Transition Zone Anisotropy Beneath Deep Slabs and the Transport of Water into the Lower Mantle
Abstract:To first order, the Earth exhibits seismic anisotropy (the variation of wave speed with direction) only in the uppermost and lowermost mantle, as well as the inner core. However, a growing body of evidence suggests that it is also present in the transition zone (TZ) and uppermost lower mantle (LM). We use the method of ‘source-side’ shear wave splitting to observe anisotropy in the regions of deep earthquakes distributed globally. This technique removes the effects of anisotropy near well-characterised receiver stations to infer the splitting at the source, allowing us to probe the midmantle where slabs appear to be impinging on the LM.
Over 130 observations, mainly beneath South America, Tonga and Japan, are made for earthquakes 200–650 km deep. They show shear wave splitting with mean delay time 1.0 s, but there is no trend of decreasing—or increasing—δt with depth. Because of the distribution of circum-Pacific deep earthquakes, our data are only sensitive to anisotropy in the sub-slab region and the slab itself.
Our observations reveal a consistent pattern: the data are best fit with a style of anisotropy which has a rotational symmetry axis pointing upwards along the slab. This pattern of anisotropy is typical of approximately uniaxial flattening of material which develops a lattice preferred orientation (LPO) by dislocation creep. This is consistent with the expected mechanics of slab sinking and supported by the P-axes of moment tensor solutions for the events we analyse. Because the amount of anisotropy does not appear to be related to the depth, we can confine the source region to either the slab itself, or the top of the LM.
The amount of anisotropy makes it unlikely that MgSiO3-perovskite in the LM is the source, as it would require a high-strain layer over 1500~km thick. Dense hydrous magnesium silicate (DHMS) phases which are known to become stable at the base of the TZ (the so-called ‘alphabet’ phases; such as D and superhydrous B), do however have very large single crystal anisotropy, would likely develop LPO, and if distributed over a few tens of km could produce the splitting we observe at subduction zones. If these phases are as ubiquitous as our data imply, then significant water volumes may be brought into the LM over geological time, influencing our understanding of subduction zone dynamics and lower mantle composition.