Internal deformation of the subducted Nazca slab inferred from seismic anisotropy

Withinoceanic lithospherea fossilized fabric is oftenpreserved originating from the time of plate formation. Such fabric is thought to form at the mid-ocean ridge when olivine crystals align with the direction of plate spreading1,2. It is unclear, however, whether this fossil fabric is preserved within slabs during subduction or overprinted by subduction-induced deformation. The alignment of olivine crystals, such as within fossil fabrics, cangenerateanisotropy that is sensedbypassing seismic waves. Seismic anisotropy is therefore a useful tool for investigating the dynamics of subduction zones, but it has so far proved di cult to observe the anisotropic properties of the subducted slab itself. Here we analyse seismic anisotropy in the subducted Nazca slab beneath Peru and find that the fast direction of seismic wave propagation aligns with the contours of the slab. We use numerical modelling to simulate the olivine fabric created at the mid-ocean ridge, but find it is inconsistent with our observations of seismic anisotropy in the subducted Nazca slab. Instead we find that an orientation of the olivine crystal fast axes aligned parallel to the strike of the slab provides the best fit, consistent with along-strike extension induced by flattening of the slab during subduction (A. Kumar et al., manuscript in preparation). We conclude that the fossil fabric has been overprinted during subduction and that the Nazca slab must therefore be su ciently weak to undergo internal deformation. It has long been suggested that the process of seafloor spreading at mid-ocean ridges (MOR) induces a lattice preferred orientation (LPO) of olivine in the underlying mantle that is subsequently ‘frozen-in’ to the oceanic plate during its formation1,2. For A-type olivine LPO fabrics, typical of dry oceanic lithosphere, the fast a axes of olivine tend to align with the plate spreading direction, resulting in a fossilized crystallographic fabric in the oceanic plate3,4. Such fossilized fabric will manifest itself as seismic anisotropy, the phenomenon by which seismic wave velocity is directionally dependent. Observations from surface-wave-derived azimuthal anisotropy5–7, shear wave splitting8,9, and refracted P-wave (Pn) velocities10 all indicate that seismic anisotropy within the oceanic lithosphere is consistent with the concept of fossil spreading. In subduction zones, where oceanic plates descend into the mantle, it is unclear whether this fossil spreading fabric is preserved within the slab to depth, or if the anisotropic signal is overprinted by subsequent subduction-associated deformation. Observations that directly constrain intra-slab anisotropy are limited in number11,12, and completely absent for the deep portions of slabs (below 200 km). An alternative model for slab anisotropy invokes vertically aligned, hydrated faults in the upper portion of the slab that result from bending stresses at the outer rise. This phenomenon could result in seismically inferred fast directions that are parallel to the trench, owing to a combined SPO (shape preferred orientation) and LPO effect of the serpentinized faults13, which is consistent with some P-wave and Rayleigh wave observations14–16 (down to a maximum depth of 200 km). The most direct observations of seismic anisotropy usually come from shear wave splitting, whereby the orientation of the fast polarized shear wave (φ) and the delay time (δt) between the fast and the slow polarized waves are measured. For most shear wave splitting studies targeting subduction zones, anisotropy within the slab itself is typically disregarded because the relative path length through the slab is small compared to the rest of the upper mantle. For the Nazca slab beneath Peru, however, the unique flatslab geometry, with a transition from a shallow to steeply dipping slab ∼500 km inboard from the trench (Fig. 1), allowed us to make splitting measurements on the seismic S phases with relatively long path lengths through the slab (see Methods), which would normally be difficult to observe. Using data from the PULSE and PeruSE arrays, we obtained 16 splitting measurements (out of 36 suitable arrivals) for deep local S phases (Fig. 2a), with the majority of fast directions oriented approximately N–S (mean φ:−6.3), and exhibiting substantial delay times (mean δt : 1.3 s). In the same vicinity as the local S results (for example, area encircled by dashed line in Fig. 2a), source-side measurements on downgoing S phases measured at distant stations show very similar splitting characteristics (mean φ: −1.2, mean δt : 1.6 s, number of measurements: 9), indicating that the two types of phases sample may the same anisotropic source region. Several lines of argument suggest that this main anisotropic source is within the subducting Nazca slab. First, when the ray paths are plotted in three dimensions (see Supplementary Information) and compared against the slab outline from regional S-wave tomography17, it is clear thatmany rays (both local S and source side) have long path lengths through the slab (Fig. 2a, Supplementary Figs 1 and 2 and Supplementary Movie 1). In particular, a cluster of N–S-oriented fast splitting measurements towards the centre of the study area travel through relatively fast (blue) material all the way from the mid-transition zone (555 km) to the mid-upper mantle (<200 km), representing a total path length of over 250 km through the slab (Supplementary Fig. 1). We also note that this N–S φ orientation roughly correlates with the N–S strike of the subducting Nazca Plate. Outside of the slab, for measurements that sample mainly sub-slab mantle, the fast directions are generally