Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 Receiver function analyses are performed to detect seismic velocity discontinuities in the uppermost mantle beneath the Kyushu subduction zone, Japan. The Philippine Sea slab subducting beneath Kyushu is young (26–50 Ma) and steeply dipping (at greater than 30°). We detect a seismic velocity contrast larger than 10% corresponding to the oceanic Moho down to 90 km in depth, implying that the subducting oceanic crust contains more than 3.0 wt.% water down to this depth. We also detect a discontinuity with downward decreasing seismic velocity at depths of 50–80 km, which is parallel to the oceanic Moho and 10 km shallower than it. This fact indicates that there is a sharp discontinuity between the mantle wedge and the hydrous oceanic crust. The existence of such a sharp discontinuity would require a large temperature gradient around the boundary or a permeability barrier at the upper boundary of the slab. We delineate the continental Moho with downward decreasing seismic velocity and the upper boundary of the slab with upward decreasing seismic velocity beneath the forearc region, which implies the existence of serpentinite and/or free fluid which causes high pore pressure in the forearc mantle.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Kyushu island is located on the eastern margin of the Eurasian plate, and the Philippine Sea (PHS) plate is west-northwesterly subducting at a rate of 5 cm/year beneath Kyushu (Figure 1) [Seno et al., 1993]. The distribution of intermediate-depth earthquakes shows that the PHS slab subducts down to 70–90 km in depth at dip angles of 30–40°, and extends from this depth down to 150–250 km at dip angles of greater than 65°. The subducting PHS slab produces arc volcanism, and the volcanic front transects in a NNE-SSW direction on the island (Figure 1) [e.g., Miyoshi et al., 2008]. The PHS slab subducting beneath Kyushu is separated into two parts by a remnant island arc, the Kyushu-Palau ridge, and the ages of the north and south portions are estimated to be 26 Ma and 50 Ma, respectively [Hilde and Lee, 1984; Okino et al., 1994]. Tahara et al. detected the oceanic crust of the PHS slab having a heterogeneous velocity structure corresponding to the Kyushu-Palau ridge near the east coast of Kyushu (Figure 1).
 In order to understand the mechanism of water transportation through a young and steeply dipping slab, it is important to identify where hydrated subducting oceanic crust and hydrated mantle exist beneath Kyushu. Recently, stable temperature and pressure conditions of rock facies of the hydrated oceanic crust and mantle have been revealed in detail [e.g., Hacker et al., 2003a]. It is known that hydrated mantle material is rheologically weak with a low density [Hyndman and Peacock, 2003]. Therefore, it is also useful to detect the hydrated portion for constraining the temperature structure and convection regime of the mantle in the subduction zone.
 In some subduction zones in the world, the geometry of the upper and lower boundaries of the subducting oceanic crust is estimated by detecting and mapping seismic velocity discontinuities [e.g., Yuan et al., 2000; Bostock et al., 2002; Kawakatsu and Watada, 2007; Audet et al., 2009]. Hydrous and anhydrous portions of the oceanic crust and the mantle wedge were obtained from examining the discontinuities in the uppermost mantle.
 However, the geometry of the discontinuities in the uppermost mantle has not been well examined for Kyushu, Japan. In this study, we estimate the geometry with receiver function (RF) analyses and detect where fluid is contained in the oceanic crust and in the mantle wedge.
 We use 28,732 waveforms from 586 teleseismic events whose epicentral distances are 30–90°, and magnitudes are greater than 5.5. The waveform data were obtained from 78 stations of Hi-net established by the National Research Institute for Earth Science and Disaster Prevention (NIED) [Obara et al., 2005], and from 61 stations of the J-array established by the Japan Meteorological Agency (JMA), Kyushu University, Kagoshima University and Kyoto University [Morita, 1996] (Figure 1). We use waveform data from Hi-net and the J-array, recorded from June 2001 to May 2010 and from August 1996 to October 2009, respectively.
 A RF is calculated by deconvolving the vertical component of a waveform of a teleseismic P-wave from its horizontal component, and phases converted at seismic velocity discontinuities in its coda can be detected with RFs [Langston, 1979]. We calculate RFs with the extended-time multitaper method [Helffrich, 2006] which was improved by Shibutani et al. , with a 0.56 Hz low-pass Gaussian filter.
 In order to estimate the geometry of steeply dipping discontinuities with RFs, refraction of the seismic rays at the dipping discontinuities should be taken into account. For this purpose, we apply a method for stacking RFs using the multi-stage fast marching method developed byAbe et al. . This stacking method enables estimation of the interface geometry of discontinuities dipping at more than 70°. First, we stack transverse component RFs whose backazimuths are 118–178°, using a 1-d velocity distribution and 2 horizontal velocity discontinuity interfaces (Conrad, 20 km; Moho, 35 km in depth), based on the ak135 model [Kennett et al., 1995], and no dipping interfaces. Then, we iteratively stack them with a model containing an additional dipping interface estimated from the former iteration. In the iteration process, we include the dipping interface (the oceanic Moho) that is detected most clearly and assume that the other dipping interfaces are parallel to it. We repeat this process until we obtain RF sections whose peaks coincide with the geometry of the assumed interfaces. After four iterations, such sections are obtained, and we conclude that the interface geometry is correctly estimated. We finally rotate radial and transverse component RFs into another horizontal component which is parallel to the horizontal component of the vibration direction of P-to-S converted wave estimated from the assumed dip angle and dip azimuth of the interface. This method, which is called vectorial RF imaging introduced byKawakatsu and Yoshioka , enhances the RF peaks corresponding to both the dipping and horizontal discontinuities. We apply this method to the RFs whose backazimuths are 0–360° and stack them.
 Although we obtain several RF sections in our analyses, we show and discuss just one cross section beneath the black box in Figure 1, in which a discontinuity interpreted as the oceanic Moho is detected more clearly and down to a greater depth than in any of the others. RF peaks stacked beneath the black box are projected on a cross section along X-X′ (Figure 2). The RF section consists of 2 km by 2 km cells and each cell shows the amplitude of RFs that is projected on the cell. When two or more RFs are projected on the same cell, the amplitudes are averaged. Discontinuities that have upward (downward) decreasing seismic velocities are detected by positive (negative) peaks of RFs, and shown by warm (cool) colors. In Figure 2a, dots indicating hypocenters in the region beneath the black box shown in Figure 1 are superimposed on the RF section. To enhance cells with larger amplitudes, cells with absolute amplitudes less than 0.05 are represented by a shade of grey, and those with absolute amplitudes larger than 0.07 are represented by a pure color in Figure 2c. Using generalized ray theory [Helmberger, 1974], we have confirmed that shear waves which are converted at a discontinuity dipping at 30–50° with 10% of S-wave velocity contrast make RF peaks with amplitudes of 0.06 ± 0.01. InFigure 2c, thus, discontinuities dipping at 30–50° indicated by chromatic cells have an S-wave velocity contrast greater than 10%.
 Positive RF peaks corresponding to the continental Moho appear at depths of 30–40 km, and distances of 70–200 km from the point X′ (Figure 2a). These depths are comparable to those observed in a previous study with the travel times of local earthquakes by Oda and Ushio . The polarity of RF peaks corresponding to the continental Moho is inverted at distances of 50–60 km (Figure 2a).
 We obtain positive peaks corresponding to the oceanic Moho of the PHS slab which extends down to 90 km in depth (Figure 2a), although their amplitudes are small at distances of 40–60 km (Figure 2c). Negative peaks corresponding to the upper boundary of the PHS slab are also found about 10 km shallower than the oceanic Moho, at depths of 50–80 km (Figure 2a), although the boundary is detected by positive RF peaks at depths of 40–50 km. At depths of 50–90 km, the hypocenters of most of the intermediate-depth earthquakes are distributed above the detected oceanic Moho (Figure 2a). At depths of 60–70 km, the absolute amplitudes of the RF peaks corresponding to the oceanic Moho and the upper boundary of the PHS slab become larger than those in the surrounding region, which is coincident with an increase in intermediate-depth seismicity (Figures 2a and 2c).
 We detect the oceanic Moho with downward increasing S-wave velocity, and the oceanic crust is estimated to have 10% lower S-wave velocity than the slab mantle down to 90 km in depth. According toHacker et al. [2003a], the subducting oceanic crust of the PHS slab would maintain an S-wave velocity reduction of at least 10% from the surrounding anhydrous mantle until the stable oceanic crustal rock facies changes from lawsonite + amphibole + eclogite (3.0 wt.% H2O) to amphibole + eclogite (0.6 wt.% H2O) or to zoisite + eclogite (0.3 wt.% H2O). At 90 km depth, therefore, this type of phase transition is interpreted to be complete. Hacker et al. [2003b] and Yamasaki and Seno indicated that dehydration embrittlement is responsible for the occurrence of intermediate-depth earthquakes. Based on the hypothesis, seismic activity in the oceanic crust (Figure 2a) would imply the occurrence of dehydration reactions in the low velocity oceanic crust.
 In our RF section, at distances of 50–60 km, the occurrence of negative RF peaks corresponding to the continental Moho implies that the mantle has lower velocity than the continental crust (Figure 2a). This unusual occurrence of “inverted Moho” has been also detected beneath Cascadia and northern Chile, and regarded as representing the existence of highly serpentinized mantle [Bostock et al., 2002; Sodoudi et al., 2011]. As shown in Figure 2c, the minimum value of the amplitudes of RF peaks corresponding to the inverted Moho is from −0.17 to −0.10, and this fact implies that the mantle has 20–30% lower S-wave velocity than the continental lower crust. When S-wave velocity of the continental lower crust regarded as 3.85 km/s based on the ak135 model [Kennett et al., 1995], the S-wave velocity of the mantle is estimated to be less than 2.7–3.1 km/s. Considering measurements byChristensen , lizardite-chrysotile serpentinites whose S-wave velocity is 2.3 km/s should exist beneath the inverted Moho rather than antigorite serpentinites whose S-wave velocity is 3.6 km/s. If lizardite-chrysotile serpentinites are the sole cause to reduce velocity, the degree of serpentinization is estimated to be 60–80%. However, while stable temperature of lizardite-chrysotile serpentinites is lower than 300°C [Christensen, 2004], the forearc mantle would not have such a low temperature according to the results of temperature calculations by Yoshioka et al. . High pore pressure significantly reduces seismic velocities of rocks according to measurements by Christensen , and might be one of the causes to reduce velocity of the forearc mantle.
 At depths of 50–80 km, negative RF peaks, which can be regarded as the upper boundary of the PHS slab, are detected about 10 km shallower than the oceanic Moho (Figure 2a), and the mantle above it is expected to have little or no water content. We roughly estimate the S-wave velocity of the oceanic crust from the amplitudes of observed RF peaks corresponding to the oceanic Moho, and confirm that the oceanic crust can be estimated 3–4 km thicker at a maximum. Although the negative peaks are larger than −0.06 (Figure 2c), they are clearly seen in Figures 2a and 2b, and the portion below the discontinuity is estimated to have a little less than 10% lower S-wave velocity than the portion above it. The upper portion of the oceanic crust is expected to have a higher temperature than the lower portion [e.g.,Hacker et al., 2003b; Yoshioka et al., 2008]. Therefore, the upper portion would dehydrate earlier, and seismic velocity contrast of the upper boundary is favorable to become smaller. Rock facies of the hydrated oceanic crust, jadeite + lawsonite + blueschist or lawsonite + amphibole + eclogite, are stable up to 420°C or 480°C, respectively, and serpentinized peridotite is stable up to 620°C [Hacker et al., 2003a]. Therefore, a temperature transition from 420–480°C to 620°C would exist around the discontinuity, and serpentinized mantle and/or the anhydrous oceanic crust may exist between the anhydrous mantle and the hydrous oceanic crust. We estimate that the thickness of this transition zone is about 4 km with forward modeling of RFs. From the estimation, a large temperature gradient (35–50°C/km) that is subperpendicular to the PHS slab is expected to exist around the boundary. We can also expect such a large temperature gradient from the results of temperature calculations by Yoshioka et al. . However, Abers indicated another possible cause for existence of the sharp discontinuity at the upper boundary of a slab that, in steeply dipping slabs, fluid dehydrated from the oceanic crust may travel up-dip through permeability fabric oriented parallel to the slab surface. If fluid does not flow vertically to the mantle wedge, it will not become serpentinized and a sharp velocity contrast would exist at the upper boundary of the slab, regardless of temperature gradient. Therefore, a large temperature gradient and/or permeability barrier on the upper boundary of the PHS slab are expected to be the causes of the sharp discontinuity at the upper boundary of the slab.
 At depths of 60–70 km, seismic activity in the low velocity layer is higher than that in shallower and deeper portions (Figure 2a). At these depths, both negative and positive peaks corresponding to the upper and lower boundaries of the oceanic crust become larger, and the S-wave velocity of the oceanic crust is expected to be lower than that in the surrounding portion of the oceanic crust. This decrease in S-wave velocity, coincident with an increase in seismic activity, may reflect the abundance of free fluid from dehydration reactions. High heat flow from the mantle wedge to the PHS slab would cause a rapid temperature increase in the subducting oceanic crust, and a large amount of fluid is expected to be generated by dehydration reactions. Alternatively, fluid dehydrated from a deeper portion of the oceanic crust might flow up-dip in the slab and concentrate there.
 We show the geometry of detected discontinuities and interpretations of them in relation to water transportation in Figure 3. The subducting hydrated oceanic crust of the PHS slab whose water content is more than 3.0 wt.% exists beneath Kyushu, and it causes a velocity contrast at the oceanic Moho. While the oceanic crust is subducting, dehydration reactions would gradually occur within it with increasing pressure and temperature, and cause the seismic activity in the oceanic crust. Almost all fluid in the oceanic crust would have been lost at a depth of 90 km, which is coincident with the disappearance of velocity contrasts at the upper and lower boundaries of the oceanic crust. Dehydrated fluid from the oceanic crust would flow to the mantle wedge. Beneath the forearc, some of the fluids would hydrate the mantle material, others may exist in pores without hydrating it. Both of them may decrease S-wave velocity of the forearc mantle to 2.7–3.1 km/s and cause a reverse in the polarity of the velocity contrast at the continental Moho and the upper boundary of the slab. Unless a permeability barrier exists at the upper boundary of the slab, fluid dehydrated from the oceanic crust would flow vertically. In this case, a large temperature gradient (35–50°C/km) that is subperpendicular to the slab may exist around the upper boundary of the slab deeper than 50 km. Abundant free fluid may exist in the oceanic crust at depths of 60–70 km, and cause a decrease in velocity in the high seismicity portion.
 We are grateful to Martha Savage, two anonymous reviewers, and the editor Michael Wysession for many suggestions which are very important to improve this article. We use seismic data observed by NIED, JMA, Kyushu University and Kagoshima University. We also use hypocentral data of JMA. We use the multi-stage 3-D fast marching code [de Kool et al., 2006]. We use GMT (Generic Mapping Tools) by Wessel and Smith to generate figures. This research has been supported by a Grant-in-Aid for Science Research (B) (20340119) from MEXT.
 The Editor thanks Martha Savage and two anonymous reviewers for their assistance in evaluating this paper.