GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L01302, doi:10.1029/2007gl031658, 2008 Topography of the 660-km discontinuity beneath northeast China: Implications for a retrograde motion of the subducting Pacific slab Juan Li, 1,2 Qi-Fu Chen, 3 Elizabeth Vanacore, 4 and Fenglin Niu 4 Received 21 August 2007; revised 29 October 2007; accepted 9 November 2007; published 8 January 2008. [1] Clear S to P converted waves at the 660-km discontinuity (S 660 P) are observed at stacked seismograms of deep earthquakes occurring in northeast China and Japan Sea recorded by broadband seismic arrays in North America and Europe. Differential travel times between the S 660 P and P waves are used to constrain depth variations of the 660 beneath northeast China. A rapid change in the depth of the 660 is observed within a narrow longitudinal range of 130.8 E 131.4 E, where the lower boundary of the subducting slab encounters the 660. Towards the west, the 660 deepens steadily as it approaches the coldest core of the slab. The maximum depression occurs at the west end of the studied region with an amplitude of 20 km. To the east, the 660 appears to be flat, showing no obvious effect of the subducting slab. These observations are consistent with a scenario for a retrograde moving Pacific slab lying over the upper and lower mantle boundary progressively from west to east. Citation: Li, J., Q.-F. Chen, E. Vanacore, and F. Niu (2008), Topography of the 660-km discontinuity beneath northeast China: Implications for a retrograde motion of the subducting Pacific slab, Geophys. Res. Lett., 35, L01302, doi:10.1029/2007gl031658. 1 State Key Laboratory of Lithosphere Evolution, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China. 2 Now at Department of Earth Science, Rice University, Houston, Texas, USA. 3 Institute of Earthquake Science, China Earthquake Administration, Beijing, China. 4 Department of Earth Science, Rice University, Houston, Texas, USA. Copyright 2008 by the American Geophysical Union. 0094-8276/08/2007GL031658 1. Introduction [2] Whether slabs of old oceanic lithosphere penetrate the 660-km discontinuity (hereafter referred to as the 660) directly into the lower mantle, or are trapped in the middle mantle is an important issue in order to understand scales of mantle circulation. The fate of the subducting slabs, however, remains an area of active debate, partly because of the complicated tomographic images obtained from various subduction zones [e.g., van der Hilst et al., 1997; Fukao et al., 2001; Grand, 2002]. Along the west margin of the Pacific plate, slabs beneath northern Kurile and Mariana appear to sink directly into the lower mantle; while beneath southern Kurile, Japan and Izu-Bonin arcs, slabs are found to be lying horizontally over the 660 [e.g., Fukao et al., 2001]. [3] While major subduction characteristics can be described by the thermal parameter of the subducting slab [e.g., Stern, 2002], the shape of the slab appears to be affected by many other processes and factors, such as trench migration, resistance from the 660, rheology of subducting lithosphere. Numerical modeling found that under different circumstances, it is possible to generate various styles of slab behavior shown in seismic tomography images [Christensen, 1996; Tagawa et al., 2007]. In general, trench migration tends to prevent slab penetration into the lower mantle and to facilitate the slab flattening above the phase boundary [Christensen, 1996; Tagawa et al., 2007]. Besides trench migration, nature of the 660 and viscosity contrast across it also seem to play an important role in controlling slab morphology. It is now generally believed that the 660-km seismic discontinuity is caused by the temperature-sensitive phase transition from ringwoodite to perovskite and magnesiowustite [Ito and Takahashi, 1989]. Because the phase change has a negative Clapeyron slope, a temperature induced depression of the 660 is expected within a cold subducting slab [Vidale and Benz, 1992; Wicks and Richards, 1993; Niu and Kawakatsu, 1995; Collier and Helffrich, 1997]. [4] One ideal location to study the interaction between a subducting slab and the 660 is northeast China, where the subducted Pacific slab seems to be deflected subhorizontally around the 660 [Fukao et al., 2001]. A large-scale depression of the 660 is observed from SS precursor data [Shearer and Masters, 1992; Flanagan and Shearer, 1998]. On the other hand receiver function study by Li and Yuan [2003] found a deeper-than-normal 660 over an area of 400 km centered by a localized depression where the 660 reaches as deep as 695 km. Both the SS and receiver function analyses require a good velocity model of the upper mantle in order to position the reflected and converted wave field to proper depths and locations. Here we use the S to P wave converted at the 660 to map fine-scale lateral variations in the depth of the 660. The method depends only on the velocity structure between a deep earthquake and the conversion point, and subsequently is less sensitive to the velocity of the heterogeneous upper mantle. 2. Data and Analysis [5] One component of P-coda wave of a teleseismic record is S to P conversion waves at mantle discontinuities located below the source, for example, the S 660 P, which starts as a down-going S wave and is subsequently converted to a P wave at the 660. Because the S 660 P and P travel along almost identical ray paths after the S to P conversion, the differential travel time of the two phases, t S660P-P, is a function of distance between the source and conversion point. Compared to P wave, the S 660 P has a smaller slowness, and thus is received at a steeper incident angle. L01302 1of7
Figure 1. Maps showing (a) the source-receiver geometry, (b) station locations of GRSN, (c) eight deep earthquakes beneath northeast China and Japan Sea, and (d) station locations of SCSN, UW, and RISTRA. Conversion points of S 660 P directing to GRSN and the three US arrays are indicated by solid circles and open squares, respectively, with a solid line connecting conversion point pair of the same event. The Wadati-Benioff zone is indicated by dashed lines. Inset shows the initial portion of ray paths of P and S 660 P schematically. [6] Seismic arrays used in this study include the Southern California Seismic Network (SCSN), the Pacific Northwest Regional Seismic Network (UW), the Colorado Plateau/Rio Grande Rift Seismic Transect Experiment (RISTRA), and the German Regional Seismic Network (GRSN). The source-receiver geometry allows investigating the subduction zone beneath northeast China from both eastward and westward directions (Figure 1). Both the UW and SCSN networks have 100 200 stations, while the RISTRA and GRSN arrays consist of 50 and 30 stations, respectively. [7] We selected a total of eight deep events with a source depth >350 km and a simple source time function that occurred beneath northeast China and Japan Sea (Table 1). These earthquakes are distributed approximately between 129 E and 136 E, covering a wide area that has either significant or no presence of the subducting slab right above the 660. We visually examined every vertical-component seismogram and excluded those with low SNR, bad records or low waveform similarity to other stations. We handpicked the depth pp phase and used the differential travel time, t pp-p, to redetermine the source depth of each event. A comparison of focal depths redetermined from the GRSN and US of the same events suggested that uncertainty in the relocated depth is < 3 km. [8] The selected seismograms were first low-pass filtered with a corner frequency of 0.5 Hz. We then manually picked the first peak of the P waves and aligned them at the picked times. All the seismograms were normalized by their maximum amplitude in the time window 10 140 s relative to P. To enhance the amplitude of the small later arrivals and estimate their slowness, we employed a slant stacking method. We varied slowness within the range of ±2 s/deg with an increment of 0.02 s/deg. Time correction for each trace is calculated with a plane wave assumption before each stacking. We have a total of 101 stacked traces which are converted to amplitude envelopes using Hilbert transform. The maximum amplitude is chosen from all the stacked traces and is used to normalize the traces in a unit of decibel. The final result is indicated by a color contour map in which hotter color clusters represent greater energy and may indicate possible phase arrivals. An Nth root stacking algorithm was employed to perform the summation of seismograms [Muirhead, 1968]. The 2nd root stacking results of the eight events recorded by different arrays are shown in Figure 2. Events are sorted by focal depth with the left and right panel showing ray paths directing to the west and east, respectively. We also made linear stacks of all the available event-array pairs and used them to measure t S660P-P, as linear stack preserves the shape of waveform while Nth root stack can severely distort the original waveform. [9] We used the iasp91 velocity model [Kennett and Engdahl, 1991] to convert t S660P-P to the conversion point 2of7
Table 1. Source Parameters and Observed Depth of the 660-km Discontinuity Event Origin Date, (Yr Mo Da) Lat., Lon., ( N, E) Magnitude, Mw Depth (km), PDE, pp-p Array Distance, (deg) Conversion Point, Lat. ( N), Lon. ( E) 660 Depths (km) 1 1981 11 27 42.91, 131.08 5.8 543, 544 GRSN 73.99 43.21, 130.78 671.8 SCSN 81.14 43.12, 131.44 667.5 2 1983 10 08 44.23, 130.74 6.3 558, 566 GRSN 72.79 44.50, 130.46 674.4 3 1990 05 11 41.82, 130.86 6.3 578, 577 GRSN 74.78 42.03, 130.65 672.4 4 2000 02 13 42.85, 131.57 6.0 513, 520 RISTRA 86.24 43.12, 131.94 667.8 5 2002 09 15 44.83, 129.92 6.4 586, 588 GRSN 71.50 45.03, 129.71 679.5 UW 69.33 45.02, 130.18 673.6 6 2003 08 31 43.39, 132.27 6.1 481, 489 GRSN 73.81 43.85, 131.82 667.2 SCSN 79.50 43.71, 132.83 668.1 7 2006 09 16 41.36, 135.70 5.9 369, 377 SCSN 79.11 41.83, 136.57 665.5 8 2007 03 09 43.22, 133.55 6.1 441, 442 GRSN 74.22 43.79, 133.02 666.6 SCSN 78.57 43.61, 134.26 664.6 depth. We also used PREM [Dziewonski and Anderson, 1981] for the time-depth conversion and found no significant difference (<1 km) between the two models. We further estimated effects of 3D velocity structure on the time-depth conversion. To do this, we assumed a 2D raypath and counted the cumulative travel time anomalies caused by 3D velocity structure along the raypath for both P and S 660 P. We used the P-wave velocity model of Fukao et al. [2001] and an S-wave velocity model converted from the P model with a scaling relationship of dlnv s /dlnv p = 1.6 [Karato and Karki, 2001]. In general, the 3D corrections on t S660P-P are trivial, usually less than 0.3 s except for the shallowest event for which the correction is 0.5 s. The depths listed in Table 1 include this 3D correction. We also calculated the 3D corrections with other global models [van der Hilst et al., 1997; Grand, 2002] and found the corrections from different models agree quite well (<0.2 s). As mentioned above, ray paths of S 660 P and P differ only within the source region before the S to P conversion and become almost identical after the conversion. The source region of the western Pacific is well resolved in most of the global models because of the large amount of earthquake data. This may explain the consistence in the 3D corrections among different models. In general the magnitude of lateral heterogeneity in the transition zone is much smaller than the upper mantle, so the error in t S660P-P associated with unmodeled 3D structure should be very trivial. Moreover, this unmodeled 3D structure error should apply to all the depth measurements, and thus should not affect our results on the depth variation of the 660. Our synthetic tests indicated that the slant stacking with a plane wave assumption causes almost no anomaly on t S660P-P. We used a cross correlation method to measure t S660P-P, which in principle permits sub-sample precision. Thus the processing error including stacking and picking is expected to be less than one sampling interval (0.05 s). Including all the errors discussed above, the uncertainty on the 660 depth measurements is anticipated to be less than 5 km. 3. Results and Discussion [10] The subducting Pacific slab beneath Japan Sea and northeast China between 41 N and 45 N shows an approximately north south trend as indicated by the Wadati- Benioff zone [Gudmundsson and Sambridge, 1998]. We thus projected the estimated depths of the S 660 P conversion points along latitude 43 N (Figure 3a). There is a rapid change in the depth of the 660 within the longitudinal range of 130.8 E 131.4 E (red symbols in Figure 3a). The location appears to coincide with the place where the lower boundary of the subducting Pacific lithosphere encountered the 660 (Figure 3b). Although our data have no resolution on whether the transition occurred gradually or abruptly within this longitudinal range, the differential depth in the 660 between the red square and red circle must be robust as they are determined from the same earthquake and thus should not be biased by earthquake depth. East of the rapid transition, the depth of the 660 appears to be flat, with no statistically significant variation. Towards the west, the 660 deepens steadily as it approaches the cold core of the subducting slab. The maximum depression, 20 km, is observed at the most west sample point. [11] Among the total of eight events, we are able to observe evident S 660 P simultaneously in the GRSN array and one of the three US arrays from four events, resulting in four-pair estimates of the 660 depth. As discussed above the relative depth in a pair is well determined since it is not affected by source depth. Because of the overlap in the conversion points, the topographic relief shown in Figure 3a should be a robust observation. [12] In general, our estimate of the depth of the 660-km discontinuity is consistent with previous results of receiver function analysis [Li and Yuan, 2003] and SS precursor studies [Shearer and Masters, 1992; Flanagan and Shearer, 1998]. Because of the lack of coverage of our data towards the further west (Figure 1), we are not able to confirm the localized area with very large depression identified by Li and Yuan [2003]. Due to the low lateral resolution of the SS precursor, it has been argued that the real anomaly might be much smaller in scale [Neele et al., 1997]. The good agreement between our measurements and the SS precursor results suggests that the large-scale and low-amplitude topography of the 660 in this area is probably a true feature, as demonstrated by Shearer et al. [1999]. [13] The 20 km depression of the 660 beneath northeast China is much smaller than those observed in Tonga (70 km) [Niu and Kawakatsu, 1995] and Izu-Bonin (40 km) [Collier and Helffrich, 1997] subduction zones. We noticed that both regions are characterized by steepangle subduction. The deepest conversions usually come from ray paths sampling the coldest slab core. Because of the horizontal deflection, the coldest core of the subducting 3of7
Figure 2. Color contour maps of the stacked traces of all the event-array pairs. Events are sorted by focal depth. Left and right panels show ray paths directing to the west and east, respectively. Hotter color clusters represent greater energy and may indicate a phase arrival. S 660 P arrivals are indicated by red arrows. For comparison, reflections from the 410-km discontinuity, p410p and s410p, predicted by iasp91 model are shown in white crosses. 4of7
Figure 3. (a) Depth cross section showing the conversion points of S 660 P. Solid circles and open squares represent conversion points of ray paths traveling to GRSN and the three US arrays, respectively. Red circle and square mark the occurrence of a sharp transition from a normal to depressed 660. Conversion point pair of the same event is connected by a solid line. To show the undulation of the 660 more clearly, the vertical scale is exaggerated. (b) Depth section of the 3D P-wave velocity model of Fukao et al. [2001]. Blue and red colors indicate fast and slow velocity anomalies. Red stars and white dots represent the eight events and the deep seismicity, respectively. Pacific slab beneath northeast China may either never reach to the 660 or lies further west to our study area. This might explain the small amplitude of the depression observed here. [14] The observed depression beneath northeast China may be converted to a temperature deficit within the subducting slab if we assume that the perturbation to the 660-km discontinuity depth is thermal in origin. The conversion depends on the values of the Clapeyron slope of the phase boundary. Early estimates of the Clapeyron slopes from experimental studies are in the range of 4 to 2 MPa/K [Ito and Takahashi, 1989; Ito et al., 1990; Akaogi and Ito, 1993]. Bina and Helffrich [1994] re-evaluated earlier experimental data and found that the implied value of Clapeyron slope of the 660 is more nearly consistent between 3.0 to 2.0 MPa/K, which also agrees with the late laboratory measurements [Irifune et al., 1998; Shim et al., 2001]. The temperature reduction with the above Clapeyron slope values is 222 333 K for a 20 km depression. Most recent studies, however, found that the Clapeyron slope of the post-spinel phase transition of anhydrous ringwoodite is rather small, between 2 to 0.4 MPa/K [Katsura et al., 2003; Fei et al., 2004; Litasov et al., 2005]. Litasov et al. [2005] suggested that a significant portion of the topography of the 660 is caused by water content rather than temperature. [15] Our observations of the 660 topography can be used to constrain subduction dynamics and rheology of the subducting lithosphere. It is unlikely to explain the observed topography with the simple slab-penetration model (Figure 4a). In this model slabs penetrate the 660 and sink into the lower mantle continuously. Because of the increase of viscosity across the boundary, a significant broadening occurs to the slabs before the penetration. Distance between the deepest event and the seaward depression corner (X in Figure 4) is expected to be larger than the normal slab thickness (Figure 4a). The observed distance beneath northeast China is around 100 km, roughly similar to the lithosphere thickness of the subducting Pacific plate. [16] A recent experiment on the grain growth of ringwoodite [Yamazaki et al., 2005] indicated that ringwoodite in a cold subducting slab has fine grain size and deforms dominantly by diffusion creep. Viscosity of the slab subducting through the transition zone is likely to be 4 6 orders of magnitude lower than that of the surrounding mantle when the temperature effect is not considered. Due to the increased viscosity in the lower mantle, such a soft slab could easily be deformed and pile up above the 660 (Figure 4b). The pileup could lead to a localized broad depression of the 660 that is more likely to be symmetric in both the seaward and landward directions. This local symmetry is, however, not supported by the data. 5of7
4. Conclusions [18] We have analyzed several array data for deep focus earthquakes that occurred beneath northeast China and Japan Sea to constrain the depth variation of the 660 adjacent to the subducting Pacific slab. We found a depression of the 660 which starts from the intersection of the lower boundary of the Pacific slab and the 660 and extends to west with a steady increase of amplitude. This observation is consistent with a scenario for a soft Pacific slab lying over the 660-km discontinuity progressively from west to east due to the eastward retreat of the Japan trench. The interaction between a subducting slab and the upper and lower mantle boundary need to be understood in an integrated dynamic system consisting of relative plate motion, trench migration, rheology of the subducting slab, and nature of the 660-km seismic discontinuity. [19] Acknowledgments. We thank the IRIS and GRSN data centers for providing waveform data. Discussion with S. Grand, H. Kawakatsu, and M. Obayashi were very helpful in preparing the manuscript. Andrea Morelli provided very constructive review that improved the manuscript significantly. This work is supported by the NSFC under grants 40404004, 40674020, and NSF EAR-063566 and an international travel grant from Rice University. Figure 4. A schematic figure showing three different scenarios when the subducting Pacific slab encounters the 660-km phase boundary: (a) continuous penetration, (b) local symmetric pileup, and (c) asymmetric deflection when significant trench retreat is present. Dashed lines indicate the expected topography of the 660. Stars and crosses indicate deep earthquakes and the seaward corner of the depression, respectively. [17] The symmetric pileup, however, can be avoided when significant trench retreat is present. In this case, one would expect the soft Pacific slab lying over the 660 progressively from west to east, resulting in an asymmetric deflection as observed in the data. This scenario was well demonstrated by numerical simulations [Christensen, 1996; Tagawa et al., 2007]. Miller et al. [2006] found that position of the trench under Japan-Kurile arc has been retreating from Miocene, resulting in a curved shape of the trench along the north central Japanese islands. References Akaogi, M., and E. Ito (1993), Refinement of enthalpy measurement of MgSiO 3 perovskite and negative pressure-temperature slopes for perovskite-forming reactions, Geophys. Res. Lett., 20, 1839 1842. Bina, C. R., and G. Helffrich (1994), Phase transition Clapeyron slopes and transition zone seismic discontinuity topography, J. Geophys. Res., 99, 15,853 15,860. Christensen, U. (1996), The influence of trench migration on slab penetration into the lower mantle, Earth Planet. Sci. Lett., 140, 27 39. Collier, J., and G. Helffrich (1997), Topography of the 410 and 660 km seismic discontinuities in the Izu-Bonin subduction zone, Geophys. Res. Lett., 24, 1535 1538. Dziewonski, A. D., and D. L. Anderson (1981), Preliminary reference Earth model, Phys. Earth. Planet. Inter., 25, 297 356. Fei, Y., J. Van Orman, J. Li, W. van Westrenen, C. Sanloup, W. Minarik, K. Hirose, T. Komabayashi, M. Walter, and K. Funakoshi (2004), Experimentally determined postspinel transformation boundary in Mg 2 SiO 4 using MgO as an internal pressure standard and its geophysical implications, J. Geophys. Res., 109, B02305, doi:10.1029/2003jb002562. Flanagan, M., and P. Shearer (1998), Global mapping of topography on transition zone velocity discontinuities by stacking SS precursors, J. Geophys. Res., 103, 2673 2692. Fukao, Y., S. Widiyantoro, and M. Obayashi (2001), Stagnant slabs in the upper and lower mantle transition region, Rev. Geophys., 39, 291 323. Grand, S. P. (2002), Mantle shear-wave tomography and the fate of subducted slabs, Philos. Trans. R. Soc. London, Ser. A, 360, 2475 2491. Gudmundsson, O., and M. Sambridge (1998), A regionalized upper mantle (RUM) seismic model, J. Geophys. Res., 103, 7121 7136. Irifune, T., et al. (1998), The postspinel phase boundary in Mg 2 SiO 4 determined by in situ X-ray diffraction, Science, 279, 1693 1700. Ito, E., and E. Takahashi (1989), Postspinel transformations in the system Mg 2 SiO 4 -Fe 2 SiO 4 and some geophysical implications, J. Geophys. Res., 94, 10,637 10,646. Ito, E., M. Akaogi, L. Topor, and A. Navrotsky (1990), Negative pressuretemperature slopes for reactions forming MgSiO 3 perovskite from calorimetry, Science, 249, 1275 1278. Karato, S., and B. Karki (2001), Origin of lateral variation of seismic wave velocities and density in the deep mantle, J. Geophys. Res., 106, 21,771 21,784. Katsura, T., et al. (2003), Post-spinel transition in Mg 2 SiO 4 determined by high P-T in situ X-ray diffraction, Phys. Earth Planet. Inter., 136, 11 24. Kennett, B. L. N., and E. R. Engdahl (1991), Travel times for global earthquake location and phase identification, Geophys. J. Int., 105, 429 465. Li, X., and X. Yuan (2003), Receiver functions in northeast China: Implications for slab penetration into the lower mantle in northwest Pacific subduction zone, Earth Planet. Sci. Lett., 261, 679 691. Litasov, K., E. Ohtani, A. Sano, A. Suzuki, and K. Funakoshi (2005), Wet subduction versus cold subduction, Geophys. Res. Lett., 32, L13312, doi:10.1029/2005gl022921. Miller, M., B. L. N. Kennett, and V. Toy (2006), Spatial and temporal evolution of the subducting Pacific plate structure along the western Pacific margin, J. Geophys. Res., 111, B02401, doi:10.1029/2005jb003705. Muirhead, K. (1968), Eliminating false alarms when detecting seismic events automatically, Nature, 217, 533 534. Neele, F., H. De Regt, and J. VanDecar (1997), Gross errors in uppermantle discontinuity topography from underside reflection data, Geophys. J. Int., 129, 194 204. Niu, F., and H. Kawakatsu (1995), Direct evidence for the undulation of the 660-km discontinuity beneath Tonga: Comparison of Japan and California array data, Geophys. Res. Lett., 22, 531 534. 6of7
Shearer, P., and T. Masters (1992), Global mapping of topography on the 660-km discontinuity, Nature, 355, 791 796. Shearer, P., M. Flanagan, and M. Hedlin (1999), Experiments in migration processing of SS precursor data to image upper mantle discontinuity structure, J. Geophys. Res., 104, 7229 7242. Shim, S.-H., T. S. Duffy, and G.-Y. Shen (2001), The post-spinel transformation in Mg 2 SiO 4 and its relation to the 660 km seismic discontinuity, Nature, 411, 571 574. Stern, R. J. (2002), Subduction zones, Rev. Geophys., 40(4), 1012, doi:10.1029/2001rg000108. Tagawa, M., T. Nakakuki, and F. Tajima (2007), Dynamical modeling of trench retreat driven by the slab interaction with the mantle transition zone, Earth Planet Space, 59, 65 74. van der Hilst, R. D., S. Widiyantoro, and E. R. Engdahl (1997), Evidence for deep mantle circulation from global tomography, Nature, 386, 578 584. Vidale, J., and H. Benz (1992), Upper-mantle seismic discontinuities and the thermal structure of subduction zones, Nature, 356, 678 680. Wicks, C., and M. Richards (1993), A detailed map of the 660-kilometer discontinuity beneath the Izu-Bonin subduction zone, Science, 261, 1424 1427. Yamazaki, D., T. Inoue, M. Okamoto, and T. Irifune (2005), Grain growth kinetics of ringwoodite and its implication for rheology of the subducting slab, Earth Planet. Sci. Lett., 236, 871 881. Q.-F. Chen, Institute of Earthquake Science, China Earthquake Administration, Beijing 100036, China. J. Li, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, 100029, China. (juanli@mail.iggcas.ac.cn) F. Niu and E. Vanacore, Department of Earth Science, Rice University, Houston, TX 77005, USA. 7of7