An Extended Calibration of the Olivine–Spinel Aluminum Exchange Thermometer: Application to the Melting Conditions and Mantle Lithologies of Large Igneous Provinces

Abstract
The application of the olivine–spinel aluminum exchange thermometer to natural samples is limited by the restricted experimental data set on which it was calibrated. Here, we present a new data set of 46 high-temperature crystallization experiments and 21 reanalyzed published experiments, which we used to extend the calibration to higher and lower temperatures. The final calibration data set spans a range of conditions relevant to crustal and upper mantle processes: 1174–1606°C, 0.1–1350 MPa, QFM − 2.5 to QFM + 7.2 (oxygen fugacity, fO2, reported in log units relative to the quartz–fayalite–magnetite buffer, QFM), and 0–7.4 wt % H2Omelt. We propose three new models. The first is thermodynamically self-consistent, based on spinel Fe, Mg, Al, and Cr compositions and Al exchange between olivine and spinel. The second and third are empirical models that consider fewer elemental exchanges: the second uses only Al exchange and spinel compositions, whereas the third considers olivine–spinel Al and Cr exchange. All models include the modest effect of pressure on olivine-spinel equilibrium chemistry, whereas fO2 and water content have negligible effects. In general, as fewer elements are considered in the olivine–spinel exchange, the fit to experimental data worsens. Conversely, the associated decrease in model complexity improves their robustness against systematic errors when applied to natural crystal pairs: the thermodynamic model may underestimate crystallization temperatures in natural samples due to spinel subsolidus re-equilibration, whereas the empirical models (independent of Fe and Mg in spinel) are less sensitive to re-equilibration but yield temperatures with larger uncertainties. We applied a statistical test to select the most appropriate model for application to natural samples. When applied to lavas from mid-ocean ridges, Iceland, Skye, Emeishan, Etendeka, and Tortugal, our new temperature estimates are 30–100°C lower than previously proposed. The lower temperature estimates cause a lower mantle melting temperature and significant impacts on the mantle lithology constraints.