A density model for high-pressure carbonate-rich melts applied to carbonatitic magmatism in the upper mantle

Abstract

Carbonatitic melts (<15 wt% SiO2) are widespread in the Earth's upper mantle and major conveyors of trace and volatile elements. Their migration through the mantle thus shapes the deep volatiles cycle and modifies its geochemical and geophysical signatures. Quantitative modelling of these processes has long been limited by the lack of continuous, predictive models for the physical properties of carbonate-rich melts, namely the density and viscosity, over a broad pressure-temperature-composition (P-T-X) range. Here we present a continuous model for the density of carbonate-rich melts in the system MgO-CaO-Na2O-K2O-Li2O-H2O-CO2 based on data compiled from the literature, including recent high pressure experimental and theoretical results. The model is calibrated on an extensive database of more than 880 data points over a temperature range of 800–2300 K and pressures from 10−4 GPa to 30 GPa, i.e. spanning most conditions of interest for the crust, upper mantle and transition zone. We have adopted a simple, volume-explicit Murnaghan EOS to describe the volumetric properties of each oxide component using 5 adjustable parameters optimized over the entire database and assume linear mixing in volume between the different oxide components in the calibration interval. The density model reproduces the calibration dataset with an average residual of ±0.05 g cm−3 and predicts densities with an average relative error of ±2%. Used in conjunction with reasonable constraints on the volumetric properties of SiO2 and FeO, the model permits density and seismic P-wave velocity VP predictions for complex carbonatitic melts experimentally equilibrated with various mantle lithologies (e.g. peridotite, eclogite, pelite). Our calculations demonstrate the first-order role of pressure by increasing both density and VP, with only a secondary role ascribed to temperature and composition. The role of H2O is contrasted and concealed by the interplay between the P-T-X conditions for the few experimental data available. Our predictions also show that both density and VP contrasts between carbonatitic melts and the host solid mantle diminish with depth, which implies larger effect of carbonate-rich melts on the seismic velocity signature of the Earth's upper mantle at shallower depths. Lastly, we pinpoint that carbonate melts are seismically faster than silicate melts, which together with their higher electrical conductivity, are promising diagnostic features to identify molten phases at depth and to discriminate between carbonate and silicate melts in the Earth's upper mantle using geophysical observations.