Shear-Wave Splitting and Implications for Mantle Flow Beneath the MELT Region of the East Pacific Rise

See allHide authors and affiliations

Science  22 May 1998:
Vol. 280, Issue 5367, pp. 1230-1232
DOI: 10.1126/science.280.5367.1230


Shear-wave splitting across the fast-spreading East Pacific Rise has been measured from records of SKS and SKKSphases on the ocean-bottom seismometers of the Mantle Electromagnetic and Tomography (MELT) Experiment. The direction of fast shear-wave polarization is aligned parallel to the spreading direction. Delay times between fast and slow shear waves are asymmetric across the rise, and off-axis values on the Pacific Plate are twice those on the Nazca Plate. Splitting on the Pacific Plate may reflect anisotropy associated with spreading-induced flow above a depth of about 100 km, as well as a deeper contribution from warm asthenospheric return flow from the Pacific Superswell region.

Observations of shear-wave splitting by the ocean-bottom seismometer (OBS) network of the MELT Experiment (1) describe the seismic anisotropy of the upper mantle beneath the southern East Pacific Rise. A shear wave passing through a homogeneous anisotropic layer splits into two waves with orthogonal polarizations and different wave speeds. Measurements of shear-wave splitting yield the direction of polarization φ of the fast shear wave and the delay time δ t between the fast and slow shear waves (2, 3). Splitting parameters for the core phases SKS and SKKS reflect the path-integrated effects of upper mantle anisotropy beneath the receiving seismometer (4) and provide information on the orientation of the anisotropy as well as the combined effects of the thickness of the anisotropic region, the degree of anisotropy, and the isotropic velocity (5).

Because olivine, the most abundant mineral in the upper mantle, is anisotropic and develops lattice-preferred orientation in response to finite strain, splitting across the East Pacific Rise may be used to infer the pattern of flow-induced alignment of olivine grains (6-8) associated with mid-ocean ridge spreading. Both passive and dynamic models (9) of upwelling and divergence beneath spreading centers predict that flow in the upper 100 km of the mantle (8) should align the crystallographic a axes of olivine grains and produce measurable anisotropy (Fig.1). Asthenospheric shear induced by absolute plate motion, large-scale mantle flow, or asthenospheric return flow may impart additional anisotropy to deeper portions of the upper mantle. Splitting is also sensitive to the presence of mantle melt in vertically aligned cracks (Fig. 1), although a random distribution of melt in cracks or tubules would be isotropic and produce no splitting (7, 8).

Figure 1

Schematic of the expected shear-wave splitting patterns. (A) If anisotropy is dominated by the preferred alignment of olivine crystals, there should be no observable splitting on or near the rise axis, because mantle upwelling in this region induces a vertically oriented olivine a axis; farther off axis, mantle flow is predominantly horizontal and shear-wave splitting should be observed with fast direction φ parallel to spreading (arrows). (B) If anisotropy is dominated by melt in crack faces aligned parallel to the rise axis, splitting should be observed near the axis with φ parallel to the rise; because melting is confined to a region straddling the rise axis, the contribution from melt should decrease to zero far off axis.

We observe clear shear-wave splitting (Fig.2) of SKS and SKKS phases from several earthquakes in the Kurile Islands and Indonesia (Table 1) recorded by the MELT network. Our splitting analysis follows standard methodology (3), as adapted (10) for deriving optimum splitting parameters from records of multiple events. Shear waves have good signal-to-noise ratio on the horizontal component records (1, 11) at frequencies of 0.04 to 0.1 Hz, which are equivalent to seismic wavelengths of 50 to 100 km. Splitting results were obtained for 18 instruments at 6 to 376 km from the rise axis on both the Pacific and Nazca Plates (Table2 and Fig.3). The splitting measurements combine data from one to six observations at each site.

Figure 2

Example SKKS waveforms from the earthquake of 25 December 1995, on radial (solid line) and transverse (dashed line) horizontal-component seismograms (bandpass filtered from 0.04 to 0.15 Hz); waveforms are aligned on the SKKS phase on the radial component. The consistent presence of energy on the transverse component is indicative of shear-wave splitting (2,3); in the absence of anisotropy, the SKKS phase should be polarized in the radial direction with no energy on the transverse component. The distance from the rise axis, in kilometers, is shown; P denotes stations on the Pacific Plate, and N denotes stations on the Nazca Plate.

Figure 3

Shear-wave splitting across the MELT array. Dots denote the position of the OBSs, and solid lines indicate the orientation of the fast direction φ, with the line length proportional to the delay time δt between fast and slow shear waves (Table 1). The rise axis is plotted as a solid line. Arrows indicate the magnitude and direction of absolute plate motion (20); the directions of absolute plate motion (20) and relative plate motion (12) are identical to within a few degrees in this region.

Table 1

Earthquakes used for measuring shear-wave splitting parameters. BAZ is the back azimuth; that is, the direction of approach of the waves as seen at the MELT network, measured clockwise from north. M S is the surface wave magnitude. Phases utilized included SKS for all earthquakes andSKKS phases for events 2 through 4.

View this table:
Table 2

Shear-wave splitting parameters at MELT OBS sites. N, Nazca Plate; P, Pacific Plate. The direction φ of shear-wave polarization is measured clockwise from north. Uncertainties for φ and δt are at one standard deviation.

View this table:

For all observations, the fast polarization axis φ is consistently oriented nearly along the spreading direction of 103° measured clockwise from north (12). The mean value of φ at the MELT sites is 95°, and all but one estimate are within 15° of this value, an agreement consistent with the combined standard errors for φ and for instrument orientation (13). The delay time δ t between fast and slow shear waves is 0.8 to 1.4 s for instruments within 25 km of the rise axis. The delay time increases off axis on the Pacific Plate to values near 2 s but remains at about 1 s for instruments off axis on the Nazca Plate (Fig. 3 and Table 2).

The shear-wave splitting displayed in Fig. 3 is contrary to the predictions from simple flow and melting models (7, 8). Near axis, the fast direction φ of 95° and delays of about 1 s are inconsistent with anisotropy dominated by the presence of melt in crack faces aligned parallel to the ridge axis, which would produce a fast direction of approximately 5° (7). The near-axis splitting observations also differ from the negligible splitting predicted for a vertically oriented olivine a axis induced by mantle upwelling. The asymmetric splitting times off axis, with δ t values as large as 2 s on the Pacific side, differ as well from model predictions (8) of smaller values (1 s) and symmetry about the rise axis. Furthermore, it is unlikely that the largest splitting delays on the Pacific Plate can be the result of anisotropy confined to the uppermost 100 km of the mantle. Typical values of percent anisotropy (5) from petrofabric measurements on mantle samples (14) are 3 to 5% (for a horizontal olivine a axis), yielding about 1 s of splitting for a 100-km-thick anisotropic layer. In addition, larger SKS anisotropy in the shallowest mantle would be inconsistent with Rayleigh-wave phase velocity measurements across the MELT network, which yield values for peak-to-peak anisotropy at a 25-s period of no more than 3 to 5% (15).

These arguments suggest that the largest splitting values reflect both a component of spreading-induced flow above a depth of 100 km and a deeper contribution, probably arising from upper mantle return flow (16) to the rise axis from the Pacific region. Both body-wave travel-time delays (11) and surface-wave phase velocities (15) across the MELT network display lower velocity values for the mantle on the Pacific side of the East Pacific Rise than on the Nazca Plate side. Asthenospheric return flow of anomalously warm mantle fed from the Pacific Superswell region (17) would explain the lower Rayleigh-wave velocities and more delayed body arrivals, as well as lesser rates of sea-floor subsidence with age (18) and a greater density of seamounts on the Pacific side of the rise axis (19). This scenario can also account for the splitting patterns seen in Fig. 3. On the Pacific side of the rise axis, greater asthenospheric shear imparted by the faster absolute motion of the Pacific Plate (20) and lower viscosities associated with higher temperatures, together with a component of return flow, can produce shear-wave splitting times that are 1 s greater than for the Nazca Plate side of the rise. We propose that splitting at the MELT region reflects two anisotropic layers, both with φ oriented parallel to spreading. The upper layer, above a depth of about 100 km, is dominated by spreading-induced flow, which produces no splitting near the axis, where flow is predominantly vertical, but contributes to off-axis splitting as the flow diverges to a predominantly horizontal direction. The lower layer on the Pacific side consists of material experiencing return flow toward the ridge, but comparable return flow is more modest or absent on the Nazca side. Such an asthenospheric return flow could account for the 1-s splitting times seen on the rise axis, because the SKS phases in this study from sources to the west and northwest (Table 1) may sample the anisotropy induced by this flow at depths greater than 100 km.


View Abstract

Navigate This Article