The lowermost mantle plays a critical role in the dynamics and chemistry of our planet. Also known as D", this region of the Earth reveals a complex and puzzling seismic structure. The characterization of 2D and 3D heterogeneity at the CMB has so far been mostly qualitative and few studies have , let alone 3D forward modeling of body wave observations to quantitatively constrain the horizontal and vertical extent of prominent structures. This work is an attempt to obtain an improved local 3D picture of heterogeneity in the lowermost mantle beneath the Central Pacific.
We considered the variations with distance of S-SKS and SKKS-SKS travel time residuals computed with respect to PREM reference model (Dwiewonski and Anderson, 1981) for a fixed source or station and along narrow azimuthal corridors. This technique was used previously to document that strong lateral variations in S velocity must locally be present in D" (Vinnik et al., 1998; Bréger et al., 1998). We analyzed residuals for several large events in the Fiji-Tonga-Kermadec Islands source region recorded on a large number of North American stations (Figure 23.1). By using differential travel times between S and SKS, or SKS and SKKS, we minimize errors due to source mislocation, uncertainty on earthquakes origin times, and heterogeneity in the crust and upper-mantle beneath sources and receivers.
As noted previously for a few profiles in the same region (Vinnik et al., 1998; Bréger et al., 1998), differential travel time residuals show systematic trends and little scatter, which indicates a complex 3D velocity structure at a scale of several hundred km at the base of the mantle.
A starting 3D model was chosen by computing predicted differential travel time residuals for several recent tomographic models and selecting the model which gave the closest match to the observed trends (model SAW12D, (Li and Romanowicz, 1996)). Then, by forward modeling of travel time residuals, we progressively perturbed the starting model in two ways: 1. by preserving the spatial distribution of anomalies and only modifying the amplitude of velocity fluctuations in selected high or low velocity regions of the model and 2. by slightly shifting their positions laterally when necessary to fit the observations. We were able to achieve fits to all our data within 1-2 sec, which is within the accuracy of the travel time measurements.
Our modeling reveals two prominent features: a large slow region at the bottom of the mantle south of Hawaii in the last 1000 km of the mantle (S in Figure 23.1), that resembles in shape that of the starting tomographic model, and where the velocity anomaly needs to be increased to 3.5-4%. A second fast region (F in Figure 23.1), adjacent to the large slow anomaly, where the velocity anomaly reaches 4 to 5% in a small domain that had only a mildly fast anomaly of about 1 to 2% in the original model. The exact shape and strength of those two required heterogeneous domains can vary slightly from one possible model to another, but by no more than 100-200km in position and 1% in amplitude.
The very large velocity anomaly could be in part related to partial melt, as proposed to explain the ultra low velocities observed in some parts of this region (Williams and Garnero, 1996), but some chemical heterogeneity may also be present.
The high velocity region documented by our modeling represents a contrast of about 7-8% with respect to adjacent "hot" mantle and cannot be readily explained by thermal effects alone. A portion of ancient slab lying at the CMB seems unlikely to be responsible for this velocity contrast since reconstructions of ancient subductions do not predict the presence of remnant lithosphere in this part of D" (Lithgow-Bertelloni and Richards, 1998), and since such a fossil slab may not produce a sufficient velocity contrast. Whether it represents a high velocity product of the decomposition of perovskite at temperatures and pressures corresponding to the lowermost mantle (Dubrovinsky et al., 1998), or of core-mantle reactions (Knittle and Jeanloz, 1991; Stixrude, 1998), is still nclear. This block of fast material lying at the core-mantle boundary may also be responsible for the intermittent reflections observed at the top of D" in this very region (Garnero et al., 1988; Mori and Helmberger, 1995).
We thank Ed Garnero for making his dataset available to us.
Bréger, L., B. Romanowicz, and L. Vinnik, Test of tomographic models of D" using differential travel time data, Geophys. Res. Lett., 25, 5-8, 1998.
Dubrovinsky, L.S., S.K. Saxena, P. Lazor, R. Ahuja, O. Eriksson, J.M. Wills, and B. Johansson, Experimental and theoritical identification of a new high-pressure phase of silica, Nature, 388, 362-365, 1998.
Dziewonski, A. M., and D. L. Anderson, Preliminary Reference Earth Model, Phys. Earth Planet. Int., 25, 297-356, 1981.
Garnero, E.J., D.V. Helmberger, and G. Engen, Lateral variations near the core-mantle boundary, Geophys. Res. Lett., 15, 609-612, 1988.
Knittle, E., and R. Jeanloz, Earth's core-mantle boundary: Results of experiments at high pressure and temperature, Science, 251, 1438-1443, 1991.
Li, X.-D., and B. Romanowicz, Global mantle shear velocity model developed using nonlinear asymptotic coupling theory, J. Geophys Res., 101, 22245-22272, 1996.
Lithgow-Bertelloni, C., and M.A. Richards, The dynamics of Cenozoic and Mesozoic Plate, Rev. Geophys., 36, 27-78, 1998.
Mori, J., and D.V. Helmberger, Localized boundary layer below the mid-Pacific velocity anomaly identified from a PcP precursor, J. Geophys. Res., 100, 20359-20365, 1995.
Stixrude, L., Elastic constants and anisotropy of MgSiO3, perovskite, periclase, and SiO2 at high pressure, in The Core-Mantle boundary, (eds Gurnis, M., Buffett, B.A., Knittle, E., and Wysession, M.)(Am. Geophys. Union), 1998.
Vinnik, L., L. Bréger, and B. Romanowicz, Anisotropic structures at the base of the Earth's mantle, Nature, 564-569, 1998.
Williams, Q., and E.J. Garnero, Seismic evidence for partial melt at the base of the Earth's mantle, Science, 273, 1528-1530, 1996.