Pressure dependence of partition coefficients ...

Pressure dependence of partition coefficients between olivine and peridotite melt Takamasa Imai１, E. Takahashi1, T. Suzuki2 and T. Hirata1 1,Tokyo Institute of Technology ; 2, IFREE-JAMSTEC Senior Researcher Research Center for the Evolving Earth and Planets, Also at Department of Earth & Planetary Sciences Graduate School of Science & Engineering, Tokyo Institute of Technology e-mail: [email protected]

In the beginning of Earth history, crystallization of a peridotite magma ocean would have produced the primitive mantle. Therefore, it is important to study precise phase relations and element partitioning during crystallization of global magma ocean. In this study we determined PC-IR diagram for 32 elements between olivine and peridotite melt at 5, 10 and 15 GPa using Kawai-type multi anvil apparatus with EPMA and LA-ICP-MS for analysis. We compare our new results with those of Taura et al. (1998) determined with SIMS.

1 Introduction In the late stage of planet building process, the growing planets may have covered by thick magma ocean. In the case of the Earth, even the entire planet may have molten due to the Moon-forming giant impact. Walter et al.(2004) suggested that magma ocean may crystallized from the bottom up because dT/dP for the liquidus curve is steeper than magma ocean adiabat and subsequent solid state convection may have re-homogenized the initial stratification. In order to understand early history of our planet, precise phase relations and element partitioning during crystallization of global magma ocean are important. Taura et al. (1998) investigated pressure dependence of partition coefficient between olivine and coexisting melts for trace element by SIMS analysis. They reported that Dolivine/melt for monovalent cations and trivalent cations may be strongly affected by pressure. For example Dolivine/melt for Li, Na, Al have positive dependence of pressure and for Cr, V, Sc, and Y have negative dependence of pressure up to 14.4GPa. Because they used natural peridotite sample KLB-1 (Takahashi, 1986) and many trivalent cations are incompatible for olivine, the partition coefficients of other trace elements are still unknown. In this study we determined pressure dependence on partition coefficients for 32 elements between olivine and peridotite melt using LA-ICP-MS using starting materials doped with some trace elements. We discuss the result in comparison with Taura et al,(1998).

2 High Pressure Experiments In order to determine partition coefficient between olivine/melt, we carried out experiments up to 14.5GPa using Kawai-type multi-anvil apparatus

SPI-1000 and SAKURA at the Magma Factory, Tokyo Institute of Technology. We used two starting materials KLB-1* and KLB-1D. KLB-1* is a mixture of peridotite KLB-1(Takahashi 1986) and 3% of natural basalt (to increase LIL elements). KLB-1* was used in experiments at 2GPa, 5GPa, 10GPa, 14.5GPa. Another starting material KLB-1D used at 5 and 10GPa is also KLB-1 based but was doped with 27 trace elements (~200ppm). These samples were charged in graphite capsule and heated by a LaCrO3 tube heater.

Fig. 1: A backscattered-electron image of charge at 5GPa and 1950°C. To avoid the erroneous measurement of to temperature >2200℃ by means of the thermocouple, temperature >2200℃ were estimated based on an electric power vs. temperature calibration line which was established at temperatures below 2000 ℃ . Because the resulting ablation pit size achieved by the present LA-ICP-MS technique was 30μm, grain size

- 17 -

larger than 30µm was needed for respective phase. To obtain large enough crystals, the samples were heated up to about 100 ℃ higher than the nominal run temperatures. Then samples were cooled down slowly to the final temperature and then kept at the final temperature for one hour to achieve chemical equilibrium. After the heating procedure, the run samples were quenched.

3 Analysis Major and minor element contents were measured using JEOL8800 electron microprobe. A beam current of 1.2×10−8A and accelerating voltage of 15kV were used. Focused beam was used for crystal phase and a beam of 30μm diameter was used for melt phase. Melt phase was very heterogeneous because dendritic quenched crystals were grown. We used average value of large area for the composition in the melt portion. Trace element contents were measured using LA-ICP-MS at the Laboratory for Planetary Science, Tokyo Institute of Technology. These analyses were performed with ArF excimer laser and a quadrupole mass spectrometer. Laser ablation was operated with a laser diameter of 30μm for respective phase. The NIST610 glass was used as a calibration standard. Variations in ablation yields were corrected by reference to Si and Cr concentrations measured by EPMA. Laser firing time for one spot was 50 seconds. During the 50 seconds, analysis for 32 elements was repeated for 80 times. In the case of experiments using KLB-1* where trace element concentration is lower, analysis was repeated for 400 times for only 6 elements in order to obtain reliable analysis.

4.2 PC-IR diagram Fig.2 is PC-IR diagram (Onuma et al. 1968) at 5GPa and 10GPa in which partition coefficients are plotted against ionic radius for six coordinated sites by Shannon and Prewitt (1969, 1970) and Shannon (1976). A fitted curve was calculated using lattice-strain model (Blundy and Wood, 1994). This result is in agreement with Taura et al. (1998) except for K. At both 5 and 10 GPa, partition coefficients for divalent cations and tetravalent cations decrease monotonously with ionic radius. This means that the smaller cation is more compatible for olivine in this ionic radius range. Accordingly, our results indicate that ionic radii of Mg and Si are most compatible for

4 Result & Discussion We defined partition coefficient, Diolivine/melt, using concentration of elements determined by both EPMA and LA-ICP-MS. Diolivine/melt is defined as the mol ratio of elemental concentration, Ciol/Cimelt, where Ciol and Cimelt represent the concentration of element i in olivine and melt, respectively.

4.1 About analysis In the experiment using doped starting material KLB-1D, many elements (which were unable to analyze with SIMS by Taura et al.(1998) for natural peridotite KLB-1) were detected and the shape of the PC-IR diagram is well defined (Fig.2, 3). In the case of KLB-1*, the number of replication for each element was longer but number of analyzed elements were limited.

Fig. 2: PC-IR diagram of olivine/melt system at 5GPa. Solid symbol and solid line are the result of this study. The results reported by Taura et al. (1998) were also shown as the open symbol and dashed line. (a: monovalent(■,□), divalent(●,○) and tetravalent cations(◆,◇). b: trivalent cations.)

- 18 -

Fig. 3: The PC-IR diagrams of olivine/melt system under the high pressure (10GPa). Solid symbol and solid line are the result of this study. The data by Taura et al. (1998) were also shown as the open symbol and dashed line. (a: monovalent(■,□), divalent(●,○) and tetravalent cations(◆,◇). b: trivalent cations.)

Fig. 4: Pressure dependence of partition coefficient for (a): monovalent and tetravalent cations, and (b): divalent cations. Where symbols are duplicated, data for KLB-1* and KLB-1D are plotted. All data are by this study.

the M site and the T site, respectively in the olivine structure M2TO4 (Birle et al. 1968). Trivalent cations form a parabolic curve. Although the shape of the PC-IR diagram was unchanged from that of the Taura et al. (1998) (open symbols in Fig.2, 3), the number of analyzed cations are substantially increased and the parabolic curve is very well defined in this study (solid symbols in Fig.2, 3). Partition coefficient of trivalent cations have a relatively constant peak at around 65pm to 70pm at each pressure, indicating that this ionic radius is most compatible for olivine regardless of pressure in trivalent cations.

4.2 Pressure dependence of partition coefficient Partition coefficients are plotted against pressure in Fig.4 and Fig.5. Di of Tetravalent cations Si, Ti and Ca are mostly constant. However, Di of Al and Na increase with pressure (Fig.4a, Fig.5a). The important factors of element substitution are radius of cations and a sum of valence of cations (Ozawa, 1991). Thus, following element substitutions may take place with increasing pressures (Taura et al., 1998).

- 19 -

(Mg2+, Mg2+)⇔(Na+, Al3+) (Si4+, Mg2+)⇔(Al3+, Al3+)

Dis of trivalent cations except for Al seems to decrease with pressure (Fig.5). These results suggest that DAl is increased by the substitution of Al for Si at the tetrahedral site in olivine. Di of divalent cations except for Ca slightly decreases with pressure. Among divalent cations, degree of pressure dependence decreases in the following order; Ni>Mg>Fe>Mn which may correlate with the number of d-electrons.

Most available data about the partition coefficient between olivine and peridotite melt were obtained by Taura et al. (1998) with SIMS. The number of elements were limited and the geometry of the high-pressure PC-IR diagram was not well defined. In this study, we obtained new data set for 32 elements using LA-ICP-MS and precisely determined the PC-IR diagrams at 5, 10 and 15 GPa. In order to discuss the crystallization of deep magma ocean, we will further extend this study for liquidus phases of β -spinel, majorite, and silicate perovskites.

Acknowledgements We thank Dr. K. Sato for support in piston- cylinder experiments and EPMA analysis. Thanks are also due to Y. Kon and T. Yokoyama for supporting in LA-ICP-MS analysis.

References

Fig. 5: Pressure dependence of the partition coefficient for trivalent cations. All data are by this study.

Birle .J, Gibbs .G.V, Moore .P.B, Smith .J.V, 1968. Crystal structures of natural olivines, American Mineralogist 53. Blundy.J and Wood .B.J, 1994. Prediction of crystal–melt partition coefficients from elastic moduli, Nature 372, 452−454. Onuma.N, Higuchi.H, Wakita.H, Nagasawa.H, 1968. Trace element partitioning between two pyroxenes and the host lava. Earth Planet Sci Lett 5, 47−51. Ozawa .S, 1991. Trivalent cations in olivine and their implication to the upper mantle tectonics as inferred from the high pressure experiments. Doctoral thesis, Univ. of Tokyo. Shannon .R.D, 1976. Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallographica Section A, Crystal Physics, Diffraction, Theoretical and General Crystallography, Volume 32, Part 5, 751-767. Shannon .R.D and Prewitt .C .T, 1960, 1970. tructural Crystallography and Crystal Chemistry. Acta crystallographica. Section B, Structural crystallography and crystal chemistry, 0567−7408. Takahashi .E, 1986 Melting of dry peridotite KLB-1 up to 14GPa:implications on the origin of peridotitic upper mantle. J. Geophys Res. 91, 9367−9382. Taura. H, Yurimoto. H, Kurita.K, Sueno.S, 1998. Pressure dependence on partition coefficients for trace elements between olivine and the coexisting melts, Phys Chem Minerals 25, 469−484. Walter .M.J, Nakamura .E, Tronnes .R .G, Frost .D.J, 2004. Experimental constrains on crystallization in a deep magma ocean. Geochimica et Cosmochimica Acta. 68, 4267−4284.

- 20 -

The Electrical Conductivity of Post-Perovskite in Earth’s D” Layer

Kei Hirose Professor Research Center for the Evolving Earth and Planets, Also at Department of Earth & Planetary Sciences Graduate School of Science & Engineering, Tokyo Institute of Technology e-mail: [email protected]

Recent discovery of a phase transition from perovskite to post-perovskite suggests that the physical properties of the Earth’s lowermost mantle, called the D” layer, may be different from those of the overlying mantle. We 2 report that the electrical conductivity of (Mg0.9Fe0.1)SiO3 post-perovskite is >10 siemens per meter and does not vary greatly with temperature at the conditions of D” layer. A post-perovskite layer above the core-mantle boundary would, by electromagnetic coupling, enhance the exchange of angular momentum between the fluid core and solid mantle, which can explain the observed changes in length of a day on decadal timescales. Heterogeneity in the conductivity of the lowermost mantle is likely to depend on changes in chemistry or mineralogy of the boundary region, not fluctuations in temperature.

1 Introduction Geomagnetic observations have shown that the electrical conductivity of the upper to middle part of the lower mantle is 1 to 10 S/m, consistent with the laboratory measurements of the conductivity of silicate perovskite (Katsura et al. 1998). A possible existence of a highly conductive layer has been often suggested in the deepest mantle from geophysical modeling (Holme 1998), where recently discovered post-perovskite phase is dominant (Murakami et al. 2004). Here we report direct measurements of the conductivity of (Mg,Fe)SiO3 post-perovskite at pressures and temperatures corresponding to the conditions at the D” layer of the mantle.

2 Results We measured the DC electrical conductivity in a laser-heated diamond-anvil cell (DAC) (Fig. 1). In the first run, we first compressed (Mg0.9Fe0.1)SiO3 amorphous sample to 37 GPa in a DAC at room temperature and then heated it to about 1800 K to synthesize perovskite. The resistance of 9 untransformed material was ~ 10 Ω, and it dropped by two orders of magnitude after the synthesis of perovskite. Subsequently the sample resistance was measured with increasing temperature; it decreased

from 21 MΩ at 300 K to 20 kΩ at 1940 K. The conductivity was estimated to be about 1 S/m at high temperatures (Fig. 2). This value is reasonably consistent with the measurements on (Mg0.93Fe0.07)SiO3 perovskite at 23 GPa in a multi-anvil apparatus by Katsura et al. (1998). Next, we further compressed this sample to 58 GPa, and carried out the measurements between 1580 and 2290 K (Fig. 2). Similarly in the second set of experiments, perovskite was first synthesized, and then the high-temperature conductivity measurements were performed up to 2700 K at 104 GPa and to 2660 K at 117 GPa (Fig. 2). The conductivity was about 10-1 to 10−2 S/m at high temperatures, considerably lower than that measured at 37−58 GPa in the first run. We further squeezed this sample to 143 GPa and reheated it to ~ 2000 K for 20 min. During laser heating, the sample resistance drastically dropped by four orders of magnitude. This pressure is well within the stability field of iron-bearing post-perovskite, suggesting that such a drastic change in the resistance was a result of the perovskite to post-perovskite phase transformation. The sample resistance remained almost constant at ~ 7 kΩ as temperature was increased from 300 to 3000 K, which corresponds to a conductivity of 1.4 × 102 S/m (Fig. 2). Pressure should have increased at higher temperatures because of thermal expansion of the sample. If such an increase is considered, temperature

- 21 -

may have a small negative effect on the electrical conductivity of post-perovskite. In a third run, we first synthesized post-perovskite by heating the starting material to 2000 K for 30 min at 129 GPa. The x-ray diffraction pattern indicated that the sample was dominated by post-perovskite but included lesser amount of perovskite. The sample resistance changed little from 31 kΩ at 300 K to 24 kΩ at 2500 K. The conductivity was estimated to be 4.9 × 101 S/m at 2500 K with a small temperature dependence (Fig. 2). During decompression of this sample to 102 GPa at room temperature, the resistance increased slightly. After the complete pressure release, the sample was examined under both transmission electron microscope (TEM) and field-emission-type scanning electron microscope (FE-SEM, HITACHI SU-70) with a spatial resolution of 1.0-nm. No metal-like phase was observed throughout the sample. In addition, the chemical composition of post-perovskite was analyzed by energy-dispersive spectroscopy (EDS) and electron energy-loss near-edge structure (ELNES) spectroscopy equipped with the TEM. The results show that post-perovskite has Mg/(Mg+total Fe) molar ratio of 0.89 and Fe3+/(Fe2++Fe3+) ratio of 0.13±0.10, consistent with those of the starting material. This indicates that the Fe3+ abundance in the sample remained unchanged during experiments, although we did not control the oxygen fugacity specifically. A fourth experiment corroborated the low conductivity of perovskite. We synthesized perovskite from the amorphous sample by heating to 2000 K for 30 min and then performed the conductivity measurements at 121 GPa. The conductivity was low between 1450 and 1950 K, pretty consistent with the results of run #2 (Fig. 2).

3 Discussion These results indicate that the electrical conductivity of (Mg, Fe)SiO3 post-perovskite is much higher than that of perovskite (Fig. 2). The post-perovskite phase has a stacked SiO6-octahedral sheet structure with interlayer (Mg, Fe) ions. The high conductivity likely reflects the short Fe-Fe distance in the (Mg, Fe) layer, which is shorter than that in perovskite. The lower conductivity of perovskite at 104−121 GPa than at 37−58 GPa could be due to the high-spin to low-spin transition of iron in perovskite (Badro et al. 2004). The conduction in the high-spin perovskite is dominated by a small-polaron process of electron hopping between ferrous and ferric iron sites.

- 22 -

Fig. 1: (A) Microscopic image of the sample and electrodes. The quasi-four-terminal electrical resistance measurements were performed at high temperature in runs #1−3 (direct current of 1mA was applied through I+ to I−, and the voltage drop between V+ and V− was recorded). Only in run #4, sample resistance was measured by a two-terminal method. (B) Schematic drawing of the configuration in a DAC. The Au electrodes were sandwiched between the samples and connected to Pt leads outside the sample hole. The sample and electrodes were insulated against Re gasket by a cubic-BN layer. (C) Close-up view of the sample during heating by laser. About 30 or 50-µm areas of the sample and Au electrodes were heated from both sides. Only high-temperature part of the sample transformed to high-pressure phase.

Fig. 2: Electrical conductivity (σ) of perovskite and post-perovskite as a function of reciprocal temperature. Squares, run #1; circles, run #2; triangles, run #3; diamonds, run #4. Open and closed symbols indicate measurements of perovskite and post-perovskite, respectively. Previous data on perovskite by Katsura et al. (1998) are also presented by crosses. The measured variations in temperature between the electrodes are shown by error bars.

The unpaired electrons in the 3d orbital play important roles in this process, but the number of unpaired electrons of ferric iron decreases from five to one at this spin-pairing transition, thus resulting in a significant reduction in the conductivity. A layer with a high electrical conductivity above the core-mantle boundary would enhance the electromagnetic (EM) coupling between the fluid core and solid mantle. It has been suggested that if the conductance of this layer is >108 S (at least 3 × 107 S), the resultant exchange of angular momentum between the core and mantle would be sufficient to change the length of a day on decadal timescales by a few milliseconds, as has been observed (Holme 1998). Post-perovskite is a primary mineral of the D” layer below about 2600-km depth. Our measurements on (Mg, Fe)SiO3 post-perovskite indicate that the conductance of the D” layer may be 4 × 107 S (the conductance is related to the conductivity and thickness of the layer). The post-perovskite phase in the D” region actually includes certain amount of Al2O3, which increases the Fe3+/(Fe2++Fe3+) ratio in post-perovskite and hence should further enhance the

electrical conductivity. The conductance of the D” layer therefore may be high enough (> 108 S) to account for the decadal variations in length of a day. In addition, the EM coupling also affects the periodic precession of the Earth’s axis of rotation (nutation). The high conductance of the post-perovskite-rich D” layer may explain the retrograde 18.6- and 1.0-year nutations that have been observed (Buffett et al. 2000). A 200-m thick metallic solid layer at the top of the core has been proposed as the source of these changes (Buffett et al. 2000), but a ~ 300-km thick post-perovskite layer provides an alternative explanation. Heterogeneity in the electrical conductivity has been also inferred in the deep lower mantle from observations of geomagnetic jerks (Nagao et al. 2003). The inferred higher conductivity underneath Africa and southern Pacific has been attributed to higher temperatures there, corresponding to the regions with large low shear-wave velocity anomalies in the D” layer. Our measurements, however, show a minimal or even negative temperature dependence of the electrical conductivity of post-perovskite (Fig. 2).

- 23 -

Thus such conductivity heterogeneities in the D” layer, if any, could have chemical or mineralogical origins rather than a thermal origin.

References Badro, J. and 5 colleagues, 2004. Electronic Transitions in Perovskite: Possible Nonconvecting Layers in the Lower Mantle, Science 305, 383−386. Buffett, B. A., Garnero, E. J. & Jeanloz, R., 2000. Sediments at the top of Earth’s core, Science 290, 1338−1342. Holme, R., 1998. in The Core-Mantle Boundary Region, M. Gurnis, M. E. Wysession, E Knittle, B. A. Buffett, Eds. (American Geophysical Union, Washington, DC), pp. 139−151. Katsura, T., Sato, K. & Ito, E., 1998. Electrical conductivity of silicate perovskite at lower-mantle conditions, Nature 395, 493−495. Murakami, M., Hirose, K., Kawamura, K., Sata, N. & Ohishi, Y., 2004. Post-perovskite phase transition in MgSiO3, Science 304, 855−858. Nagao, H., Iyemori, T., Higuchi, T. & Araki, T., 2003. Lower mantle conductivity anomalies estimated from geomagnetic jerks, J. Geophys. Res. 108, 2254−2271.

- 24 -

Molecular dynamics simulation of hydrotalcite-water systems Katsuyuki Kawamura Professor Research Center for the Evolving Earth and Planets, Also at Department of Earth and Planetary Sciences Graduate School of Science & Engineering, Tokyo Institute of Technology e-mail: [email protected]

Smectite is one of the most effective materials as cation adsorbents in the potential engineering barrier for radioactive waste disposal. On the other hand, anion mobility seems to be fairly large in smectite clays. Hydrotalcite is one of the most effective candidate for the anion barrier. In this study, behavior of hydrotalcite is investigated by means of molecular dynamics method. Cl- and I-hydrotalcite - water systems were simulated for various mineral/water ratio. The structure and dynamic properties are predicted.

1 Introduction In engineering barrier systems for high level radio active waste disposal, smectite clays may be used for the major component of buffer materials. Smectite is expected to work as cation adsorbent together with some other functions, whereas semctite can not adsorb anions such as iodide ion. Hydrotalcite may be one of potential candidates of anion adsorbent. Hydrotalcite is a mineral name relating to brucite and gibbsite having a sheet structure of edge sharing octahedra (Table 1). Because of of the solid solution of Al and Mg in the octahedral site, and uncertainty of species and amounts of ions and molecule in the interlayer region, the crystal chemistry is not known completely yet. In this paper, the structure and properties of hydraotalcite and water in the vicinity of the surface are investigated by means of molecular dynamics method.

2 Molecular dynamics simulations (MD) and interatomic potential model The Ewald method was used for the summations of Coulomb interactions. Integration of equation of atom motions was performed by the velocity Verlet

algorithm with a time increment of 0.4 fs. The NVT (in cases of including vacuum space in the systems) and NPT ensembles were employed, where N is the number of atoms in a simulation cell, V the cell volume, T temperature, and P pressure. Temperature and pressure were controlled by scaling of atom velocities and cell edges. MD simulations were carried out at 293K and 0.1 MPa. The MD simulation codes, MXDORTO /MXDTRICL (Hirao and Kawamura 1994) were used. At least, a 100,000 step calculation was performed for the initial relaxation for each system. 100,000 to 2,000,000 step simulations for each system were carried out to obtain ensemble averaged properties. Interatomic potential model used in this study was the pair atom central force model with full freedom of atom motions, and described in Kumagai et al. (1994), Nakano et al. (2003), and Kawamura (2005). Three body force term was added for H2O molecule. Before the MD calculation of surface-water drop system, water drops in vacuum were prepared and equilibrated for 2000 to 5000 H2O molecule systems. Molecular dynamics calculations were started form the system containing a sheet of clay mineral and a water drop. In some cases, two clay sheets and two water drops systems were used for initial structure. The interatomic potential model used in this study is composed of two body central force and three body

Table 1. Hydrotalcite and related minerals. (Struntz Mineralogical Tables, 2001) Mineral Chemical formula System S.G. (a,b,c(Å),β(°)) Z Brucite Mg(OH)2 Trigonal P-3m1 (3.15 4.77) 1 Gibbsite γ-Al(OH)3 Monoclinic P21/n (8.66 5.07 9.72 94.5) 8 Bayerite α-Al(OH)3 Monoclinic P21/n (5.05 8.67 9.42 90.3) 8 Meixnerite Mg6Al2(OH)18.4H2O Trigonal R-3m (3.05 22.93) 3/8 Hydrotalcite Mg6Al2[(OH)16|CO3]4H2O Trigonal R-3m (3.05 22.81) 3/8 Manasseite Mg6Al2[(OH)16|CO3].4H2O Hexagonal P63/mmc (6.12 15.34) 1 Quintinite-2H Mg4Al2[(OH)12|CO3].3H2O Hexagonal P6322 (10.57 15.14) 4 Quintinite-3T Mg4Al2[(OH)12|CO3].3H2O Trigonal P3112 (10.80 22.71) 6

- 25 -

force terms. The two body terms represent Coulomb, van der Waarls, non-bonding repulsive, and radial covalent terms. The tree body term represents the angular part of covalent interaction. The parameters appeared in these formula, z, c, a, and b, for atoms, D1, D2, D3, β1, β2, β3, and r3 for atom pairs, fk, θ0, gr, and rm. for tree atoms, were presented for the molecular simulations of systems relating to H2O. CO32-, and clay minerals (Table 2).

3 Crystal structure of hydrotalcite Although the the rather simple chemical formula of hydrotalcite, Mg6Al2[(OH)16|CO3]4H2O, the relation between crystal structure and the chemistry has complex features. The crystal data of hydrotalcite of Allmann and Jepsen (1969) and Bellotto et al. (1996) are shown in Table 2. The chemical formula of these two report are different each other. Both of the crystal structure data have occupancy parameters for Al, Mg, CO32-, and H2O. For performing molecular dynamics calculations, all the atom positions in the “true” unit cell should be given properly. These data can not be used for MD calculations as they stand. When Mg:Al=3:1, the octahedral sheet may have ordered structure. Fig 1. is one of the ordered model, and another models is that of Al at 3 and Mg at 2. When Mg:Al=2:1, the structure include adjoining of Al octahedra. In our model, the unit cell was taken as 0.61 nm x 1.06 nm x 2.28 nm which is 8 times larger than the reported ones (Z=3 for Mg6Al2[(OH)16|CO3]4H2O). We calculated for three types of hydrotalcite, CO32-, Cl-, and I-hydrotalcites. The (NPT)-MD calculations of model hydrotalcite at

Fig. 1: Unit structure of hydrotalcite octahedral sheet. 293K and 0.1MPa, shows the average unit cell parameters of a=3.0693, c=23.218 giving the density of 2.00362x103 kg/m3 which is well reproducing the reference parameters. The structure snapshot is shown in Fig. 2. The octahedral sheet structure (horizontal 3 layers) is stably exist, and interlayer sheet also formed with CO32- and H2O molecules. The single sheet of interlayer structure is shown in Fig. 3. While the CO32- units line regularly, the arrangement of H2O molecules is irregular.

4 Wetting on hydrotalcite surface Wetting of water on the Cl- and I-hydrotalcite surface was investigated using molecular dynamics calculations. A water drop was put just in the vicinity of the surface, and MD calculations ware

Table 2. The crystal data of hydrotalcite reported by Allmann and Jepsen (1969) and Belloto et al. (1996). SOF: site occupancy factor, ITF: isometric temperature factor. ((Mg4,Al2)(OH)12(CO3)(H2O)3)0.5 Allmann and Jepsen (1969) Atom SITE x y z SOF Trigonal Mg 3a 0 0 0 0.6667 R-3m Al 3a 0 0 0 0.3333 a=3.054 A O 6c 0 0 0.3771(5) 1. b=3.054 A H 6c 0 0 0.4144(86) 1. c=22.81 A H 6c 0 0 0.4144(86) 1. Z=3 O 18h 0.092(25) -0.092(25) 0.5 0.167 C 6c 0 0 0.167 0.083 (Mg0.833,Al0.167)(OH)2(CO3)0.083(H2O)0.75 Bellotto et al. (1996) Atom SITE x y z SOF ITF(U) Trigonal Mg 3a 0 0 0 0.8333 0.0319(8) R-3m Mg 3a 0 0 0 0.8333 0.0319(8) a=3.0808(3) A Al 3a 0 0 0 0.1667 0.0319(8) b=3.0808(3) A O 6c 0 0 0.3754(2) 1. 0.036(1) c=23.784(4) A H 6c 0 0 0.432(1) 1. 0.059(7) Z=3 O 18h 0.1260(9) 0.8740(9) 0.5 0.1667 0.021(3) C 6c 0.3333 0.6667 0.5 0.0416 0.059(7) H 6c 0.3333 0.6667 0.5 0.75 0.059(7)

- 26 -

Fig. 4: Structure snap shots of sheet of Cl-(upper) and I-hydrotalcite (lower) surface – water drop systems, where the number of H2O molecules are 3000 for both systems.

5 Local structure and properties of water between hydrotalcite surfaces

Fig. 2: Snap shot of structure of hydrotalcite, Mg6Al2[(OH)16|CO3]4H2O by MD simulation.

Hydrotalcite surface – water systems ware used to investigate the local structure, density, diffusion coefficient, and viscosity of water as functions of distance from hydrotalcite surface. First, we investigate the cases that relatively small amounts of H2O are inserted in interlayer regions of hydrotalcie. Mg3Al(OH)8[Cl or I].nH2O where n=1 to 20 are investigate by MD calculations.The structure of these two series are slightly different because of the difference of anion size. In the structure of n=2 and 3 in I-hydrotalcite, the raw of H2O molecules are clearly separated in the single interlayer region.wo dimensional diffusion coefficients of anions and H2O molecule were calculated (Table 3.) The diffusion

Fig. 3: The structure of interlayer CO32- and H2O from middle layer of Fig.2. started. The water drop is composed of 3000 H2O molecules. The structural snap shots of steady state structures are displayed in Fig. Both Cl- and Ihydrotalcite show good wetting behavior where wet angle s are about 30 degrees. In both structure, anions attach the hydrotalcite sheet surfaces directly unlike the case of smectite surfaces where ions are fully hydrated.

Table 3. Two dimension diffusion coefficients Dxy of anion and H2O molecule in interlayer region of Cl- and I-hydrotalcite at 293K, 0.1 MPa. The nH2O means Mg3Al(OH)8[Cl or I].nH2O nH2O diffusion coefficient / cm2/s 1H2O Cl 2.7E-7 H2O 9.1E-7 I 17.7E-7 H2O 45.3E-7 2H2O Cl 42.9E-7 H2O 69.1E-7 I 48.6E-7 H2O 74.7E-7 3H2O Cl 26.3E-7 H2O 47.0E-7 I 52.9E-7 H2O 134.1E-7 10H2O Cl 94.2E-7 H2O 165.2E-7 I 88.1E-7 H2O 179.6E-7 20H2O Cl 108.3E-7 H2O 162.8E-7 I 91.4E-7 H2O 180.6E-7

- 27 -

hydrotalcite – intergranular water systems, and calculated local density, diffusion coefficient, and viscosity of water. The behavior of surface water resembles to that of brucite surface.

References

Fig. 5: Structure Mg3Al(OH)8.Cl.nH2O Mg3Al(OH)8.I.nH2O (right).

snapshots (left)

of and

coefficients increase with n for both anions and H2O molecule.T In the systems of 10H2O and 20H2O, the diffusion coefficients of H2O molecules are almost same each other and same with bulk water. As the hydrotalcite – intergranular water systems, we simulated Mg3Al(OH)8[Cl or I]-nH2O, n=100 and 200 systems. The results for the Mg3Al(OH)8Cl-200H2O system shows that electrical double layer was formed by one molecular layer of surface water where the molecular orientations are strongly biased. The density profile of water show the decrease at the surface two molecular layers of water unlike the case of smectite. Diffusion coefficients of water shows the maximum at the surface, and decrease with distance from the surface to three molecule distance, and increase again with distance. The viscosity shows the reverse behavior to that of diffusion coefficient. The behavior is very similar to that of brucite (Sakuma, et al., 2003).

Allmann, R, Jepsen, H.P (1969) Die Struktur des Hydrotalkits. Neues Jahrbuch fuer Mineralogie. Monatshefte (Band=Jahr), 1969, 544-551 Bellotto, M. Rebours, B. Clause, O. Lynch, J. Bazin, D.;Elkaim, E. (1996) A reexamination of hydrotalcite crystal chemistry. Journal of Physical Chemistry, 100, 8527-8534 Hirrao, K, Kawamura, K, 1994, “Material Design using Personal Computers” (Japanese), Shokabou, 217p. Kawamura, K. (2005) Molecular H2O model with total freedom of motion applied to molecular simulations of water, ice, etc. (Japanese) Low Temperature Science, 65, 3-11. Kumagai, N., Kawamura, K. and Yokokawa, T. (1994) An interatomic potential model for H2O systems and the molecular dynamics applications to water and ice polymorphs, Molecular Simulation 12, pp177- 186. Nakano, M., Kawamura, K., and Ichikawa, Y. (2003) Local structural information of Cs in smectite hydrates by means of an EXAFS study and molecular dynamics simulations. Applied Clay Science 23, pp15-23 Sakuma, Hiroshi, Taku Tsuchiya, Katsuyuki Kawamura, Kenshiro Otsuki (2003) Large self-diffusion of water on brucite surface by ab initio potential energy surface and molecular dynamics simulations. Surface Science, 536 (2003) L396-L402 (2003) Struntz Hugo. and Nickel Ernest H. 2001, “Structz Mineralogical Tables. Chemical-Structural Mineral Classification System”, 9th ed. Stuttgart, Schweizerb

Conclusion Crystal structure of hydrotalcite was reasonably reproduced using molecular dynamics simulations with the appropriate interatomic potential model of hydrotalcite sheet, H2O molecule, halogen anion, and CO32- ion. Wetting of water drop on hydrotalcite surfaces show fairly good wetting behavior with wetting angle ca. 30 degrees. We performed

- 28 -

Scale separation in convection-driven MHD dynamos at low Ekman number Futoshi Takahashi1, Masaki Matsushima2, and Yoshimori Honkura3 1. Post Doctoral Fellow, 2. Assistant Professor, 3. Professor 1, 2, 3. Research Center for the Evolving Earth and Planets, Also at Department of Earth & Planetary Sciences, Graduate School of Science & Engineering, Tokyo Institute of Technology e-mail: [email protected], [email protected], [email protected]

We have examined convection-driven MHD dynamos in a rapidly rotating spherical shell with the Ekman number, E, down to 2 × 10 and the magnetic Prandtl number, Pm, down to 0.2, focusing on the characteristic length scales of the flow and the magnetic field. Smaller-scale convection vortices responsible for generating the magnetic field appear at lower Ekman numbers, whereas the scale of the magnetic field less varies in comparison with the flow. As a result, we have found scale separation between the flow and the magnetic field as the Ekman number is decreased. Scale separation helps dynamos to maintain the magnetic field at Pm < 1 through increase in the effective value of the magnetic Reynolds number. Scaling laws for the viscous and Ohmic dissipations are derived from the scales of the flow and magnetic field, and these dissipations in the Earth’s core are estimated through extrapolation of the scaling laws.

1 Introduction The intrinsic magnetic field of the Earth is produced by a dynamo action working in the metallic fluid outer core. Considerable numerical simulations and laboratory experiments have been carried out to investigate the generation mechanism of the Earth’s magnetic field since the last decade. The Ekman number (E = ν/ΩL2, where ν is the kinematic viscosity, Ω the rotation rate of the mantle, and L the thickness of the outer core) in the Earth’s core is very low, e.g. E ~ 10−9, even for the eddy diffusivity, νT ~ 1 m2/s. Such a low value of E indicates that very small-scale columnar vortices dominate the fluid motion in a rotating spherical shell. Therefore it is still impossible to attain the realistic value of the Ekman number even in recent numerical simulations. The same argument is applicable to the magnetic Prandtl number (Pm = ν/η, where η is the magnetic diffusivity). It should be noted, however, that the threshold of Pm for dynamo action strongly depends on E (Christensen et al. 1999), while the properties of the dynamo are nearly independent of Pm, because the magnetic field strength is determined by available power input (Christensen & Aubert 2006). Decrease in the Ekman number leads to the dominance of smaller-scale convection. However, how the scale of the magnetic field depends on the Ekman number is uncertain. We investigate how dynamo action varies with the Ekman number

focusing on the scales of the flow and the magnetic field in the core. For this purpose we take advantage of our extensive set of MHD dynamo simulation results with E ranging from 2 × 10−6 to 2 × 10−4.

2 Model and Numerical Method We consider time-dependent, three-dimensional magnetic field generation by thermal convection in a rotating spherical shell filled with a Boussinesq fluid of constant density, ρ, and electrical conductivity, σ. The ratio of the inner to the outer radius, ri/ro, is fixed at 0.35 (Fig. 1). The angular velocity of the shell is aligned with the z-axis, Ωez, where ez is the unit vector in the z-direction. Thermal convection is driven by fixed temperature difference, ∆T, between the inner and outer boundaries. Using the following scales as the thickness of the

- 29 -

Fig. 1: Schematic picture of the model.

spherical shell L = ro – ri for length, L2/ν for time, ν/L for velocity u, ∆T for temperature, and (ρΩ/σ)1/2 for the magnetic field B, we solve the equation of motion, the induction equation, and the heat transport equation with the equations of continuity for the velocity and magnetic fields in non-dimensional form. The solid inner core and the mantle are treated as electrical insulators and assumed to be in co-rotation. No-slip boundary condition for the velocity and fixed temperature boundary conditions are adopted at the inner and outer spherical shells. Non-dimensional parameters are the Ekman number, E, the magnetic Prandtl number, Pm, the Prandtl number, Pr = ν/κ, and the Rayleigh number, Ra = αgo∆TL3/κν, where κ is the thermal diffusivity, α the thermal expansion coefficient, and go gravitational acceleration rate at the outer boundary. We fix the value of Pr at unity for simplicity and to make comparison with the previous studies easy, while we vary other three parameters, E, Pm and Ra: E = 2 × 10−4, 8 × 10−5, 2 × 10−5, 8 × 10−6, and 2 × 10−6, Pm = 0.2–2, and Ra up to 30 times supercritical are used for calculations. Spherical harmonic expansion is made up to degree and order lmax = 255. The maximum numbers of grid points in the radial, latitudinal and longitudinal directions are (nr, nθ, nφ) = (256, 384, 768) based on 3/2-rule of the grid points in the angular directions for dealiasing. The radial grid points are distributed non-uniformly so that they concentrate within thin boundary layers formed around the inner and outer boundaries with the thickness of E1/2 (Takahashi et al. 2001).

3 Results We have obtained 27 new successful dynamos out of 40 simulations in addition to 12 dynamos reported before (Takahashi & Matsushima 2005; Takahashi et al. 2005). The dipole magnetic field is dominant in most dynamo solutions, while its fraction is dramatically reduced and non-dipole fields are stronger in high-Ra dynamos at E = 2 × 10−4, 8 × 10−5, and 2 × 10−5 (Takahashi & Matsushima 2005). Fig. 2 shows the flow and magnetic field structures in the equatorial plane. In Fig. 2(a), locations of negative axial vorticity in the equatorial plane well correlate with five spots of the magnetic field, which are generated through flux concentration due to flow convergence there. On the other hand, as shown in Fig. 2(c), the strong magnetic field is confined near the inner boundary of the spherical shell and the magnetic field structure is nearly axisymmetric on the equatorial plane, while many columnar convection vortices distribute over the fluid

Fig. 2: The axial components of (left) the vorticity and (right) the magnetic field in the equatorial plane as viewed from the north. (a) for E = 2×10−4, Pm = 2, Ra = 1.5×106, (b) for E = 2×10−5, Pm = 1.5, Ra = 2×107, and (c) for E = 2×10−6, Pm = 0.5, Ra = 5×108 (modified from Takahashi et al. 2008).

shell. The flow and magnetic field structures in Fig. 2(b), showing about eight flux spots correlating with negative axial vorticity, seem to be in the intermediate between those in Fig. 2(a) and 2(c). The flow and the magnetic field tend to show less correlation with decreasing E, suggesting scale separation at low E and low Pm. Next we estimate the mean harmonic degree and order, as characteristic length scales, for the flow and the magnetic field. The mean values for the velocity field are calculated from lu=Σllul2/2Ek and mu = Σmmum2/2Ek, where ul2 and um2 are |u|2 per unit volume at degree l and order m, respectively. The kinetic energy density in the spherical shell, Ek, is given by Ek = (2V)−1∫Vu2dV, where V is the volume of the spherical shell. The mean harmonic degree and order for the magnetic field are calculated in a similar way in terms of the magnetic energy density Em = (2EPmV)−1 ∫VB2dV. Fig. 3 shows lu and lb for the successful dynamos as a function of the supercriticality of the Rayleigh number, Ra/Rac, where Rac is the critical Rayleigh number. The mean harmonic degree for the flow, lu, systematically increases with decrease in the Ekman number. It is nearly flat with respect to Ra/Rac for E = 2×10−4, 8×10−5 and 2×10−5, whereas it slightly increases for E = 8×10−6 and 2×10−6. On the other

- 30 -

Fig. 3: The mean values of the wavenumber in terms of spherical harmonic degree l based on (a) the kinetic energy spectrum lu, and (b) the magnetic energy spectrum lb, as a function of the supercriticality of the Rayleigh number Ra/Rac for E = 2×10−4 (inverted triangle), E = 8×10−5 (circle), E = 2×10−5 (triangle), E = 8×10−6 (square), and E = 2×10−6 (gray diamond). Reversing and non-dipolar dynamos are denoted by symbols filled with black (modified from Takahashi et al. 2008).

hand, the mean harmonic degree for the magnetic field, lb, varies only a little with E. Steep increase of lb with respect to Ra/Rac corresponds to transition from the stable dipolar dynamo regime to the reversing dynamo regime. Takahashi & Matsushima (2005) introduced the effective magnetic Reynolds number Rm* adopting different length-scales for the flow and the magnetic field instead of the magnetic Reynolds number, Rm = Lu/η = Pm(2Ek)1/2: Rm* =

| (B ⋅ ∇)u | | η∇ B | 2

~ Rm

Lb Lb L Lu

where Lu and Lb are the length scales of the flow and the magnetic field, respectively. It is clear from the definition that Rm* is large when the length scale of the magnetic field is much larger than that of the flow. We calculate Rm* using π/lu as Lu and π/lb as Lb and show in Fig. 4(a) as a function of Ra/Rac. It is clear that Rm* is raised with decreasing E in the regime of dipolar dynamo. Large value of Rm* enables us to obtain dynamos at low Pm. Since Rm* is proportional to Lb/Lu, that is, to lu/lb, scale separation is an important factor to obtain dynamos at low Pm. Qualitatively similar plot is obtained by using the mean harmonic order for the scales instead of degree, as shown in Fig. 4(b). We derive scaling for the viscous and Ohmic dissipations over the spherical shell respectively given by Dv = V−1∫V(curl u)2dV and Do = (EPm2V)−1 ∫V(curl B)2dV on the basis of the scales for the flow and magnetic field. We adopt δm = 1/mu as the characteristic length of the flow and the Reynolds number, Re = (2Ek)1/2 as the flow. Then the scaling for Dv is written as Dv ~ (Re/δm)2, which can be checked in Fig. 5(a). In fact, the linearity is very good for all the dynamos with wide variety of parameters. Similarly, we derive a scaling for Do by adopting δm =

Fig. 4: The effective magnetic Reynolds number based on (a) l and (b) m as a function of the supercriticality of the Rayleigh number Ra/Rac. Symbols are the same as in Fig. 3 (modified from Takahashi et al. 2008).

1/mb and the Elsasser number, Λ = 2EPmEm as the square of the magnetic field; that is, Do ~ Λ/(EPm2δm2). Fig. 5(b) shows the good linearity. These suggest that the characteristic dissipation length scale is proportional to the flow and magnetic field scales.

4 Discussion and Concluding Remarks We have obtained self-sustaining dynamos with the Ekman number ranging over two orders of magnitude. The strong rotational constraint on the flow imposed by a low Ekman number prefers small-scale columnar convection, whereas the scale of the magnetic field is not as much varied with the Ekman number as the flow scale. The magnetic field induced by the small-scale flow rapidly diffuses, and thus the scale of the magnetic field does not decrease as the flow scale does. This is obviously due to low magnetic Prandtl number. That is, at low magnetic Prandtl number the magnetic field is smoothed at scales of small eddies in the flow, whereas at high magnetic Prandtl number advection in small eddies can create magnetic field structures that may even finer than the eddy scale. As a result, scales are separated between the flow and the magnetic field. An aspect of scale separation is described by the effective magnetic Reynolds number, which is proportional to the scale ratio. Under the scale separation condition at a low Ekman number, the effective magnetic Reynolds number is large enough to sustain the magnetic field even at a low magnetic Prandtl number. With insufficient scale separation at a moderate Ekman number, dynamo action cannot maintain the magnetic field at a low magnetic Prandtl number, even if the conventional magnetic Reynolds number is seemingly large enough for successful dynamo. Assuming that the present results can be extrapolated to the condition of the Earth’s core, we attempt to estimate the viscous and Ohmic dissipations in the Earth’s core using the scale separation approach. We have E ~ 10−9 and Pm ~ 1 for eddy diffusivities in the core, and take mu = 100 – 500.

- 31 -

(a) Dv

separation helps dynamos to maintain the magnetic field at low Pm with relative ease through increase in the effective magnetic Reynolds number.

(b) Do

Acknowledgements The numerical simulations were carried out on the Earth Simulator under support of JAMSTEC and SX-6 at Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency. FT was supported by the Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists. This work was partly supported by the Japan Society for the Promotion of Science under the Grant-in-Aid for Scientific Research, No. 18340131.

Fig. 5: (a) The viscous dissipation, Dv, as a function of the scaling parameter (Re/δm)2. The solid line represents the best-fit line Dv = 3.2(Re/δm)2. (b) The Ohmic dissipation, Do, as a function of the scaling parameter Λ/(EPm2δm2). The solid line represents the best-fit line Do = 8.8(Λ/EPm2δm2). All symbols are the same as in Fig. 3 (modified from Takahashi et al. 2008).

The kinematic Reynolds number, Re is about 500 using a relation Re = Rm/Pm. Then we obtain Dv ~ 3.2×5002×(1002 – 5002) ~ (0.08 – 2) ×1011. As for the Ohmic dissipation, the magnetic field scale in terms of the harmonic order does not vary systematically with the Ekman number. We then roughly take mb ~ 10. Since a likely estimate of the Elsasser number in the core is Λ = 0.1 – 1, we have Do ~ 8.8×(0.1 – 1) ×102×109 ~ (0.9 – 9) ×1011. These estimates leave the possibility that the Ohmic dissipation is reasonably large to neglect the viscous one in the core, or that they are comparable. In the dimensional form, we obtain total values for the viscous and Ohmic dissipations in the core as 0.06 – 1 TW and 0.5 – 5 TW, respectively. This estimate is not inconsistent compared with the estimates by Roberts et al. (2003) and Christensen & Tilgner (2004), who respectively give 1 – 2 TW and 0.2 – 0.5 TW for the Ohmic dissipation. It should be noted, however, that applying these estimates to the geodynamo is speculative at present, because this argument relies on assumptions that are not rigorously proved yet. More improvement is inevitably required for application to the geodynamo. We have found that effects of low magnetic Prandtl number on dynamo action keep the magnetic field structure large-scale compared with the flow scale. That is, scale separation is a result of combined effects of decreasing the Ekman and magnetic Prandtl numbers. Although much wider parameter survey with respect to Pm is needed to confirm the robustness of the Pm-dependence, trends similar to the present results are found in data base of Christensen & Aubert (2006), which covers broader range of the magnetic Prandtl number. In conclusion, as the Ekman number is decreased, the flow scale becomes smaller, whereas the magnetic field scale shows less variation compared with the flow. As a result, scales of the flow and magnetic field are separated at low Ekman number. This scale

References Christensen, U. R., & Aubert, J., 2006. Scaling properties of convection-driven dynamos in rotating spherical shells and application to planetary magnetic fields. Geophys. J. Int. 166, 97−114. Christensen, U. R., & Tilgner, A., 2004. Power requirement of the geodynamo from ohmic losses in numerical and laboratory dynamos. Nature 429, 169−171. Christensen, U. R., Olson, P., & Glatzmaier, G. A., 1999. Numerical modelling of the geodynamo: A systematic parameter study. Geophys. J. Int. 138, 393−409. Roberts, P. H., Jones, C. A., & Calderwood, A. R., 2003. Energy fluxes and ohmic dissipation in the earth’s core. In: Jones, C. A., Soward, A. M., & Zhang, K. (Eds.), Earth’s Core and Lower Mantle, Taylor & Francis, London, pp. 100−129. Takahashi, F., Katayama, J. S., Matsushima, M., & Honkura, Y., 2001. Effects of boundary layers on magnetic field behavior in an MHD dynamo model. Phys. Earth Planet. Inter. 128, 149−161. Takahashi, F., & Matsushima, M., 2005. Dynamo action in a rotating spherical shell at high Rayleigh numbers. Phys. Fluids 17, 076601. Takahashi, F., Matsushima, M., & Honkura, Y., 2005. Simulations of a quasi-Taylor state geomagnetic field including polarity reversals on the Earth Simulator. Science 309, 459−461. Takahashi, F., Matsushima, M., & Honkura, Y., 2008. Scale variability in convection-driven MHD dynamos at low Ekman number. Phys. Earth Planet. Inter. 167, 168−178.

- 32 -

Ocean Bottom Magnetotelluric Measurements for Imaging of the North Anatolian Fault Zone at the Marmara Sea, Turkey Yasuo Ogawa Professor Research Center for the Evolving Earth and Planets, Also at Volcanic Fluid Research Center, Tokyo Institute of Technology e-mail: [email protected]

Fluid distribution in the mid crust plays an important role in crustal deformation and earthquake generation process. Electrical resistivity is one of the most sensitive physical parameters for existence and connectivity of fluids. Magnetotelluric method is a natural electromagnetic method to image the resistivity structure of the crust and upper mantle and thus is now widely used to study fluids for the seismogenesis. The North Anatolian Fault in Turkey is a 1200km long right-lateral strike-slip fault and westward migrations of rapture of fault segments are known in the last 60 years. The most recent large earthquakes took place at Izmit and Durce in 1999, and future raptures are expected further west, i.e., under the Marmara Sea. In our experiment, we deployed newly-developed portable OBEM (ocean bottom electromagnetic) instruments across the fault trace offshore Istanbul. We had successful recovery of 7 stations and data processing and resistivity modeling are underway. Together with future deployments in the Marmara Sea, we aim to image the fault segmentation and asperity as a resistive body.

1 Introduction to Electromagnetic Studies in Active Deformation Zones Magnetotelluric (MT) method can image resistivity structure of the crust, which is mainly controlled by the existence and connectivity of the conductive fluids in the pore spaces, fractures and grain boundaries rather than by the host rock resistivity itself. Thus the MT method has been successfully used for imaging the damaged zones in the upper crust (Unsworth et al., 1999; Ritter et al., 2003) and for imaging the fluid distribution at the mid-crustal depth (e.g., Ogawa et al. 2001, 2002; Mitsuhata et al. 2001; Tank et al. 2003, 2005; Ogawa and Honkura 2004). Distribution of fluids is important in the framework of the earthquake generation processes (e.g., Sibson et al., 1988; Iio and Kobayashi, 2002). The brittle-ductile boundary can be imaged mechanically by the cut-off depth of the earthquakes, which is controlled by the geotherm (e.g., Ito 1999). On the other hand, many magnetotelluric studies have shown that the brittle ductile boundary often corresponds to the top of the mid-crustal conductors (e.g., Ogawa et al. 2001; Mitsuhata et al. 2001; Ogawa & Honkura 2004) and have suggested that the fluids distribute under the brittle-ductile boundary in the seismically active zones. The existence of the fluids at the brittle-ductile transition is also directly supported by the geological evidences at the exhumed deep crustal sections (Cox 2002; Fujimoto et al. 2002).

2 The North Anatolian Fault Zone and previous studies

The main strand of the North Anatolian fault Zone (NAFZ) in Turkey is a 1,200 km long right–lateral strike–slip fault that extends in the east–west direction, separating the Anatolian block from the Eurasian plate. The events of a series of large earthquakes have been propagating westwards along NAFZ since 1939. The western part of NAFZ breaks into several strands. For the recent Izmit earthquake in 1999, the rupture extended over 120 km in the east–west directions. The crustal and upper mantle resistivity structure of the focal area of the Izmit earthquake was investigated intensively by wide–band and long–period magnetotelluric measurements (Tank et al. 2003, 2005). The important features of the resistivity model were (1) that the hypocenter was located at the resistive side of the resistive–conductive boundary and (2) that there was a distinct conductor in the lower crust to upper mantle depth beneath the fault zone. Assuming the conductor as a ductile shear zone and the resister as a brittle “asperity”, the stress will be accumulated at the resistive–conductive boundary and the edge of the resistive block may start to rupture. Further westward migration of the earthquakes along the NAFZ is expected towards Istanbul. Then the future earthquakes will be off-shore in the Marmara Sea. The architecture of NAFZ at the Marmara Sea remains controversial and deep crustal structure is not well resolved. Based on the GPS study, Le Pichon et al.(2003) claim that there is one fault segment in the Marmara Sea. On the other hand, Sato et al.(2004) showed along-strike variation of fault mechanical properties. The western part of the Marmara Sea has strike-slip fault mechanism while that of the eastern one as normal fault component in

- 33 -

addition. In this study, we aimed at imaging the asperities and segmentations in terms of resistivity. We will have multiple profiles across the fault traces and will have a three-dimensional resistivity image under the Marmara Sea. We will see the fault segmentations (as along-strike resistivity variations) and distribution of asperities (as resistive blocks).

Fault system in Sumatra, Indonesia (Nurhasan et al., 2008). Comparative electro-magnetic studies of these 1000km long strike-slip systems will be another future work.

3 Ocean Bottom Magnetotelluric Measurements and Data Processing In February to March 2008, we deployed 5 units of ocean bottom electromagnetic (OBEM) equipments (Kasaya et al. 2006) along profiles which cross the ANFZ off-shore in the eastern part of the Marmara Sea as shown in Fig.1. The OBEMs have three-component fluxgate sensors and four-component (where two of them are redundant) electrode arms (Fig. 2). The bathymetry ranges of the OBEMs were from 70m to 1225m. We had 100% recovery of the instruments and data. At each site, OBEM recorded the electromagnetic data continuously with 8Hz sampling for three-four weeks. Preliminary robust time series analyses using Chave’s code produced impedance estimates in the period range of 50s-20,000s. With the typical apparent resistivity of the order of 1-10ohmm, the corresponding skin depths are 3km-300km. The sites near the sharp fault scarp, where there is a sharp 1000m gap in bathymetry, the sounding curves seem heavily disturbed by the topography.

Fig. 1: Distribution of OBEM (ocean bottom electromagnetic) sites. The blue and red stars denote stations deployed during February to March 2008 and during March to April 2008, respectively. The bathymetry ranges from 70m to 1225m. The fault trace is a sharp cliff in the northern part of the Marmara Sea.

4 Preliminary Results and Perspectives Fig. 3 shows the data as pseudo-sections using the five sites in the NE-SW directions. The upper two figures are apparent resistivity and phase pseudo-sections for TM mode, whereas the lower two ones are those for TE mode. The most important feature is the low resistivity detected by TE mode (as C1 and H1) at periods longer than 1,000 seconds. Detail structural features still needs to wait for two-dimensional and three-dimensional numerical modeling studies. The suggested conductor will be a localized conductor at mid-crust to upper mantle depth. We also deployed on–shore MT stations using wide-band and long–period instruments. These new measurements will elucidate the asperity and shear zones of the NAFZ in the Marmara Sea near Istanbul. Our group is also engaged in other electromagnetic imaging of 1000km-long strike slip faults systems. One is the Alpine Fault system in Marlborough, South Island, New Zealand (Wannamaker et al., 2008). The other is a Sumatra - 34 -

Fig. 2: Launching operation of the OBEM from the boat. The four long plastic bars have electrodes at their ends to measure orthogonal electric fields. The magnetometer is attached to the yellow plastic case, in which a glass sphere contains the amplifier and recorder. The Bottom gray plastic plate is a load, which will be detached by electric erosion for the recovery.

Fig. 3: Pseudo-sections for the five sites in NE-SW directions. The upper two figures are apparent resistivity and phase pseudo-sections for TM mode, whereas the lower two ones are those for TE mode.

Acknowledgements The author thanks T. Kasaya, S.B. Tank, M.K. Tuncer, Y. Honkura,T. Kaya, E. Tolak, N. Oshiman, M. Matsushima1, and S. Nagaoka. This research was supported by the Ministry of Education, Science, Sports and Culture, Grant-in-Aid for Scientific Research (A) 19253002.

References Cox, S., 2002. Fluid flow in mid- to deep crustal shear systems: Experimental constraints, observations on exhumed high fluid flux shear system, andimplications for seismogenic processes, Earth Planets Space, 54, 1121–1125. Fujimoto, K., T. Ohtani, N. Shigematsu, Y. Miyashita, T. Tomita, H.Tanaka, K. Omura, and Y. Kobayashi, 2002. Water-rock interaction observed in the brittle-plastic transition zone, Earth Planets Space, 54, 1127–1132. Iio, Y. and Y. Kobayashi, 2002. A physical

understanding of large intraplate earthquakes, Earth Planets Space, 54, 1001–1004. Ito., K., 1999. Seismogenic layer, reflective lower crust, surface heat flow and large inland earthquakes, Tectonophysics, 306, 3-4, 423-433. Kasaya T., T. Goto, and R.Takagi 2006. Marine electromagnetic observation technique and its development-For crustal structure survey, Butsuri-Tansa, 59, 585-594 (in Japanese). Le Pichon, X., N. Chamot-Rooke, C. Rangin, and A. M. C. Sengör, 2003, The North Anatolian fault in the Sea of Marmara, J. Geophys. Res., 108(B4), 2179, doi:10.1029/2002JB001862. Mitsuhata, Y., Y. Ogawa, M. Mishina, T. Kono, T. Yokokura and T. Uchida, 2001. Electromagnetic heterogeneity of the seismogenic region of 1962 M6.5 Northern Miyagi Earthquake, northeastern Japan, Geophy. Res. Lett., 28, 4371–4374. Nurhasan, Y. Ogawa, D. Sutarno, D. S Widarto, D. Sugiyan, 2008. Imaging fault zone by MT Method at Sumatra area in Indonesia, presented at 19th Electromagnetic Induction Workshop, Beijing, China. Ogawa Y., and Y. Honkura, 2004. Mid-crustal electrical conductors and their correlations to

- 35 -

seismicity and deformation at Itoigawa-Shizuoka Tectonic Line, Central Japan, Earth Planets Space, 56, 1285-1291. Ogawa, Y., M. Mishina, T. Goto, H. Satoh, N. Oshiman, T. Kasaya, Y. Takahashi, T. Nisitani, S. Sakanaka, M. Uyeshima, Y. Takahashi, Y. Honkura, and M. Matsushima, 2001. Magnetotelluric imaging of fluids in intraplate earthquakes zones, NE Japan back arc, Geophy. Res. Lett., 28, 3741–3744. Ogawa, Y., S. Takakura, and Y. Honkura, 2002. Resistivity structure across Itoigawa-Shizuoka tectonic line and its implications for concentrated deformation, Earth Planets Space, 54, 1115–1120. Ritter O., T. Ryberg, U. Weckmann, A. Hoffmann-Rothe, A. Abueladas, and Z. Garfunkel 2007. Geophysical images of the Dead Sea Transform inJordan reveal an impermeable barrier for fluid flow, Goephys. Res. Lett., 30, Article Number: 1741, 2003. Sato T, Kasahara J, Taymaz T, et al., 2004. A study of microearthquake seismicity and focal mechanisms within the Sea of Marmara (NW Turkey) using ocean bottom seismometers (OBSs), Tectonophysics, 391, 1-4, 303-314. Sibson, R. H., F. Roberts, and K. H. Paulson, 1988. High-angle reverse faults, fluid-pressure cycling and mesothermal gold deposits, Geology, 16, 551–555. Tank, S. B., Y. Honkura, Y. Ogawa, M. Matsushima, N. Oshiman, M. K. Tuncer, C. Celik, E. Tolak, and A. M. Isikara, 2005. Magnetotelluric imaging of the fault rupture area of the 1999 Izmit (Turkey) earthquake, Phys.Earth Planet Interior, 150, 213-225. Tank, S. B., Y. Honkura, Y. Ogawa, N. Oshiman, M. K. Tuncer, M. Matsushima, C. Celik, E. Tolak, and A. M. Isikara, 2003. Resistivity structure in the western part of the fault rupture zone associated with the 1999 ˙Izmit earthquake and its seismogenic implication, Earth Planets Space, 55, 437–442. Unsworth, M., G. Egbert, and J. Booker, 1999. High-resolution electromagnetic imaging of the San Andreas fault in Central California, J. Geophys. Res., 104, 1131–1150. Wannamaker, P.E. , T.G. Caldwell, G.R. Jiracek, V. Maris, G.J. Hill, Y. Ogawa, H.M. Bibby, S.L. Bennie and W. Heise, 2008. Strain localisation and ductile shear zones: Implications of magnetotelluric results from the Marlborough Fault system, New Zealand, presented at 19th Electromagnetic Induction Workshop, Beijing, China.

- 36 -

High pressure stress measurement using Kawai-type multi-anvil apparatus combined with synchrotron radiation and its application to stress relaxation test of olivine under Earth’s deep upper mantle conditions Yu Nishihara Assistant Professor Research Center for the Evolving Earth and Planets, Also at Department of Earth & Planetary Sciences Graduate School of Science & Engineering, Tokyo Institute of Technology, Now at Senior Research Fellow Center, Ehime University e-mail: [email protected]

We have recently developed a system for stress measurement under high pressure at a beam line BL04B1, SPring-8, Japan. In the system, sample is pressurized with Kawai-type multi-anvil apparatus SPEED-1500, and stress in the sample is determined from lattice distortion measured by two-dimensional X-ray diffraction using monochromatic synchrotron X-ray. Using the system, polycrystalline KCl was pressurized to 9.9 GPa using mechanically anisotropic cell assembly in which KCl is sandwiched between Al2O3 pistons, and variation of deviatoric stress was measured during compression. Deviatoric stress increased with increasing pressure, and phase transition in KCl (from B1 to B2 at pressure of 2.3–2.4 GPa) found to reduce the deviatoric stress significantly. We have also carried out stress relaxation test on polycrystalline olivine at 6.6–9.0 GPa and up to 1273 K using a similar cell assembly. Relaxation of stress in olivine was clearly observed. The application of this new system to precise experimental studies on deep Earth rheology is expected.

1 Introduction Knowledge of rheological property of the deep Earth material is important for understanding of dynamics in the Earth’s interiors. However, due to technical difficulty of precise experiments at high pressure, not much is known about rheological property in the Earth’s deep interiors. Among upper mantle minerals, olivine is the most abundant phase and is considered to be the weakest phase. Therefore, rheological property in the upper mantle is most probably dominated by rheological property of olivine (e.g. Karato 2008). However, effect of pressure on the rheological property is still controversial (e.g. Li et al. 2006; Durham et al. 2009; Kawazoe et al. 2009). Recently, we have developed a system for stress measurement under high pressure at a beam line BL04B1, SPring-8, Japan. In the new system, sample is pressurized with a Kawai-type multi-anvil apparatus SPEED-1500 and stress in the sample is determined from lattice distortion measured by two-dimensional X-ray diffraction using monochromatic synchrotron X-ray. The strain measurement can be done by X-ray radiography simultaneously. Since achievable maximum pressure using Kawai-type apparatus is ~30 GPa (e.g. Wang and Takahashi 2000), this new system enables us to conduct precise rheological measurement up to lower mantle condition (> 23 GPa). In this study, by using the high pressure stress

measurement system, we have conducted two series of experiments: (1) high pressure stress measurement of polycrystalline KCl at room temperature and (2) stress relaxation test of olivine at high pressure and high temperature. The experiment (1) was conducted in order to test ability of the new system. The experiment (2) was conducted to investigate pressure dependence of rheological property of olivine in the deep upper mantle conditions.

Fig. 1: Schematic illustration of the system of two-dimensional X-ray diffraction at high pressure and high temperature at BL04B1, SPring-8. Sample is packed within an octahedral pressure medium which is placed at the center of 8 cubic anvils.

2 Experimental methods In-situ stress measurements under high pressure and high temperature were conducted using a

- 37 -

Kawai-type multi-anvil apparatus SPEED-1500 at BL04B1, SPring-8 in conjunction with a two-dimensional X-ray diffraction system. A schematic illustration of the high pressure stress measurement system is shown in Figure 1. An octahedral Cr2O3-doped MgO pressure medium with 10 mm edge length was compressed with six WC and two cBN anvils with 5 mm truncated edge length. Since cBN anvil is transparent for X-ray, the two cBN anvils were used as windows for diffracted X-ray. In order to avoid diffraction from gaskets and pressure medium, materials along X-ray path were replaced by mixture of amorphous B and epoxy resin. Two-dimensional X-ray diffraction pattern was taken using monochromatic X-ray (collimated to 200 × 200 µm2) and an imaging plate as a detector. The observable diffraction angle using the present system is 2θ ~ 10º at the maximum. In experiment (2), strain measurement was done simultaneously by X-ray radiography using the monochromatic X-ray with typical exposure time of 1 min. For the high pressure stress measurement of KCl, a polycrystalline KCl rod (2.1 mm diameter and 1.5 mm height), which was formed by cold-press, was used. A detailed design of cell assembly for this experiment is shown in Figure 2a. The sample was inserted between dense Al2O3 pistons to yield high stress by compression of the pressure medium. Energy of used monochromatic X-ray was 50 keV. Stress and pressure were determined from two-dimensional diffraction data of the KCl sample using theoretical equation (e.g. Singh et al. 1998) with known thermoelastic parameters (Walker et al. 2002; Haddadi et al. 2008). For stress relaxation test of olivine, pre-sintered San Carlos olivine sample (1.2 mm diameter and 1.1

mm height) was used. A design of cell assembly for this experiment is shown in Figure 2b. The sample was inserted between dense Al2O3 pistons and surrounded by hBN cylinder to yield high stress by compression. In order to determine sample length strain) by in-situ X-ray radiography, 5 µm thick Au foils were put on the end surfaces of sintered olivine. Heating was done by using tubular graphite heater and temperature was measured using W5Re-W26Re thermocouple. Energy of used monochromatic X-ray was 40 keV. Thermoelastic parameters reported by Abramson et al. (1997), Isaak (1992) and Liu and Li (2006) were used for calculation of stress and pressure.

Fig. 2: Schematic cross section of cell assemblies for (a) room-temperature experiments and for (b) high-temperature experiments. Large arrows show compression direction in the assemblies (Ψ = 54.7º).

Fig. 3: Variation of normalized d-spacing (d/d0) of B1 phase of KCl as a function of Ψ angle upon compression. Lattice distortion increases with increasing pressure for each diffraction plane at lower pressures (< 2 GPa), and the lattice distortion of the (220) and (222) planes is much larger than that of the (200) plane.

- 38 -

3 Results 3.1 High pressure stress measurement of KCl at room temperature In the stress measurement of KCl, pressure was increased monotonously up to 9.9 GPa at room temperature. The low-pressure B1 phase was transformed to high-pressure B2 phase during compression, and the coexistence of the two phases was observed at P = 2.3 and 2.4 GPa. Due to significant volume reduction of this phase transformation (~12%, Walker et al. 2002), pressure generation was less efficient at the two phase region. Significant lattice distortion was observed during the compression by two-dimensional X-ray diffraction. Figure 3 shows variation of normalized d-spacing (d/d0) of B1 phase as a function of Ψ angle, where Ψ angle is orientation of specific lattice plane relative to bottom direction (see Figure 2). The minimum d/d0 value was observed at near compression direction (Ψ = 54.7º) of cell assembly (Figure 2a) and maximum was at near dilatation direction (144.7º). Lattice distortion increases with increasing pressure both in B1 and B2 phases except for at P = ~2–3 GPa where the two phases coexist due to the phase transition. Magnitude of lattice distortion in both phases were significantly different depending on diffraction planes (as shown in Figure 3 for B1 phase). This is due to stress heterogeneity and elastic anisotropy.

Fig. 4: Variation of deviatoric stress in KCl upon compression to 9.9 GPa. The deviatoric stress increased with increasing pressure both in B1 and B2 phases except for at P = ~2–3 GPa where the two phases coexist due to pressure induced phase transition. Coexistence of B1 and B2 phases were observed at the pressure range between two dotted lines.

Figure 4 shows variation of deviatoric stress in KCl to up 9.9 GPa. The deviatoric stress increased with increasing pressure both in B1 and B2 phases except for two phase region at P = ~2–3 GPa. Difference of deviatoric stress between diffractions was more significant in B2 phase as far as we observed. Through the measurements of KCL, we could confirm the ability of accurate stress measurement using present system.

Fig. 5: Variation of stress in polycrystalline olivine during heating at P = 8.1–6.6 GPa.

3.2 Stress relaxation test of olivine under high pressure and high temperature In the stress relaxation test of olivine, pressure was first increased to 9.0 GPa at room temperature, and then temperature was increased. Temperature was kept at 673, 873, 1073 and 1273 K for 33–241 min, and changes of stress and strain were observed during heating at constant temperature. The stress and pressure were calculated using the following five diffractions; (021), (101), (130), (131) and (112). Figure 5 shows variation of stress at high temperatures as a function of elapsed time. Calculated values of deviatoric stress are significantly different depending on diffraction index. The stress values by (130) and (101) were generally the maximum and minimum, respectively, and the difference was as much as factor of ~2–4. Deviatoric stress [based on (130) diffraction] was significantly decreased from 4.7 to 2.6 GPa by increasing temperature from 300 to 673 K. Changes of stress value at constant temperatures were 2.6 → 2.2 GPa at 673 K (in 154 min), 1.7 → 1.2 GPa at 873 K (241 min), 0.63 → 0.56 GPa at 1073 K (112 min) and 0.20 → 0.20 GPa at 1273 K (33 min). Result of strain measurements by radiography suggests that most of sample strain is elastic and plastic strain is undetectably small (plastic strain rate of less than ~3 × 10–7 s–1) during present stress relaxation test. Further experiments at wider range of pressure conditions are considered to be important to

- 39 -

assess effect of pressure on rheological property of olivine.

4 Concluding remarks In this study, we have established technique of high pressure stress measurement using Kawai-type multi anvil apparatus and two-dimensional X-ray diffraction with monochromatic synchrotron X-ray at a beam line BL04B1, SPring-8. Although pressure and temperature conditions in this study was 10 GPa and 1273 K at the maximum, similar measurement is potentially possible up to ~30 GPa and ~2000 K. On the other hand, we realized that magnitude of plastic strain obtained by stress relaxation test at high pressure and high temperature is too small for quantitative rheological measurement. Therefore, in order to investigate deep Earth rheology, it is very important to conduct rheological experiments using high pressure deformation apparatus such as RDA (e.g. Nishihara et al. 2008), D-DIA (e.g. Wang et al. 2003) and KATD (Nishihara, 2007) in conjunction with synchrotron radiation.

Int. in press. Li, L. and 6 colleagues, 2006. Deformation of olivine at mantle pressure using D-DIA. Eur. J. Mineral. 18, 7−19. Liu, W. & Li, B., 2006. Thermal equation of state of (Mg0.9Fe0.1)2SiO4 olivine. Phys. Earth Planet. Int. 157, 188−195. Nishihara, Y., 2007. High-pressure and high –temperature deformation experiments using Kawai-type apparatus for triaxial deformation (KATD). Ann. Report Research. Center Evolv. Earth Planets 37−44. Nishihara, Y. and 6 colleagues, 2008. Plastic deformation of wadsleyite and olivine at high-pressure and high-temperature using a rotational Drickmer apparatus (RDA). Phys. Earth Planet. Int. 170, 156−169. Singh, A.K. and 4 colleagues, 1998. Analysis of lattice strains measured under nonhydrostatic pressures. J. Appl. Phys. 83, 7567−7575.

Acknowledgements

Walker, D. and 4 colleagues, 2002. Thermal equation of state for B1 and B2 KCl. Am. Mineral. 87, 805−812.

The author thanks Ken-ichi Funakoshi, Yuji Higo, Hidenori Terasaki, Norimasa Nishiyama, Tomoaki Kubo, Akira Shimojuku and Noriyoshi Tsujino for technical assistance in in-situ X-ray diffraction experiments at SPring-8.

Wang, W. and Takahashi, E., 2000. Subsolidus and melting experiments of K-doped peridotite KLB-1 to 27 GPa: Its geophysical and geochemical implications. J. Geophys. Res. 105, 2855−2868.

References Abramson, E.H. and 3 colleagues, 1997. The elastic constants of San carlos olivine to 17 GPa. J. Geophys. Res. 102, 12253−12263.

Wang, Y. and 3 colleagues, 2003. The deformation-DIA: A new apparatus for high temperature triaxial deformation to pressures up to 15 GPa. Rev. Sci. Inst. 74, 3002−3011.

Durham, W.B. and 4 colleagues, 2009. New measurements of activation volume in olivine under anhydrous conditions. Phys. Earth Planet. Int. 172, 67−73. Haddadi, K., Louail, L. & Maouche, D., 2008. Elastic properties of potassium halides under pressure. J. Mater. Sci. Technol. 24, 241−244. Isaak, D.G., 1992. High-temperature elasticity of iron-bearing olivines. J. Geophys. Res. 97, 1871−1885. Karato, S., 2008. Deformation of Earth Materials. Cambridge University Press, Cambridge, UK. Kawazoe, T. and 4 colleagues, 2009. Shear deformation of dry polycrystalline olivine under deep upper mantle conditions using a rotational Drickamer apparatus (RDA). Phys. Earth Planet.

- 40 -

Crustal evolution of continent collisional orogens with ultrahigh-temperature extreme metamorphism Kei Sato COE Post doctoral fellow Research Center for the Evolving Earth and Planets, Also at Department of Earth & Planetary Sciences Graduate School of Science & Engineering, Tokyo Institute of Technology e-mail: [email protected]

The evolution process of continental crust in collisional orogens with ultrahigh-temperature (UHT) extreme metamorphism was studied by the combined method of fieldwork geology, petrography, experimental petrology and geochronology. The fieldworks were performed at the major UHT metamorphic belts in southern India and southern Africa. The high P-T experiment was carried out on the Mg-rich staurolite composition in the system FeO-MgO-Al2O3-SiO2-H2O. Also, the solubility of Na2O in sodicgedrite was investigated experimentally.

1 Introduction The author’s study was aimed at demonstrating the prograde high-pressure metamorphic history of highgrade metamorphic belts, particularly in collisional orogens with crustal ultrahigh-temperature (UHT) metamorphism (i.e. T ≥ 900°C; Brown, 2007; Kelsey, 2008) that are linked to the formation and evolution of supercontinents in Earth history. Several natural occurrences of unusual moderate-Mg staurolite inclusions (Mg# = 0.5) in poikiloblastic garnets have been reported from UHT metamorphic regions in some of the major continent collisional orogens of the world. The Southern Indian Granulite Terrane constituted a central position in the Gondwana supercontinent assembly in the Late NeoproterozoicCambrian (e.g. Collins and Pisarevsky, 2005; Meert and Lieberman, 2008; Santosh et al., 2009). The Palghat-Cauvery Suture Zone (PCSZ) in the terrane is the trace of the collisional suture of the supercontinent. Since the first report of moderate-Mg staurolites from the PCSZ (Shimpo et al., 2006), such relic staurolite inclusions have been found in several UHT metamorphic localities along the suture zone (Collins et al., 2007; Tsunogae et al., 2008; Kanazawa et al., 2009; Sato et al., 2009). On the other hand, the Limpopo Belt in southern Africa is a classic Archaean high-grade metamorphic province sandwiched between the Zimbabwe Craton and the Kaapvaal Craton (e.g. van Reenen et al., 1987; Armstrong et al., 1991). Moderate-Mg staurolite inclusions within poikiloblastic garnets were also reported from the Limpopo Belt (Schreyer et al., 1984; Tsunogae and van Reenen, 2006). Regardless of the ages of the rocks (i.e. Late Neoproterozoic-Cambrian in southern India; Late Archaean in southern Africa), the

collisional orogens have common staurolite-bearing mineral texture. In the quartz-free (or -undersaturated) and MgO-Al2O3-SiO2-H2O system, pure-Mg-staurolite is a stable phase at pressures of about 10-60 kbar and temperatures of about 600-900°C (Schreyer, 1968; 1988; Fockenberg, 1998). Also, the natural occurrences of Mg-rich staurolites with Mg/(Mg+Fe+ Zn,Co) > 0.5 have been found in several metamorphic regions of the world that were buried in lower crust deep. For example, Enami and Zhang (1988) reported the symplectitic Mg-rich staurolite (Mg# = 0.7) in eclogite-facies rock from eastern China. Therefore, these experimental evidences and natural occurrences suggest the possibility for prograde high-pressure metamorphism in the PCSZ and the Limpopo Belt before UHT thermal peak. Some previous experiments related to Mg-rich and/or moderate-Mg staurolites have been performed at more than 20 kbar (e.g. Hellman and Green, 1979; Lattard and Bubenik, 1995). However, the high P-T stability field on such staurolite compositions had not been demonstrated systematically.

2 Geologic and petrologic background In recent years, the author and collaborative researchers (Prof. M. Santosh, Dr. T. Tsunogae et al.) have performed geological surveys in southern India (March 2007; March 2008; December 2008) and southern Africa (July 2008). The Southern Indian Granulite Terrane is composed of a mosaic of polymetamorphic terranes that is characterized by the major lithology of charnockite and hornblende-biotite orthogneiss. Al-Mg granulite and UHT metapelite are not dominant in the terranes, but they are found as layer, pod and/or patches in

- 41 -

various localities within the terranes. The PCSZ is expanded suture zone between the high-grade metamorphic region (Madurai Granulite Block in the south) and the Archaean granite-greenstone terrane (Dharwar Craton in the north). The staurolite-bearing textures have been found within Al-Mg granulites in UHT localities along the PCSZ (see Introduction). The Limpopo Belt passing across the border in northern South Africa and Zimbabwe is subdivided into the Central Zone, the Southern Marginal Zone and the Northern Marginal Zone by major shear zones and/or lithology. The Central Zone is characterized by the occurrences of pure metaquartzite, marble, calc-silicate rocks and extensive leucocratic gneisses, which suffered remarkable deformation. Also, UHT and staurolite-bearing assemblies were reported as inclusions in poikiloblastic garnets from this zone (see below). When the Fe-Mg exchange thermometers between minerals are applied to UHT metamorphic rocks, such thermometry often provides retrograde metamorphic condition (i.e. lower temperature condition) but not thermal peak condition (e.g. Kawasaki and Sato, 2002; Sato and Kawasaki, 2002; Sato et al., 2006; Sato et al., 2008a). In some cases, UHT condition exceeding 1000°C can be estimated by using the ternary feldspar solvus geothermometer (e.g. Hokada, 2001; Miyamoto et al., 2004). Also, the mineral assemblies of sapphirine + quartz and/or spinel + quartz, which were found in the PCSZ and the Limpopo Belt (Schreyer et al., 1984; Tsunogae and van Reenen, 2007; Tsunogae et al., 2008; Sato et al., 2009), are useful as robust evidences for UHT condition (Friedman, 1953; Hensen and Green, 1971). For example, Sato et al. (2009) reconstructed primary mineral assemblages from various inclusion minerals (such as sapphirine, spinel, quartz and staurolite) in poikiloblastic garnets within granulites from the PCSZ, and inferred that the granulites experienced UHT metamorphism within the stability field of sapphirine + quartz assemblage. UHT metamorphic rocks from some of major collisional orogens of the world often contain remarkable CO2-rich fluid inclusions (e.g. Sato and Santosh, 2007; Santosh et al., 2008). Although the relationship between CO2 fluids and UHT metamorphic process is still uncertain, the occurrence of CO2 fluids under UHT metamorphic condition might have an influence on the decreasing of H2O activity in lower crust (Santosh and Omori, 2008; Kanazawa et al., 2009; see the next chapter). The possibility for prograde high-pressure metamorphism was proposed based on the occurrences of kyanite (but not sillimanite) and moderate-Mg staurolite inclusions (Mg# = 0.5) in poikiloblastic garnets both in the PCSZ and in the Limpopo Belt (e.g. Shimpo et al., 2006; Tsunogae and van Reenen, 2006;

Tsunogae et al., 2008; Kanazawa et al., 2009; Sato et al., 2009). However, the question whether the rocks carrying the moderate-Mg staurolite inclusions experienced the prograde high-pressure metamorphism had been still uncertain due to the absence of detailed experimental data for the high P-T phase relation on the system of moderate-Mg staurolite composition.

3 High P-T experiments The author experimentally investigated the high P-T phase relation on the Mg0.7Fe0.3-staurolite composition with excess water in the system FeO-MgO-Al2O3SiO2-H2O at 12-19 kbar and 850-1050°C, particularly focusing on breakdown reaction around the lower pressure limit of the stability field of Mg-rich staurolite (Sato et al., submitted). The experimental results show that the stability field of staurolite in the system shifts to high-temperature side, as compared to that of pure-Mg-staurolite (Fig. 1). The results also demonstrate that the lower pressure limit of the stability at 950°C is situated between 15 kbar and 14 kbar, and that staurolite is decomposed with isothermal decompression by a series of reactions: (1) staurolite ⇒ orthopyroxene + corundum + Melt, and (2) staurolite + orthopyroxene ⇒ sapphirine + Melt. When the synthetic staurolites coexisted with the orthopyroxene in run products, the Mg# of staurolites was decreased to 0.6-0.5. As it has been mentioned above, the natural occurrences of moderate-Mg staurolite inclusions (Mg# = 0.5) in poikiloblastic garnets were reported in the major collisional orogens with crustal UHT metamorphism. Therefore, the author’s experimental data supports strongly the possibility that the staurolite-bearing rocks had experienced high-pressure metamorphism and/or UHT extreme metamorphism on the prograde stage during collisional orogeny. Also, the author(s) experimentally investigated the solubility of Na2O in gedrite under UHT condition (Kanazawa et al., 2009). In natural occurrences from UHT metamorphic belts, gedrite often contains considerable amount of Na2O (2.3 wt % in maximum). In the experiment, the mixture of 90 wt% natural sodicgedrite and 10 wt% garnet was kept at 12 kbar and 1000°C. As results, the mixture was molten partially and small amount of garnet was preserved as metastable phase, and gedrite was confirmed as stable phase. Also, the most of Na within the closed system was incorporated into molten phase rather than gedrite. The experimental result is inconsistent with the natural occurrences. Therefore, the author(s) proposed the possibility that the natural sodicgedrites might be products caused due to the low H2O activity under UHT condition. This implies the possible occurrence

- 42 -

of abundant primary high-density CO2 fluid inclusions in natural UHT rocks, because the low H2O activity can result from the increasing of CO2 fluids. The author’s experiment supports the numerical simulation of Santosh and Omori (2008), where the carbon-cycle model between upper mantle, crust and atmosphere relating to the UHT metamorphic process was proposed.

Acknowledgements The author thanks Prof. M. Santosh (Kochi Univ.) and Dr. T. Tsunogae (Univ. Tsukuba) for collaboration. Also, the author thanks Profs. E. Takahashi, S. Maruyama, T. Kawasaki (Ehime Univ.), Drs. T. Hirata, S. Omori, Y. Nishihara (Ehime Univ.), S. Yamamoto and Y. Kon for valuable discussions.

References Armstrong, R. A. and 4 colleagues 1991. Zircon ion microprobe studies bearing on the age and evolution of the Witwatersrand triad. Precam. Res. 53, 243−266. Bohlen, S. R., A. Montana & D. M. Kerrick 1991. Precise determinations of the equilibria kyanite ⇔ sillimanite and kyanite ⇔ andalusite and a revised triple point for Al2SiO5 polymorphs. Am. Mineral. 76, 677−680. Brown, M. 2007. Metamorphic conditions in orogenic belts: A record of secular change. Inter. Geol. Rev. 49, 193−234.

Fig. 1: P-T trajectories of moderate-Mg staurolitebearing and sapphirine + quartz assemblage-bearing UHT metamorphic rocks from the PCSZ in southern India and/or the Limpopo Belt in southern Africa (Sato et al., submitted). Data sources: Stability field of staurolite in the system MASH, Fockenberg (1998); Stability field of aluminium silicate, Holdaway (1971) and Bohlen (1991); petro- genetic grids in the system FMAS, Kelsey (2008).

4 Remarks The author’s geologic, petrologic and experimental studies provide the series of reaction process of minerals in lower crust, which experienced prograde high-pressure metamorphism and/or even UHT extreme metamorphism during continent collisional orogenies. Recently, the author performed the preliminary geochronologic study of U-Pb age dating for zircon grains from southern India (Sato et al., 2008b) by laser ablation-ICP-mass spectrometry technique, in order to demonstrate systematically the tectonic ages/setting of the Southern Indian Granulite Terrane. As it has mentioned above, the detailed relationship between CO2 fluids and crustal UHT metamorphic process is still uncertain. Further systematic studies including field geology, petrography, experimental petrology and geochronology are necessary for demonstrating the evolution process of collisional orogens with UHT extreme metamorphism.

Collins, A. S. & S. A. Pisarevsky 2005. Amalgamating eastern Gondwana: The evolution of the Circum-Indian Orogens. Earth-Sci. Rev. 71, 229−270. Collins, A. S. and 5 colleagues 2007. Passage through India: the Mozambique Ocean suture, high- pressure granulites and the Palghat-Cauvery shear zone. Terra Nova 19, 141−147. Enami, M. & Q. Zhang 1988. Magnesian staurolite in garnet-corundum rocks and eclogite from the Donghai distinct, Jiangsu province, east China. Am. Mineral. 73, 48−56. Fockenberg, T. 1998. An experimental investigation on the P-T stability of Mg-staurolite in the system MgO-Al2O3-SiO2-H2O. Contrib. Mineral. Petrol. 130, 187−198. Friedman, G. M. 1953. The spinel-silica reaction succession: a study of incompatible mineral phases. Jour. Geol. 62, 366−374. Hellman, P. L. & T. H. Green 1979. The high-pressure experimental crystallization of staurolite in hydrous mafic compositions. Contrib. Mineral. Petrol. 68, 369−372. Hensen, B. J. & D. H. Green 1971. Experimental study of the stability of cordierite and garnet in pelitic compositions at high pressures and temperatures. I: Compositions with excess alumino-silicate. Contrib. Mineral. Petrol. 33, 309−330. Hokada, T. 2001. Feldspar thermometry in ultrahightemperature metamorphic rocks: Evidence of crustal

- 43 -

metamorphism attaining ∼1100°C in the Archean Napier Complex. Am. Mineral. 86, 932−938. Holdaway, M. J. 1971. Stability of andalusite and aluminium silicate phase diagram. Am. Jour. Soc. 271, 97−131. Kanazawa, T., K. Sato and 2 colleagues 2009. The stability and origin of sodicgedrite in ultrahightemperature Mg-Al granulites: a case study from the Gondwana suture in southern India. Contrib. Mineral. Petrol. 157, 95−110. Kawasaki, T. & K. Sato 2002. Experimental study of Fe-Mg exchange reaction between orthopyroxene and sapphirine and its calibration as a geothermometer. Gondwana Res. 5, 741−747. Kelsey, D. E. 2008. On ultrahigh-temperature crustal metamorphism. Gondwana Res. 13, 1−29. Lattard, D. & W. Bubenik 1995. Synthetic staurolites in the system Mg-Fe-Al-Si-O-H. Euro. Jour. Mineral. 7 931−947. Meert, J. G. & Lieberman, B. S. 2008. The Neoproterozoic assembly of Gondwana and its relationship to the Ediacaran-Cambrian radiation. Gondwana Res. 14, 5−21. Miyamoto, T., K. Sato and 4 colleagues 2004. Occurrences of metamorphosed ultramafic rock and associating rocks in Howard Hills, Enderby Land, East Antarctica: Evidence of partial melting from geochemical and isotopic characteristics. Polar Geosci. 17, 88−111. Santosh, M. & S. Omori 2008. CO2 flushing: A plate tectonic perspective. Gondwana Res. 13, 86-102. Santosh, M., K. Sato and 4 colleagues 2008. Carbonic metamorphism at ultrahigh-temperatures: Evi- dence from North China Craton. Earth Planet. Sci. Lett. 266, 149−165. Santosh, M., S. Maruyama & K. Sato 2009 (in press). Anatomy of a Cambrian suture in Gondwana: Pacific-type orogeny in southern India? Gondwana Res. Sato, K. & T. Kawasaki 2002. High-pressure and high-temperature experiments on the phase relations in the system of Mg-rich garnet composition (Prp75Alm25): Implication for the Fe-Mg partitioning between garnet and orthopyroxene. Polar Geosci. 15, 66−79. Sato, K. & M. Santosh 2007. Titanium in quartz as a record of ultrahigh-temperature metamorphism: the granulites of Karur, southern India. Mineral. Mag. 71, 143−154. Sato, K., T. Miyamoto & T. Kawasaki 2006. Experimental calibration of sapphirine-spinel Fe2+-Mg exchange thermometer: Implication for constraints on P-T condition of Howard Hills, Napier Complex, East Antarctica. Gondwana Res. 9,

398-408. Sato, K., T. Miyamoto & T. Kawasaki 2008a. Fe2+-Mg partitioning experiments between orthopyroxene and spinel using ultrahigh-temperature granulite from the Napier Complex, East Antarctica. Geol. Soc. London, Spec. Publ. 308, 431−447. Sato, K., M. Santosh, T. Tsunogae, Y. Kon, S. Yamamoto & T. Hirata 2008b. U-Pb dating of zircons in UHT granulites and associated granulite from Achankovil Shear Zone in southern India. International Association for Gondwana Research, Conference Series 7, 55−56. Sato, K., M. Santosh & T. Tsunogae 2009 (in press). A petrologic and laser Raman spectroscopic study of sapphirine-spinel-Mg-staurolite inclusions in garnet from Kumiloothu, southern India: Implications for extreme metamorphism in a collisional orogen. Jour. Geodynamics. Sato, K., M. Santosh & T. Tsunogae (submitted). High P-T phase relation of magnesian staurolite in the system FeO-MgO-Al2O3-SiO2-H2O: Implications for high-pressure metamorphism in collisional orogens. Schreyer, W. 1968. A reconnaissance study of the system MgO-Al2O3-SiO2-H2O at pressures between 10 and 25 kbar. Carnegie Inst. Washington Year 66, 380−392. Schreyer, W. 1988. Experimental studies on metamorphism of crustal rocks under mantle pressures. Mineral. Mag. 52, 1−26. Schreyer, W., P. C. Horrocks & K. Abraham 1984. High-magnesium staurolite in a sapphirine-garnet rock from the Limpopo Belt, Southern Africa. Contrib. Mineral. Petrol. 86, 200−207. Shimpo, M., T. Tsunogae & M. Santosh 2006. First report of garnet-corundum rocks from southern India: implications for prograde high-pressure (eclogite-facies?) metamorphism. Earth Planet. Sci. Lett. 242, 111−129. Tsunogae, T. & D. D. van Reenen 2006. Corundum + quartz and Mg-staurolite bearing granulite from the Limpopo Belt, southern Africa: Implications for a P-T path. Lithos 92, 576−587. Tsunogae, T. & D. D. van Reenen 2007. Carbonic fluid inclusions in sapphirine + quartz bearing garnet granulite from the Limpopo Belt, southern Africa. Jour. Mineral. Petrol. Sci. 102, 57−60. Tsunogae, T., K Sato and 2 colleagues 2008. High-pressure and ultrahigh-temperature metamorphism at Komateri, northern Madurai Block, southern India. Jour. Asian Earth Sci. 33, 395−413. van Reenen, D. D. and 4 colleagues 1987. Deep crustal response to continental collision: the Limpopo Belt of southern Africa. Geology 15, 11−14.

- 44 -

Chemical equilibrium between perovskite and molten iron to 146 GPa and implications for chemically-distinct buoyant layer at the top of the core Haruka Ozawa COE Research Assistant Research Center for the Evolving Earth and Planets, Also at Department of Earth & Planetary Sciences Graduate School of Science & Engineering, Tokyo Institute of Technology e-mail: [email protected]

Introduction The core-mantle boundary (CMB) is the largest thermal and chemical boundary inside the Earth and has significant roles in the evolution of this planet. It has long been recognized that the outer core is less dense by 5 to 10% than pure iron at core pressure and temperature[1]. Such core density deficit indicates the incorporation of light element(s). The bulk outer core liquid may be chemically homogeneous because of a vigorous convection, and is likely in direct contact with the mantle silicates and oxide at the CMB. The lowermost mantle comprises of ferropericlase and perovskite if the temperature is higher than 3500 K[2,3]. Chemical equilibrium between perovskite, ferropericlase, and molten iron may thus provide essential information for constraining the composition of the liquid outer core[4]. Here we investigated the reaction between liquid iron and (Mg,Fe)SiO3 perovskite up to 146 GPa and 3500 K. The chemical equilibrium between molten iron, perovskite, and ferropericlase is discussed on the basis of available high pressure and temperature (P-T) partitioning data. Experiments High P-T conditions were generated in a laser-heated diamond anvil cell. Three starting materials were prepared as fine powder mixtures of iron metal (99.9% purity) and gels (Table 1). Chemical compositions of co-existing quenched molten iron and perovskite were determined with analytical transmission electron microscope (TEM). The chemical analyses were made by energy-dispersive X-ray spectrometry attached to the TEM except for oxygen in iron-rich metal. Oxygen in quenched

molten iron was analyzed by the electron energy loss spectroscopy attached to the TEM.

Results Three separate runs were conducted at 3000 to 3500 K in a pressure range from 124 to 146 GPa. The chemical compositions of co-existing quenched liquid metal and perovskite were given in Table 1. The O and Si contents in liquid iron can be treated in terms of the equilibrium[5]; FeSiO3 (perovskite) = Fe + Si + 3O (metal) (1) We calculate the distribution coefficient, Kd, between perovskite and liquid iron; 3

a Fe aSi a O (2) a FeSiO3 where aFe, aSi, aO, and aFeSiO3 are the activities of Fe, Si, and O in metal, and FeSiO3 in perovskite, respectively. Present data were fitted to the following equation, together with the previous experimental results[5,6] (Fig. 1); Kd =

RTlnKd = -∆Hr + T∆Sr - P∆Vr (3) where ∆Hr, ∆Sr, and ∆Vr are the changes in enthalpy, entropy, and volume, respectively, for the reaction (1), and R is a gas constant. Fitting results gave ∆Hr = 476 (±63) kJ mol−1, ∆Sr = 74.8 (±23.9) J K−1 mol−1, and ∆Vr = −0.739 (±0.156) cm3 mol−1. The solubilities of O and Si in liquid iron co-existing with perovskite exhibit strong temperature dependence as well as a positive pressure effect (Fig. 1).

Discussions The chemical equilibrium between perovskite, ferropericlase, and molten iron at the P-T conditions of the CMB was calculated in the Mg-Fe-Si-O system

Table 1: Experimental conditions and results Starting material Duration Quenched liquid iron Perovskite Run# P(GPa) T(K) Silicate composition (min) O (wt%) Si (wt%) XFeSiO3 1 (Mg0.8Fe0.2)2SiO4 124 3150 1 4.9(16) 4.4(2) 0.195(21) 2 (Mg0.9Fe0.1)2SiO4 146 3500 1 19.3(22) 1.0(1) 0.019(7) 3 Mg0.9Fe0.1SiO3 130 3000 0.5 3.9(20) 3.5(1) 0.025(3) Numbers in parentheses are standard deviations in the last digit. The temperature uncertainty was ±10%.

- 45 -

Fig. 1: Plot of lnKd as functions of pressure and temperature. Solid lines represent the fitting results for 2500, 3000, and 4000 K.

from these experimental results and previous data[7,8]. We found that molten iron should include oxygen and silicon more than required to account for the core density deficit (