Volume 26, Issue 1 e2024GC011846
Research Article
Open Access

Nanolite Crystallization in Volcanic Glasses: Insights From High-Temperature Raman Spectroscopy and Low-Temperature Rock-Magnetic Analysis

Dmitry Bondar

Corresponding Author

Dmitry Bondar

Bayerisches Geoinstitut, University of Bayreuth, Universitätsstraße 30, Bayreuth, Germany

Technology and Sustainability for Ceramics (ISSMC), National Research Council (CNR), Institute of Science, Faenza, Italy

Correspondence to:

D. Bondar and D. Di Genova,

[email protected];

[email protected]

Contribution: Conceptualization, Formal analysis, ​Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization

Search for more papers by this author
Aurèlien Canizarès

Aurèlien Canizarès

CNRS, CEMHTI UPR3079, Univ. Orléans, Orléans, France

Contribution: Formal analysis, Writing - review & editing

Search for more papers by this author
Dario Bilardello

Dario Bilardello

Department of Earth and Environmental Sciences, Institute for Rock Magnetism, University of Minnesota, Minneapolis, MN, USA

Contribution: Formal analysis, Writing - original draft, Writing - review & editing

Search for more papers by this author
Pedro Valdivia

Pedro Valdivia

Bayerisches Geoinstitut, University of Bayreuth, Universitätsstraße 30, Bayreuth, Germany

Technology and Sustainability for Ceramics (ISSMC), National Research Council (CNR), Institute of Science, Faenza, Italy

Contribution: Formal analysis

Search for more papers by this author
Alessio Zandonà

Alessio Zandonà

Institute of Non-Metallic Materials, Clausthal University of Technology, Clausthal-Zellerfeld, Germany

Contribution: Writing - review & editing

Search for more papers by this author
Claudia Romano

Claudia Romano

Dipartimento di Scienze, Università degli Studi di Roma Tre, Rome, Italy

Contribution: Writing - review & editing

Search for more papers by this author
Mathieu Allix

Mathieu Allix

CNRS, CEMHTI UPR3079, Univ. Orléans, Orléans, France

Contribution: Writing - review & editing

Search for more papers by this author
Danilo Di Genova

Corresponding Author

Danilo Di Genova

Technology and Sustainability for Ceramics (ISSMC), National Research Council (CNR), Institute of Science, Faenza, Italy

Correspondence to:

D. Bondar and D. Di Genova,

[email protected];

[email protected]

Contribution: Conceptualization, Data curation, Writing - original draft, Writing - review & editing, Supervision, Project administration, Funding acquisition

Search for more papers by this author
First published: 14 January 2025
Citations: 1

Abstract

High-temperature Raman spectroscopy offers a cost-effective alternative to extensive infrastructure and sensitive instrumentation for investigating nanolite crystallization in undercooled volcanic melts, a key area of interest in volcanology. This study examined nanolite formation in anhydrous andesite melts in situ at high temperatures, identifying distinct Raman peaks at 310 and 670 cm−1 appearing above the glass transition temperature. The initial amorphous glass remained stable up to 655°C, beyond which Fe-Ti-oxide nanolites progressively formed at higher temperatures, as also confirmed by complementary XRD analysis. The evolution of the 310 cm−1 peak depends only on the magnitude of nanolite crystallization, while the intensity of the 670 cm−1 peak is temperature-dependent and challenging to observe above 500°C. Complementary low-temperature rock-magnetic analyses confirmed Fe-Ti-oxide nanocrystallization with nanolites around 20 nm in diameter. The study tested lasers of different wavelengths (from 355 to 514 nm) and found the green laser to be the most effective for collecting spectra at both room and high temperature. However, above 720°C, black body radiation significantly hinders Raman observation with the green laser when using a non-confocal setup and analyzing poorly transparent samples. If higher temperature measurements are desired, switching to a confocal setup and using lower wavelength lasers should be considered. This research offers a protocol for studying nanolite formation and melt dynamics at high temperatures, providing a foundation for future studies of volcanic processes.

Key Points

  • Nanocrystallization in Fe-bearing volcanic melts has been observed for the first time using in situ Raman spectroscopy at high temperatures

  • Fe-Ti-oxide nanolites form and progressively grow with increasing temperature and time, reaching 2.6 vol% and 20 nm in mean diameter

  • The 310 cm−1 peak solely depends on the degree of nanolite crystallization, while the 670 cm−1 peak is strongly temperature-dependent

Plain Language Summary

Understanding how tiny crystals, known as nanolites, form in volcanic melts is crucial for studying volcanic processes and predicting eruptions. However, observing these crystals at high temperatures typically requires sophisticated and expensive equipment. Our research demonstrates that Raman spectroscopy, a relatively accessible technique, can effectively study nanolite formation in volcanic melts, specifically in anhydrous andesite. In our experiments, we heated andesite samples to temperatures above the glass transition, where the material begins to soften. As the temperature increased beyond 655°C, we observed the formation of nanolites through specific Raman signals. These observations show that nanolites continue to develop both as temperatures rise and as the duration of exposure to heat increases. This study presents a reliable methodology for investigating high-temperature nanolite formation, offering valuable insights into volcanic melt dynamics and providing a foundation for future research in volcanology.

1 Introduction

The study of crystallization timescales in volcanic melts is central to understanding magma transport, since they play a pivotal role in degassing, ascent dynamics and ultimately the eruptive style of volcanoes (Arzilli et al., 2019, 2022; Cashman & Blundy, 2000; Di Genova, Brooker, et al., 2020; Julia E. Hammer & Rutherford, 2002; Knafelc et al., 2022; Matsumoto et al., 2023; Mujin & Nakamura, 2014; Noguchi et al., 2006; Yoshida et al., 2022). Furthermore, melt crystallization timescales are also crucial in experimental studies that measure melt viscosity, particularly around the glass transition temperature (Tg) (Liebske et al., 2003; Neuville et al., 1993; Okumura et al., 2022; Richet et al., 1996; Valdivia et al., under review; Vetere et al., 2006). This is because the formation of nanocrystals can significantly hinder the accurate measurement of the true melt viscosity (Valdivia et al., under review).

Of particular interest here is the in situ Raman spectroscopic analysis of nanocrystal formation in an andesite melt around the glass transition temperature. As a major player in the continental crust formation and a frequent product of volcanic arcs (Reubi & Blundy, 2009), andesite magma physical properties have to be taken into account for both Earth's differentiation models (Melekhova et al., 2017; Reubi & Blundy, 2009) and volcanic hazard assessments (Pallister et al., 1992; Rampino & Self, 1982). Andesite magmas are often rich in micrometric crystals (Blundy & Cashman, 2001; J. E. Hammer et al., 2000; Reubi & Blundy, 2009) and, interestingly, smaller—nanometric—crystals (Mujin & Nakamura, 2014) known as nanolites (Mujin et al., 2017; Sharp et al., 1996).

Currently, in situ observation of nanolite crystallization in undercooled volcanic melts presents challenges. It often requires large infrastructures such as synchrotron light sources (Di Genova, Brooker, et al., 2020; Okumura et al., 2022) or expensive and sensitive instrumentation such as transmission electron microscopes combined with a heating stage (Valdivia et al., under review). Conversely, Raman spectroscopy has been established as a fast and cost-effective tool to identify nanocrystals in erupted and synthetic glasses at room temperature (Allabar et al., 2020; Cáceres et al., 2020; Di Genova, Brooker, et al., 2020; Di Genova, Kolzenburg, et al., 2017; Di Genova, Sicola, et al., 2017; Di Genova, Zandona, & Deubener, 2020; Di Genova et al., 2018; Giordano et al., 2021; González-García et al., 2021; Kleest et al., 2020; Scarani et al., 2022; Schiavi et al., 2018; Valdivia et al., 2023). Compared to other methods, Raman spectroscopy offers several advantages: it requires minimal sample preparation, allows for sample recovery, and provides high spatial resolution, simultaneously acquiring chemical information on glass composition (Di Genova et al., 2015; González-García et al., 2020), iron oxidation state (Di Genova et al., 2016; Di Muro et al., 2009; Le Losq et al., 2019), H2O (Behrens et al., 2006; Di Genova, Sicola, et al., 2017; Le Losq et al., 2012; Schiavi et al., 2018), SO2 and CO2 content (Morizet et al., 2013, 2017).

Despite these advantages, no studies have yet focused on high-temperature, in situ observation of nanolite formation in volcanic melts using Raman spectroscopy. This work aims to bridge this gap. We used a Raman microscope coupled with a heating stage to investigate nanolite formation in anhydrous andesite melts at high temperature. Our study was complemented by high-temperature X-ray diffraction (HT-XRD) as well as room and low-temperature rock-magnetic analyses, which have been long-known within the rock-magnetic community to investigate magnetic nanoparticles (e.g., Mullins & Tite, 1973; Néel, 1949), further supporting the tendency of the andesite melt to undergo Fe-Ti-oxide nanocrystallization. This is evidenced by distinct Raman features that may be influenced by temperature. This work offers an optimal protocol for studying the timescale of nanolite formation in volcanic melts using Raman spectroscopy, laying the foundation for further studies of nanolite formation and melt dynamics at high temperatures.

2 Materials and Methods

2.1 Synthesis of Starting Material

The starting material (AND100) is a synthetic anhydrous andesitic glass (Table 1) that mirrors the melt composition of Sakurajima volcano (Okumura et al., 2022; Valdivia et al., under review). The glass was obtained by mixing powdered reagents (SiO2, TiO2, Al2O3, Fe2O3, MnO, MgO, CaCO3, Na2CO3, K2CO3, and P2O5). The mixture was placed in an alumina crucible and subjected to an overnight heat treatment at 900°C to remove CO2 from the carbonate compounds. After decarbonization, the material was transferred to a Pt crucible and melted for 24 hr at 1400°C. The resulting melt was rapidly cooled in water to prevent crystallization. The resulting quenched glass was crushed into powder using a stainless steel percussion mortar and manually mixed before undergoing a second melting step to achieve chemical homogenization. The second round of melting at 1400°C lasted for 4 hr, after which the crucible was immersed in water for rapid cooling.

Table 1. Chemical Composition of AND100 Synthetic Anhydrous Andesitic Glass (wt.%) by Electron Microprobe Analyzer
SiO2 TiO2 Al2O3 FeOtot MnO MgO CaO Na2O K2O P2O5
n = 10 60.38 0.79 16.69 6.77 0.17 3.00 6.62 3.50 1.58 0.18
σ 0.36 0.04 0.18 0.12 0.03 0.07 0.09 0.15 0.07 0.05

2.2 Electron Microprobe Analyses (EMPA)

The chemical composition was determined utilizing a JEOL JXA-8200 electron microprobe at the Bayerisches Geoinstitut (University of Bayreuth, Germany). Measurements were conducted at 15 kV voltage, 5 nA current and 20 s of counting time under a defocused 10 μm beam. A total of 10 points were collected to account for possible heterogeneities. Synthetic wollastonite (Ca, Si), periclase (Mg), hematite (Fe), spinel (Al), orthoclase (K), albite (Na), manganese titanate (Mn, Ti), and apatite (P) served as standards. Notably, sodium and potassium were analyzed first to mitigate alkali migration effects, as outlined by Hughes et al. (2019).

2.3 High-Temperature Raman Spectroscopy (HT Raman)

High-temperature Raman spectroscopy analyses were conducted using a Renishaw InVia Qontor Raman spectrometer at CEMHTI (Orleans, France). Spectra were obtained in UV configuration at 355 and 405 nm, employing a 3,600 mm−1 grating and a 15x NA 0.3 objective within the range of 180–2,030 cm−1. Additionally, spectra were acquired in the visible range at 488 and 514 nm, utilizing a 20x NA 0.35 objective, covering the ranges of 150–2,250 cm−1 and 150–2,000 cm−1, respectively. The lasers used were a Cobolt Zouk™ 355 nm laser delivering 20 mW (12 mW under the objective), Ondax 405 nm LM series (40 mW), Coherent Sapphire 488 nm (50 mW), and Cobolt Fandango™ 514 nm laser delivering 50 mW (30 mW under the objective). Acquisition parameters involved an integration time of 120 s at room temperature and 60 s at high temperature, and a laser power ranging from 20 to 50 mW. A non-confocal configuration with a slit size of 65 microns was employed to gain signal intensity from a broader sample area. A Savitzky–Golay filter was employed (11 points; second-order) to smooth the Raman data presented here in order to improve data visualization. A Linkam TS1500 heating stage was used to control the sample temperature in an air atmosphere. No specific sample preparation or mounting was required, allowing the samples (∼1 mm3 size) to be analyzed directly. The heating rate applied was typically 20°C min−1, except for the Experiment 3, where it was 10°C min−1 and except for the last 100°C before reaching the target temperature in experiments 4 and 5 (Table 2), where it was set at 5°C min−1. The sample was maintained at the target temperature for 60 min in all experiments, except for Experiment 3, where the isothermal hold lasted 30 min. The cooling rate employed was consistently 10°C min−1 for all experiments. For measurements conducted during heating or cooling, the temperature was not held constant; spectra acquisition proceeded concurrently, representing an average temperature range of 5–20°C depending on the applied rate.

Table 2. A Summary of In Situ High Temperature Experiments
Experimenta Target T, °C Laser λ, nm Source laser power, mW
1 655 488 (blue) 20–50
514 (green) 20–50
2 723 514 (green) 20
3 723 514 (green) 20
4 808 514 (green) 50
5 808 355 (HF UV) 20
  • Note. The Experiment 3 was the only one with nanolite-bearing starting material, produced in Experiment 2.
  • a The actual order of experiments was 4, 5, 1, 2, 3. Note that we decreased the green laser power from 50 to 20 mW (see Section 3.4 for details) and tested the blue laser power of 20 and 50 mW (see Section 3.5 for details) at high temperature in Experiment 1.

2.4 High-Temperature X-Ray Diffraction (HT-XRD)

The crystallization sequence of sample AND100 was characterized by HT-XRD using a Malvern Panalytical Empyrean diffractometer (Cu Kα1/2 radiation) in Bragg-Brentano geometry, equipped with an Anton Paar HTK 1200N heating chamber. In order to perform the measurement, the starting glass was polished into a 2-mm thick plane-parallel platelet. The time-temperature protocol involved heating at 20°C min−1 and isothermal holdings at 650°C, 715°C, and 800°C, during which diffractograms were recorded for a total duration of 1 hr in the range 17–70°2θ (acquisition step: 0.026°2θ). Data analysis was performed using the software HighScore Plus (Panalytical), using suitable references from the Crystallography Open Database (COD, see Gražulis et al., 2009), provided in the text using their respective COD ID numbers.

2.5 Fourier-Transform Infrared Spectroscopy (FTIR)

A visible range spectrum was collected at room temperature using a Bruker IFS 125 high-resolution spectrometer with a Bruker IR microscope employing a xenon arc light source, quartz beam-splitter and Si detector at BGI (Bayreuth, Germany). The double polished section with a thickness of 8 μm made from sample produced in Experiment 5 (Table 2) was analyzed by FTIR. The spectrum was recorded in the 10,000–30,000 cm−1 range with an 8 cm−1 resolution. Five hundred scans were accumulated for the background and two thousand scans were accumulated for the sample.

2.6 Differential Scanning Calorimetry

We conducted a Differential Scanning Calorimetry (DSC) measurements at TU Clausthal (Germany) to determine the glass transition temperature (Tg) of AND100. Approximately 15 mg of glass was placed in a Pt80Rh20 crucible under a constant N2 flow rate of 20 ml min−1. Calibration of the DSC instrument was achieved using melting temperatures and enthalpies of fusion of reference materials, specifically pure metals such as In, Sn, Bi, Zn, Al, Ag, and Au.

The experimental procedure followed the methodology outlined by Stabile et al. (2021). To eliminate the thermal history of the glass, a two-step thermal treatment (i.e., matching protocol) was applied. This involved an initial upscan at a rate of 20°C min−1 until the heat flow indicated the attainment of the glass transition interval. Subsequently, the sample was cooled to 100°C at a rate (qc) of 10 min−1. The determination of Tg was then carried out during the subsequent upscan with qh = 10°C min−1, maintaining thereby matching rates (qh = qc). Characteristic temperatures Tonset (655°C) and Tpeak (715°C) were extracted from the measured heat flow during the matching upscan (Figure 1).

Details are in the caption following the image

Measured heat flow as a function of temperature for AND100 glass at a heating rate of 0.17°C s−1 (10°C min−1) after cooling at 0.17°C s−1 through the glass transition region. The characteristic temperatures Tonset and Tpeak are shown.

Following theoretical principles discussed elsewhere (Al-Mukadam et al., 2020; Di Genova, Zandona, & Deubener, 2020; Di Genova et al., 2023; Yue, 2009; Zheng et al., 2017), viscosity values were derived from DSC data using the relationship between the heating rate (qh) of the measurement and the shift factors Konset and Kpeak (Di Genova, Zandona, & Deubener, 2020; Stabile et al., 2021), expressed in Equation 1:
log 10 η T o n s e t , p e a k = K o n s e t , p e a k log 10 q h ${\mathit{log}}_{10}\,\eta \left({T}_{onset,peak}\right)={K}_{onset,peak}-{\mathit{log}}_{10}\left({q}_{h}\right)$ (1)
where Konset = 11.20 ± 0.15 and Kpeak = 9.84 ± 0.20 (Di Genova, Zandona, & Deubener, 2020; Stabile et al., 2021). When qh is 10°C min−1, η(Tonset) = 1011.98 Pa s and consequently, Tonset ≈ Tg (for the glass measured here, Tonset = 655°C). Similarly, for qh = 10°C min−1, η(Tpeak) = 1010.62 Pa s at Tpeak = 715°C.

2.7 Magnetic Analyses

Magnetic analyses were conducted at the Institute for Rock Magnetism, University of Minnesota. Room Temperature (300 K) hysteresis measurements were performed on a Lake Shore Cryotronics (Westerville, OH) 8600 Vibrating Sample Magnetometer in 1.8 T maximum inductive fields B. Low-Temperature magnetic measurements were conducted on a Quantum Design (San Diego, CA) Magnetic Properties Measurement System XL. A hysteresis loop was acquired at 5 K in 2.5 T maximum B fields. Remnant magnetization curves were measured (in zero field) upon warming from 10 to 300 K in 5 K increments after field cooling the sample in a 2.5 T DC saturating field (FC magnetization), and after zero-field cooling the sample and applying a 2.5 T saturating isothermal remanent magnetization (SIRM) at 10 K (ZFC magnetization). In-field FC-ZFC experiments were conducted as above, but using 5 mT DC fields and measuring the FC and ZFC magnetization in applied 5 mT fields. Room temperature cooling and warming isothermal remanent magnetization curves (RTSIRM) were acquired after applying a 2.5 T SIRM at 300 K and measuring the remanent magnetization while 300-10-300 K temperature cycling in 5 K increments. All magnetization are reported in SI units of Amkg−1.

In-phase (χ’) and out-of-phase (χ”) frequency and temperature-dependent AC mass susceptibilities, χ(FT), units of mkg−1, were acquired between 10 and 300 K every 5 K in an applied magnetic field strength H of 238.7 Am−1 and frequencies (f) of 1, 10, 100 and 1,000 Hz. After preliminary analyses, the experiment was repeated between 5 and 80 K in 2 K increments to best-capture the low-temperature details; χ’ and χ” as a function of amplitude and temperature (χ(HT)) were acquired between 10 and 300 K using a measurement frequency of 100 Hz and applied field strengths of 19.9, 39.8, 79.6, 159.2, and 314.3 Am−1.

The “π/2 law” (Mullins & Tite, 1973; Néel, 1949) was calculated as χ”visc = −(π/2) × (δχ’/δln(f)). For very small viscous superparamagnetic SD grains governed using thermal activation processes, χ” and χ”visc are expected to be comparable. To better determine the χ” peak temperatures to be used for magnetic grain-size analysis, the χ” data were fitted and interpolated using positively skewed gaussian distributions between 5 and 80 K. Following Néel (1949) theory for thermally activated particles as described by Worm and Jackson (1999), particle blocking volumes Vb were calculated as
V b = 2 k b T B k M s ln τ τ 0 ${V}_{\mathrm{b}}=\frac{2{k}_{\mathrm{b}}T}{{B}_{\mathrm{k}}{M}_{\mathrm{s}}}\hspace*{.5em}\mathrm{ln}\left(\frac{\tau }{{\tau }_{0}}\right)$ (2)
where: kb is Boltzmann's constant (1.38649 × 10−23 in SI units); T is the temperature (in K); Bk is the grain microcoercivity (in T) calculated as 2.09 × Bc (Stoner & Wohlfarth, 1948; Worm & Jackson, 1999), where Bc is the coercive force (in T) determined from the 5 K hysteresis loop; Ms is the volume-normalized saturation magnetization (in Am−1) at 5 K, estimated as Ms-5K (Am2 kg−1)/92 (Am2 kg−1) × 480 (kAm−1), where Ms-5K is the mass-normalized saturation magnetization determined from the 5 K hysteresis loop, and 92 Am2 kg−1 and 480 kAm−1 are the mass- and volume-normalized saturation magnetization for stoichiometric magnetite at 300 K, respectively. The relaxation time τ and the time constant τ0 were determined from the linear Arrhenius relation between ln(1/f) (Hz−1) versus 1/T (K−1), where T is the temperature (K) of the χ” frequency-dependent blocking peaks: the activation energy of the particles (in eV) is determined as Ea = slope × kb (8.617333262 × 10−5 eV), while the intercept is ln(τ0) (s) (e.g., Carter-Stiglitz et al., 2006; Skumryev et al., 1999; Özdemir et al., 2009).

3 Results

The first four Sections 3.1-3.4 are based on measurements during which the green laser (514 nm) was employed, whereas the comparison of different laser wavelengths is provided in Section 3.5. Sections 3.6, 3.7 and 3.8 present HT-XRD, visible spectroscopy and magnetic analysis data, respectively.

3.1 Room-Temperature Raman Spectra

Figure 2 displays the Raman spectrum of the starting material AND100 utilizing a green laser (514 nm). Inspection of this spectrum verified the crystal-free amorphous nature of our sample (Di Genova et al., 2015, 2016; Di Muro et al., 2009; Schiavi et al., 2018). This observation stems particularly from the absence of distinctive features in the 660–690 cm−1 range, which is typically associated with the most prominent signal of Fe-Ti-oxides (Di Genova et al., 2018; Valdivia et al., 2023).

Details are in the caption following the image

Room temperature (RT) Raman spectra of the starting material (AND100) and subsequent spectra following the heating of the sample at 655, 723, and 808°C. Note that the sample heated at 723°C was later reheated at 723°C (723°C (R)): see text for details. As the heating temperature and duration (as seen in the 723°C (R) spectrum) increased, there was a noticeable enhancement in the Raman spectral features at 300–330 cm−1 and 660–690 cm−1, accompanied by a reduction at 850–1,150 cm−1.

Additionally, four room-temperature Raman spectra of AND100 were obtained after heating at Tg (655°C) as well as at 723 and 808°C (Figure 2). The Raman intensities at 300–330 cm−1 and 660–690 cm−1 increased as a function of temperature. Notably, the room temperature Raman spectrum after heating at Tg (655°C, Experiment 1) exhibited no significant changes in its spectral features. However, after heating at 723°C (Experiment 2), there was a moderate intensity increase in the regions 300–330 cm−1 and 660–690 cm−1 accompanied by a decrease in the Raman signal at 850–1,150 cm−1. Furthermore, the reheating (Experiment 3) of the sample produced in Experiment 2 to the same temperature (723°C) also resulted in a slightly increased Raman signal in the 660–690 cm−1 region and a slightly decreased Raman signal at 850–1,150 cm−1 (Figure 2). Finally, the room temperature Raman spectrum after heating at 808°C (Experiment 4) showed a substantial intensity increase in the regions 300–330 cm−1 and 660–690 cm−1 and a sizable decrease at 850–1,150 cm−1.

3.2 High-Temperature Raman Spectra

Raman spectra of AND100 were also collected using the green laser at high temperatures, namely during heating, isothermal hold and cooling segments. In Figure 3, we present Raman spectra of the AND100 sample during an isothermal hold at 723°C for 60 min (Experiment 2). The figure inspection reveals an enhancement in the intensity of the 300–330 cm−1 and 660–690 cm−1 regions, coupled with a reduction in the intensity of the 850–1,150 cm−1 region—its centroid shifted to lower wavenumbers, transitioning from 940 to 880 cm−1. It is noted that while the broad spectral feature centered around 300–330 cm−1 clearly increases in intensity, the emergence of the feature centered around 660–690 cm−1 appears less evident by simple visual inspection, resulting in a general intensity increase (also referred to as spectrum flattening) without a distinct peak (Figure 3). However, after subtracting from all spectra the first spectrum collected at the target temperature, both features become visible with comparable broadness (Figure 3).

Details are in the caption following the image

Raman spectra of the AND100 sample at 723°C for 60 min (Experiment 2). The spectra were normalized at around 500 cm−1. A linear baseline fit to the spectrum within the 1,940–1,430 cm−1 interval was subtracted. The inset covering the 180–850 cm−1 interval in the lower left corner shows the result of subtracting the initial spectrum collected at the target temperature from all subsequent spectra, isolating changes in spectral features from the underlying spectral shape.

The Raman spectra collected at room temperature after heating at 723°C and 808°C showed the presence of nanocrystals (Figure 2), exhibiting a distinctly sharp spectral feature at 660–690 cm−1 that was however broader or absent in the in situ spectra at the same temperature of 723°C (Figure 3). Given this apparent inconsistency, we decided to conduct additional spectral acquisitions during the heating process from room temperature (Experiment 3), using as starting material the same nanolite-bearing sample produced in Experiment 2 (Table 2, Figure 2). In this scenario, we observed a gradual decrease in the intensity of the peak within the 660–690 cm−1 range with increasing temperature from 25°C until its complete disappearance at around 500°C (Figure 4a). Interestingly, upon cooling, the aforementioned peak re-emerged similarly (Figure 4b). Note that, as mentioned above, its overall room-temperature intensity slightly increased as a result of being subjected to a second temperature hold at 723°C for 30 min. This phenomenon, namely the effect of temperature on the intensity of the 660–690 cm−1 peak, is further examined below in the discussion section.

Details are in the caption following the image

Raman spectra of AND100 nanolite-bearing glass showcasing the heating to 723°C (a) and subsequent cooling (b) in Experiment 3. This sequence demonstrates the vanishing/broadening and subsequent reappearance/sharpening of the 660–690 cm−1 spectral feature. The spectra were normalized at around 750 cm−1. A linear baseline fit to the spectrum within the 1,940–1,430 cm−1 interval was subtracted.

Conversely, the intensity of the 300–330 cm−1 region appears to be less sensitive to such temperature-related effects. We clearly observed the formation of this peak in Experiment 2 during the isothermal hold at 723°C (Figure 3) and at 720°C while increasing the temperature in Experiment 4 (Figure 5). In Experiment 3, where this peak was present before heating, its intensity remained essentially constant during heating, as well as during the temperature hold and cooling processes (Figure 4). It is noted, however, that upon increasing temperature in Experiment 4 (Figure 5), a spectrum flattening over a broad range (ca. 500–800 cm−1) centered around 660–690 cm−1 can also be recognized, likely corresponding to the 660–690 cm−1 peak observed at room temperature (Figures 3 and 4).

Details are in the caption following the image

Raman spectra of AND100 during heating from 710 to 740°C (Experiment 4). Spectra were normalized to the maximum intensity at ∼500 cm−1; some vertical offset was added for clarity.

3.3 Black Body Radiation Effect

Using the green laser in a non-confocal configuration, the background intensity increases when heating beyond temperatures of approximately 720°C (Figure 6) due to black body radiation (Planck, 1901). While keeping the sample at a target temperature, both the relative and absolute background intensity persistently remain at the same levels. Upon cooling, the background intensity decreases similarly.

Details are in the caption following the image

Raman spectra of AND100 during heating from 710 to 810°C (Experiment 4). Spectra were normalized to the maximum intensity at ∼500 cm−1.

3.4 Glass Alteration Due To Laser Irradiation

The 300–330 cm−1 and 660–690 cm−1 spectral contributions were not observed in the Raman spectra of sample AND100 collected at room temperature using the green laser after heating to Tg (655°C, Experiment 1) (Figure 2). However, spectra collected in situ at a constant temperature at Tg revealed the appearance of the broad 300–330 cm−1 and 660–690 cm−1 spectral contributions (Figure 7a), with a clear intensity increase of 300–330 cm−1 feature and a spectrum flattening over a broad range (ca. 500–800 cm−1) centered around 660–690 cm−1, respectively, similar to what was observed at 723°C (Figure 3). During the isothermal hold at Tg, the laser was relocated to another spot and Raman spectra were collected for another 10 min, observing the same phenomenon: the material transitioned from the seemingly unmodified spectrum of the starting material to that of a nanocrystallized melt/glass. Subsequent optical examination, however, revealed signs of sample modification (burning) caused by the laser (Figure 7c). Consequently, the changes in the Raman features (Figure 7a) can be attributed to laser-induced overheating of the sample, as was previously reported for other volcanic glasses (Di Genova, Sicola, et al., 2017). Indeed, after reducing the laser power from 50 to 20 mW, relocating to another spot, and collecting seven Raman spectra, we observed that these spectra exhibited spectroscopic features remaining constant over time (Figure 7b), resembling those of the starting material (Figure 2). The reason why a 50 mW green laser power led to glass burning in Experiment 1 (655°C), but did not alter the glass structure in Experiment 4 (808°C), as evidenced by the absence of a laser-induced pit and the absence of nanocrystallization at temperatures below 720°C (Figure 5), is not immediately apparent. We speculate that it may be associated with variations in sample focus, surface quality, geometry and/or thickness.

Details are in the caption following the image

Raman spectra of AND100 collected at 655°C during Experiment 1 using a green laser: (a) 50 mW laser power, seven spectra collected during 30 min; (b) 20 mW laser power, seven spectra collected during 10 min (red color) and the spectra from “a” (gray color). All the spectra were normalized to the maximum intensity at ∼500 cm−1. Additionally, a linear baseline fit to each spectrum in the 1,940–1,430 cm−1 range was subtracted. (c) Surface of AND100 altered by the green laser at 50 mW and by the blue laser at 50 and 20 mW in Experiment 1 (655°C).

3.5 Effect of Laser Wavelength

In addition to utilizing the green laser (514 nm) for the data collection, we explored the effects of three other laser frequencies on Raman features (Figure 8) at room and high temperatures: blue (488 nm) and two UV lasers (355 and 405 nm).

Details are in the caption following the image

Comparison between Raman spectra of AND100 collected with different lasers before (dashed lines) and after (solid lines) heating to 808°C in Experiment 5. The spectra were collected using four different laser frequencies.

Comparing the sample spectra collected with the green Raman laser before and after the experiment, we observed a significant decrease in the intensity of the silicate framework broad bands at 400–600 cm−1 and 850–1,200 cm−1 after treatment at 808°C. In contrast, highly intensive peaks at 300–330 cm−1 and 660–690 cm−1 became apparent (Figure 8).

The blue laser yielded similar results to the green laser, except that after the heating treatment, the initial 400–600 cm−1 and 850–1,200 cm−1 features were more distinctly observed, and the 660–690 cm−1 peak showed significantly lower intensity (Figure 8). On the other hand, the newly formed peak at 300–330 cm−1 was more pronounced than that in the green-laser spectrum. The higher-frequency UV laser (355 nm) also produced very similar results, with the exception that the 300–330 cm−1 and 660–690 cm−1 peaks were extremely weak and no nanolite signal was visible (Figure 8). However, spectra obtained with the lower-frequency UV laser (405 nm) exhibited dramatically different spectroscopic features compared to the other lasers (Figure 8). Given these observations, we decided to limit our high-temperature testing to the green laser (514 nm), the higher-frequency UV laser (355 nm) and the blue laser (488 nm).

A higher-frequency UV (HF UV) laser was employed to gather spectra throughout Experiment 5 (Figure 9). In contrast to the green laser (Figure 6), despite heating the sample to the same temperature of 808°C as in Experiment 4, we did not observe an increase in background intensity due to black body radiation. This observation aligns with the principles of black body radiation, as described by Planck (1901), where significant ultraviolet emission occurs only at much higher temperatures. Additionally, according to Wien's displacement law, the peak emission of black body radiation shifts toward shorter wavelengths, including the UV range, as temperature increases. Therefore, using a UV excitation source helps to minimize interference from thermal emissions, which are stronger in the visible and infrared regions at this temperature. Nevertheless, spectra obtained using the HF UV laser did not display significant variations, suggesting that the HF UV laser is relatively insensitive to the appearance of features related to Fe-Ti-oxides at 300–330 cm−1 and 660–690 cm−1 (Figures 8 and 9).

Details are in the caption following the image

Raman spectra collected with the higher-frequency UV (HF UV) laser during Experiment 5. Representative Raman spectra were selected from data collected at intervals of approximately every 200°C during heating from room temperature to 808°C, during the following cooling to room temperature, as well as every 20 min while maintaining the temperature at 808°C for one hour. The exact temperatures were 55, 235, 435, 635, 808, 808, 808, 808, 715, 515, 325, and 155°C, which led to 12 spectra in total. Nine spectra exhibit no discernible difference except for three labeled spectra, where the difference is not obvious. A linear baseline fit to the spectrum in the 1,940–1,430 cm−1 range, was subtracted, and the resulting spectra were normalized to the maximum intensity at around 300 cm−1.

The blue laser was tested in Experiment 1 (Table 2) at Tg (655°C). Initially, three spectra were collected from different spots (Figure 10): the first two spots with a laser power of 50 mW and the third one with 20 mW. The first spot coincided with the location where spectra were collected with the green laser for 30 min (Figure 7a), indicating some prior glass modification (burning). Subsequently, we moved to spot number two (Figure 10); however, after collecting the spectrum, we observed the formation of another pit (Figure 7c). Following this, we reduced the laser power from 50 to 20 mW and obtained one more spectrum at a spot number three (Figure 10). No burning of the glass occurred after collecting this spectrum, prompting us to initiate the collection of one spectrum per minute with the blue laser at 20 mW at the same spot. However, after several spectra, an increase in background intensity was noticed, and upon optical examination, a newly formed pit was observed, confirming glass burning even at 20 mW with a blue laser. In summary, the blue laser exhibited greater destructiveness compared to the green laser and proved to be much less sensitive to changes in the 660–690 cm−1 region both at room temperature (Figure 8) and high temperature (Figures 3 and 10). Although further test experiments are desired, the green laser appears to be a more suitable choice for investigating the evolution of Raman spectra of volcanic glasses at high temperatures.

Details are in the caption following the image

Raman spectra of AND100 acquired at 655°C (Tg) using the blue laser (see text for details). The extent of glass modification (burning) is reflected in the elevated background intensity attributable to black body radiation (see Section 3.3 for details). The spectra were normalized to the maximum intensity at approximately 1,100 cm−1.

3.6 High-Temperature X-Ray Diffraction Data

To gain further insight into the crystallization sequence of the sample AND100, an HT-XRD experiment was performed by collecting diffractograms during heating to 800°C and subsequent cooling (Figure 11). Due to the substantially longer time necessary for data acquisition (1 hr for each isothermal measurement), only a qualitative comparison with the above-presented Raman spectroscopy data is possible; nevertheless, the HT-XRD results exhibit very good agreement with Raman observations. The initially amorphous sample developed the first resolvable broad diffraction peaks at 715°C during heating, corresponding to the formation of titanomagnetite (COD ID: 1010109; Barth & Posnjak, 1932) nanocrystals with an average size of 3.5(5) nm, as estimated by Rietveld refinement. Further heating to 800°C induced an apparent growth and sharpening of the peaks related to titanomagnetite and the additional precipitation of diopside-like pyroxene solid solution crystals (COD ID: 9000336; Cameron et al., 1973). The phase assemblage remained essentially unaltered during cooling to room temperature.

Details are in the caption following the image

High-temperature X-ray diffraction results were obtained from the AND100 sample during heating up to 800°C and subsequent cooling down to room temperature. Labels: m for titanomagnetite (COD ID: 1010109; Barth & Posnjak, 1932) and d for diopside-like pyroxene solid solution (COD ID: 9000336; Cameron et al., 1973); * marks an artifact due to the corundum sample holder.

3.7 Optical Absorbance Data

Figure 12 displays the optical absorption spectrum of the andesitic glass after heat treatment at 808°C from Experiment 5. The spectrum exhibits a broad absorption band stretching across the visible region up to 700 nm. Collecting meaningful spectrum data below 400 nm is not possible due to the limitations of the detector, beam, and filter used. The obtained optical absorbance data were further analyzed in the discussion section.

Details are in the caption following the image

The absorption optical spectrum of AND100 from Experiment 5 is represented by a black solid line and polarized absorption spectrum of glaucophane with a peak corresponding to intervalence charge transfer (IVCT) transition centered at 620 nm from Burns (1993) is represented by a gray dashed line. The frequencies of three lasers used in the present study as well as the frequency of the red laser (633 nm) used in Di Genova et al. (2018) are shown by colored dashed lines.

3.8 Magnetic Characterization

The sample for the magnetic analyses was prepared in the furnace following an analogous treatment as for the sample produced in Experiment 5 (Table 2) except for the shorter dwell time at 808°C (30 vs. 60 min). Additionally, the cooling rate was determined by the furnace's heat conductivity upon shutdown and thus was not precisely measured.

The room temperature hysteresis loop of the sample is closed throughout, so that after saturation there is no measurable saturation remanent magnetization (Mrs) and coercivity (Bc) in zero field and the remanent hysteretic magnetization Mrh (Fabian, 2003), denoting coercivity distribution within the loop (pink curve in Figure 13a), is zero. The (ferromagnetic s.l.) saturation magnetization (Ms) is 2.4 Am2 kg−1 after correcting for the paramagnetic high-field slope using an approach to saturation technique (Jackson & Solheid, 2010) (Figure 13a). The 5 K hysteresis loop after paramagnetic correction displays appreciable Mrs (1.39 Am2 kg−1), Bc (104.7 mT), Mrh and increased Ms (3.8 Am2 kg−1, Figure 13b) in comparison to the room temperature loop.

Details are in the caption following the image

Magnetic hysteresis for the AND100 sample after heating to 808°C: (a) room temperature (300 K) hysteresis loop before (orange) and after (blue) paramagnetic correction, while the pink Mrh curve is the remnant hysteretic magnetization. (b) 5 K hysteresis loop, color-coded as above. Note the increased magnetization at 5 K (both 300 and 5 K loops are plotted using the same vertical scale) and width of the loop, accompanied by non-zero Mrh (see text for details).

The FC-ZFC remanence curves show low-temperature magnetization that decrease rapidly between 10 and ∼50 K, after which there is no remanence up to room temperature (300 K) (Figure S1a in Supporting Information S1). Likewise, the RTSIRM cooling curve remains flat and near-zero between 300 and ∼80 K, at which point the remanence increases to 10 K, with the sharpest slope <50 K (Figure S1b in Supporting Information S1). The RTSIRM upon warming largely retraces the cooling curve, with minimal separation between the two curves between ∼30 and 130 K and again above ∼200 K, which is attributed to measurement noise (Figure S1b in Supporting Information S1). The 5 mT in-field FC-ZFC curves differ from their remnant counterparts in that the 5 mT FC decreases more gently toward 300 K, following a ∼1/T relation. The 5 mT ZFC instead increases rapidly from 10 to 35 K, where it peaks and rapidly begins decreasing past 40 K until 300 K, merging with the FC curve >55 K (Figure S2 in Supporting Information S1).

In-phase (χ’) and out-of-phase (χ”) χ(FT) susceptibility curves reveal low-temperature peaks between ∼30 and 60 K, which decrease in amplitude and increase in temperature with increasing measurement frequency. Above the peak temperatures, the susceptibilities decrease following a ∼1/T behavior for χ’ (Figure 14a), whereas χ” decreases somewhat more abruptly (Figures 14b and 14c, Figure S3b in Supporting Information S1). The higher resolution χ” data in Figure 14b were interpolated after best-fitting a positively skewed gaussian distribution for magnetic grain size analysis (see Discussion). χ”visc was calculated from the differences of the χ’ data for the measured frequencies (see methods section for details), and is in agreement with the corresponding measured χ” data (Figure 14c). χ(HT) measured at 100 Hz possesses similar curve shapes for both χ’ and χ”, but no amplitude-dependence is present, so that only one peak is observed between ∼30 and 50 K in the raw, unfitted χ’ and χ” data, respectively (Figure S4 in Supporting Information S1), equivalent to the 100 Hz χ(FT) data.

Details are in the caption following the image

Frequency-dependent susceptibility as a function of temperature χ(FT) of sample AND100 after heating to 808°C. (a) In-phase susceptibility χ’ acquired between 10 and 300 K in 5 K increments, using AC frequencies of 1, 10, 100, and 1000 Hz, revealing frequency-dependent blocking peaks located between ∼30 and 60 K, and a constant 1/T decrease up to 300K. (b) Out-of-phase susceptibility χ” acquired between 5 and 80 K in 2 K increments for the same four measurement frequencies (labeled “χ” data”) and fitted with skewed Gaussian curves (labeled “interpolated fit”), revealing frequency-dependent blocking peaks identified between 27.7 K (1 Hz) and 38.3 K (1000 Hz), above which χ” goes to zero. (c) Comparison between the incremental averages of χ” acquired at the four measurement frequencies (χ” Av. 1–10 Hz, χ” Av. 10–100 Hz, χ” Av. 100–1000 Hz) and the corresponding χ”visc curves calculated from the χ’ high-resolution data (χvisc. 1–10 Hz, etc.), and respective residuals. (d) Arrhenius relation between the logarithm of the inverse of the measurement frequencies over the inverse of the χ” blocking peak temperatures determined from the fitted data in (b). From the slope of the linear fit, the activation energy of the particles (in eV) is determined as Ea = slope × Kb (8.617333262 × 10−5 eV), while the intercept is ln(τ0) (s) (e.g., Carter-Stiglitz et al., 2006; Özdemir et al., 2009; Skumryev et al., 1999).

4 Discussion

4.1 Thermal Behavior and Nanocrystallization of AND100 Glass Through HT Raman

The Raman spectrum of the AND100 starting glass reveals two distinctive features: one within the 300–630 cm−1 range and another between 830 and 1,250 cm−1, both linked to the aluminosilicate network (Rossano & Mysen, 2012). The 300–630 cm−1 region is attributed to vibrations of bridging oxygens (BO) within Si-O-Si bridges and ring structures—three-membered, four-membered, five-membered, six-membered or higher-membered rings of tetrahedra present in silicate networks (Barrio et al., 1993; Galeener & Geissberger, 1983; P. McMillan & Piriou, 1982; P. F. McMillan & Wolf, 1995; B. Mysen, 2003; Pasquarello et al., 1998; Seifert et al., 1982; Umari et al., 2003). The 830–1,250 cm−1 range offers insights into the vibration of T-O stretching in different Qn tetrahedral units, in which a network-forming element such as Si4+, Ti4+, Al3+, and Fe3+ is surrounded by four O2− atoms. The Qn species are defined by the number (n) of bridging oxygens, ranging from 0 to 4 (P. McMillan, 1984a; B. Mysen, 2003; Rossano & Mysen, 2012). Additionally, extensive literature indicates that in this region, the structural influence of network-modifying or charge-balancing cations can be deduced (Bell & Dean, 1972; Bondar et al., 2022; Furukawa et al., 1981; P. McMillan, 1984b; B. Mysen, 2003; B. O. Mysen & Toplis, 2007).

The AND100 Raman spectrum (Figure 2) exhibits the same two amorphous features observed in andesite glasses from the literature. Our spectrum was compared with that from Montserrat andesite (MSA) (Di Genova et al., 2015), characterized by virtually the same chemical composition (SiO2 = 60 ± 0.4 wt.%, Al2O3 = 17.3 ± 0.7 wt.%, and FeOtot. = 6.5 ± 0.2 wt.%). The relative intensity of the two bands is similar, typical for mildly polymerized glass (NBO/TAND100 = 0.34 after B. O. Mysen (1988)). Additionally, we compared the Raman spectrum of AND100 with that of another andesite documented in the literature (A040417) (Schiavi et al., 2018), sharing similar SiO2 and Al2O3 content but with a lower FeOtot. (4.8 wt.%). We observed that the 830–1,250 cm−1 region for A040417 exhibited lower intensity, as anticipated for glasses relatively depleted in FeOtot (Di Genova et al., 2016; Di Muro et al., 2009), assuming that the iron speciation remains relatively unchanged. Finally, as the AND100 Raman spectrum lacks the established “nanocrystal (i.e., nanolite)” peak at ∼670 cm−1 (Di Genova, Sicola, et al., 2017), we conclude that the AND100 Raman spectrum represents the reference spectrum for an andesite free of nanolites.

The room temperature Raman spectra of AND100 glasses, after exposure to high temperatures (Figure 2), provide valuable insights into the temperature-induced alterations in melt structure within the timeframe of our experiments. Upon comparison with the reference spectrum of nanolite-free andesite (Figure 2), an examination of the Raman spectrum from the sample subjected to 655°C for 60 min indicates no observable changes in the glass structure. This observation aligns with the measured Tg of 655°C of the AND100 melt. Additionally, we employed the chemically-based viscosity models proposed by Giordano et al. (2008) (GRD) and Langhammer et al. (2022) (LDS) to calculate Tg of AND100. The GRD model yields a Tg of 699.6°C and the LDS model provides 687.3°C. Recent literature, as suggested by Valdivia et al. (2023) and Scarani et al. (2022), indicates that these models may tend to overestimate the viscosity of volcanic melts due to the impact of nanocrystallization (Di Genova, Sicola, et al., 2017; Di Genova, Zandona, & Deubener, 2020) during the viscosity measurements used to develop these models. Valdivia et al. (under review) have provided an experimentally constrained viscosity model for AND100 and it was found that at 699.6°C and 687.3°C, the melt viscosities were 1010.62 and 1010.97 Pa s, respectively. Thus, the literature-derived Tg may not accurately reflect the Tg of the AND100 melt.

Our results show (Figure 5) that nanocrystallization of the AND100 melt occurs within minutes between 655 and 720°C. In line with our observations, Valdivia et al. (under review) reported a time-dependent viscosity increase at ∼660°C, from 1012 to 1012.70 Pa s over 90 min. Here, this is mirrored by the appearance (Figures 3 and 5) of vibrational features assignable to Fe-Ti-oxides such as titanomagnetite (Di Genova, Zandona, & Deubener, 2020; Jubb & Allen, 2010; Shebanova and Lazor, 2003a, 2003b; Stabile et al., 2021), which became distinctly identifiable upon cooling our samples to room temperature (Figure 4). HT-XRD results closely align with Raman observations, revealing the first resolvable broad diffraction peaks at 715°C during heating, which indicate the formation of titanomagnetite (Figure 11). The nanocrystallization of titanomagnetite is illustrated by the emergence of Raman contributions at 310 and 670 cm−1 (Figures 2-5). However, while the peak at 670 cm−1 exhibits strong temperature dependence (Figures 3-5) and becomes challenging to observe above 500°C, the 310 cm−1 peak appears to be less sensitive to temperature (Figures 3-5).

4.2 Temperature and Laser Frequency Dependence of the 660–690 cm−1 Raman Feature

To explain the strong temperature dependence (Figure 4) and laser frequency dependence (Figure 8) of the peak intensity at 670 cm−1 and to explore whether these dependencies are interconnected, we considered four hypotheses. Below, we present these hypotheses and discuss the measurements we conducted to test two of them.

The first hypothesis was loss of magnetic ordering above the (titano)magnetite Curie temperature (580°C for stoichiometric magnetite, Fe3O4, and decreasing with increasing titanium content, for example to 500°C for a composition of ∼Fe2.85Ti0.15O4 (e.g., Dunlop & Özdemir, 1997; Lattard et al., 2006)), causing the 670 cm−1 Raman peak to only be observable in the nanolite-bearing samples below their Curie temperature. For this purpose, we conducted magnetic characterization experiments to determine the nanolite grain size and concentration, and, if appropriate, their Curie Temperature. Magnetic hysteresis at room temperature (Figure 13a), as well as the remnant FC-ZFC and RTSIRM curves (Figure S1 in Supporting Information S1) indicate that the iron oxides present are superparamagnetic, not capable of holding any magnetic remanence at room temperature and conversely decaying rapidly after a low-temperature saturation magnetization is removed (Figures S1a and S1b in Supporting Information S1). Conversely, at 5 K, the hysteresis loop possesses discrete coercivity Bc, indicating that the particles are blocked (Figure 13b), and the marked increase in saturation magnetization Ms and Bc between room temperature and 5 K (cf. Figures 13a and 13b) is consistent with superparamagnetic behavior. Assuming a magnetite composition for the small grain sizes expected (Di Genova, Brooker, et al., 2020; Di Genova, Sicola, et al., 2017), and using the Ms value for stoichiometric magnetite at room temperature (92 Am2 kg−1), a ∼2.6 vol.% magnetite content was estimated from the sample's Ms at room temperature.

The low-temperature increase of the 5 mT ZFC curve is consistent with a blocking peak for a distribution of superparamagnetic particles, which become paramagnetic ≥55 K where the 5 mT FC and ZFC curves converge and follow a ∼1/T decrease toward room temperature, consistent with Curie-Weiss paramagnetic behavior (Figure S2 in Supporting Information S1).

The frequency and temperature-dependence of the χ(FT) curves confirm the superparamagnetic nature of the particles (<25–30 nm, e.g., Dunlop, 1973). Frequency-dependent out-of-phase susceptibility χ” (Figure 14), in the absence of amplitude-dependence in the sample (Figure S4 in Supporting Information S1), can be entirely attributed to magnetic viscosity with relaxation times comparable to the AC field reversal interval. Likewise, the near zero χ” curves above ∼80 K for both χ(FT) and χ(HT) (Figures S3 and S4 in Supporting Information S1) indicate that no particles larger than the superparamagnetic (SP)—single domain (SD) threshold (∼25–30 nm, e.g., Dunlop, 1973) are present. Such an interpretation is confirmed by the agreement between the calculated χ”visc and the corresponding χ” intervals determined from the measured data (Figure 14c). Furthermore, the shapes of the χ’ and χ” curves, bearing a single, well-defined peak for each frequency, suggest a unimodal and narrow distribution of grain sizes of the same composition.

From the fitted χ”(FT) data (Figure 14b) and the derived Arrhenius relation (Figure 14d), an activation energy Ea of 0.06 eV was estimated for the modal grain distribution, as well as a time constant τ0 of 1.14 × 10−11 s, which is in agreement with estimates reported elsewhere for (titano)magnetite nanoparticle-bearing samples, 10−8–10−13 s ((Berndt et al., 2015; Worm & Jackson, 1999) and references therein). Characteristic relaxation times τ were estimated for 5 and 300 K, yielding 7.1 × 1049 s and 4.4 × 10−11 s, respectively, indicating that while the grains are fully blocked approaching absolute zero, they are superparamagnetic at room temperature. Assuming a magnetite composition, and using saturation magnetization values for stoichiometric magnetite, a modal grain size radius of 10.2 nm was estimated. Increasing titanium substitution would lower Ms (e.g., Dunlop & Özdemir, 1997), so that an increased grain size would be estimated from Equation 2. However, because the observed particle distribution is entirely superparamagnetic above ∼80 K, the grain sizes are below the SP-SD threshold, and thus the composition must be close to stoichiometric. Regardless, these analyses fully support a superparamagnetic grain size of the nanolites and thus invalidate our ability to verify the hypothesis of magnetic disordering above the Curie temperature through magnetic measurements.

The second hypothesis explains these dependencies in terms of the intervalence charge transfer (IVCT) between Fe2+ and Fe3+ in the magnetite structure. If this highly intensive band coincides with the frequency of the green laser, it would cause a Resonance Raman effect, thus multiplying the intensity of the 670 cm−1 peak, collected employing the green laser. In an attempt to determine the position of this band, we collected an absorption optical spectrum of the AND100 sample produced from Experiment 5 (Table 2). Unfortunately, the broad absorption band observed (Figure 12) is more likely to represent the glass matrix (e.g., Khalil et al., 2010; Wahab et al., 2020) and, if the IVCT band is present, it is likely hidden by this absorption. It is also extremely difficult to measure the absorption spectrum of a very opaque mineral such as magnetite in the visible region (Burns, 1993). However, the maximum of this IVCT band in other minerals reported in the literature corresponds to 600–850 nm ((Burns, 1993) and references therein)—Figure 12 reports this band in glaucophane as an example, and, due to the broadness of this band, it could potentially explain the more efficient resonance Raman effect while using the green laser. In fact, this finding aligns with and could also explain (Figure 12) a similar observation of Di Genova et al. (2018), where the intensity of the 670 cm−1 peak decreased as a function of increasing laser source frequencies: red, green and blue. Finally, for most of the minerals, the intensity of the Fe2+ → Fe3+ IVCT band decreases with increasing temperature (Table 9.3 in Burns (1993)), which should lead to a decrease in the resonance Raman effect and could potentially explain the observed decrease in the intensity of the 670 cm−1 peak with increasing temperature (Figure 4). However, further research, and in particular high-temperature visible range measurements are required to test this hypothesis.

The third hypothesis suggests that the high-temperature dependence of the peak intensity at 670 cm−1 (Figure 4) is due to changes in the oxidation state of magnetite caused by electron hopping between magnetite crystals and the glass matrix. When silicate glasses containing nanosized magnetite crystals are heated, the redox state of the system can be influenced by the temperature and the intrinsic properties of the materials. For instance, silicate melts heated under atmospheric conditions tend to become more reduced as the temperature increases ((Borisov et al., 2018) and references therein). If this holds true for silicate glasses, Fe3+ ions in the glass matrix are reduced to Fe2+, with a corresponding oxidation of the magnetite crystals to preserve redox balance, for example, forming maghemite. It is also noted that although the 670 cm−1 peak assigned to magnetite corresponds to Fe-O vibrations in tetrahedral sites, which are typically occupied by Fe3+ in perfectly ordered crystals, the cationic distribution in spinel structures, including magnetite, is strongly temperature dependent (Wiβmann et al., 1998) and complete ordering between tetrahedral and octahedral positions is rarely observed. Moreover, the nanocrystals are expected to contain non-negligible amounts of Ti3+/4+, which certainly affect charge balancing and preferential site occupation of iron species in the magnetite structure. Finally, crystals formed at deep supercooling by heating of glasses typically experience even higher degrees of metastable disorder than usual, which may contribute to the observed behavior. This oxidation process would not represent a conventional phase transition but rather a shift in electronic structure through electron hopping, occurring without structural reorganization and well below the melting temperature. This change would also be reversible, as evidenced by the reappearance of the magnetite peak at 670 cm−1 upon cooling (Figure 4). When comparing the Raman features of magnetite and maghemite (see Table 2 and Figures 4 and 6 in Jubb and Allen (2010)), it becomes evident that the magnetite peaks at 310 cm⁻1 and 672 cm⁻1 correspond to peaks at 365 and 700 cm⁻1 in maghemite, respectively. Additionally, the latter peak in maghemite becomes noticeably broader. Considering the possibility of a continuous solid solution between magnetite and maghemite, the nanocrystals in this study may exhibit intermediate spectra between these two endmembers depending on the temperature, rather than being fully oxidized to Fe₂O₃. This behavior could explain the changes in peak broadness and the smooth spectral transitions observed with temperature changes (Figure 4). If this hypothesis is correct, it suggests that nanolites may initially form as maghemite-like crystals at high temperature and then reduce to magnetite-like crystals upon cooling. However, further investigations, such as in situ high-temperature synchrotron measurements, are necessary to validate this hypothesis.

The fourth hypothesis suggests that the high-temperature dependence (Figure 4) and laser frequency dependence (Figure 8) of the peak intensity at 670 cm−1 can be attributed to the absorption edge of the glass matrix. As discussed previously, the 670 cm−1 peak intensity is significantly reduced when using the blue laser compared to the green or red lasers (Di Genova et al., 2018), which might suggest that Resonance Raman efficiency decreases when the excitation laser is closer to the UV absorption edge. At room temperature, the blue laser's wavelength is nearer to the UV absorption edge of the glass matrix, leading to reduced peak intensity. As the temperature increases, the absorption edge shifts to higher wavelengths (e.g., Higazy et al., 1988), gradually approaching the green laser wavelength, which could consequently diminish the 670 cm−1 peak's intensity at higher temperatures. Upon cooling, the absorption edge shifts back, possibly allowing the band to reappear. Further high-temperature spectral analyses of the glass matrix are necessary to validate this hypothesis, focusing on the dynamic changes in absorption behavior.

While these hypotheses provide potential explanations for the observed phenomena, no particular hypothesis is favored at this stage, and further research is necessary to determine which, if any, is correct.

4.3 Black Body Radiation in HT Raman: Challenges and Solutions

Finally, it was also demonstrated using a green laser in a non-confocal configuration that black body radiation could pose a challenge in discerning glass-related features at temperatures above 720°C (Figure 6). This low-temperature appearance of black body emission could be attributed to several factors, such as the weak cross-section of glass samples and the low numerical aperture of the long-distance microscope objective used. Additionally, compared to other studies where black body emission was observed at much higher temperatures (e.g., Daniel et al., 1995; Neuville & Mysen, 1996), the samples in this study were poorly transparent and black, leading to higher sample emissivity and the necessity to use significantly lower laser power to avoid sample modification, both resulting in a reduced sample signal to thermal background ratio. Furthermore, the non-confocal configuration employed in this study may have contributed to the lower temperature at which thermal emission interferes with the Raman signal, as noted by Daniel et al. (1995), who reported a temperature difference of 600°C between non-confocal and confocal setups (727°C vs. 1327°C). If investigations at higher temperatures are desired, one option is to use a different laser wavelength. As the laser frequency increases, the temperature at which black body radiation affects the spectrum also rises (see Section 3.5 for details). However, it is important to note that other laser wavelengths may not clearly discern glass and Fe-Ti-oxide-related features, as demonstrated using a higher-frequency UV laser (Figure 9). Switching to a confocal configuration is another option, though it may lead to lower signal intensity, which could be problematic given the low Raman scattering efficiency of glassy samples. Another factor to consider is sample heating by the laser: it was demonstrated in Sections 3.4 and 3.5 how highly energetic laser radiation could locally increase the sample temperature and lead to nanocrystallization. Daniel et al. (1995) also observed additional heating of glassy samples through absorption of the incident laser light, leading to excess background and an estimated temperature increase of about 20°C. Interestingly, the increase in the black body radiation in the present study (Figure 6) coincides with an appearance of a peak in the 300–330 cm−1 range (Figure 5). Possibly, growing crystals absorb the laser more efficiently compared to the glass, which could lead to an increasing temperature and more efficient crystallization, thus further promoting heating by the laser. Therefore, the temperature at which the intensity of the black body radiation increases could also depend on the investigated material. Hence, the laser power should be chosen to provide sufficient resolution for spectroscopic features without altering the glass structure by overheating the sample.

4.4 Contribution of HT Raman, HT-XRD, and Magnetic Analysis to Reactive Melt Studies

This HT Raman study underscores the challenges of investigating reactive volcanic melts, where rapid nanocrystallization and iron oxidation during viscosity and DSC measurements can affect results. It highlights the need for a rigorous, multifaceted approach, with careful sample checks both before and after measurements using techniques like SEM, Raman spectroscopy, and TEM, to ensure accurate viscosity data without interference from undetected crystallization (e.g., Scarani et al., 2022; Valdivia et al., 2023). In combination with these methods, HT Raman spectroscopy—supplemented by HT-XRD—offers a powerful tool for monitoring nanolite crystallization in volcanic melts in situ, tracking its evolution over time and temperature, and visualizing melt changes alongside viscosity or DSC measurements under a consistent time-temperature protocol. Magnetic measurements, as used in this study, also hold potential for detecting and quantifying nanolites and evaluating their size, especially when other techniques are unavailable. Together, these tools support a holistic approach, yielding more reliable conclusions and reducing the risk of misinterpretation from methodological shortcuts.

5 Conclusions

By subjecting an anhydrous andesite glass to high-temperature Raman spectroscopy measurements, we found the following:
  1. The Raman features of the initial amorphous glass remain unchanged from room temperature up to 655°C (Tg). However, as the temperature increased to 723 and 808°C, the glass underwent progressive crystallization, forming Fe-Ti-oxide nanolites, which was also confirmed by HT-XRD. These nanolites are estimated from low-temperature rock-magnetic experiments to be ∼2.6 vol% and ∼20 nm in mean diameter after crystallization at 808°C. Thus, Raman spectroscopy is a valuable tool for detecting nanocrystallization in volcanic melts. This is crucial for modeling volcanic processes that require the viscosity of crystal-free melt.

  2. Nanolite formation is marked by distinct peaks at 310 and 670 cm−1 in the Raman spectra. The evolution of the 310 cm−1 peak is solely dependent on the degree of nanolite crystallization, while the appearance and evolution of the 670 cm−1 peak are temperature-dependent, being challenging to observe above 500°C.

  3. Repeated heating of a nanolite-bearing glass above Tg induces further nanocrystallization.

  4. When using a green laser in a non-confocal configuration above 720°C to analyze poorly transparent glass, the intensity of black body radiation significantly increases, hindering the observation of Raman features. Potential solutions are to switch to a confocal setup or to use lasers with lower wavelengths.

  5. The performance of lasers at different wavelengths (HF-UV—355 nm, LF-UV—405 nm, blue—488 nm, and green—514 nm) was tested. The green laser proved most effective for collecting informative spectra at both room and high temperature, while UV lasers provided the least informative spectra. Additionally, the blue laser demonstrated the highest tendency to damage the sample.

Acknowledgments

DDG acknowledges the funding from Deutsche Forschungsgemeinschaft (DFG) project DI 2751/2-1, European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (NANOVOLC, ERC Consolidator Grant 101044772), and MUR-PRIN Project (CRYSTALlKIN, 2022L2APNR). D. Bilardello acknowledges the US National Science Foundation (NSF)-EAR 2153786 Instrumentation and Facilities grant to the Institute for Rock Magnetism. This is IRM publication #2406. We thank H. Keppler for the help with the visible range measurement, G. Helsch for the help with HT X-ray diffraction analysis, A. Withers for discussion and R. Njul for preparing double polished glass samples. Open access publishing facilitated by Consiglio Nazionale delle Ricerche, as part of the Wiley - CRUI-CARE agreement.

    Data Availability Statement

    Magnetic analyses, DSC, FTIR, Raman, and XRD data are archived through Zenodo data repository via Bilardello (2024) and Bondar (2024a, 2024b, 2024c, 2024d), respectively.