Mid-mantle anisotropy in subduction zones and deep water transport

The Earth's transition zone has until recently been assumed to be seismically isotropic. Increasingly, however, evidence suggests that ordering of material over seismic wavelengths occurs there, but it is unclear what causes this. We use the method of source-side shear wave splitting to examine the anisotropy surrounding earthquakes deeper than 200 km in slabs around the globe. We find significant amounts of splitting (≤2.4 s), confirming that the transition zone is anisotropic here. However, there is no decrease in the amount of splitting with depth, as would be the case for a metastable tongue of olivine which thins with depth, suggesting this is not the cause. The amount of splitting does not appear to be consistent with processes in the ambient mantle, such as lattice-preferred orientation development in wadsleyite, ringwoodite, or MgSiO<inf>3</inf>-perovskite. We invert for the orientation of several mechanisms - subject to uncertainties in mineralogy and deformation - and the best fit is given by updip flattening in a style of anisotropy common to hydrous phases and layered inclusions. We suggest that highly anisotropic hydrous phases or hydrated layering is a likely cause of anisotropy within the slab, implying significant water transport from the surface down to at least 660 km depth.