The glass transition and non-Arrhenian viscosity of extremely depolymerized silicate melts

Properties of extremely depolymerized melts and glasses have always been elusive due to their inaccessibility in experiments. Information on viscosity, for example, was previously restricted to very high temperatures. Yet, theory and experimental details about peridotites suggest that highly depolymerized melts exhibit exceptional properties, such as extremely non-Arrhenian viscosity-temperature relationships. Studies at low temperatures, i.e., just above the glass transition temperature, required for investigating such phenomena have been impossible for many compositions due to their rapid crystallization and thus extremely poor glass-forming ability. Here, we have developed a novel method combining synthesis by pulsed laser deposition with characterization by fast differential scanning calorimetry (FDSC) to deposit glassy thin films, from a plasma directly onto FDSC sensors, combined with calorimetric analysis at very high heating-cooling rates (3500–9500 °C s −1 ). In this way we have been able to explore the glass transition of the glasses (at or near olivine stoichiometry): Mg 1.30 SiO 3.3 , Fe 0.34 Mg 1.67 SiO 4.0 , Fe 0.49 Mg 1.53 SiO 4.0 , Fe 1.21 Mg 0.91 SiO 4.1 , and Fe 2.26 SiO 4.3 . The peak glass transition temperature at 5000 °C s −1 increases from 550 °C for Fe 2.26 SiO 4.3 with decreasing (Fe + Mg)/Si and with increasing Mg/(Fe + Mg) to 736 °C for Mg 1.30 SiO 3.3 . Viscosities derived through the shift factor approach range from 10 5.61 to 10 6.04 Pa∙s in that temperature range. These results thus extend the experimentally available viscosity range by ca. seven log units and were employed here, together with high-temperature superliquidus viscosity data to fit the Vogel-Fulcher-Tammann (VFT) equations for these melts. The viscosity-temperature relationships are all strongly non-Arrhenian (m = 82–123 ± 15). We provide a comparison with other experimental studies of ultrabasic melts and with predictions from viscosity models. The inability of the latter to predict the observed viscosities highlights the necessity of direct experimental studies such as that presented here.