Subduction-linked entrainment and counterflow within plume-fed asthenosphere

Morgan et al. in 2007 suggested that subduction-driven entrainment is likely to be relatively inefficient at pulling down a buoyant low-viscosity asthenospheric layer at the base of a subducting slab; it will only drag down a ∼10–20 km-thick layer of asthenosphere at the base and top of ∼100-km-thick subducting oceanic lithosphere. If so, buoyant and low viscosity asthenosphere will mostly be accreted into overlying oceanic lithosphere, first at mid-ocean ridges where a ∼70-km compositional lithosphere forms by melt-induced dehydration of upwelling peridotitic mantle, and later off-axis when the thermal cooling of overlying oceanic lithosphere penetrates deeper than 70 km. This study uses 2-D time-dependent numerical experiments to revisit this question with much improved computational resources. For conceptual simplicity we only explore stable subduction geometries, including the effects of both non-Newtonian and Newtonian temperature-dependent flow rheologies. For a range of asthenosphere flow-laws, we find that a 5–20 km-thick entrained layer will be dragged down along the lower-side of a subducting slab, with an associated volume flux per unit length of 100–1600 k m 2 / Ma This limited subduction-driven entrainment is linked with vigorous asthenospheric counterflow. When mantle viscosities also increase with depth, maximum entrainment is found to occur when mantle viscosity increases take place at a relatively ‘shallow’ 410-km depth instead of at greater depth. A non-Newtonian rheology also leads to less entrained asthenosphere. Finally, if there is a vertical density stratification in the asthenosphere due to variable amounts of plume melt-extraction, this leads to similar predicted entrainment by subducting slabs within a more complex pattern of buoyantly layered asthenospheric counterflow and a local ‘pile-up’ of the most buoyant asthenosphere beneath the trench axis.