Compressional (Vp) and shear (Vs) wave velocities of mid-ocean-ridge basalt (MORB) was investigated atin situ high pressure and high temperature conditions of the mantle transition region by using a combination of ultrasonic and in situx-ray measurements. Both Vp and Vs of MORB are lower than the previously predicted velocities of the major mantle mineral phases. We found that the Vp and Vs of MORB along a typical geotherm are lower by about 2 and 5%, respectively than those of seismological models, and do not match any global and regional seismological models in the deeper parts of the mantle transition region. Thus, the existence of a basaltic layer in this region is unlikely, suggesting that the oceanic crust materials are transported into the lower mantle.
 Recent laboratory measurements of sound velocities (P-wave velocity, Vp and S-wave velocity, Vs) of high-pressure minerals have demonstrated that the upper to middle part of the MTR consist mainly of pyrolite [Irifune et al., 2008]. However, the Vs of pyrolite is significantly lower than those of the seismological models [Dziewonski and Anderson, 1981; Kennett et al., 1995] in the bottom part of the MTR, suggesting the presence of other lithologies in this region. Subducted oceanic crust is one of the potential candidates for such materials as stated above.
 Mid-ocean-ridge basalt (MORB) has a representative chemical composition of oceanic crust, which constitutes the upper part of subducting slabs. High-pressure and high-temperature experiments demonstrate that MORB transforms to eclogite (composed of clinopyroxene + garnet) in the shallow upper mantle, and to garnetite (majorite garnet + stishovite) under the P-T conditions near the 410 km seismic discontinuity [e.g.,Irifune et al., 1986]. Majorite garnet (hereafter majorite) partially transforms to CaSiO3-perovskite (Ca-Pv), and garnetite eventually transforms to perovskitite (Ca-Pv and MgSiO3-perovskite + stishovite + Al-rich phases) in the upper part of the lower mantle (at depths between 660 and 2890 km) [e.g.,Irifune and Ringwood, 1993; Hirose et al., 1999; Perrillat et al., 2006]. Here we investigated the Vp and Vs of garnetite and Ca-Pv bearing garnetite with a MORB composition atin situ high pressure and high temperature conditions corresponding to those of the mantle transition region, and examined the possibility of the existence of MORB in this region.
2. Sample Descriptions and Experimental Methods
 Table S1 in Text S1 of the auxiliary material shows major chemical compositions of the MORB sample used in this study, which is identical to those of earlier phase relation studies [Irifune et al., 1986; Irifune and Ringwood, 1993]. We used a sintered body of garnetite produced from a synthetic MORB glass for the simultaneous ultrasonic and in situ X-ray measurements at high pressure and temperature conditions within the stability field of garnetite. The sample recovered from the run was found to maintain the same phase assemblage of majorite + stishovite with grain sizes in a range of 0.5–1.0μm in diameter (Figures 1a and 1b). For higher pressure experiments where Ca-Pv forms as a stable phase, we used a MORB glass rod as the starting material, which was converted to the stable phase assemblages at about 21 and 24 GPa and 1800 K prior to the ultrasonic measurements. This is because preparing well sintered bodies of Ca-Pv bearing garnetite under such conditions is difficult due to the amorphization of Ca-Pv upon pressure release. X-ray diffraction profiles and a TEM image of one of these samples recovered after the ultrasonic measurements are shown inFigures 1a and 1c.
 Simultaneous ultrasonic measurements, X-ray radiography observations, and diffraction experiments were carried out using a multi-anvil apparatus at the BL04B1 beamline, SPring-8. We used cell assemblies with basically the same design as those adopted in our earlier studies [Irifune et al., 2008; Higo et al., 2009; Kono et al., 2010]. Tungsten carbide anvils with a truncated edge length (TEL) of 7 mm were used for the experiments up to ∼18 GPa, while the anvils with TEL = 5 mm were used for higher pressure runs. MgO and NaCl sleeves were used as the sample container to realize a quasi-hydrostatic environment. In order to further reduce the effects of deviatoric stress, all the ultrasonic measurements were performed only during the process of decreasing temperature after anealing at high pressure. Temperature was monitored with a W3%Re-W25%Re thermocouple, and pressure was determined from the unit-cell volume of Au in a mixture of NaCl + Au + BN, which was placed adjacent to the sample, using an equation of state [Tsuchiya, 2003]. Energy-dispersive X-ray diffraction measurements of sample and Au pressure standard were carried out at a fixed diffraction angle of 5°. Ultrasonic Vp and Vs measurements were conducted using the pulse reflection method, and the sample length was determined by X-ray radiography technique. The overall uncertainty in the present Vp and Vs determination is less than ±0.5%. Further technical details of the combined ultrasonic and in situ X-ray measurements are described in our previous studies [Higo et al., 2009; Kono et al., 2010].
Figures 2ashows the P-T paths of the present sound velocity measurements on the MORB samples, together with the phase transition boundaries. A single run was made within the stability field of majorite + stishovite by changing pressure and temperature up to 18.5 GPa and 1800 K, while other two runs were conducted in the fields of majorite + stishovite + Ca-Pv ± an Al-rich phase. These phase assemblages are consistent with the results of earlier experimental studies [Irifune et al., 1986; Irifune and Ringwood, 1993]. No changes in the phase assemblages were observed by in situ X-ray diffraction during the ultrasonic measurements performed while decreasing temperature along the P-T paths.
Figure 2b shows the variations of the sound velocities of MORB as functions of pressure and temperature (original data: Table S2 in Text S1). Both Vp and Vs increase with increasing pressure, while decrease with temperature. The temperature effect seems to become greater at higher temperatures for the majorite + stishovite region, which is consistent with the earlier results on majorite with the pyrolite minus olivine composition [Irifune et al., 2008]. Two-dimensional linear fittings for the observed Vp and Vs were made, where we used the data sets only at temperatures of 1500–1800 K because they are more closely compared to the realistic geotherm in this depth of the mantle. We separately fitted the data sets in the two regions with and without Ca-Pv, and obtained Vp = 9.44(12) + 5.7(6) × 10−2× P-5.7(7) × 10−4× (T-300) and Vs = 5.27(12) + 4.1(6) × 10−2× P-5.8(7) × 10−4× (T-300) for garnetite without Ca-Pv, and Vp = 9.44(6) + 5.2(2) × 10−2× P-4.1(3) × 10−4× (T-300) and Vs = 5.26(5) + 2.8(2) × 10−2× P-4.0(3) × 10−4× (T-300) for Ca-Pv bearing garnetite (P, pressure in GPa; T, temperature in K). The velocity changes along a typical geotherm [Brown and Shankland, 1981] based on these equations, yielding a small (∼1.5%) Vp increase around 20 GPa where Ca-Pv starts to form, while Vs shows no notable changes and exhibits a smaller pressure dependency at the higher pressures.
 Some earlier studies predicted that the formation of Ca-Pv from majorite at the lower part of the MTR would cause significant increases in Vp and Vs, due to the plausible high velocities of Ca-Pv [Karki and Crain, 1998]. However, our direct measurements of the sound velocities of MORB demonstrated only very small increases in Vp and Vs through the formation of Ca-Pv. To understand the reason for the subtle changes in the sound velocities, we quantitatively evaluated the effect of the Ca-Pv formation on the velocities of garnetite at ambient conditions using available data on the end-member minerals (Table S3 inText S1) [Weidner et al., 1982; O'Neill et al., 1989; Bass et al., 1989; Karki and Crain, 1998; Sinogeikin and Bass, 2002; Tsuchiya, 2011], where the mineral proportions and chemical compositions of the garnetite with and without Ca-Pv were derived from previous high-pressure phase transition studies [Irifune et al., 1986; Irifune and Ringwood, 1993].
 The results show that the velocities of majorite (Vp0 = 9.06 km/s and Vs0 = 5.17 km/s) decrease to Vp0 = 8.98 km/s and Vs0= 5.08 km/s after the formation of Ca-Pv. This is mainly due to the decrease of the relative proportion of the grossular (Ca3Al2Si3O12) component in majorite, which has significantly higher sound velocities than those of other end-member garnets [Bass, 1989]. Ca-Pv is predicted to have higher sound velocities than those of garnets based onab initio calculations [Karki and Crain, 1998; Tsuchiya, 2011]. We calculated the velocity changes in garnetite through the formation of Ca-Pv by using two differentab initiopredictions of the sound velocities of Ca-Pv with taking into account the observed velocity changes in majorite. The results show that the velocities of garnetite change from Vp0 = 9.34 km/s and Vs0 = 5.37 km/s to Vp0 = 9.49 km/s and Vs0= 5.47 km/s, respectively, when the higher velocity values for Ca-Pv [Karki and Crain, 1998] are used. In contrast, the velocities of Ca-Pv bearing garnetite are calculated as Vp0 = 9.45 km/s and Vs0= 5.40 km/s when the lower velocities for Ca-Pv are adopted [Tsuchiya, 2011]. The present experimental results on the velocity changes upon the Ca-Pv formation in garnetite are consistent with the latter case, where only a small increase in Vp and no notable increase in Vs are observed, suggesting that the Ca-Pv may have lower sound velocities than those previously estimated [Karki and Crain, 1998]. This result is consistent with an ultrasonic measurement study of Ca-Pv at ∼8–12 GPa and room temperature [Li et al., 2004] and a Brillouin scattering measurement of Ca-Pv at 32–133 GPa and room temperature [Kudo et al., 2012], in which experimentally observed Vp and/or Vs of Ca-Pv are markedly lower than those ofKarki and Crain , but is closer to those of Tsuchiya . On the other hand, recent ab initio calculations suggest that the Vs of stishovite would stay constant at pressures above ∼20 GPa and even decrease at higher pressures due to the shear softening prior to the ferroelastic transition of stishovite to the CaCl2 structure near 60 GPa [Tsuchiya, 2011], which may also contribute to the absence of observable increase of Vs in Ca-Pv bearing garnetite. The phase transition pressure and elastic properties of stishovite may be influenced by the incorporation of aluminum and/or water [e.g.Lakshtanov et al., 2007], although those effects have not yet been well understood. However, the amount of stishovite in MORB is only 10 vol.% at 25 GPa [Irifune and Ringwood, 1993], and therefore it probably does not significantly influence the Vs change. Since garnet is the most abundant mineral phase at 25 GPa (70 vol.% [cf. Irifune and Ringwood, 1993]), the decrease in relative proportion of the grossular component would rather play an important role to interpret the absence of Vs change in Ca-Pv bearing garnetite.
Figure 3 shows a comparison of the Vp and Vs of MORB depicted in Figure 2b with those of pyrolite and representative seismological models [Dziewonski and Anderson, 1981; Kennett et al., 1995]. Both Vp and Vs of MORB are substantially lower (about 2 and 5%, respectively) than those of the reference models in the middle to lower part of the MTR. Our recent study using the same method as adopted in the present measurements demonstrated that pyrolite is a reasonable composition in the upper to middle part of the MTR. However, it also shows that the Vs of pyrolite is substantially lower than those models in the lower part of the MTR, suggesting the presence of a layer with distinct chemical compositions. The present study clearly demonstrates that both Vp and Vs of MORB are further lower than those of pyrolite, showing that the oceanic crust material is unsuitable as the candidate for the material in this region of the MTR. Although a recent regional seismological study [Shen and Blum, 2003] suggested the presence of oceanic crust material just above the 660 km seismic discontinuity beneath south Africa, the Vs of ∼5.4–5.5 km/s reported in this region is approximately 2–3% higher than that of MORB (Vs = 5.3 km/s) and is therefore difficult to reconcile with the present laboratory sound velocity data. Thus, the Vp and Vs of MORB are lower than those of global and regional seismological models in the lower part of the MTR. In addition, the presence of water in subducting slab may cause further deviation between the observed Vp and Vs of MORB and those of seismological models, because it is expected to decrease the Vp and Vs of minerals [e.g. Jacobsen, 2006]. Therefore, our observed data strongly suggest that the existence of a basaltic layer in the lower part of the MTR is unlikely.
 Seismic tomographic imaging demonstrates that many slabs subducted into the MTR are stagnant to form megalith structures or spread over along the 660 km discontinuity, while some others subduct further deep into the lower mantle [van der Hilst et al., 1997; Fukao et al., 2001]. The oceanic crust entrained in the subducted slabs should be delivered to the lower mantle in the latter case. Most of the stagnant slabs (or megaliths) also seem to penetrate partly into the lower mantle, and the oceanic crust material would become denser than the surrounding mantle if they reach the depth of ∼800 km, where the density-crossover ends. Thus, the most of the oceanic crust materials should eventually reach the bottom regions of the lower mantle, while the rest of these materials may reside at depths of ∼660–800 km in the uppermost lower mantle. This is consistent with the recent finding of the “calcium ferrite (CF)” and “new aluminum silicate (NAL)” phases in some natural diamonds of lower mantle origin [Walter et al., 2011], which are shown to be formed under the pressures of upper parts of the lower mantle [Irifune and Ringwood, 1993].
 On the other hand, a recent study suggested that the extremely slow kinetics of eclogite to garnetite transformation may result in the relatively low density of subducted ocean crust materials compared to the surrounding mantle [Nishi et al., 2009], in contrast to the density relation based on the thermodynamic equilibrium model. In this case, subducted oceanic crust may become less dense than the mantle material when it reaches the 410 km discontinuity, where a large density jump of the mantle material occurs due to the olivine to wadsleyite transition, and would be trapped at around this depth. In addition, there is some seismological evidence that shows the presence of a low velocity region at the top of the MTR [Song et al., 2004]. The partial trapping of oceanic crust materials, which consist mainly of untransformed pyroxene and garnet, near the 410 km discontinuity may be another reason for the absence of the oceanic crust material in the MTR.
 We acknowledge two anonymous reviewers for valuable comments and suggestions. This study was conducted based on research proposals (proposal 2007B1648 and 2008B1245 to Y. Kono) to SPring-8 with Grant-in-aid for scientific research from Japanese government (to T. Irifune).
 The Editor thanks the anonymous reviewers for their assistance in evaluating this paper.