Slab dehydration and fluid migration at the base of the upper mantle: implications for deep earthquake mechanisms

SUMMARY Water enters the Earth's mantle via subduction of oceanic lithosphere and sediments. A lot of this water immediately returns to the atmosphere through arc volcanism, but part, retained in Dense Hydrous Magnesium Silicates (DHMSs) and Nominally Anhydrous Minerals (NAMs) like olivine, is expected to be subducted as deep as the bottom of the upper mantle (660 km depth). Then, due to its low solubility in lower mantle minerals, water is likely to be released as a hydrated fluid during the spinel–post‐spinel phase change. The dynamics of this fluid phase is investigated through a 1‐D model of compaction, in which a source term has been introduced to take the fluid precipitation into account. The competition between the advective transport by the descending slab and the buoyant rise of the fluid results in three distinct situations, depending on the properties of the solid and the fluid phases. Low matrix permeability and high fluid viscosity inhibit compaction and favour the entrainment of fluid towards the deep mantle. In this case, the entire slab water content would enter the lower mantle and would be mixed at large scale. However, realistic values of the fluid viscosity and matrix permeability make this possibility unlikely. When effective, compaction results in an accumulation of fluid at and below the phase boundary. Then, depending on the value of the matrix viscosity, the situation evolves differently. Above 10 20 Pa s, accumulation of fluid extends below the phase boundary and the pressure difference between the fluid and the matrix increases continuously, exceeding the yield strength of rocks. As a result, cracks would form and evolve towards the formation of dykes. In case of very low mantle viscosity, possibly due to strong grain size reduction during phase change, compaction becomes very efficient and the fluid remains confined within the phase change horizon, without increasing pressure. In the long term, this last situation appears unstable and would also evolve towards the formation of dykes. Thus, it is expected that water returns to the upper mantle by dykes propagating in the direction of the maximum compressive stress. Since maximum compressive stress appears to follow the dip of the slab below 410 km depth, we predict the formation of dykes extending from the 660 km phase change to 410 km depth. The possible existence of such dykes in slabs offers the necessary conditions for strong double couple component earthquake in the deepest part of the upper mantle.