Water-rich Melt Channel Beneath the Cocos Plate Imaged With Petrologically-constrained Bayesian Inversion of Magnetotelluric Data

Abstract:

Plate tectonics – a salient feature of the Earth as a geodynamical system – requires a low viscosity layer at the base of the rigid lithosphere. The origin of this low viscosity at the lithosphere-asthenosphere boundary (LAB) is still debated. Beneath oceanic plates, possible explanations include the weakening effects of temperature, mineral hydration, and/or the present of partial melt. These factors also affect the electrical resistivity of the mantle, especially hydration and melt, so knowledge of subsurface resistivity at the LAB should shed light on the degree of melt and hydration there. We use Bayesian inversion of seafloor magnetotelluric (MT) data from the Cocos plate to derive probabilistic estimates of a low resistivity layer at the LAB. Since the volatiles CO₂ and H₂O can greatly reduce the resistivity of the melt phase, there is a trade-off between volatile concentration and melt fraction in explaining the low resistivity, which can be explained by more melt with less volatiles or less melt with more volatiles. To resolve this ambiguity, we apply petrological modeling constraints to the Bayesian uncertainties in resistivity so that only petrologically stable melt and volatile contents are allowed. Our results require a water-rich partially molten channel beneath the Cocos plate. Fully 100% of all resistivity models that fit the MT data require enough bulk mantle hydration for some degree of melt to be stable -- depending on volatile content, up to 10 vol% melt. We also find that the LAB beneath the Cocos plate must be either very warm (consistent with a mantle potential temperature of 1440­­­–1500 C) or else highly hydrated (equivalent to several hundred ppm or more bulk mantle water). Data from mid-ocean ridge basalt (MORB) glasses suggests that if the melt present is left over from mid-ocean ridge melting, then high mantle temperatures and high melt fractions are required to explain our findings. If instead this melt channel is enriched in volatiles, then less elevated temperatures and lower melt fractions are required.­­