Multicomponent Diffusion in Experimentally Cooled Melt Inclusions

Glassy olivine-hosted melt inclusions are compositionally zoned, characterized by a boundary layer depleted in olivine-compatible components that extends into the melt inclusion from its wall. The boundary layer forms in response to crystallization of olivine and relaxes with time due to diffusive exchange with the interior of the inclusion. At magmatic temperatures, the time scale for homogenization of inclusions is minutes to hours. Preservation of compositional gradients in natural inclusions results from rapid cooling upon eruption. A model of MgO concentration profiles that couples crystal growth and diffusive relaxation of a boundary layer can be used to solve for eruptive cooling rates [1]. Controlled cooling-rate experiments were conducted to test the accuracy of the model. Mauna Loa olivine containing >80 µm melt inclusions were equilibrated at 1225°C in a 1-atm furnace for 24 hours, followed by linear cooling at rates of 102 - 105 °C/hr. High-resolution concentration profiles of 40 inclusions were obtained using an electron microprobe. The model of [1] fits the experimental data with low residuals and the best-fit cooling rates are within 30% of experimental values. The initial temperature of 1225 °C is underestimated by 65°C. The model was modified using (i) MELTS to calculate the interface melt composition as a function of temperature, and (ii) a concentration-dependent MgO diffusion coefficient using the functional form of [2]. With this calibration the best-fit starting temperatures are within 5°C of the experimental values and the best-fit cooling rates are within 20% of experimental rates. The evolution of the CaO profile during cooling is evidence for strong diffusive coupling between melt components. Because CaO is incompatible in olivine, CaO concentrations are expected to be elevated in the boundary layer adjacent to the growing olivine. Although this is observed at short time scales, as the profile evolves the CaO concentration near the crystal interface becomes increasingly depleted. The drawdown in CaO can be explained by non-ideal mixing that leads to increases in the CaO activity coefficient in the melt. A regular solution model [3] can be used to describe the evolution of the CaO profiles. [1]Newcombe et al (2014) CMP 168 [2] Zhang (2010) RevMineralGeochem 72 [3] Ghiorso & Sack (1995) CMP 119