The Upper Mantle Transition Zone in the Western Pacific

A little more than a decade ago, a whole mantle convection or a layered convection was still a matter of debate. Mineral physics predicted an eventual whole mantle convection, but seismic modeling was not able to constrain such a feature even using similar data sets. In the 1990's seismic tomography models captured a large-scale subhorizontal high velocity anomaly that suggested a large volume of stagnant cold slab as well as down going slab images into lower mantle at the bottom of the upper mantle. While such tomographic approaches are robust for modeling the Earth's deep structure, the gap between the conjecture based on large scale tomographic images and the reality of mantle property measurements in mineral physics is enormous. In an effort to reduce such a gap we have carried out broadband waveform modeling using regional body waveform data with support from NSF (EAR-9526678, EAR-9805006, EAR-9996301). The seismic models developed have shown some correspondences to mineral physics experimental results or modeling [Tajima, 2000].

Volume of Stagnant Slab vs. Subducted Slab?

We have shown various features associated with flattened cold slab in the back arc of southern Kuriles, central Japan and Izu-Bonin subduction zones. The velocity models M3.11 and M2.0 developed for the structure with stagnant slab in the transition zone are characterized by high velocity anomaly in the deeper part of the transition zone ( $\sim$ 525 to 660 km) with its maximum intensity in a depth range $\sim$ 100 km above the 660 km discontinuity. The volume of stagnant slab estimated from the waveform modeling is much smaller than that in the tomographic images, and comparable to the subducted slab in the past $\sim$ 17 Ma. There is variation of the discontinuity depth in the range between 660 and 690 km from the Kuriles to Izu-Bonin subduction zones where the subduction and back arc opening history is more or less similar to each other in the past $\sim$ 17 Ma. Whether the discontinuity depth is depressed or not suggests the temperature condition right beneath the flattened slab, i.e., the temperature may be normal possibly due to the higher thermal conductivity at the bottom of the flattened slab, or the supply of cold slab over the flattened slab is less in the northern Philippine Sea than in the Kuriles. While the subducting slabs in the southern Kuril-Japan-Izu Bonin arcs are bent to subhorizontal above the "660" km discontinuity, the slab-like high velocity zone penetrates this discontinuity with considerable spreading in the Java arc. We are also intrigued by the image of high velocity anomalies that extend into lower mantle from a stagnant zone beneath the transition zone from place to place. This image may indicate that a stagnant slab eventually descends into lower mantle and is favorable for a mantle-wide convective model.

**Figure:** Map of the study areas. Solid triangles show broadband station locations excluding stations in Japan where the network coverage is dense. Epicenters of recent intermediate to deep-focus ( $\geq 300 km$ ) events (M $\geq 5.2$ ) are shown with stars. Two deep focus events #18 and #19 in the Sunda arc are shown with large stars (see the text). Lines A to H illustrate cross sections of the tomographic model perpendicular to the trench line.
$\begin{figure} \begin{center} \epsfig{file=fumiko00_2_1.ps, width=7cm}\end{center}\end{figure}$

Does stagnant slab eventually descend into lower mantle?

If so, what controls the slab behavior for stagnation and how much of the subducted slab can stagnate in the transition zone? These issues are being explored. Tajima et al. [1996] identified early high-frequency phases which are multiple arrivals of P waves at some stations on the Eurasian continent and the surrounding regions from deep focus events in the Java Sea to Flores Sea depending on the relative locations of the ray paths to the high velocity zone beneath Kalimantan (formerly Borneo) (see event and station locations Fig. 32.1).

**Figure 32.2:** Waveforms observed for the deep focus events #18 and #19 in the Sunda arc. The focal mechanisms with station locations are also shown. Waveforms in the two center columns were recorded at stations to which the rays propagated through or near the high velocity zone beneath Kalimantan (formerly Borneo) in the tomographic model. On the other hand the propagation paths for the waveforms in the left column for Event 19, or those in the right column for Event 18 did not pass though that zone. While these P wave trains indicate a simple on set of the source process, those in the center columns commonly include high frequency phases in the beginning.
$\begin{figure} \begin{center} \epsfig{file=fumiko00_2_2.ps, width=7cm} \par\end{center}\end{figure}$

Figure 32.2 shows an example of waveform sets from Events #18 (depth=596 km, M=5.9, 08/30/94) and #19 (depth=638 km, M=5.9, 09/28/94) that took place in the Java Sea region. Event #19 is located southwest off Borneo and its ray paths to QIZ, CHTO and LSA are outside the high velocity zone beneath Kalimantan. The waveforms at these stations show simple P-wave onsets while those at stations in the direction to Kalimantan show substantial high frequency arrivals (see the caption for Event #18 in Fig. 32.2). Comparison of the waveforms recorded in different directions show structural effects specific for propagation path such as apparent multi-paths waves that propagated beneath Kalimantan in which a low velocity zone outside the slab of high velocity anomaly may need to be accounted for.

Generally we often encounter an equivocal situation for deep-focus events that the waveforms could be modeled either with structural effects associated with the deep slab or with a source of multiple subevents. Waveforms need to be analyzed accounting for both effects. In addition to the previously used reflectivity code we are using a 3-D finite difference code [Larsen, 1995] to model structure along profiles between earthquake and station locations focusing on the boundary structure (see the illustration in Fig. 32.1).

Java Region - Slab Penetration and Possible "920" km Discontinuity?

The issues of "mid-mantle discontinuities" at around 920 to 1100 km are also being investigated. Most of the resulting images suggest the existence of a possible mid-mantle seismic discontinuity ("920" km discontinuity) with about 3 % of velocity increase associated with the descending slabs into the lower mantle (see also Kawakatsu and Niu, 1994). Niu and Kawakatsu [1997] identified some near-source S-P converted waves that were interpreted to have occurred at the mantle transition discontinuities in the depth range between 900 and 1100 km. The depth variation of the mid mantle discontinuity beneath the Indonesia arc appears to be well correlated with the location of the high velocity anomalies in the recent tomographic models [Kawakatsu and Niu, 1997]. The "discontinuities" may not be flat as previously interpreted (e.g., Niu and Kawakatsu [1997]) but the anomalous arrivals could be due to S-P conversions from more vertical structure associated with the deep anomaly seen in the tomographic models. Using the 3-D finite difference modeling of wave propagation we are investigating other possibilities for the "mid-mantle discontinuities" not associated with flat layering. If the complicated arrivals are due to structure, they should provide constraints on the amplitude and gradients associated with this lower mantle (or "mid-mantle") anomaly.

We will eventually compare the results between the regions concerning slab behaviours with a certain degree of flattening in the transition zone or descending into lower mantle and attempt to understand the apparent difference. At present the 3-D modeling is especially important for the Java region as our preliminary check of the waveforms indicate the substantial effects of the deep subduction zone below the "660" km discontinuity. Our starting models along the cross section between a hypocenter and a given station based on the images from the ISC P tomographic model [Obayashi et al., 1997] (see the illustration of cross sections in Fig. 32.1).

Acknowledgements

We thank S. P. Grand for useful suggestions for this work, S. Larsen for the use of his 3-D finite difference code, Y. Fukao and M. Obayashi for the use of their P-wave tomographic model, and H. Kawakatsu for discussion.

References

Engebretson, D. C., K. P. Kelley, H. J. Cashman, and M. A. Richards, 180 millioan years of subduction, GSA Today, 2, 93-100, 1992.

Kawakatsu, H., and F. Niu, Seismic evidence for a 920-km discontinuity in the mantle, Nature, 371, 301-305, 1994.

Larsen, S., 3D MPP simulations in the earth sciences: seismic applications, LLNL symposium on distributed computing and massively parallel processing, 1995.

Niu, F., and H. Kawakatsu , Depth variation of the mid-mantle seismic discontinuity, Geophys. Res. Lett., 24, 429-432, 1997.

Obayashi, M., T. Sakurai, and Y. Fukao, Comparison of recent tomographic models, Abstract of International Symposium on New Images of the Earth's Interior through Long-term Ocean-floor Observations, 29, 1997.

Tajima, F., F.-L. Niu, H. Kawakatsu, and Y. Fukao, Velocity anomalies and discontinuity depths in the mantle transition zone from broad-Band waveform modeling: (II) beneath Java, Abstracts of the 1996 Japan Seismological Society Meeting, Fall 1996 (full text in preparation).

Tajima, F., Seismic probing of the upper mantle transition zone with stagnant cold slab and mineral physics correspondences, Eos Trans., AGU, 81 (22), WP187, 2000 (full text in preparation).

Tajima, F., and S. P. Grand, Variation of transition zone high velocity anomalies and depression of the "660"km discontinuity associated with subduction zones from the southern Kuriles to Izu-Bonin, J. Geophys. Res., 103, B7, 15015-15036, 1998.

Tajima, F., Y. Fukao, M. Obayashi, and T. Sakurai, Evaluation of slab images in the Northwestern Pacific , Special issue of the Ocean Hemisphere Project symposium, Earth Planet. Space, 50, no. 11 & 12, 953-964, 1998.