Submitted to Science, July 29, 2004 1

Variable azimuthal anisotropy in Earth’s lowermost mantle

Edward J. Garnero1, Valérie Maupin2, Thorne Lay3, Matthew J. Fouch1

1 Department of Geological Sciences, Arizona State University, Tempe, AZ 85287-1404, USA

2 Department of Geosciences, University of Oslo PO Box 1047, Blindern, Oslo 0316, Norway

3 Institute of Geophysics and Planetary Physics and Earth Sciences Department, University of California, Santa Cruz, CA 95064, USA

A subtle yet persistent reversed polarity onset precursory pulse in for vertically polarized shear waves (SV) that graze or diffract the lowermost mantle (the D" layer) in many regions around the globe initiates at the arrival time of horizontally polarized shear waves (SH). Such SV-SH coupling cannot be explained by the current paradigm for D" anisotropy, ; transverse isotropy with a vertical axis of symmetry (VTI). Full waveform modeling of the SV precursorssplit shear waves for D"paths beneath the Caribbean, a region with the most abundant high quality broadband data, indicates requires azimuthal anisotropy must be present at the base of Earth’s mantle. Data are well fit by models with laterally coherent patterns of transverse isotropy with the hexagonal symmetry axis tilted from the vertical by up to 20º. Small-scale patterns in convection of the overlying mantle may cause the observed variations by inducing laterally variable crystallographic or shape-preferred orientation in mineralogical components in the boundary layer.

The deepest mantle likely plays as critical of a role in Earth’s dynamical systems as its near-surface counterpart, the lithosphere-asthenosphere system (1,2). Both of these major internal boundary layers possess extensive seismic wave anisotropy (3-5), which holds promise to reveal mineral and structural fabric alignment (6). High-resolution investigations of D" anisotropy are restricted to a few key regions due to geographical limitations in the global distribution of earthquake sources and seismic sensors; these include beneath the Caribbean Ocean, Alaska, the central Atlantic, the central and southern Pacific, central Eurasia, and southern India. Figure 1 shows representative seismograms from several regions. Ground velocity recordings (Fig 1A,B) allow picking of an impulsive SV downward arrival having a polarity opposite to that of SKS (as predicted by VTI or isotropic models) and arriving seconds after the SH onset (7,8). However, on the displacement trace there is clear SV energy with the same polarity as SKS, arriving at the same time as the initial SH motion. This is consistent with a projection of the fast polarization arrival onto SH and SV components, a clear indication of azimuthal anisotropy (9). The SH arrival involves two impulsive ground velocity signals, and a correspondingly complex displacement record. This feature can be accounted for by triplication of the S wave due to an abrupt 2-3% velocity discontinuity at the top of D" (8,10); the first arrival involves energy refracting below the discontinuity, the second is energy reflected from the discontinuity (10-11). Figure 1C shows examples of similar reversed onsets in SV precursory behaviorsignals (shaded upwards pulse) for other D" regions, all ofnoted which have been characterized in past studies as having VTI anisotropy (12).

Source complexity or deconvolution problems are ruled out as the cause of the precursors SV waveform complexity since SKS signals show no corresponding reversed onsets. SV-to-P conversions at the receiver are precluded by the rotation to the incident wavefront and the lack of such any precursors for SKS arrivals. Models with complex D" lamellae and VTI structures (12),and with 3D isotropic heterogeneities (13), show that it is very difficult to produce reversed SV onsets without introducing azimuthal anisotropy. We explore models where in which the D" discontinuity coincides with the onset of anisotropy 250 km above the core-mantle boundary, since such that azimuthal anisotropy affects the first arrival, which penetrates into the D" layer.

We focus on D" structure beneath the Caribbean, a the region with the greatest sampling by high quality broadband data, afforded provided by S waves from intermediate and deep focus South American earthquakes recorded by the Canadian National Seismographic Network (CNSN). The epicentral distance range (87-112°) is such that waves graze or diffract horizontally through the D" layer. Broadband data were deconvolved by the instrument responses to obtain ground displacement traces, and then rotated into the reference frame of the incoming S wavefront to obtain SH and SV motion, minimizing crustal SV-to-P conversions (14). Data with simple, impulsive signals and favorable signal-to-noise ratios for both SH and SV energy were retained, reducing our initial dataset from 120 to 16 earthquakes recorded at 19 broadband CNSN stations, comprising 89 individual high-quality recordings. Data were corrected for upper mantle anisotropy using shear wave splitting parameters published for CNSN stations (15,16), and were discarded when corrections did not adequately remove SKS energy from the SH component.

Past studies (7,8) in this region have reported SV delays relative to SH. This type of splitting, with purported decoupling of SV and SH components, has led to the interpretation of extensive VTI in the D" layer (7,8,17,18). VTI can arise from preferred orientation of hexagonal crystal lattices with vertical symmetry axes or preferred orientation in the horizontal plane of azimuthally random crystals with a more general anisotropy. Alternatively, it can arise from fine-scale lamellae of strongly contrasting material properties (12,19).

A full wave theory method (9)was used to construct synthetic seismograms for general forms of D" anisotropy in a the first comprehensive modeling of a large data set. Introduction of azimuthal anisotropy dramatically increases the model space; limited azimuthal sampling restricts unique resolution to two anisotropic parameters: the direction of polarization of the fast and slow waves in the plane perpendicular to the propagation direction, and the amplitude of the anisotropy. We also have clear observations of simply delayed SV signals, with no apparent SH-SV coupling (similarly with other regions). These features led us to reduce adopt a the model space to involving transverse isotropy where the symmetry axis ranges from vertical to tilted, essentially setting VTI as a the reference state.

Synthetic seismograms are shown in Figure 2 for an isotropic reference model, a VTI model, and tilted transverse isotropy (TTI) models where the tilt angle is measured from the vertical and azimuth varies in all directions. TTI models clearly produce reversed onset SV signals for suitable orientations. The precursory behavior is dependent on initial S wave polarization upon entry into the anisotropic D" layer, thus every record was individually modeled using the correct source mechanism. We exclude waveforms for which uncertainty in the source radiation pattern gives rise to uncertainty in the predicted SV waveform. While the overall strength of anisotropy is fairly well resolved, as it basically controls the timing of the flip in polarity of the SV arrivalsarrival time difference between fast and slow polarizations, the depth distribution of D" anisotropy is less constrained. The difference in overall SV and SH wave shape is partly partly a consequence of interference with the ScSV and ScSH core-reflections arriving a few seconds after S, which are opposite in polaritycause destructive interference on the SV components and constructive interference on the SH components.

Individual waveform modeling for 6 example observations is shown in Figure 3 for a range of isotropic, VTI and TTI models. These are compared to both VTI and isotropic models. Small tick marks indicate SH arrival times and ‘correctly polarized’ SV waves, which guides the choice of anisotropy strength for the VTI models. Due to the mixing of fast and slow polarization energy onto the SV and SH waveforms, the relative amplitudes are very dependent on the anisotropic properties, and not explained by isotropic models. Two of the data are relatively well modeled by the VTI model, but those two, as well as the other four data are well modeled by TTI models. In the latter four cases there is clear preference for either east or west tilting symmetry axes. We focus on precise modeling of the early part of the waveforms; the latter portions depend strongly on the poorly constrained details of the discontinuity structure and the individual receiver crusts.

The map in Figure 4 summarizes our modeling. The first-order result is the predominance of east dipping TTI in the southeast of the study area, and west dipping TTI toward the northwest. There is an intermingling of observations consistent with VTI throughout the study area, with numerous waveforms being compatible with both TTI and VTI predictions. Only a few waveforms are best explained by isotropy. A reasonable interpretation is that there are spatial fluctuations in the macroscopic fabric of the boundary layer, deviating from a VTI average as a consequence of complexity in the flow field.

Prior studies have proposed the existence of azimuthal anisotropy in localized regions of D", primarily below the Pacific (9,20,21), based on early arrivals of SV or ScSV signals. Those observations can be explained by TTI models with horizontal symmetry axes (90° tilt) sampled if theby raypaths are at azimuths perpendicular to the symmetry axes. The current lack of azimuthal sampling for any one region of of any D" region prevents resolution of more complex anisotropy systems such as those expected for lattice preferred orientation (LPO) of likely lower mantle minerals. MgO has been shown to have anisotropic properties and dominant slip systems that should favor SH velocities being faster than SV velocities in horizontally sheared portions of D", but with a 4 azimuthal variation in relative velocities(22-25). We calculated synthetics for the cubic symmetry of ferropericlase magnesiouwüstite (Mg,Fe)O using published elastic parameters (22), finding that .This structure has a 4 azimuthal variation in velocities and polarizations and specific crystallographic orientations can be found that reproduce the waveform behavior we observe here. The azimuthal dependence rapid variation with azimuth of the anisotropy for this system could may explain why our data show variations at a rather small scale, but lacking good azimuthal coverage this cannot be robustly resolved.

Deep mantle anisotropy may be related to mantle circulation, with mid-mantle downwelling currents beneath subduction zones possibly inducing LPO in cooled, high stress regions of D" (such as beneath the Caribbean), while hot upwelling regions, such as beneath groupings of surface hot spots, could give rise to shape preferred orientation (SPO) by, for example, aligning melt inclusions (6,12,18,19). Three-dimensional flow patterns could account for the spatial pattern in tilting of the symmetry axis imaged here, whether the fundamental cause of the anisotropy is LPO or SPO. Further modeling of data for other regions is needed to assess the overall variability in azimuthal anisotropy.

The apparent onset of anisotropy with the velocity jump at the top of D" suggests a link between the two, which could arise if there is a chemically distinct layer (1), a transition to a sheared zone of enhanced velocity heterogeneity(26), a change in iron partitioning associated with pressure-induced change in spin-state (27), or a major phase change(28,29). Improved azimuthal constraints will be required to distinguish between these possibilities, but we now have unequivocal evidence that extensive regions of D" anisotropy contain more complex structure than simple VTI, with high resolution seismic analysis being required to detect and map the structure. , andThe lateral variations in the observed azimuthal anisotropy are undoubtedly coupled to small scale dynamics in the deepest mantle.

References

1. Lay, T., Garnero, E. J., Williams, Q., Partial melting in a thermo-chemical boundary layer at the base of the mantle, Phys. Earth Planet. Int., in press (2004).

2. Garnero, E. J., A new paradigm for Earth’s core-mantle boundary, Science,304, 834-836. (2004).

3. Ekström, G., Dziewonski, A. M., The unique anisotropy of the Pacific upper mantle, Nature 394, 168-172, (1998).

4. Lay, T., Williams, Q., Garnero, E. J., The core-mantle boundary layer and deep Earth dynamics, Nature392, 461-468 (1998).

5. Dziewonski, A.M., Anderson, D.L., Preliminary Reference Earth Model, Phys. Earth Planet Int. 25, 297-356 (1981).

6. McNamara, A. K., van Keken, P. E., Karato, S.-I., Development of anisotropic structure in the Earth’s lower mantle by solid-state convection, Nature 416, 310-314 (2002).

7. Kendall, J.-M., Silver, P. G.,Constraints from seismic anisotropy on the nature of the lowermost mantle, Nature 381, 409-412 (1996).

8. Garnero, E. J., Lay, T., D" shear velocity heterogeneity, anisotropy, and discontinuity structure beneath the Caribbean and Central America, Phys. Earth Planet. Int. 140, 219-242 (2003).

9. Maupin, V. On the possibility of anisotropy in the D'' layer as inferred from the polarization of diffracted S-waves, Phys. Earth Planet. Inter.87, 1-32 (1994).

10. Kendall, J.-M., Nangini, C., Lateral variations in D" below the Caribbean, Geophys. Res. Lett.23, 399-402 (1996).

11. Lay, T., Helmberger, D. V., The shear-wave velocity-gradient at the base of the mantle, J. Geophys. Res.88, 8160-8170 (1983).

12. Moore, M. M., Garnero, E. J., Lay, T., Williams, Q., Shear wave splitting and waveform complexity for lowermost mantle structures with low-velocity lamellae and transverse isotropy, J. Geophys. Res. 109, B02319, doi:10.1029/2003JB002546 (2004).

13. Emery, V., Maupin, V., Nataf, H.C., Scattering of S waves diffracted at the core-mantle boundary: forward modeling, Geophys. J. Int. 137, 325-344 (1999).

14. Bostock, M.G., Ps conversions from the upper mantle transition zone beneath the Canadian landmass, J. Geophys. Res. 2, 8393-8402 (1996).

15. Bostock, M.G., Cassidy, J.F., Variations in SKS splitting across western Canada. Geophys. Res. Lett.22, 5-8 (1995).

16. Currie, C. A., Cassidy, J. F., Hyndman, R. D., Bostock, M. G., Shear wave anisotropy beneath the Cascadia subduction zone and western North American craton, Geophys. J. Int. 157, 341-353 (2004).

17. Rokosky, J. M., Lay, T., Garnero, E. J., Russell, S. A., High-resolution investigation of shear wave anisotropy in D" beneath the Cocos plate, Geophys. Res. Lett. 31, L07605, doi:10.1029/2003Gl018902 (2004).

18. Panning, M., B. Romanowicz, Inferences on flow at the base of Earth’s mantle based on seismic anisotropy, Science303, 351-353 (2004).

19. Kendall, J.-M., Seismic anisotropy in the boundary layers of the mantle, in Earth's Deep Interior: Mineral Physics and Tomography From the Atomic to the Global Scale, S. Karato, A.M. Forte, R.C. Liebermann, G. Masters, and L. Stixrude (eds.), pp. 133-159. American Geophysical Union, Washington, D.C., U.S.A. (2000).

20. Russell, S.A., Lay, T., Garnero, E. J., Seismic evidence for small-scale dynamics in the lowermost mantle at the root of the Hawaiian hotspot, Nature396, 255-258 (1998).

21. Pulliam, J., Sen, M. K., Seismic anisotropy in the core-mantle transition zone, Geophys. J. Int.135, 113-128 (1998).

22. Yamazaki, D., Karato, S.-I., Fabric development in (Mg,Fe)O during large strain, shear deformation: implications for seismic anisotropy in Earth's lower mantle, Phys. Earth Planet. Int.131, 251-267 (2002).

23. Stixrude, L., Elastic constants and anisotropy of MgSiO3 perovskite, periclase, and SiO2 at high pressure, In The Core-Mantle Boundary Region, eds. M. Gurnis, M. Wysession, E. Knittle, and B. Buffet, pp. 83-96, AGU, Washington D.C. (1998).

24. Karki, B. B., Wentzcovitch, R., de Gironcoli, S., Baroni, S., First-principles determination of elastic anisotropy and wave velocities of MgO at lower mantle conditions, Science 286, 1705-1709 (1999).

25. Mainprice, D., Barruol, G., Ben Ismail, W., The seismic anisotropy of the Earth’s mantle: from single crystal to polycrystal, In Earth's Deep Interior: Mineral Physics and Tomography From the Atomic to the Global Scale, eds. S. Karato, A. M. Forte, R. C. Liebermann, G. Masters, and L. Stixrude, pp. 237-264, AGU Washington, D.C. (2000).

26. Cormier, V. F., Anisotropy of heterogeneity scale lengths in the lower mantle from PKIKP precursors, Geophys. J. Int. 136, 373-384 (1999).

27. Badro, J., Fiquet, G., Guyot, F., Rueff, J.P., Struzhkin, V. V., Vankó, G., Monaco, G., Iron partitioning in Earth’s mantle: Toward a deep lower mantle discontinuity, Science 300, 789-791 (2003).

28. Murakami, M., Hirose, K., Kawamura, K., Sata, N., Ohishi, Y., Post-perovskite phase transition in MgSiO3, Science 304, 855-858 (2004).

29. Shim, S.-H., Duffy, T. S., Jeanloz, R., Shen, G., Stability and crystal structure of MgSiO3 perovskite to the core-mantle boundary, Geophys. Res. Lett. 31, L10603, doi:10.1029/2004GL019639 (2004).

30. Grand, S.P., Mantle shear-wave tomography and the fate of subducted slabs, Phil. Trans. R. Soc. Lond. A 360, 2475-2491 (2002).

31. Acknowledgements. We thank Michael Bostock for assistance in accessing some of the Canadian National Seismic Network waveform data, and the IRIS Data Management Center for all other waveform data. This work was supported by the U.S. National Science Foundation. VM was on leave to Institut de Physique du Globe de Paris while part of this work was done.

Figure captions

Figure 1. Broadband shear wave data from station FCC for the events of (A) 29 April 1994 and (B)10 May 1994. SV and SH ground velocity recordings (Vel.) are shown along with ground displacements (Disp.). SKS is cleanly isolated on the SV components. The predicted polarity of the SV component of the S arrival is downward in the figure, and the short vertical line indicates an arrival that could be picked as a late SV on the velocity records (delay times noted on records), but in the displacement record there is clearly SV energy arriving at the same time as the strong SH onset (shaded, peak noted by inverted triangle). The SH waveforms show two arrivals, which correspond to interactions with a shear velocity increase at the top of D". (C). Examples from other D” regions. D" sampling of data are indicated by thick black lines in map, with shading indicating past D" anisotropy study regions.

Figure 2. (top) Tilt (T) and azimuth (AZ) parameters of the symmetry axis (dashed line), for models of tilted transverse isotropy (depicted by sheets). (below) Full wave theory synthetics (SV-solid line; SH-dashed line) for varying T (left column) and AZ (right column) values computed for the parameters of station FRB at a distance of 91.8° for the event 29 April 1994 (see corresponding data in Figure 3). ISO indicates isotropic reference model, with a 2.7% D" discontinuity 250 km above the CMB. AZ=90° for the varying T calculations, and T=20° for the varying AZ calculations. Several examples of SV signals having a reversed polarity onset (shaded) are present. The strength of the anisotropy for all TTI models is 2.2%.

Figure 3. Observed SH (blue) and SV (red) displacement traces for six event-station pairs (columns a through f). The data are shown with original polarities, normalized to the peak SH amplitudes, with the top row having small markers at the onset of SH and correctly polarized SV arrivals. The data are compared with synthetics (dotted) for four models: top row, W(20) synthetics are for tilted transverse isotropy with T=20°, Az=270°; second row, E(20) synthetics are for tilted transverse isotropy with T=20°, Az=90°; third row, VTI synthetics are for transverse isotropy with vertical symmetry axis; bottom row, ISO synthetics are for isotropic D” structure. Asterisks (*) indicate models that fit the data very well.