Interaction between iron and post-perovskite at core-mantle boundary and core signature in plume source region



[1] Interaction between the lower mantle and core is essential for understanding the nature of D″ layer at the core-mantle boundary (CMB). Here, we report the reaction between post-perovskite (PPv) and metallic iron under the condition of the CMB, for example, 139 GPa and 3000 Kelvin. Analytical transmission electron microscope (ATEM) analysis revealed that significant amount of oxygen up to 6.3 weight percent (wt.%) and silicon up to 4.0 wt.% can be dissolved into molten iron. The dihedral angle between PPv and molten iron is 67 degrees. Thus, a small amount of core metal of about 2 volume percent (vol.%) can be trapped without separation in the PPv region at the CMB. The amount of core metal trapped by this mechanism can produce the isotopic signature of the outer core in the plume source at the base of the lower mantle.

1. Introduction

[2] A chemical reaction between (Mg, Fe)SiO3 perovskite and metallic iron has been experimentally investigated so far, using laser-heated diamond anvil cell (LHDAC) to find out possible light elements in the core and to figure out the anomalous seismic structure at the core-mantle boundary (CMB). Knittle and Jeanloz [1989, 1991] and Goarant et al. [1992] reported that (Mg, Fe)SiO3 perovskite and metallic iron reacted to form SiO2 and FeO, FeSi alloy. However, Hillgren and Boehler [1998] denied significant amount dissolution of silicon and oxygen into the metal up to 100 GPa. More recently, Takafuji et al. [2004, 2005] reported that increase in solubility of silicon and oxygen into molten iron with pressure up to 97 GPa at around 3000 K. They also showed that dihedral angle between perovskite and molten iron, which indicates wetting property, decreases with increasing pressure to 47 GPa. However, as perovskite phase transforms to post-perovskite phase (PPv) at the base of mantle [Murakami et al., 2004], it is essential to clarify the chemical reaction and textural relationship between iron and PPv at the CMB condition.

2. Experimental Method

[3] We performed reaction experiments at 130(5) GPa and 2000 K (Run FeEn19), 139(10) GPa and 3000 K (Run f70En07) using LHDAC technique [Shen et al., 1996]. We used a Mao-Bell type DAC with the bevelled diamond anvils of inner culet size of 150 μm. Powdered natural (Mg0.987, Fe0.013)SiO3 enstatite from Sri Lanka and iron powder (99.5% pure) were used as the starting material. A mixture of metallic iron and enstatite powder (Run f70En07) or thin iron foil (Run FeEn19) embedded in powdered enstatite was loaded into the sample hole with a diameter of 70 μm of a pre-compressed rhenium gasket. The pressures before and after laser heating were measured at room temperature by the ruby fluorescence method [Mao et al., 1978] and the raman shift of the T2g mode at the culet of the diamond anvil [Akahama and Kawamura, 2004]. Both measurements were consistent with each other.

[4] Before heating to a target temperature at high pressure, the sample was annealed by laser scanning over the whole sample areas at about 1800 K for 30 minutes in order to complete the transformation to high pressure phase and minimize the deviatoric stress in the samples. Then, the sample was heated at the target temperature for 10 min (Run f70En07), which is enough to achieve textural steady state [Takafuji et al., 2004]. The samples were heated for 10–35 minutes by the double-sided laser heating method [Shen et al., 1996] using a high power multimode Nd:YAG laser (λ = 1.064 μm, LEE LASER8100MQ) with the laser spot of around 50 μm. Temperatures were determined by fitting the emission spectra from the heated sample to the grey body formula. Fluctuation of temperature was around ±200 K at a constant laser power.

[5] PPv coexisted with hcp-iron and small amount of magnesiowüstite were confirmed by in-situ X-ray diffraction studies after laser heating at high pressure. Angle dispersive in situ X-ray diffraction at high pressure was carried out using a synchrotron radiation and an imaging plate (IP) at BL-13A of KEK-PF in Tsukuba, Japan. A monochromatized X-ray to about 30 keV was collimated to 30 μm in diameter, and introduced to the sample through the diamond. The lattice parameters of PPv at 139 GPa were a = 2.451(15) Å, b = 8.083(41) Å, c = 6.078(16) Å.

[6] We made further detailed TEM observations of the texture and composition of the coexisting phases for Runs FeEn19 and f70En07, in which metallic iron and PPv coexisted in the run products. A thin film with the size of 20 μm × 20 μm and the thickness around 100 nm was prepared from the recovered sample in a rhenium gasket by using Focused Ion Beam System (FIB, JEM-9310FIB) [Suzuki et al., 2004]. FIB has a big advantage for TEM observation of the wide sample area (i.e., to measure dihedral angle) because the sample can be processed to have a uniform thickness of about 100 nm and without selective thinning of small grains or thin films of iron at the PPv grain boundary during the nanofabrication. Micro-textures and chemical compositions of the product phases were examined by a 200 kV ATEM (JEOL JEM-2100F and JEM-2500SE) equipped with an energy dispersive X-ray spectrometer (EDS). Quantitative chemical analysis was carried out in STEM (scanning TEM)-EDS mode with an electron beam current of 5 nA using k-factors determined by the procedure of Cliff and Lorimer [1975].

[7] The TEM images of the whole area of the sample (Run f70En07) are shown in Figure 1a, in which we can see clear images of metallic iron sandwiched by PPv. Figures 1b and 1c show the high magnification of the TEM images of the same run product, where we can see the metallic melt confined at a triple junction of PPv grains (Figure 1c).

Figure 1.

Bright field TEM images of the whole area of the sample recovered from 139 GPa and 3000 K (Run f70En07). (a) A whole view of the recovered sample which was prepared by FIB (Focused Ion Beam) technique. Metallic iron (MI) exists as sheets in silicate. Top and bottom in Figure 1A are the contact with the diamond anvils. (b and c) High magnification images of coexistence of metallic iron and PPv.

3. Results and Discussion

[8] The compositions of the metallic iron and PPv determined using ATEM were listed in Table 1, indicating that a significant amount of oxygen up to 6.3 wt.% (18.3 at.%) and silicon up to 4.0 wt.% (6.6 at.%) are dissolved in metallic iron. Recent reevaluation of the outer core density has been lower by 7–10% than that of pure iron [Anderson and Isaak, 2002; Shanker et al., 2004]. The solubility of silicon and oxygen in metallic iron observed in this study can readily account for 10% of the core density deficit. The present results strongly imply that both oxygen and silicon are the plausible candidates of the light elements in the core. Figure 2a shows solubility of oxygen and silicon in molten iron coexisting with PPv and perovskite at various pressures and about 3000 K [Takafuji et al., 2004, 2005]. The extrapolated solubility of these elements in molten iron coexisting with perovskite [Takafuji et al., 2005] to CMB condition is consistent with those with PPv in this study.

Figure 2.

(a) Solubility of oxygen and silicon in molten iron coexisting with PPv (this work) and those with perovskite as a function of pressure at about 3000 K [Takafuji et al., 2004, 2005]. Open and solid circles, silicon and oxygen solubility of this study. Open and Solid diamonds, silicon and oxygen solubility reported by Takafuji et al. [2004, 2005]. (b) Relation between dihedral angle and oxygen content in metallic iron. Solid circles, dihedral angle between PPv and Fe-O-Si melt, this study; Solid diamond, dihedral angle between Pv and Fe-O-Si melt, [Takafuji et al., 2004]; Open squares, dihedral angle between olivine or ringwoodite and Fe-O-S melt, [Terasaki et al., 2005]; Rightward-triangle, dihedral angle between olivine and Fe-Ni-O-S melt, [Minarik et al., 1996]; Upward-triangles, dihedral angle between olivine and Fe-O-S melt, [Rose and Brenan, 2001]; Downward-triangle, dihedral angle between olivine and Fe-O-S melt, [Gaetani and Grove, 1999]. Angle error of this study corresponds to a 95% confidence interval around the population median [Stickels and Hucke, 1964]. Solid line and doted line indicate the critical wetting angle with the melt fraction of <1%, 2%, 4%, 6%, respectively [von Bargen and Waff, 1986]. Inset: The distribution of the apparent dihedral angles between metallic iron and PPv. The median value of the dihedral angle is 67 degrees. The number of angles measured is 114.

Table 1. Compositions of Metallic Iron and Coexisting Post-Perovskite at 139 GPa and 3000 K (Run f70En07), and 130 GPa and 2000 K (Run FeEn19)
Elementf70En07af70En07 err.FeEn19bFeEn19 err.Oxidef70En07cf70En07 err.FeEn19dFeEn19 err.
  • a

    Av. n = 4.

  • b

    Av. n = 2.

  • c

    Av. n = 3.

  • d

    Av. n = 7.

Mg0.36 wt.%0.20n.d.-MgO39.48 wt.%3.8040.78 wt.%0.85
Si4.03 wt.%1.390.13 wt.%0.02SiO258.91 wt.%6.5757.44 wt.%1.46
Fe89.29 wt.%2.8599.11 wt.%0.06FeO1.60 wt.%0.711.78 wt.%0.55
O6.33 wt.%1.350.71 wt.%0.09     
Total100.00 wt.% 100.00 wt.% Total100.00 100.00 
Mg0.67 at.%0.360.16 at.%0.03MgO49.41 at.%4.7650.78 at.%1.06
Si6.61 at.%2.050.24 at.%0.04SiO249.46 at.%5.5247.98 at.%1.22
Fe74.44 at.%5.3897.19 at.%0.24FeO1.13 at.%0.501.24 at.%0.39
O18.28 at.%3.162.41 at.%0.31     
Total100.00 at.% 100.00 at.% Total100.00 at.% 100.00 at.% 

[9] On the other hand, the ATEM analysis of PPv and metallic iron at the subsolidus condition of 130 GPa and relatively low temperature conditions around 2000 K (Run FeEn19) indicates that no reaction occurs between the phases, for example, the oxygen and silicon contents in metallic iron were less than 1 wt.% that is below the detection limit of the ATEM analysis (Table 1). Very low oxygen and silicon contents in iron quenched from 130 GPa and 2000 K indicate that metallic iron-alloy did not melt below the melting point at this pressure [Knittle and Jeanloz, 1991].

[10] Reaction products, such as FeO and FeSi suggested by previous authors [Knittle and Jeanloz, 1991; Goarant et al., 1992], were not identified from X-ray diffraction pattern of the sample after laser heating and from TEM observation of recovered sample. It seems that these reaction products appeared in past study due to the difference of cooling rate and/or laser heating technique.

[11] We have measured apparent dihedral angles between molten iron and PPv in Run f70En07 from TEM images (Figures 1b and 1c). Distribution of apparent dihedral angles is shown in inset of Figure 2b as a histogram. The median value of apparent dihedral angle, which corresponds to the true dihedral angle, was 67 degrees. Although the dihedral angle is slightly smaller than that between perovskite and molten iron-alloy determined at 25 GPa and 2373 K, 71 degrees [Shannon and Agee, 1998], it is larger than 51 degrees determined at 47 GPa and 3000 K [Takafuji et al., 2004], and is still larger than the critical wetting angle of 60 degrees. The result is plotted in Figure 2b as a function of oxygen content in molten iron together with the previous results performed at lower pressures. The dihedral angle decreases significantly with increasing oxygen content in iron-alloy. The present result is consistent with that of the previous studies [Minarik et al., 1996; Gaetani and Grove, 1999; Terasaki et al., 2005] in spite of the difference in both the pressure conditions and coexisting silicate phases. The dihedral angle between molten iron-alloy and crystalline silicates may be mainly controlled by the solubility of oxygen in molten iron which corresponds to a surface active element. The dihedral angle around 67 degrees between the PPv and molten iron indicates that the percolation of metallic melt occurs effectively at the base of the mantle in the case of existence of the metallic melt fraction greater than 2 vol.% [von Bargen and Waff, 1986], and simultaneously indicating that a small amount of iron around 2 vol.% could remain in the regions of the PPv lithology at the CMB.

[12] Geodynamic studies imply that some of mantle plumes are originated from the CMB region. Walker et al. [1995] pointed out that enrichment of 187Os/188Os in some plume derived magmas shows the outer core signature formed by the inner core crystallization. Brandon et al. [1998] and Brandon and Walker [2005] reinforced the core signature by observation of a coupled enrichment of 186Os/188Os and 187Os/188Os, and estimated that addition of 0.5–1 wt.% of core metal to the D″ layer can produce the signature observed in the plumes, although there are some debates on the origin of the coupled isotopic enrichment [e.g., Brandon and Walker, 2005].

[13] Strong dynamic instabilities are expected at the CMB due to sedimentation of the slab materials and upward flow of the hot mantle plumes [e.g., Brandon and Walker, 2005; Lay et al., 1998; Garnero, 2004]. Consequently, the metallic melt and lower mantle silicates can be mixed in the contact region of the CMB. Since dihedral angle between perovskite and molten iron is enough small [Takafuji et al., 2005] to proceed complete separation of molten iron from the lower mantle with the perovskite lithology. On the other hand, a small amount of molten iron, about 2 vol.% (1 wt.%), can coexist with crystalline silicates [von Bargen and Waff, 1986] in the PPv region at the CMB, since the present result indicates complete percolative separation of the core metal does not occur. The amount of core metal trapped by this mechanism can produce the core signature of the plume source at the base of the lower mantle [Brandon and Walker, 2005]. Existence of the other phases such as magnesiowüstite may not affect the arguments since the dihedral angle of magnesiowüsitite and molten iron is greater than 100 degrees at high pressure [Agee and Shannon, 1997]. The trapped metallic iron at the bottom of the lower mantle will be oxidized during ascent of the plumes due to relatively higher oxygen fugacity conditions in the shallower mantle.

[14] The stability of the PPv in the real earth is a matter of debate [Ono and Oganov, 2005; Hirose et al., 2006]. Although we need more careful studies on the pressure scale corresponding to the CMB conditions, the present experiments on the measurement of the dihedral angle between PPv and molten iron revealed that PPv may play an essential role for producing the core signature in the plume source region by trapping a small amount of core material at the base of the mantle.


[15] We thank N. Hirao and M. Miyahara for useful discussion and experimental assistance. This work was partly supported by the Grants-in-aid for scientific research from Ministry of Education, Culture, Science, and Sport and Technology of Japanese Government to E. Ohtani (14102009 and 16075202) and to T. Kondo (16330164). The work was conducted as a part of the 21st COE program Advanced Science and Technology Center for the Dynamic Earth at Tohoku University.