Glass Composition-Dependent Silicate Absorption Peaks in FTIR Spectroscopy: Implications for Measuring Sample Thickness and Molecular H2O

Abstract:

Fourier-transform infrared spectroscopy (FTIR) is often used to measure the H₂O and CO₂ contents of volcanic glasses. A key advantage of FTIR over other analytical techniques is that it can reveal not only total H₂O concentration but also H₂O speciation, i.e. how much H₂O is present as molecular H₂O (H₂O_m) and how much as hydroxyl groups (OH) bound to the silicate network. This H₂O speciation data can be used to investigate cooling rate and glass transition temperature of volcanic glasses, and to interpret H₂O contents of pyroclasts affected by partial bubble resorption during cooling or secondary hydration after deposition. FTIR in transmitted light requires sample wafers polished on both sides of known thickness. Thickness is commonly measured using a micrometer but this may damage fragile samples and in samples with non-uniform thickness, e.g. vesicular samples, it is difficult to position at the exact location of FTIR analysis. Furthermore, in FTIR images or maps of such samples it is impractical to determine the thickness across the whole of the analysed area, resulting either in only a selection of the collected data being processed quantitatively and the rest being unused, or results being presented in terms of absorbance, which does not account for variations in thickness.

It is known that FTIR spectra contain absorption peaks related to the glass aluminosilicate network at wavenumbers of ~2000, ~1830 and ~1600 cm^-1 [1]. These have been shown to be proportional to sample thickness at the analysis location for one obsidian composition with up to 0.66 wt% H₂O [2]. We test whether this calibration can be applied more widely by analysing a range of synthetic and natural glasses (andesitic to rhyolitic) to examine how the position and relative intensities of the different silicate absorption peaks vary with composition and H₂O content. Our data show that even minor differences in composition necessitate a unique calibration. Furthermore, importantly we show how interference between the ~1600 cm^-1 silicate peak and ~1630 cm^-1 H₂O_m peak can result in significant overestimation of H₂O_m, and demonstrate how the relationship between silicate peak intensities and sample thickness can be used to assess and correct for this interference.

[1] Newman et al (1986) Am Min 71, 1527-41

[2] Miwa & Toramaru (2013) Bull Volc 75, 685-97