 8 and 9 wt%, while the glass transition temperature has been determined trough the T_g^peak by DSC measurements, as described in [17].

Table 1 Chemical composition in wt% and Fe³⁺/Fe_tot. index, obtained by wet chemistry titration and the glass transition temperature peak (T_g^peak) is also reported. See details in [17]

Raman measurements were done with a micro-Raman spectrometer (Horiba Jobin–Yvon model T-64000) in backscattering geometry. The setting consists of three holographic gratings (1800 lines/mm) in double subtractive/single configuration and CCD detector with 1024 × 256 pixels and cooled by liquid nitrogen. The possibility of polarizing the laser beam which is provided by a combined Ar–Kr ion gas laser set at 514.5 nm (Spectra Physics, Satellite 2018 RM), allowed both parallel (HH) and crossed (HV) acquisitions.

3 Results

The HH polarized Raman spectra of samples with measured T_g^peak are shown in Fig. 1. Generally, some features hold the same characteristics, in particular the broad band at 460 cm⁻¹, associated to rocking and symmetric bending motions of bridging oxygen (typical of silica-rich glasses [31, 32]) suggests that the tetrahedral T-O-T bond angle is not markedly affected by the subtle chemical changes [33]. However, some slight differences can be appreciated on the component at ~ 490 cm⁻¹ and the shoulder at 600 cm⁻¹. These features are associated to the in-phase O-bending motion of bridging oxygen in 4- and 3-membered rings [34]. In particular, the shoulder at ~ 600 cm⁻¹ decreases as the Fe³⁺/Fe_tot. ratio increases suggesting a slight decrease in 3-memberd rings. The 800 cm⁻¹ band, ascribed to the threefold degenerate “rigid cage” vibrational mode of SiO₂ units [35] or to the T-O stretching vibrations along the T-O-T plane [36], does not reveal appreciable changes or shift. Since this band seems insensitive to compositional changes, its integrated area has been used to normalize both the HH and HV spectra. By contrast, moving toward the high-frequency region (850–1250 cm⁻¹), the spectra significantly change. In particular, as the Fe³⁺/Fe_tot. decreases, the centroid band shifts toward higher frequency (i.e., from ~ 970 cm⁻¹ for the FSP2 with Fe³⁺/Fe_tot. = 0.74; to ~ 1100 cm⁻¹ for FSP9 with Fe³⁺/Fe_tot. = 0.24). This spectral region is dominated by the symmetric stretching vibrations of tetrahedrally coordinated ions bonded to non-bridging oxygens (NBO). Their variations are due to oscillations of inter-tetrahedral bond angles, distances, force constants and more generally speaking they provides a picture of the network connectivity [37]. Within this interval is furtherly defined a sub-interval that seems affected by the contribution of the Fe³⁺ stretching in fourfold coordination state (^IVFe³⁺–O bonds, [38]), although it seems affected by overlapping tetrahedral units bonded to 2 and 3 oxygens [20].

Fig. 1

Raman spectra acquired in HH polarization and corrected after a quadratic baseline. The spectra are normalized to the 800 cm⁻¹ peak

The low-frequency Raman spectra analysis has been conducted following the procedure in [20], by considering the intensity $I^{{{\text{obs}}}}$, proportional to the vibrational density of states g$(\omega )$. Considered a Stokes process the intensity follows the relation:

$$I^{{{\text{obs}}}} (\omega ) = C(\omega )g(\omega )\frac{{\left[ {n(\omega ,T) + 1} \right]}}{\omega }$$

(1)

Here, $C(\omega )$ is the light to vibration coupling function and is assumed to be $\sim \omega$ in the BP region. Dividing $I^{{{\text{obs}}}} (\omega )$ by [n $(\omega ,T)$ + 1] (where n $(\omega ,T)$ is the Bose–Einstein factor) and frequency (ω), is obtained the reduced Raman intensity $I^{{{\text{red}}}} (\omega )$ [39], which is proportional to the reduced density of vibrational states as $I^{{{\text{red}}}} (\omega ) = C(\omega ) \times \left[ {g(\omega )/\omega^{2} } \right]$. The reduced spectra are plotted in Fig. 2, and for a suitable comparison, they were normalized to the integrated area of the 800 cm⁻¹ band (as shown in the inset in Fig. 2). The BP position has been determined by fitting its asymmetric shape to a log–normal function [40]. The BP position shift from ~ 48 cm⁻¹ to ~ 54 cm⁻¹, while its maximum becomes more and more suppressed showing an intensity reduction ($I^{{{\text{red}}}}$) of ~ 40% (see Table 2). It is worth to note that the high-frequency region of the HV spectra (Fig. 2 inset) show a component at $\sim$ 960 cm⁻¹ whose centroid show the same behavior of their equivalent HH spectra. This band in HV spectra is originated by the antisymmetric coupled mode of FeO, SiO and tetrahedra. The strong depolarization of this band in the spectra of Fe³⁺-rich glasses suggests that the local symmetry of T cations is near to the T^2S band (the asymmetric component filtered in Raman spectra of silicate glasses [38]). This is an important proof that Fe³⁺/Fe_tot. affect with the same magnitude this polarized spectral interval in which symmetric contributions are virtually not expected. 

Fig. 2

Crossed (HV) polarization and reduced Raman spectra of the low-frequency region. The black dashed line is an eye guideline. The spectra normalized to the integrated area of the 800 cm⁻¹ band is shown in the inset, in which is also shown the component at 960 cm⁻¹

Table 2 Values of the boson peak frequency obtained by fitting the low frequency region to the lognormal function $I (\omega )$ $\propto \exp \left\{ { - \left[ {\ln (\omega /\omega_{{{\text{BP}}}} )} \right]^{2} /2\sigma^{2} } \right\}$ (where $\sigma$ is the BP width) and those of the squeezed frequency $\omega_{S}$ obtained as explained in the text

It has been further examined the reduced spectra I^red(${\upnu }$), by considering their rescaled frequency $\nu = \omega /\omega_{S}$ like in [41], and by performing the variable transformation $g(\nu ){\text{d}}\nu = g(\omega ){\text{d}}\omega$. The parameter $\omega_{S}$ (squeezing frequency) is the frequency at which all the spectra are expected to fall in the same master curve with the same intensity (values reported in Table 2).

The squeezed spectra with varying Fe³⁺/Fe_tot. are plotted in Fig. 3a and choosing $\omega_{S}$ to get the same peak intensity, a different behavior is clearly observed. In the low-frequency region below ~ 20 cm⁻¹, the quasi-elastic scattering exhibit some contributions by perhaps extending shortly below the BP low-frequency tail and slightly influencing its distribution. Moving forward, the high-frequency flank of the squeezed spectra results more and more tilted inward as the Fe³⁺/Fe_tot. decreases. Additionally, as the tilting progresses, the squeezed maximum results slightly shifted backward. Only samples with Fe³⁺/Fe_tot. ranging between 0.74 and 0.40 collapse on the same master curve, while for Fe³⁺/Fe_tot. $\le$ 0.35, the scaling gradually drops. Plot in Fig. 3b shows graphically the aforementioned discussions regarding the variation of the $\omega_{S}$ values normalized to those of the $\omega_{{{\text{BP}}}}$ from their lowest values. This result highlights a different behavior of the VDoS. One possible explanation for this behavior is a gradual change in the system’s coordination, which is thought to take place at Fe³⁺/Fe_tot. ~ 0.35. However, this is only a general conclusion that holds up to a significant modification to the local environment of iron.

Fig. 3

a Master curve of the boson peak, obtained as discussed in the text. The black arrows emphasize the tilting of the right flank and the backward shift of the BP maximum. b Gradient pentagons correspond to the BP frequency ($\omega_{{{\text{BP}}}}$) as function of Fe³⁺/Fe_tot. ratio obtained from reduced Raman spectra and gray diamonds are the values of the squeezing frequency $\omega_{S}$ of I^red(${\upnu }$). Gray dashed and light-green solid lines are guides for the eye

4 Discussion and conclusions

Among the different techniques used to investigate the SRO behavior of iron in silicate glasses (i.e., Mössbauer, XANES and EXAFS), data coming from similar pantelleritic glasses suggest that Fe³⁺ is mostly fourfold coordinated (^IVFe³⁺) with respect to Fe²⁺ that results fivefold, and in a small fraction, fourfold (^IV−VFe²⁺) [42]. Indeed, there are also evidence of fivefold coordinated ^VFe³⁺ suggesting that Fe³⁺ contribution in being NF is probably less relevant than expected [5, 13]. Figure 4 shows the HV reduced low-frequency spectra of samples with unknown T_g^peak FSP1, FSP5 and FSP8. When introduced in the analysis, they scale notably with the other pantelleritic glasses, as expressed by the $\omega_{{{\text{BP}}}}$ vs. Fe³⁺/Fe_tot plot in Fig. 4b. This confirm that a clear distinction of the iron coordination state (average of four- and sixfold or dominant fivefold) is not straightforward even in this case. However, the increasing tetrahedral network units (and an increasing polymerization of the structure) by adding Fe³⁺, induce a negative shift of the BP as reported in Fig. 4b pointing to an increasing network connectivity due to ^IVFe³⁺ as already reported for simple model systems [43].

Fig. 4

a HV reduced Raman spectra of the low-frequency region for samples FSP1 FSP5 and FSP8. Labels on the right of the panel report Fe³⁺/Fe_tot. b Relationships between the BP position $\omega_{{{\text{BP}}}}$ and the Fe³⁺/Fe_tot. with the glass transition temperature peak (T_g^peak) as secondary axes (gray symbols). Black diamonds are $\omega_{{{\text{BP}}}}$ vs. Fe³⁺/Fe_tot from panel (a). The gradient bar represents the fragility range estimated by the model of Ref. [27]

An equivalent trend is observed between the glass transition temperature peak (T_g^peak) and $\omega_{{{\text{BP}}}}$ Fig. 4b. This behavior suggests again a clear relationship between the yet observed polymerization of the glass structure (retrieved by adopting the band deconvolution of the high-frequency spectral region) [17]. This seems to be true for both the same glasses and similar systems [17, 44] and the effect on the MRO through the shift of $\omega_{{{\text{BP}}}}$ in basaltic glasses where iron replaces alkalis [20].

Answering the question: is the BP an indicator of the polymerizing potential of iron?

Having in mind that:

• (1) Increasing the total iron content (disregarding its ratio) the elastic medium experiences a general softening.

• (2) Increasing Fe³⁺ content within the total iron content, viscosity, fragility and glass transition temperature increases.

• (3) The strong probability that Fe³⁺ is not totally NF, the eventual presence of ^VFe³⁺ may have a nonlinear effect on the glass connectivity.

The answer is yes, the negative tendency of $\omega_{{{\text{BP}}}}$ vs. Fe³⁺ accompanied by a scaling drop around a certain ratio (Fe³⁺/Fe_tot ~ 0.35 in this case) testify a gradual lowering of tetrahedral structures. This behavior could be fostered by an excess of ^VFe-site geometries (trigonal bipyramid, square-based pyramid and distorted sites, see [13]). Thus, part of these [FeO₄] polyhedra contribute together with the [SiO₄] tetrahedra to increase the network connectivity, providing a negative charge (from different valence between Si⁴⁺ and Fe³⁺) to be balanced by NM cations. Following, increasing content of  Fe³⁺ progressively tightens the network by shortening and strengthening the bonds comparing to what Si⁴⁺ and Al³⁺ do. Conversely, Fe²⁺ increases the number of NBO, providing more ionic bonds [5] and bring the system to a chemical densification-like process through the migration of Fe²⁺ inside the void structures of the network (rings). Furthermore, optical absorption and low-T Mössbauer spectroscopies studies revealed the presence of isolated Fe³⁺ or Fe²⁺–O²⁻–Fe³⁺ and Fe³⁺–O²⁻–Fe³⁺ interactions identified as clusters in iron-bearing silicate glasses [45]. Thus, it is reasonable to assume that moving away from oxidized samples toward the reduced ones, the network would experiences a toning toward fivefold structures [5, 11, 46]. Since T_g^peak is strongly related to the mean atomic bonding energy, it is expected to increases when weakest bonds are replaced by stronger ones, which is what it might expected during the transition from five- to fourfold.

The above discussed effects are nonlinearly mirrored by the BP shift induced by a diminishing of the pseudo-Brillouin zone’ size and an enlarging of the correlation length as Fe³⁺ (as NF) increases. This trend is furtherly confirmed by the new introduced set in Fig. 4a, where the $\omega_{{{\text{BP}}}}$ of the samples FSP8 and FSP5 shows the same behavior of the master-trend, while the result of FSP1 sample suggests that by increasing the Fe³⁺/Fe_tot the $\omega_{{{\text{BP}}}}$ turns ever-more backward compared to the reduced samples. Thus, the system is capable of forming stronger connections than it currently does in the given conditions Fe³⁺/Fe_tot $\le$ 0.83.

The range within which the shift of the BP occurs, suggests also that the melt fragility m seems only marginally affected by the Fe³⁺/Fe_tot. ratio (~ 13% in 0.24 < Fe³⁺/Fe_tot. < 0.74). Such implications seem reasonable since just ~ 7% of the general chemical variation is affected. However, the remarkable variation of the BP intensity reflects changes in the elasticity of the glass [47] which is definitely related to the atomic packing density and to the bond strength [48]. These variation ultimately impact the visco-elastic response of the supercooled liquid once overcomes the T_g, influencing perhaps the intensity and magnitude at which explosive-effusive eruptions may occur and evolve [49].

To conclude, the boson peak of this pantelleritic system weakens and shifts to higher frequency as the Fe³⁺/Fe_tot. decreases mirroring a structural hardening-like process and modifications of the elastic continuum in the MRO. The same trend observed in the T_g^peak supports the hypothesis in [20] and represents a far more general result.

The scaling of the boson peak, which results in an inflexion of the high-frequency tail, may provide a different way to analyze the structural transformation of the medium-range order of natural iron-bearing glass system.