Upper mantle anisotropy beneath Peru from SKS splitting: Constraints on flat slab dynamics and interaction with the Nazca Ridge

Highlights

• Northern limit of Nazca Ridge coincides with sharp transition in seismic anisotropy.

• Ridge therefore plays key role in mantle dynamics of Peruvian flat slab system.

• Spatially coherent SKS apparent splitting observed north of the ridge.

• Pervasive null measurements found above and to the south of the ridge.

• Both anisotropic domains appear to reflect complex and multi-layered anisotropy.

Abstract

The Peruvian flat slab is by far the largest region of flat subduction in the world today, but aspects of its structure and dynamics remain poorly understood. In particular, questions remain over whether the relatively narrow Nazca Ridge subducting beneath southern Peru provides dynamic support for the flat slab or it is just a passive feature. We investigate the dynamics and interaction of the Nazca Ridge and the flat slab system by studying upper mantle seismic anisotropy across southern Peru. We analyze shear wave splitting of SKS, sSKS, and PKS phases at 49 stations distributed across the area, primarily from the PerU Lithosphere and Slab Experiment (PULSE). We observe distinct spatial variations in anisotropic structure along strike, most notably a sharp transition from coherent splitting in the north to pervasive null (non-split) arrivals in the south, with the transition coinciding with the northern limit of the Nazca Ridge. For both anisotropic domains there is evidence for complex and multi-layered anisotropy. To the north of the ridge our ${}^{⁎}\mathit{KS}$ splitting measurements likely reflect trench-normal mantle flow beneath the flat slab. This signal is then modified by shallower anisotropic layers, most likely in the supra-slab mantle, but also potentially from within the slab. To the south the sub-slab mantle is similarly anisotropic, with a trench-oblique fast direction, but widespread nulls appear to reflect dramatic heterogeneity in anisotropic structure above the flat slab. Overall the regional anisotropic structure, and thus the pattern of deformation, appears to be closely tied to the location of the Nazca Ridge, which further suggests that the ridge plays a key role in the mantle dynamics of the Peruvian flat slab system.

Introduction

Episodes of flat or shallow subduction are often associated with volcanic gaps, increased intra-plate seismic energy release, and widespread deformation of continental interiors (e.g. Gutscher, 2002). Despite the significant impact of flat subduction upon the surface, and the prominence of shallow subduction worldwide today (representing ∼10% of subduction zones globally; Gutscher et al., 2000), the formation, evolution and dynamics of flat slabs remain poorly understood. For example, there has long been a suggestion that flat slabs result from the subduction of thick oceanic crust, such as plateaus, but the global correlation between such oceanic features and shallow subduction is imperfect (Skinner and Clayton, 2013).

Even at the regional scale, for the Peruvian flat slab, open questions exist over the importance of the Nazca Ridge, which is presently subducting beneath southern Peru, considering the large difference in size between the Nazca Ridge (∼200 km wide) and the flat slab segment (∼1500 km wide). Investigation of the mantle flow field surrounding the Peruvian flat slab, particularly in the vicinity of the Nazca Ridge, has the potential to shed light on the dynamics and inner workings of this and other flat slab systems. Our goal is therefore to explore the interaction between the ridge and the flat slab from the point of view of how the surrounding mantle deforms. In this way we hope to better understand what role, if any, the Nazca Ridge plays in the development of the flat slab.

Over the last few decades shear wave splitting has become an invaluable tool for probing seismic anisotropy and the dynamics of the earth's interior. Seismic anisotropy describes the directional dependence of seismic wave speed. Shear wave splitting is the process whereby a shear wave passing through such an anisotropic medium is polarized into two quasi-S waves with distinct polarizations in a fast direction (φ) and a slow direction. The two waves travel at different velocities through the medium and therefore accumulate a delay time (δt) between their arrivals on the seismogram. The size of the delay time accrued depends on the strength of the anisotropy and the length of the path through the anisotropic material (which is proportional to the thickness of the anisotropic layer).

One of the most significant anisotropic domains within the earth is in the upper mantle, in which strain (deformation) in the dislocation creep regime results in the lattice preferred orientation (LPO) of anisotropic minerals such as olivine (e.g. Karato et al., 2008). This non-random distribution of crystallographic axes results in an effective bulk anisotropy that is detectable by seismic waves. Armed with knowledge of the relationship between mantle strain and LPO fabric geometry, we can exploit observations of seismic anisotropy to infer the orientation of deformation and flow in the upper mantle.

For olivine, the most abundant upper mantle mineral, there are several known LPO fabrics (most commonly A-, B-, C-, D-, and E-types) that occur under different physical and chemical conditions, such as temperature, stress, and water content (Jung and Karato, 2001, Jung et al., 2006, Karato et al., 2008, Katayama et al., 2004). Each fabric type represents different 3D orientations of the crystallographic axes with respect to a given strain geometry. Under conditions typical of the upper mantle, A-, C-, or E-type fabrics are most likely, and for all three fabrics the fast splitting direction tends to align with the direction of shear (Karato et al., 2008). Under low temperature, high stress and water rich conditions, as expected in the fore-arc mantle wedge (Kneller et al., 2007, Kneller et al., 2005), B-type fabric likely dominates. In this regime, the fast axis aligns within the shear plane, orthogonal to the shear direction (Jung and Karato, 2001, Jung et al., 2006).

Shear wave phases that pass through the core, such as SKS, are particularly well suited for probing upper mantle anisotropy. These phases are polarized in the radial direction upon conversion from a P wave in the liquid outer core, thus removing any effect from seismic anisotropy on the source side. They are sensitive to anisotropy anywhere along their paths from the core–mantle boundary to the receiver. In most cases, however, contributions from $D^{″}$, the lower mantle, and the crust are thought to be minimal, with upper mantle anisotropy making the dominant contribution to the signal. This is evidenced by, for example, a first order global correlation between SKS splitting observations and predictions based on surface plate motions and geodynamic models of mantle flow (e.g. Conrad et al., 2007, Long and Becker, 2010, Walpole et al., 2014).

Occasionally SKS phases arrive at seismic stations displaying no signs of having been split, in what is known as a null measurement. This can occur when the initial SKS polarization is aligned with the fast or slow axis of anisotropy; therefore, null measurements can be useful in constraining the anisotropic orientation. When stations are dominated by null measurements over a wide range of backazimuths, however, this can signify an isotropic (or weakly anisotropic) upper mantle, destructive interference of two or more layers of anisotropy, or a vertical axis of (transversely isotropic) symmetry. The last option is typically interpreted as vertical mantle flow, either in the form of active mantle upwelling, e.g., beneath Bermuda (Benoit et al., 2013), or down-welling, e.g., for a proposed lithospheric drip beneath Nevada (West et al., 2009).

Simple correlations between SKS splitting (or nulls) and flow in the upper mantle break down, however, around subduction zones (e.g. Long and Silver, 2008, Long and Silver, 2009), where mantle flow is likely more complex and the interpretation of SKS splitting becomes increasingly difficult. A factor in this difficulty is the possibility of multiple domains of anisotropy, with potential contributions from the sub-slab mantle, slab, mantle wedge, and overriding plate. Additional complications may also arise from the effects of slab dip and volatile influx. In particular, the presence of water can significantly affect the olivine LPO fabric geometry and the distribution of highly anisotropic hydrous phases such as serpentine. Challenges in interpreting SKS splitting measurements in subduction systems are confounded when using data from temporary seismic arrays, which are usually deployed for ∼2 yr. In this case, it is often difficult to acquire sufficient backazimuthal coverage to discriminate among single-layered, multi-layered and/or dipping anisotropy (e.g. Eakin et al., 2010).

In this study we use data from the temporary PULSE (PerU Lithosphere and Slab Experiment) deployment located over the Peruvian flat subduction zone (Fig. 1). This location provides several advantages that ameliorate some of the challenges inherent in interpreting SKS splitting in a subduction zone. First, in the flat subduction regime, once the slab approaches ∼100 km depth the slab dip angle is very shallow (∼10°) (Cahill and Isacks, 1992, Hayes et al., 2012), so dipping anisotropy is less likely. Second, there is no substantial mantle wedge to consider; at most only a thin layer of mantle (∼30 km) remains between the flat slab and overlying continental crust, and directly above the Nazca Ridge this layer is absent (Bishop et al., 2013, Phillips and Clayton, 2014). Third, previously published studies of local direct S and source-side teleseismic S phases independently constrain the supra-slab and sub-slab anisotropy, respectively (Eakin and Long, 2013, Eakin et al., 2014). Fourth, a detailed analysis of SKS splitting at the long-running GSN station NNA, located in our study area, has provided sufficient sampling in backazimuth and frequency to constrain multiple layers of anisotropy (Eakin and Long, 2013). This permanent station therefore provides us with a valuable benchmark with which to compare and contrast SKS splitting measured at our temporary stations.

SKS splitting provides a depth-integrated view of upper mantle anisotropy, and is therefore particularly useful for looking at lateral variations in anisotropic structure. Complementary to this, splitting from local and source-side S phases provide important information about anisotropy within different sub-domains of the subduction zone (i.e. sub-slab vs. wedge anisotropy), but their interpretation can be hampered by the heterogeneous and often biased distribution of slab seismicity. In contrast, the nearly vertical propagation of SKS phases ensures efficient sampling of the upper mantle beneath seismic stations and thus provides a more uniform sampling of regional anisotropic patterns. SKS splitting thus provides an ideal tool for exploring lateral variations in upper mantle anisotropy beneath southern Peru and understanding how those variations relate spatially to Nazca Ridge subduction, thus shedding light on the dynamic interactions between the ridge and the flat slab. In this study we document a sharp contrast in anisotropic structure, and therefore mantle deformation, in association with the Nazca Ridge. This is illuminated by the SKS splitting patterns, which show apparent splitting north of the ridge (mostly likely reflecting a combination of multiple layers of anisotropy), while nulls are pervasive directly over the ridge and to its south.