This article was published in an Elsevier journal.

The attached copy is furnished to the author for non-commercial research and education use, including for instruction at the authors institution, sharing with colleagues and providing to institution administration. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elseviers archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy

Available online at www.sciencedirect.com

Earth and Planetary Science Letters 267 (2008) 107 117 www.elsevier.com/locate/epsl

Phase transitions in pyrolite and MORB at lowermost mantle conditions: Implications for a MORB-rich pile above the coremantle boundary
Kenji Ohta a , Kei Hirose a,b,, Thorne Lay c , Nagayoshi Sata b , Yasuo Ohishi d
b

Department of Earth and Planetary Sciences, Tokyo Institute of Technology, Meguro, Tokyo 152-8551, Japan Institute for Research on Earth Evolution, Japan Agency for MarineEarth Science and Technology, Yokosuka, Kanagawa 237-0061, Japan c Department of Earth and Planetary Sciences, University of California, Santa Cruz, Santa Cruz, CA 95064, USA d Japan Synchrotron Radiation Research Institute, Sayo, Hyogo 679-5198, Japan Received 31 January 2007; received in revised form 17 October 2007; accepted 19 November 2007 Editor: R.D. van der Hilst

Abstract Subduction of mid-oceanic ridge basalt (MORB) gives rise to strong chemical heterogeneities in the Earth's mantle, possibly extending down to the coremantle boundary. Phase relations in both pyrolite and MORB compositions are precisely determined at high pressures and temperatures corresponding to lowermost mantle conditions. The results demonstrate that the post-perovskite phase transition occurs in pyrolite between 116 and 121 GPa at 2500 K, while post-perovskite and SiO2 phase transitions occur in MORB at 4 GPa lower pressure at the same temperature. Theory predicts that these phase changes in pyrolite and MORB cause shear wave velocity increase and decrease, respectively. Near the northern margin of the large low shear velocity province in the lowermost mantle beneath the Pacific, reflections from a negative shear velocity jump near 2520-km depth are followed by reflections from a positive velocity jump 135 to 155-km deeper. These negative and positive velocity changes are consistent with the expected phase transitions in a dense pile containing a mixture of MORB and pyrolitic material. This may be a direct demonstration of the presence of accumulations of subducted MORB crust in the deep mantle. 2007 Elsevier B.V. All rights reserved.
Keywords: D"; post-perovskite; core-mantle boundary; perovskite; superplume

1. Introduction The 6-km thick basaltic oceanic crust (MORB) is injected into the mantle in significant quantities as oceanic lithosphere sinks at subduction zones. The average oceanic crust production rate during Mesozoic to present time is estimated to be about 25 km3 per year (Reymer and Schubert, 1984). Assuming this production (= subduction) rate has been constant during the last 4 billion years (a much larger production rate has been suggested for the Archean (Komiya, 2004)), the total amount of oceanic crust that has subducted corresponds to at least 11% of

Corresponding author. Department of Earth and Planetary Sciences, Tokyo Institute of Technology, Meguro, Tokyo, 152-8551, Japan. Tel.: +81 3 5734 2618; fax: +81 3 5734 3538. E-mail address: kei@geo.titech.ac.jp (K. Hirose). 0012-821X/$ - see front matter 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2007.11.037

the Earth's mantle in volume. It is thus probable that there are large quantities of MORB residing in the mantle. Subducting MORB crust undergoes phase transitions as pressure and temperature increase with depth, and MORB's distinct chemistry from the balance of the slab and surrounding mantle produces chemical and density heterogeneities over the depth range of subduction, possibly extending down to the core mantle boundary (CMB). The chemical composition of subducted MORB crust is not significantly modified in the Earth's deep interior because the solid-state diffusion rate is very slow especially in the lower mantle (Holzapfel et al., 2005), although the upper part of MORB crust is hydrated before subduction and loses minor amounts of SiO2 and Al2O3 upon dehydration (Kessel et al., 2005). Previous density measurements (Ono et al., 2005; Hirose et al., 2005) and geodynamical simulations (Christensen and Hofmann, 1994; Tackley, 1998; McNamara and Zhong, 2004; Nakagawa and Tackley, 2005) suggest that dense MORB

Author's personal copy

108

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

Fig. 1. (a) S-wave velocity variations in the lowermost 200-km of the mantle in the seismic tomography model of Grand (2002). The large low shear velocity provinces (LLSVP) regions appear to be dense and chemically distinct from surrounding mantle (Lay and Garnero, 2004). (b) Schematic cross-section through an LLSVP, suggesting that a mix of dense, separated MORB and pyrolitic mantle comprises the pile. Phase transitions within the pile give rise to the reflectivity (Fig. 8) shown for the northern edge of the Pacific LLSVP (see small box in (a) corresponds to Fig. 2c).

above the CMB. Subsequent high-pressure experimental studies have shown that the post-perovskite phase transition also occurs in pyrolite and MORB compositions for lowermost mantle conditions (Ono et al., 2005; Hirose et al., 2005; Murakami et al., 2005; Ono and Oganov, 2005). However, the exact locations of the phase transition boundary and its width have not yet been determined for pyrolite and MORB. In addition, a phase transition in SiO2 takes place in MORB at similar high PT conditions (Murakami et al., 2003; Hirose et al., 2005). If the effects of these phase changes in pyrolite and MORB can be related to observed deep mantle seismic velocity discontinuities, there may be seismological indicators of the presence of any MORB accumulations near the base of the mantle. In this study, we examine the deep mantle phase relations in both pyrolite and MORB compositions based on X-ray diffraction measurements in-situ at high PT conditions appropriate for the D region. We also analyze new seismic data to further resolve the seismic velocity discontinuity structure in the lowermost mantle near the northern margin of the Pacific large low shear velocity province (LLSVP) (Fig. 1a) (Avants et al., 2006; Lay et al., 2006). The LLSVPs under the Pacific and Africa appear to be chemically distinct from the surrounding deep mantle, and it has been proposed that they may involve dense debris from ancient oceanic slabs (Fig. 1b) (see Tackley, 1998; McNamara and Zhong, 2004; Lay and Garnero, 2004; Nakagawa and Tackley, 2005). The possible accumulation of subducted MORB crust in a pile in the D region is discussed, based on joint consideration of our mineral physics and seismological findings. 2. Methods

crust may plausibly have accumulated in chemically distinct piles in the D region at the bottom of the mantle. While strong seismic velocity heterogeneities in the lowermost mantle have been detected (see Lay and Garnero, 2004 for a review), direct seismological evidence for the presence of MORB piles has not been presented, in part because the properties of MORB are not well characterized for lowermost mantle conditions and in part because there are alternate interpretations of seismic heterogeneities in the lower mantle. Nishihara (2003) calculated seismic velocities of pyrolite and MORB, showing that they are very similar to each other at lower mantle conditions except around 1600-km depth where the velocities significantly decrease in MORB materials due to ferroelastic-type phase transition in SiO2 phase; however, absolute seismic velocities do not uniquely resolve specific mineralogical structure in the deep mantle because there are strong trade-offs with temperature. The non-uniqueness of interpretation of seismic structure can be reduced if phase transitions in the medium are detected and can be compared to the expectations for a given mineralogy. MgSiO3-rich perovskite is expected to be the most abundant mineral in both pyrolite and MORB materials under lower mantle conditions down to great depths (e.g., Kesson et al., 1994, 1998). However, the perovskite to post-perovskite phase transition recently discovered in MgSiO3 (e.g., Murakami et al., 2004; Tsuchiya et al., 2004; Oganov and Ono, 2004) is expected for pressuretemperature (PT) conditions a few hundred kilometers

2.1. High-pressure experiments High PT conditions were generated in a laser-heated diamondanvil cell (LHDAC). Starting materials were prepared as amorphous gels with chemical compositions of natural KLB-1 peridotite (Takahashi, 1986), which is similar in composition to pyrolite, and normal MORB (Hirose et al., 1999) (Table 1). The starting materials, sample configurations, and high-pressure experimental techniques are the same as those used in our previous studies (Hirose et al., 2005; Murakami et al., 2005). The samples were mixed with fine gold powder and loaded into a hole in a rhenium gasket together with thermal insulation layers of NaCl or
Table 1 Compositions of starting materials KLB-1 SiO2 TiO2 Al2O3 FeO a MgO CaO Na2O K2O Total
a

MORB 49.80 1.65 14.93 11.47 8.54 10.58 2.91 0.12 100.00

44.80 0.16 3.62 8.16 39.50 3.46 0.30 100.00

Total Fe as FeO.

Author's personal copy

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117 Table 2 Experimental conditions and results Run no. Pressure a (GPa) KLB-1 #1 108.4 110.3 113.5 111.1 112.4 116.3 97.6 106.9 122.1 128.5 131.9 1.3 0.9 1.5 1.0 1.1 0.7 0.3 1.9 0.2 1.9 2.8 2.1 2.5 3.4 2.2 2.5 2.5 2.9 2.6 3.5 3.1 3.4 Error1 b Error2 c Volume (Au) ( ) 51.907 51.929 52.263 51.702 51.747 51.426 53.433 52.334 51.561 50.790 50.724 0.106 0.073 0.121 0.083 0.088 0.055 0.025 0.165 0.014 0.137 0.202
3

109

Error

Temperature d (K) 1780 1960 2540 1800 1950 1940 2300 2070 2550 2250 2450

Assemblage e

#2 #3 #4 #5 #6 #7 #8 MORB #9 #10 #11 #12 #13 #14 #15

MgPv + Mw + CaPv MgPv + Mw + CaPv MgPv + Mw + CaPv MgPv + Mw + CaPv + PPv(trace) MgPv + Mw + CaPv + PPv(trace) PPv + MgPv + Mw + CaPv MgPv + Mw + CaPv MgPv + Mw + CaPv PPv + Mw + CaPv PPv + Mw + CaPv PPv + Mw + CaPv

109.5 112.0 114.1 117.5 88.4 103.8 117.4 129.2 132.0

0.5 0.6 1.2 0.7 0.0 1.2 1.3 0.1 1.2

2.7 2.6 2.9 3.2 2.4 2.6 3.0 3.0 3.2

52.142 51.882 51.883 51.779 53.968 52.598 51.630 50.664 50.564

0.045 0.049 0.097 0.059 0.002 0.105 0.100 0.005 0.085

2100 2050 2220 2390 1980 2060 2240 2170 2280

MgPv + CaCl2-type SiO2 + CaPv + CF + PPv(trace) MgPv + CaCl2-type SiO2 + CaPv + CF + PPv PPv + MgPv + -PbO2-type SiO2 + CaPv + CF PPv + MgPv + -PbO2-type SiO2 + CaPv + CF MgPv + CaCl2-type SiO2 + CaPv + CF MgPv + CaCl2-type SiO2 + CaPv + CF PPv + -PbO2-type SiO2 + CaPv + CF PPv + -PbO2-type SiO2 + CaPv + CF PPv + -PbO2-type SiO2 + CaPv + CF

Runs #48 and #1315 were reported by Murakami et al. (2005) and Hirose et al. (2005), respectively. a Pressures were calculated from PVT EOS of Au (Hirose et al., submitted for publication) that is consistent with MgO pressure scale (Speziale et al., 2001) This new pressure scale is represented by a third-order BirchMurnaghan EOS: 3 P KT;0 2 " " 7  5 # ( 2 #) VT;0 3 VT;0 3 3 VT ;0 3 1 1 4 K V 4 V V V

where isothermal bulk modulus K300,0 = 167 GPa at 1 bar and 300 K, and its pressure derivative K' = 5.58 and temperature dependence dK/dT = 0.028 GPa/K. The volume at ambient condition V300,0 = 67.85 3, and those at high temperatures Z T aT;0 dT ; where aT;0 3:179 105 1:477 108 T : VT;0 V300;0 exp
300
b c d e

Pressure error derived from uncertainty in the unit-cell volume of Au. Pressure difference corresponding to temperature variation in the sample. Temperature variations are less than 10%. MgPv, MgSiO3-rich perovskite; Mw, (Mg,Fe)O magnesiowstite; CaPv, CaSiO3 perovskite; PPv, post-perovskite; CF, Ca-ferrite-type Al-phase.

pure samples unmixed with gold. They were compressed with beveled diamond anvils with 200-m culet. Heating was done using a multi-mode Nd:YAG laser or TEM01-mode Nd:YLF laser using a double-sided heating technique that minimizes axial temperature gradients within the sample (Shen et al., 1996). Temperature was measured by the spectroradiometric method (Watanuki et al., 2001). Angle-dispersive X-ray diffraction spectra of the samples were collected on a CCD detector or an imaging plate at BL10XU of SPring-8. A monochromatic incident X-ray beam with a wavelength of about 0.41 was collimated to 20-m in diameter. The diffraction data were obtained in-situ at high PT during heating and after quenching temperature to 300 K. Two-dimensional X-ray diffraction images were integrated as a function of 2-theta angle in order to give a conventional one-dimensional diffraction profile using the fit-2D program (Hammersley, 1996). Recently, there has been an extensive debate on the accuracy of pressure scales for high-pressure experiments. The experimen-

tally-determined pressure estimates for the post-perovskite phase transition in MgSiO3 differ by as much as 15 GPa, depending on the internal pressure standard and its PVT equation of state (EOS) used in each study (see a review by Hirose, 2006). The MgO pressure scale may currently be the most practical because it has been extensively studied and is least controversial. Indeed, the post-spinel phase transition boundary pressure in Mg2SiO4 based on the PVT EOS of MgO proposed by Speziale et al. (2001) matches the associated 660-km seismic discontinuity pressure (Fei et al., 2004). Moreover, when Speziale's MgO pressure scale is used, the MgSiO3 post-perovskite phase transition is determined to occur at 119 GPa for a temperature of 2500 K, which is consistent with the typical observed depth of the D shear velocity discontinuity commonly attributed to this phase change (Hirose et al., 2006). In the present study, however, MgO could not be used as a direct pressure indicator because it reacts readily with the samples. Therefore, we measured the unit-cell volume of Au mixed with the sample from (111) and (200) diffraction lines,

Author's personal copy

110

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

and estimated the pressure using a new PVT EOS of Au (Hirose et al., submitted for publication) that is consistent with the Speziale's MgO pressure scale (Table 2). This new EOS of Au was obtained by the simultaneous volume measurements of Au and MgO at 0140 GPa and 3002330 K, using Speziale's MgO scale as a reference pressure scale. It predicts higher pressure than previously reported EOS of Au at high temperatures above 1500 K (e.g., Shim et al., 2002; Tsuchiya, 2003). The errors in our pressure measurements using PVT EOS are derived from errors in volume and temperature, with the absolute uncertainty due to the uncertainty in EOS itself. The errors in the unit-cell volume of Au based on X-ray diffraction measurements were relatively small, resulting in pressure uncertainty of typically about 1 GPa (Table 2). On the other hand, temperature variations are generally large in laser-heated samples; up to 10% variations within a 20-m area from which X-ray diffractions were collected (Hirose et al., 2006). Such temperature gradient causes more than 2 GPa pressure gradient due to the effect of thermal pressure; high temperature increasing the pressure and low temperature reducing the pressure (e.g., Dewaele et al., 1998). 2.2. Seismological observations If MORB is present in the deep mantle, it could either be located in recently subducted material, localized in a thin layer of a potentially very strongly contorted slab, or it could have accumulated as large quantities of dense dregs, separated over time from ancient subducted slabs and swept by deep mantle flow into piles away from current downwellings. While there have been some seismological interpretations of thin lamellae of anomalous material in the deep mantle that could correspond to a former crustal layer (Weber, 1994; Thomas et al., 1998), it is very difficult to seek such structures or to interpret them reliably.

We instead consider the possibility that large amounts of MORB material has accumulated into dense piles in the lowermost mantle, such that volumetrically the material has distinctive observable seismic properties; in particular, phase changes give rise to seismic velocity contrasts that reflect elastic waves. We focus on the shear velocity structure near the northern margin of the Pacific LLSVP (Fig. 2), as this is one of the few regions where it is possible to determine the detailed velocity structure of a chemically distinct region of D. We substantially augment previously analyzed seismic data sampling the D region beneath a localized domain under the central Pacific (Avants et al., 2006; Lay et al., 2006), to refine the regional shear velocity structure. Our data are from earthquakes in the TongaFiji region recorded at broadband stations in western North America (Fig. 2a). We focus on shear wave observations in the distance range 7080, for which there is enough time separation between the CMB reflection, ScS, and the direct wave turning in the mid-mantle, S (Fig. 2b), such that we can seek any reflectivity within the lowermost 400-km of the LLSVP. We consider S waves rather than P waves due to the greater time separation of relevant phases and the observed tendency for S-wave reflections from the deep mantle to be stronger than P wave reflections, as expected for computed post-perovskite phase transition reflectivity (Stackhouse et al., 2005; Wentzcovitch et al., 2006). The detailed data processing is identical to that described by Lay et al. (2006); we use transverse component ground displacements corrected for receiver anisotropy and deconvolved by the stacked ScS wavelet to remove source time function complexity. The deconvolved traces from all events are used in a double-array stacking procedure that determines the 1D reflectivity in the signals as a function of depth relative to the CMB. Identical analysis of corresponding synthetic seismograms for localized 1D velocity models is used to match the observed

Fig. 2. (a) Globe map showing the source locations (S) and receiver locations (R) for our seismic data, the 444 great-circle raypaths, and D velocities for the tomographic shear velocity model of Grand (2002). The coremantle boundary (CMB) reflection points for ScS phases are shown by black circles. (b) Cross section along the great-circle arc marked by white dashes (a), showing representative raypaths of S and ScS phases and the tomographic shear velocity variations in that greatcircle plane. (c) Map showing the ScS reflection points (black circles) and the tomographic variations in the lowermost 200-km of the mantle. Note that the paths sample the northern margin of the LLSVP under the Pacific shown in Fig. 1. This region has high density, strong lateral margins, and 3 to 7% low shear velocities (Lay and Garnero, 2004).

Author's personal copy

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

111

stacks, perturbing the absolute velocities and any velocity discontinuities in the structure. Our final data set involves 444 high quality SH signals sampling a very localized region of about 500750 km in horizontal extent (Fig. 2c) near the northern boundary of the Pacific LLSVP, corresponding to one of the 3 subregions studied in (Lay et al., 2006). 3. Deep mantle phase relations 3.1. Pyrolitic mantle Combining this study and that of Murakami et al. (2005), eight separate sets of experiments were conducted on pyrolitic mantle (KLB-1 peridotite) composition in the pressure range between 98 and 132 GPa (Table 2). In order to avoid kinetic hindering of phase transformation especially in such a multi-component system, each run was conducted on amorphous starting material at a single PT condition of interest, except that we changed temperature during heating in run #1. The diffraction peaks of MgSiO3-rich perovskite (Mg-perovskite), (Mg,Fe)O magnesiowstite, and CaSiO3-rich perovskite (Ca-perovskite) were ob-

served below 114 GPa and 2540 K (Fig. 3b). At higher pressures, the peaks from post-perovskite were observed together with these three phases (Fig. 3a). Mg-perovskite was not present above 122 GPa at 2550 K. These results demonstrate that at a temperature of 2500 K the post-perovskite phase transition occurs between 116 and 122 GPa, corresponding to 2550 to 2640-km depth in the mantle (Fig. 4). This is similar to the transition pressure in pure MgSiO3 (119 GPa at 2500 K for MgSiO3 (Hirose et al., 2006)). The dP/dT slope of post-perovskite-in and Mgperovskite-out curves is not tightly constrained but is best estimated to be +8(4) MPa/K, consistent with a previous report (Ono and Oganov, 2005). The post-perovskite transition occurs within a 5 GPa pressure range in pyrolitic material, corresponding to a lowermost mantle depth range of 90-km. Such a width is much narrower than for post-perovskite phase transitions in MgSiO3Al2O3 (Akber-Knutson et al., 2005) and MgSiO3FeSiO3 (Mao et al., 2004; Tateno et al., 2007). This is similar to the majorite to perovskite phase transition; the two-phase coexisting field in MgSiO3Al2O3 (Irifune et al., 1996) is much wider than that in natural MORB composition (Hirose et al., 1999). However, a

Fig. 3. X-ray diffraction patterns for pyrolite at (a) 116 GPa, 1940 K and (b) 110 GPa, 1960 K, and for MORB obtained after quenching temperature from (c) 109 GPa, 2100 K and (d) 117 GPa, 2240 K. (a)(c) are from this study and (d) is from Hirose et al. (2005). MP, MgSiO3-rich perovskite; PPv, post-perovskite; Mw, magnesiowstite; CaPv, CaSiO3 perovskite; SC, CaCl2-type SiO2 phase; SA, -PbO2-type SiO2 phase; CF, Ca-ferrite-type Al-phase; Au, gold; NaCl, pressure medium.

Author's personal copy

112

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

Fig. 4. Phase relations in a pyrolitic lowermost mantle. Open and solid symbols indicate the stabilities of Mg-perovskite and post-perovskite, respectively. Halffilled symbols show the coexistence of perovskite and post-perovskite. Magnesiowstite and Ca-perovskite were observed in all the experiments. The horizontal error bars indicate pressure uncertainty due to the error in unit-cell volume of pressure standard. The oblique bars represent a pressure gradient corresponding to a temperature gradient (10%) in the sample. The uncertainties in the locations of post-perovskite (PPv)-in and perovskite (Pv)-out curves are shown by gray. A general depth range of the D discontinuity (2600 to 2700-km depth) is indicated by a bar (Wysession et al., 1998).

5 GPa pressure interval for post-perovskite phase transition in pyrolite is still larger than estimates of the sharpness of the D seismic discontinuity. Seismological studies have inferred a depth extent of velocity increase up to 50 to 75-km (Revenaugh and Jordan, 1991; Weber et al., 1996), or less than 30-km (Lay and Young, 1989). This discrepancy with experimental results may be due to the errors in pressure determinations in the experiments. On the other hand, there has been extensive debate on a similar discrepancy for the 410-km seismic discontinuity. The effective width of the olivine to -spinel transition may be less than half of two-phase coexisting region (Stixrude, 1997), possibly this is also true for the post-perovskite phase transition. Further seismological studies are needed to better constrain the sharpness of the D discontinuity. The perovskite to post-perovskite phase transition in pure MgSiO3 is theoretically predicted to cause a 1.5% shear velocity increase (e.g., Stackhouse et al., 2005; Wentzcovitch et al., 2006), although the incorporation of Fe and Al significantly reduces the velocity contrast (Caracas and Cohen, 2005). Perovskite in a pyrolitic mantle contains relatively little Fe and Al (Murakami et al., 2005), and the phase transition to post-perovskite produces about 1% S-wave velocity increase (Tsuchiya and Tsuchiya, 2006). 3.2. MORB crust We conducted seven separate sets of experiments on normal MORB composition at 88 to 132 GPa in this study and Hirose et al. (2005) (Table 2). Similarly to the pyrolite experiments, heating was made at a single PT condition for each run except

run #13. We observed a four-phase assemblage of Mg-perovskite, Ca-perovskite, Ca-ferrite-type Al-phase, and CaCl2-type SiO2 phase below 104 GPa and 2060 K. Detailed chemical analyses are discussed in Hirose et al. (2005) and Sinmyo et al. (2006). The most intense (022) line of post-perovskite was found as a very weak peak at 110 GPa and 2100 K (Fig. 3c), indicating that the post-perovskite-in curve should be very close to this PT condition. SiO2 phase also transformed from CaCl2-type to -PbO2type structure between 112 and 114 GPa at 20502220 K (Fig. 3d). Mg-perovskite disappeared above 117 GPa at 2240 K. These results show that post-perovskite phase transition takes place in MORB material between 112 and 118 GPa (2480 and 2580-km depth) at 2500 K (Fig. 5), a range shifted by about 4 GPa lower pressure than that for a pyrolitic composition. This is a small but, we believe, sensible difference, even allowing for the pressure and temperature uncertainties in our measurements (Figs. 4 and 5). This pressure shift is supported by theory (Caracas and Cohen, 2005; Stackhouse et al., 2006; Zhang and Oganov, 2006). Perovskite and post-perovskite in MORB are enriched in both Al and Fe (Hirose et al., 2005). Theoretical calculations have demonstrated that small increase in the transition pressure by Al is negligible compared to the large reduction by Fe. Indeed, the calculations suggest much lower transition pressure in MORB than the present experimental results, due to the effect of Fe. The SiO2 phase transition in MORB occurs at an indistinguishable pressure from the post-perovskite transition. This is consistent with previous experimental results indicating that the postperovskite phase transition boundary in MgSiO3 end-member composition is very close to the CaCl2-type to -PbO2-type phase transition boundary in pure SiO2 (Murakami et al., 2003, 2004). Theory predicts that both MORB phase transitions involve decreases in shear velocity. The calculations have shown that the incorporation of Al and Fe remarkably reduces the velocity

Fig. 5. Phase relations in a subducted MORB crust at lowermost mantle conditions. The stabilities of Mg-perovskite and post-perovskite were shown by open and solid symbols, respectively. Half-filled symbols indicate the coexistence of perovskite and post-perovskite. The boundaries are assumed to have the same pressure/ temperature slope as that for pyrolite, and the gray indicates the uncertainties in their locations. The dashed line indicates the phase transition boundary between CaCl2-type (circles) and -PbO2-type (squares) SiO2 phases.

Author's personal copy

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

113

contrast between perovskite and post-perovskite (Caracas and Cohen, 2005) and results in a negative S-wave discontinuity at the phase transition in MORB (Tsuchiya and Tsuchiya, 2006). The (Al, Fe)-enriched perovskite in MORB contains 14 wt.% Al2O3 and 20 wt.% FeO (Hirose et al., 2005), and its transformation to post-perovskite produces a 1.5% shear velocity discontinuity, assuming simple linear effects of Al and Fe. The S-wave velocity change for the SiO2 phase transition from CaCl2-type to -PbO2-type structure has been also calculated as 1 to 2% (Karki et al., 1997). Note that velocity contrast for the SiO2 phase transition and the effects of Al and Fe impurities on post-perovskite phase change have been calculated only at T = 0 K. We therefore assume here that the magnitude of velocity jump is the same at high temperatures. Since both (Fe, Al)-enriched post-perovskite and SiO2 phase are important constituent minerals in MORB (our MORB composition includes 38% post-perovskite and 23% SiO2 phase at lowermost mantle conditions (Hirose et al., 2005)), a negative shear wave velocity gradient and/or jump of about 1% ( 0.6% by postperovskite phase change and another 0.5% by SiO2 phase transition) is expected around 2500 to 2550-km depth, for any subducted MORB component present. Given the experimental evidence for a possible depth shift between MORB-related phase changes that cause a velocity decrease and pyrolitic phase change that causes a velocity increase, we explore seismic data for any corresponding structures at about the right pressure conditions. 4. Seismological analysis Our seismic data display waveform complexities that indicate the presence of complex structure in the deep mantle under the central Pacific. Representative examples of the deconvolved waveforms are shown in Figs. 6 and 7. Fig. 6 shows representative waveforms from four events, aligned on the ScS arrivals. The dominant periods for all of our data are from 3 to 5 s. There are several arrivals between the S and ScS peaks, the largest being a positive amplitude arrival labeled B, which is preceded by a negative peak labeled A. Side-lobes are present on the major peaks in the traces due to the source wavelet deconvolution, but this does not account for the A or B features. The A and B arrivals are observed to shift in time relative to ScS systematically with distance, as shown in the two event profiles shown in Fig. 7, indicating that these arrivals are associated with deep mantle structure, not source or receiver reverberations. Move-out relative to S is weak over the narrow distance range spanned by our localized sampling, since the reflected waves are not turning far apart, but there are clear shifts in timing relative to S from event to event, as expected given variable source depths and distance ranges for each event. The phases have stable relative behavior, indicating that they are not scattered from out of plane, and to first-order at least, the data can be characterized by a 1D structure. Clearly, however, the arrivals are small (the weak arrivals cannot be tracked in raw displacement data) and noise and deconvolution artifacts are present in the signals, so a waveform stacking procedure is needed to quantify the stability and characteristics of the A and B arrivals, including their move-out

Fig. 6. Examples of SH ground displacement waveforms for different stations and events, indicating systematic negative (A) and positive (B) peaks between S and ScS arrivals. These signals have been deconvolved by the average ScS source wavelet for each event (Avants et al., 2006), and low pass filtered to retain periods longer than 3 s. Variations in the time differences between S and ScS, and between ScS and A and B are consequences of variable source depth (d) and epicentral distance ().

relative to S and ScS for events at different distances and source depths. Given the very localized sampling of our 444 signals, we assume that the regional structure can be represented by a 1D velocity structure, justifying the double-array stacking method that is employed (Lay et al., 2006). Earlier studies of a more distributed data set have established that there are lateral variations in the wave field over scale lengths of 1000-km in this region (Avants et al., 2006; Lay et al., 2006), but we restrict our specific data set to a subregion with lateral scale of less than 750-km, allowing for the lateral averaging effects of the grazing raypaths in the deep mantle. A single 1D stack is thus obtained, having about 40% greater sampling in our localized area than in previous work due to the inclusion of more recent earthquakes, giving a total of 26 events. The large increase in data was achieved by using several larger events than in earlier work (Lay et al., 2006), which have clear signal windows for detecting reflectivity associated with the A and B arrivals in double-array stacks. The double-array stack that is obtained is shown in Fig. 8a, with the stack amplitude plotted as a function of vertical distance from the CMB. The reference velocity model used in the stacking is the same regional model used in Lay et al. (2006), and the vertical distances are apparent depths due to dependence on this reference model. Note that the data stack has very tight uncertainty estimates, with strong peaks associated with ScS at the CMB, and clear positive reflection B and negative reflection A. In this case, the traces were aligned and normalized on ScS, so all amplitudes are relative to the stack amplitude at the CMB. Additional negative arrivals just above the CMB, labeled C and D, are inferred from the analysis by Lay et al. (2006). The additional data from this study tend to obscure features very close

Author's personal copy

114

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

Fig. 7. Distance profiles of shear waves from two deep focus TongaFiji earthquakes recorded by broadband stations in California. These data augment the large data sets previously collected for our study region by Avants et al. (2006) and Lay et al. (2006). The traces have all been deconvolved by average source wavelets and are aligned on the ScS arrivals. The arrival labeled B corresponds to the reflection from the velocity increase labeled B in Fig. 8, and A corresponds to the negative amplitude arrival from the velocity decrease labeled A in Fig. 8. Stacking of many recordings is needed to suppress incoherent noise, receiver reverberations, sidelobes from source wavelet and instrument deconvolution, and other contributions to the signals.

to the CMB due to reduced bandwidth due to some events having longer source durations, but the new data enhance and clarify the A and B features noted in Lay et al. (2006). At apparent distances more than 400-km above the CMB the stack is contaminated by strong coda of the direct S phase. The significance of the data stack features is demonstrated by processing synthetic seismograms for the smooth shear velocity model PREM (Dziewonski and Anderson, 1981) in exactly the same way as the data. This results in the dashed blue curve (Fig. 8a), which shows only the ScS peak and the S coda contamination, and defines a reference level from which the positive and negative peaks B and A can be measured. The latter features are clearly stronger than predicted by PREM (the PREM synthetics produce a small side lobe that is slightly shallower and weaker than the A trough; this is apparent because the synthetics, made for just one depth, minimize the variations in the side lobe, while the data, involving many source depths, will average it out). Thus, perturbations of the PREM model were made by trial and error, introducing positive and negative discontinuities at appropriate depths, with the synthetics being processed like the data and resulting stacks being compared. Matching the data requires introduction of velocity decreases (A, and C and D for the broader bandwidth data used in Lay et al. (2006)) and a velocity increase (B) in the lowermost 400-km of the mantle.

The precision of the estimated discontinuity depths is about 20-km. This level of precision is inferred by separately using ScS and S as reference phases for the stacking and modeling (see on-line Fig. S1), which greatly reduces trade-offs between absolute velocities and discontinuity depths. This tight bound on discontinuity depths is well-within the experimental uncertainties for mineral physics constraints on deep mantle phase transitions. The tight error bars on the data stacks yield 15% precision in the strengths of the discontinuities within the context of 1D modeling. While the velocity contrasts are wellresolved, it must be kept in mind that they may arise from phase changes in only a portion of a chemically mixed assemblage. It is harder to assess the uncertainties due to assuming localized 1D structure in what is clearly a 3D Earth model, but our constraint of very tight spatial sampling and extensive modeling with 2.5D structures (Lay et al., 2006) gives us confidence that our final model has basic validity as a local average representation of the structure in this region of the LLSVP. The preferred shear velocity model, SPAC3, predicts well the amplitude of all of the precursors to the CMB reflected phase, ScS (Fig. 8a). SPAC3 includes a shear velocity jump of 0.5% at 2520-km depth (discontinuity A), a jump of +0.6% at 2675-km depth (discontinuity B), and deeper decreases in velocity within the lowermost 76-km of the mantle (discontinuities C and D)

Author's personal copy

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

115

Fig. 8. Seismogram stack and velocity profile for the northern margin of the Pacific LLSVP. (a) Double-array stack of 444 transverse component (SH) recordings of TongaFiji earthquakes made at broadband stations in California (see example waveforms in Figs. 6 and 7). The traces were aligned and normalized on the core mantle boundary (CMB) reflected phase ScS, and stacked to determine reflectivity of energy as a function of depth relative to the CMB (solid black line). Bootstrap uncertainties on the stack are shown by dashed lines (Avants et al., 2006). The stacked data for depths more than 450-km above the core are contaminated by coda from direct S arrivals that turn at shallower depths. The blue dashed curve shows the results of stacking of similarly processed synthetics for model PREM (Dziewonski and Anderson, 1981). The red curve shows the results of stacking synthetics for model SPAC3. (b) The velocity profiles for PREM and SPAC3. SPAC3 had four velocity discontinuities (A, B, C, D) that produce reflectivity associated with the corresponding labeled peaks on the left. For the interpretation of this region as a MORBpyrolite mixed pile, discontinuity A results from post-perovskite and SiO2 phase transitions in the MORB component, discontinuity B from post-perovskite phase transition in the pyrolitic component, discontinuity C from back transformation of the post-perovskite to perovskite in the pyrolitic material due to rapid temperature increase (Lay et al., 2006), and discontinuity D from onset of partial melting just above the CMB.

(Fig. 8b). This model predicts the data well whether ScS or S are used as reference phases, and provides good fit to the ScSS differential times (Supplemental Fig. S1). The velocity changes could be distributed over transition zones extending as much as 30-km in depth, rather than being the sharp discontinuities found in model SPAC3, and still produce synthetics that match the data well. This is an intrinsic limitation of using wide-angle reflections with 35 s periods. The actual velocity jumps may be underestimated somewhat due to modeling of extensively stacked signals and the assumption of a 1D structure. The depth interval between A and B is estimated to have about 20-km uncertainty, from 135 to 155-km based on the fitting of stacks aligned on S versus ScS, and assuming simple intervening velocity structure. The predicted reflections are small (see Supplemental Fig. S2), accounting for the difficulty of readily seeing them in individual traces, and therefore stacking of many signals is required to have confident detection of the weak phases. 5. Implications for a MORB-rich pile above the CMB Thermally-equilibrated subducted MORB crust is denser than the average mantle at all depths except between 660 and 720-km (Irifune and Ringwood, 1993; Hirose et al., 1999; Ono et al., 2005; Hirose et al., 2005). The density contrast is about 3% at the base of the mantle (Hirose et al., 2005), so if subducted slabs penetrate to the D region, MORB material could be stable there. Geodynamical simulations have demonstrated that this magnitude of density contrast can induce separation and accumulation of dense MORB crust at the base of the mantle (Christensen and Hofmann, 1994; Tackley, 1998; McNamara and Zhong, 2004; Nakagawa

and Tackley, 2005). It has been further suggested that dense MORB-enriched materials are swept into piles concentrated underneath lower mantle upwellings (Christensen and Hofmann, 1994; Tackley, 1998; McNamara and Zhong, 2004; Nakagawa and Tackley, 2005). MORB material may therefore contribute to seismic heterogeneities in the D region, either within recently subducted slab material or in piles of separated MORB from ancient subduction, which possibly represents the LLSVP. Our experimental results indicate that Mg-perovskite begins to transform to post-perovskite in MORB about 70-km shallower than the post-perovskite phase transition in pyrolite at the same temperature (Figs. 4 and 5). In addition, SiO2 phase in MORB undergoes a phase transformation from CaCl2-type to -PbO2type structure at the same depth as the post-perovskite transition in MORB. If MORB has accumulated into substantial piles of mixed, but not homogenized MORB and pyrolitic material, within which phase boundaries could exist laterally over large extent, seismology may be able to detect the small velocity effects of the phase changes in MORB. The LLSVP in the lowermost mantle under the southern Pacific may involve such a large accumulation of dense MORB-enriched materials (Christensen and Hofmann, 1994; Tackley, 1998; McNamara and Zhong, 2004; Nakagawa and Tackley, 2005). If this is the case, post-perovskite and SiO2 phase transitions in the MORB component of the mixed chemical pile could account for discontinuity A, with post-perovskite transition in the pyrolitic component accounting for discontinuity B. Comparing the magnitudes of observed velocity jumps ( 0.5% at discontinuity A and +0.6% at discontinuity B) with those predicted from mineral physics ( 1% for MORB and +1%

Author's personal copy

116

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117

for pyrolite), a 50/50 mix of MORB and pyrolite can be reconciled with the seismic data, generally accounting for the depth and strength of the A and B discontinuities, although collective uncertainties in all of the measurements provide only loose bounds on the bulk composition. The chemical reaction between the MORB and pyrolite should be negligible unless partial melting is involved because the solid-state diffusion is very slow; only 1 m in 109 to 1010 yr in perovskite lithology (Holzapfel et al., 2005), so the mixed material can be very old. The depths and signs of the seismic discontinuities in model SPAC3 are generally consistent with the MORB-enriched pile scenario. The 2520-km depth for discontinuity A matches the pressure of post-perovskite and SiO2 phase transitions in MORB at about 2500 K (Fig. 5). The temperature at discontinuity B would have to be about 400500 K higher than that at discontinuity A to reconcile the depth of discontinuity B with postperovskite phase transition in pyrolite (Fig. 4), although this value strongly depends on the dP/dT slope of the phase transition boundary and may be an overestimate. This temperature increase could be attributed to a strong temperature gradient within a dense chemical pile as predicted in geodynamic models (Nakagawa and Tackley, 2005). Interpreting deep mantle seismic structure involves significant non-uniqueness, so other possible interpretations can be advanced. Lay et al. (2006) interpreted discontinuity A as the effect of a change in composition at the top of the LLSVP, possibly involving iron enrichment that causes an abrupt shear velocity decrease, with discontinuity B being a post-perovskite transformation. They further suggest that discontinuity C is the result of post-perovskite converting back to perovskite in a steep thermal boundary layer, with discontinuity D being the upper boundary of an ultra-low velocity zone, perhaps involving partial melt. The key difference relative to the current paper involves the explanation for discontinuity A. Given the new mineral physics results, the MORB-pile interpretation appears at least equally viable, and in terms of accounting for a large-scale heterogeneous composition, the MORB-pile notion is much more straightforward than accounting for a large iron-enriched domain. Comparable detailed seismic analysis of reflectivity in other regions of the Pacific and African LLSVPs is needed (see a review of recent results in Lay and Garnero, 2007), but the seismic model for the region under the central Pacific in Fig. 8 is generally consistent with expected phase transitions in a chemically distinct pile of mixed MORB and pyrolitic material. This inference is clearly right at the frontier limits of both mineral physics experiments and seismological imaging of the deep interior, but the large data sets used for both arenas at least establish the viability of this interpretation. If further mapping of the structure sustains the interpretation in terms of phase transitions in a MORB-enriched pile, geophysicists may finally have a clear indication of the fate of large quantities of subducted oceanic crust. Acknowledgments We thank M. Murakami for technical advice, and T. Nakagawa and J. Hernlund, along with three anonymous reviewers and the Editor for thoughtful comments. The synchrotron X-ray ex-

periments were conducted at SPring-8 (proposal no. 2005B6892PUl-np, 2005B0010-LD2-np, and 2006A0099). Seismic data were obtained from the IRIS, Berkeley, and Caltech/USGS TRInet data centers. This work was supported by grants from the Japan Society for the Promotion of Science and the U.S. National Science Foundation (EAR-0125595, EAR-0453884, and EAR-0635570). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.epsl.2007.11.037. References
Akber-Knutson, S., Steinle-Neumann, G., Asimow, P.D., 2005. The effect of Al on the sharpness of the MgSiO3 perovskite to post-perovskite phase transition. Geophys. Res. Lett. 32, L14303. doi:10.1029/2005GL023192. Avants, M., Lay, T., Russell, S.A., Garnero, E.J., 2006. Shear-velocity variation within the D region beneath the Central Pacific. J. Geophys. Res. 111, B05305. doi:10.1029/2004JB003270. Caracas, R., Cohen, R.E., 2005. Effect of chemistry on the stability and elasticity of the perovskite and post-perovskite phase in the MgSiO3FeSiO3Al2O3 system and implications for the lowermost mantle. Geophys. Res. Lett. 32, L16310. doi:10.1029/2005GL023164. Christensen, U.R., Hofmann, A.W., 1994. Segregation of subducted oceanic crust in the convecting mantle. J. Geophys. Res. 99, 1986719884. Dewaele, A., Fiquet, G., Gillet, P., 1998. Temperature and pressure distribution in the laser-heated diamond-anvil cell. Rev. Sci. Instrum. 69, 24212426. Dziewonski, A.M., Anderson, D.L., 1981. Preliminary reference Earth model. Phys. Earth Planet. Inter. 25, 297356. Fei, Y., Van Orman, J., Li, J., van Westrenen, W., Sanloup, C., Minarik, W., Hirose, K., Komabayashi, T., Walter, M., Funakoshi, K., 2004. Experimentally determined postspinel transformation boundary in Mg2SiO4 using MgO as an internal pressure standard and its geophysical implications. J. Geophys. Res. 109, B02305. doi:10.1029/2003JB002562. Grand, S.P., 2002. Mantle shear-wave tomography and the fate of subducted slabs. Philos. Trans. R. Soc. Lond. A360, 24752491. Hammersley, J., 1996. Publication No. ESRF98HA01T, ESRF, Grenoble, France. Hirose, K., 2006. Post-perovskite phase transition and its geophysical implications. Rev. Geophys. 44, RG3001. doi:10.1029/2005RG000186. Hirose, K., Fei, Y., Ma, Y., Mao, H.K., 1999. The fate of subducted basaltic crust in the Earth's lower mantle. Nature 397, 5356. Hirose, K., Takafuji, N., Sata, N., Ohishi, Y., 2005. Phase transition and density of subducted MORB crust in the lower mantle. Earth Planet. Sci. Lett. 237, 239251. Hirose, K., Sinmyo, R., Sata, Y., Ohishi, Y., 2006. Determination of postperovskite phase transition boundary in MgSiO3 using Au and MgO pressure standards. Geophys. Res. Lett. 33, L01310. doi:10.1029/2005GL024468. Hirose, K., Sata, N., Komabayashi, T., Ohishi, Y., submitted for publication 2007. Simultaneous measurements of Au and MgO to 140 GPa and thermal equation of state of Au based on MgO pressure scale. Phys. Earth Planet. Inter. Holzapfel, C., Rubie, D.C., Frost, D.J., Langenhorst, F., 2005. FeMg interdiffusion in (Mg,Fe)SiO3 perovskite and lower mantle reequilibration. Science 309, 17071710. Irifune, T., Ringwood, A.E., 1993. Phase transformations in subducted oceanic crust and buoyancy relationships at depths of 600800 km in the mantle. Earth Planet. Sci. Lett. 117, 101110. Irifune, T., Koizumi, T., Ando, J.I., 1996. An experimental study of the garnetperovskite transformation in the system MgSiO3Mg3Al2Si3O12. Phys. Earth Planet. Inter. 96, 147157. Karki, B.B., Stixrude, L., Crain, J., 1997. Ab initio elasticity of the three highpressure polymorphs of silica. Geophys. Res. Lett. 24, 32693272. Kessel, R., Schmidt, M.W., Ulmer, P., Pettke, T., 2005. Trace element signature of subducted-zone fluids, melts and supercritical liquids at 120180 km depth. Nature 437, 724727.

Author's personal copy

K. Ohta et al. / Earth and Planetary Science Letters 267 (2008) 107117 Kesson, S.E., Fitz Gerald, J.D., Shelley, J.M.G., 1994. Mineral chemistry and density of subducted basaltic crust at lower-mantle pressures. Nature 372, 767769. Kesson, S.E., Fitz Gerald, J.D., Shelley, J.M., 1998. Mineralogy and dynamics of a pyrolite lower mantle. Nature 393, 252255. Komiya, T., 2004. Material circulation model including chemical differentiation within the mantle and secular variation of temperature and composition of the mantle. Phys. Earth Planet. Inter. 146, 333367. Lay, T., Young, C.J., 1989. Waveform complexity in teleseismic broadband SH displacements: Slab diffractions or deep mantle reflections? Geophys. Res. Lett. 16, 605608. Lay, T., Garnero, E.J., 2004. Coremantle boundary structures and processes. In: Sparks, R.S.J., Hawkesworth, C.J. (Eds.), The State of the Planet: Frontiers and Challenges in Geophysics. Geophys. Monogr., vol. 150. American Geophysical Union, Washington, DC, pp. 2541. Lay, T., Garnero, E.J., 2007. Reconciling the post-perovskite phase with seismological observations of lowermost mantle structure. In: Hirose, K., Brodholt, J., Lay, T., Yuen, D. (Eds.), Post-perovskite: The Last Mantle Phase Transition. Geophys. Monogr., vol. 174. American Geophysical Union, Washington, DC, pp. 129154. Lay, T., Hernlund, J., Garnero, E.J., Thorne, M., 2006. A post-perovskite lens and D heat flux beneath the central Pacific. Science 314, 272276. Mao, W.L., Shen, G., Prakapenka, V.B., Meng, Y., Campbell, A.J., Heinz, D.L., Shu, J., Hemley, R.J., Mao, H.K., 2004. Ferromagnesian post-perovskite silicates in the D layer of the Earth. Proc. Natl. Acad. Sci. 101, 1586715869. McNamara, A.K., Zhong, S., 2004. Thermochemical structures within a spherical mantle: superplumes or piles? J. Geophys. Res. 109, B07402. doi:10.1029/ 2003JB002847. Murakami, M., Hirose, K., Ono, S., Ohishi, Y., 2003. Stability of CaCl2-type and -PbO2-type SiO2 at high pressure and temperature determined by insitu X-ray measurements. Geophys. Res. Lett. 30, 1207. doi:10.1029/ 2002GL016722. Murakami, M., Hirose, K., Kawamura, K., Sata, N., Ohishi, Y., 2004. Postperovskite phase transition in MgSiO3. Science 304, 855858. Murakami, M., Hirose, K., Sata, N., Ohishi, Y., 2005. Post-perovskite phase transition and mineral chemistry in the pyrolitic lowermost mantle. Geophys. Res. Lett. 32, L03304. doi:10.1029/2004GL021956. Nakagawa, T., Tackley, P.J., 2005. The interaction between the post-perovskite phase change and a thermo-chemical boundary layer near the coremantle boundary. Earth Planet. Sci. Lett. 238, 204216. Nishihara, Y., 2003. Density and elasticity of subducted oceanic crust in the Earth's mantle. Thesis, Tokyo Tech. Oganov, A.R., Ono, S., 2004. Theoretical and experimental evidence for a postperovskite phase of MgSiO3 in Earth's D layer. Nature 430, 445448. Ono, S., Oganov, A.R., 2005. In situ observations of phase transition between perovskite and CaIrO3-type phase in MgSiO3 and pyrolitic mantle composition. Earth Planet. Sci. Lett. 236, 914932. Ono, S., Ohishi, Y., Isshiki, M., Watanuki, T., 2005. In situ X-ray observations of phase assemblages in peridotite and basalt compositions at lower mantle conditions: implications for density of subducted oceanic plate. J. Geophys. Res. 110. doi:10.1029/2004JB003196. Revenaugh, J., Jordan, T.H., 1991. Mantle layering from ScS reverberations, 4. The lower mantle and coremantle boundary. J. Geophys. Res. 96, 1981119824. Reymer, A., Schubert, G., 1984. Phanerozoic addition rates to the continental crust and crustal growth. Tectonics 3, 6377. Shen, G., Mao, H.K., Hemley, R.J., 1996. Laser-heated diamond anvil cell technique: double-sided heating with multimode Nd:YAG laser. Advance Materials '96 New Trends in High Pressure Research, 3rd NIRIM ISAM Proc, pp. 149152.

117

Shim, S., Duffy, T.S., Takemura, K., 2002. Equation of state of gold and its application to the phase boundaries near 660 km depth in the Earth's mantle. Earth Planet. Sci. Lett. 203, 729739. Sinmyo, R., Hirose, K., O'Neill, H.C., Okunishi, E., 2006. Ferric iron in Al-bearing post-perovskite. Geophys. Res. Lett. 33, L12S13. doi:10.1029/2006GL025858. Speziale, S., Zha, C., Duffy, T.S., Hemley, R.J., Mao, H.K., 2001. Quasi-hydrostatic compression of magnesium oxide to 52 GPa: implications for the pressure volumetemperature equation of state. J. Geophys. Res. 106, 515528. Stackhouse, S., Brodholt, J.P., Wookey, J., Kendall, J.M., Price, G.D., 2005. The effect of temperature on the seismic anisotropy of the perovskite and postperovskite polymorphs of MgSiO3. Earth Planet. Sci. Lett. 230, 110. Stackhouse, S., Brodholt, J.P., Price, G.D., 2006. Elastic anisotropy of FeSiO3 end-members of the perovskite and post-perovskite phases. Geophys. Res. Lett. 33, L01304. doi:10.1029/2005GL023887. Stixrude, L., 1997. Structure and sharpness of phase transitions and mantle discontinuities. J. Geophys. Res. 102, 1483514852. Tackley, P.J., 1998. Three-dimensional simulations of mantle convection with a thermo-chemical basal boundary layer: D? In: Gurnis, M., Wysession, M.E., Knittle, E., Buffet, B.A. (Eds.), The Coremantle Boundary Region. Geodyn. Ser., vol. 28. American Geophysical Union, Washington, DC, pp. 231253. Takahashi, E., 1986. Melting of a dry peridotite KLB-1 up to 14 GPa: implications on the origin of peridotite upper mantle. J. Geophys. Res. 91, 93679382. Tateno, S., Hirose, K., Sata, N., Ohishi, Y., 2007. Solubility of FeO in (Mg,Fe) SiO3 perovskite and the post-perovskite phase transition. Phys. Earth Planet. Inter. 160, 319325. Thomas, C., Weber, M., Agnon, A., Hofstetter, A., 1998. A low velocity lamella in D. Geophys. Res. Lett. 25, 28852888. Tsuchiya, T., 2003. First-principles prediction of the PVTequation of gold and the 660-km discontinuity in Earth's mantle. J. Geophys. Res. 102, 2039520411. Tsuchiya, T., Tsuchiya, J., 2006. Effect of impurity on the elasticity of perovskite and post-perovskite: velocity contrast across the post-perovskite transition in (Mg,Fe,Al)(Si,Al)O3. Geophys. Res. Lett. 33, L12S04. doi:10.1029/ 2006GL025706. Tsuchiya, T., Tsuchiya, J., Umemoto, K., Wentzcovitch, R.M., 2004. Phase transition in MgSiO3 perovskite in the Earth's lower mantle. Earth Planet. Sci. Lett. 224, 241248. Watanuki, T., Shimomura, O., Kondo, T., Isshiki, M., 2001. Construction of laser-heated diamond anvil cell system for in situ X-ray diffraction study at SPring-8. Rev. Sci. Instrum. 72, 12891292. Weber, M., 1994. Lamellae in D? An alternative model for lower mantle anomalies. Geophys. Res. Lett. 21, 25312534. Weber, M., Davis, J.P., Thomas, C., Kruger, F., Sherbaum, F., Schlittenhardt, J., Kornig, M., 1996. The structure of the lowermost mantle as determined from using seismic arrays. In: Boschi, E., Ekstrom, G., Morelli, A. (Eds.), Seismic Modeling of the Earth's Structure. Istituto Nazionale di Geophysica, Rome, pp. 399442. Wentzcovitch, R.M., Tsuchiya, T., Tsuchiya, J., 2006. MgSiO3 postperovskite at D conditions. Proc. Natl. Acad. Sci. 103, 543546. Wysession, M.E., Lay, T., Revenaugh, J., Williams, Q., Garnero, E., Jeanloz, R., Kellog, L., 1998. The D discontinuity and its implications. In: Gurnis, M., Wysession, M.E., Knittle, E., Buffet, B.A. (Eds.), The Coremantle Boundary Region. Geodyn. Ser., 28. American Geophysical Union, Washington, DC, pp. 231253. Zhang, F., Oganov, A.R., 2006. Valence state and spin transitions of iron in Earth's mantle silicates, Earth Planet. Sci. Lett. 249, 436443.