Seismological support for the metastable superplume model, sharp features, and phase changes within the lower mantle

Sun et al. 10.1073/pnas.0608160104.

Supporting Information

Files in this Data Supplement:

SI Text
SI Figure 7
SI Figure 8
SI Figure 9
SI Figure 10
SI Figure 11

SI Figure 7

Fig. 7. SKS waveforms predicted for the HBMS model. The predictions with rays from both sides show a rapid delay of SKS and waveform complexity when sampling the boundary zone (shaded zone). The SKS ray paths are shown in Fig. 3a.

SI Figure 8

Fig. 8. The cross-sections from South Sandwich Island to the Ethiopia/Kenya array for the hybrid model (a) and the HBMS model (b) with background Grand's tomography model. In the hybrid model, the S velocity anomaly inside box I is -3% and -2% inside Box II. The second Box II was added to fix the relative SKS and S timing issue by slowing S as can be seen in the ray path sampling. This is not unique and any extra S delays along these particular paths will satisfy the data.

SI Figure 9

Fig. 9. The comparison with observed waveforms against predictions from the HBMS model (a) and hybrid models constructed in Fig. 6. In b, only box I in SI Fig. 8 is enhanced. Both box I and II are enhanced in c.

SI Figure 10

Fig. 10. The comparison with observed waveforms against predictions from the CM model (a) and hybrid model (b). In the CM model, the velocity is that given in Fig. 1b except the velocity in the bottom 100 km has been increased by 4% (see Fig. 5b).

SI Figure 11

Fig. 11. Display of ray paths for PKP, PKP(DF) in cyan, and PKP(AB) in magenta, for geometrics sampling the edge structures relative to PREM along with synthetic predictions. Solid traces correspond to the edge structure with dotted traces relative to PREM. Note the nearly 2-s offset caused by sampling the edge structure.

SI Text

Synthetic record sections for event 970529 are displayed in SI Fig. 7. The model is a 2D section containing the HBMS model embedded in PREM. These synthetics were generated with the WKM technique introduced in Ni et al. (1). The method is basically analytical, which satisfies the wave equation assuming tomographic-type models. A comparison of synthetic seismograms for the model displayed in Fig. 1b, using the above semianalytical code with a 2D numerical technique, is given in Ni et al. (2). A similar record section of data and synthetic for this event is discussed by Helmberger and Ni (3). The event occurred on Feb. 10, 2002 at South Sandwich Island at a depth of 200 km. This depth allows the direct arrivals to be well separated from the depth phases, sSKS, etc. The event is well recorded by the Ethiopia/Kenya array and provides ideal geometry for studying the structure directly beneath a superplume (SI Fig. 8). Both the HBMS model and Ni's model displayed in Fig. 1b (box I) can not produce the large separation between S and SKS at small distances (~ 82-84°) (SI Fig. 9 a and b). Adding the extra box II, with the 2% velocity reduction, slows the arrival time of S and has no effect on the SKS phase, and produces the observed differential time between S and SKS in SI Fig. 9c. In the HBMS model, there are some extra small arrivals between S and S_CS caused by the down-welling. All three models in SI Fig. 9 fail to explain the observed Scd phase. Moreover, the S_CS on the synthetics appear to be too late at the smaller ranges, which suggest the addition of a fast zone near the top of D". Both data and synthetics are aligned on the arrival of S by cross-correlation.

To fit the observed Scd phase, we first assume a model with an abrupt shear velocity jump across the boundary. By grid search, we derive a model (CM) with 4% jump located 100 km above the CMB (Fig. 5b). The synthetics for this model are shown in SI Fig. 10a. A significant feature in the HBMS model is the down-welling region near the center, which will cause a diffuse increase in seismic velocity. To mimic this structure, we adopted a model (hybrid model) with one linear gradient followed by a small velocity jump (1.7%) shown in Fig. 5b. Both models provide a reasonable fit to the data (SI Fig. 10). The small velocity jump in the hybrid model supports the possible PV-to-PPV transition across the boundary seen globally, whereas the CM model appears more unique. Because the down-welling in the HBMS model raises the possibility of the occurrence of phase transition in this area, we prefer the hybrid interpretation.

The position and the magnitude of the negative velocity jump beneath the boundary is not well constrained, although adjustments can be made to correct the travel time of S_CS. Because a velocity reduction is more difficult to detect than a velocity increase (4, 5), there is no constraint on the exact velocity structure below the boundary of the positive velocity jump in both models.

The P-wave velocity structures in HBMS are not anomalous on average as pointed out earlier with respect to P_d. Moreover, many phases that travel horizontally encounter both fast and slow zones, such as P, P_CP, etc. Vertically traveling phases have the best chances of detecting the abrupt lateral changes near the edges, in particular the differential branches of PKP phases. Although the edge anomalies are narrow, they are easily detected in data. As first pointed out by Shearer and Toy (6), the differential travel delays between PKPdf and PKPab is not PREM-like or well behaved in many regions. Global studies of these phases indicate that the mid-Pacific contains some of the strongest anomalies (7, 8). SI Fig. 11 displays synthetics for the two PKP branches (DF and AB) sampling the edge structure of HMBS.

In the HBMS model, the conversion from temperature (T) and composition (C) to seismic velocities is based on the following parameters:

Here, (dX/dY)_Z is the abbreviation for (∂lnX/?Y)_Z, K_S is the bulk modulus, and mis the shear modulus. The value of (dK_S/dC)_T is dictated by the dynamic model. The value of (dm/dC)_T is a free parameter and is chosen to fit the seismic observation. When scaled, the dimensional values of (dK_S/dT)_C and (dm/dT)_C used in the conversion are -3.33 ´ 10-⁵ K-¹ and -8.33 ´ 10-⁵ K-¹, respectively. More information about the conversion is described by Tan and Gurnis (9). For reference, the dimensional values of (dK_S/dT)_C and (dm/dT)_C for MgSiO₃ perovskite are calculated to be -3.29 ´ 10-⁵ K-¹ and -8.62 ´ 10-⁵ K-¹, respectively (10).

1. Ni SD, Ding XM, Helmberger DV (2000) Geophys J Int 140:71-82.

2. Ni SD, Helmberger DV (2003) J Geophys Res, 10.1029/2001JB001545.

3. Helmberger DV, Ni SD (2005) in Earth's Deep Mantle: Structure, Composition, and Evolution, eds van der Hilst RD, Bass JD, Matas J, Trampert J (Am Geophys Union, Washington, DC), Am Geophys Union Geophys Monogr 160, pp 63-81.

4. Flores C, Lay T (2005) Geophys Res Lett, 10.1029/2005GL024366.

5. Sun DY, Song TR, Helmberger DV (2006) Geophys Res Lett, 10.1029/2005GL025384.

6. Shearer PM, Toy KM (1991) J Geophys Res 96:2233-2247.

7. Song XD, Helmberger DV (1993) Geophys Res Lett 20:285-288.

8. Cormier VF, Choy GL (1986) Geophys Res Lett 13:1553-1556.

9. Tan E, Gurnis M (2006) J Geophys Res 10.1029/2006JB004505.

10. Oganov AR, Brodholt JP, Price GD (2001) Nature 411:934-937.