Low osmium solubility in silicate at high pressures and temperatures

Low osmium solubility in silicate at high pressures and temperatures

Earth and Planetary Science Letters 279 (2009) 165–173 Contents lists available at ScienceDirect Earth and Planetary Science Letters j o u r n a l h...

1MB Sizes 1 Downloads 73 Views

Earth and Planetary Science Letters 279 (2009) 165–173

Contents lists available at ScienceDirect

Earth and Planetary Science Letters j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e p s l

Low osmium solubility in silicate at high pressures and temperatures Tetsuya Yokoyama a,b,⁎, David Walker c, Richard J. Walker a a b c

Department of Geology, University of Maryland, College Park, MD 20742, USA Department of Earth and Planetary Sciences, Tokyo Institute of Technology, Ookayama, Tokyo 152-8551, Japan Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY 10964, USA

a r t i c l e

i n f o

Article history: Received 8 October 2008 Received in revised form 20 December 2008 Accepted 23 December 2008 Available online 12 February 2009 Editor: R.W. Carlson Keywords: osmium highly siderophile elements metal silicate partition coefficients

a b s t r a c t A new methodology to experimentally investigate metal/silicate partitioning for Os (Dmetal/silicate ) under high Os P–T conditions was devised that utilizes metal and silicate components with natural Os abundances and disparate 187Os/188Os. The isotopic contrast in starting components allows examination of isotopic exchange between metal and silicate. As is common for metal–silicate partitioning experiments involving highly siderophile elements (HSE), tiny nuggets of metal appeared in the silicate melts. Linear trends in the 187Os/ 188 Os versus 1/188Os diagram for different silicate chunks from individual experimental charges permit projections of data to nugget-free silicate concentrations. The Os concentrations and isotope compositions in pure silicates reveal that the nuggets remove Os from the silicate melt while percolating through the melt, but contribute limited Os back to the silicate melt. Our data suggest that the rate of chemical equilibration regarding Os concentration between metal and silicate is much faster than that for isotopic exchange. Results N 2 × 105) at for multiple experiments indicate extremely low Os solubility in silicate melt (Dmetal/silicate Os pressures as high as 2 GPa, and temperatures as high as 2400 °C. These results indicate that metal–silicate segregation during planetary differentiation at high temperatures and pressures may leave silicate that is extremely depleted in Os compared to the mantle abundances of Os observed, and that isotopic exchange of HSE between planetary cores and silicate mantles subsequent to primordial differentiation may be limited. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Highly siderophile elements (HSE: including Re, Os, Ir, Ru, Rh, Pt, Pd and Au) are characterized by extremely high metal–silicate partition coefficients (Dmetal/silicate N 104) at 1 atmospheric pressure (e.g. Kimura et al., 1974; Newsom, 1990). These elements are strongly depleted in the silicate portions of the Earth and other differentiated bodies (e.g. Moon, Mars) when compared to the bulk compositions of primitive chondrites (Becker et al., 2006; Day et al., 2007; Jones et al., 2003; Walker et al., 2004), evidently owing to the removal of HSE into planetary cores. The behavior of HSE between metal and silicate under high P–T conditions is, therefore, important for understanding planetary core formation processes (e.g. Chabot and Agee, 2003; Corgne et al., 2008; Jones and Drake, 1986; Righter, 2003; Walter and Tronnes, 2004; Wood et al., 2006). Improved understanding of elemental and/or isotopic exchange between metal and silicate could also provide new insights regarding whether chemical or isotopic signatures originating in the core can be transmitted to mantle plumes that may rise from the core–mantle boundary (Brandon and Walker, 2005; Luguet et al., 2008; Walker et al., 1995; Walker and Walker, 2005). ⁎ Corresponding author. Department of Earth and Planetary Sciences, Tokyo Institute of Technology, Ookayama, Tokyo 152-8551, Japan. Tel.: +81 3 5734 3539; fax: +81 3 5734 3538. E-mail address: [email protected] (T. Yokoyama). 0012-821X/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2008.12.046

Precise experimental determination of the metal–silicate partition coefficients for HSE at high pressures and temperatures has been hindered by analytical and interpretive difficulties including: 1) the extremely low concentrations of HSE in quenched silicate glasses are sometimes at or below the detection limits of applicable measurement methods, 2) to circumvent limitations in measurement capabilities, experiments are typically designed to contain metal components with far higher HSE concentrations than are appropriate for natural systems, so the relevance of these experiments to natural systems has sometimes been questioned, and 3) the common development of HSE micro-nuggets scattered in the silicate glass of run products can lead to complexities in the interpretation of Dmetal/silicate values. If the nuggets HSE that are rich in HSE originate from the metal, and they are present in suspension in the silicate liquid at the high P–T of the experiment, then the nuggets are contaminants in the silicate melt and should not be included in estimates for D values (Borisov and Palme, 1997; Borisov and Walker, 2000; Ertel et al., 1999, 2001; Fortenfant et al., 2006; Holzheid et al., 2000; O'Neill et al., 1995). However, if they exsolve from the silicate upon quenching of the experiments, the HSE incorporated in the nuggets should be included in the D value estimate (Cottrell and Walker, 2006). Here we present data obtained using a new methodology to examine the partitioning characteristics of Os between metal and silicate at P–T–fO2 conditions of 1–2 GPa, 1450–2400 °C, and 2 logunits below iron–wüstite buffer (IW-2), respectively. Osmium is a


T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173

particularly important HSE because of its participation in the 187Re– 187 Os and 190Pt–186Os radiogenic isotope systems, yet the Dmetal/silicate Os values are available only for low P–T experiments (1 atm, 1200– 1600 °C) (Borisov and Walker, 2000; Fortenfant et al., 2006). Unlike most prior experimental studies for HSE partitioning between metal and silicate, the approach taken in this study was to use natural materials, iron meteorites and basalts, with Os concentrations that mimic concentrations that might be present in silicates and metals during planetary core segregation. Further, samples of iron meteorites were paired with basalts having very different 187Os/188Os ratios so that isotopic exchange at high P–T could be monitored. We will show that this novel approach provides new insights to the study of HSE partitioning.

thick slab. Spacers internal to the furnace were UHP grade semisintered MgO, and thermocouple ducts for type D (W3Re/W25Re) thermocouples were high density Al2O3. Experiments at 2 GPa and 2000–2400 °C modified the assembly by insertion of an UHP MgO (0.312″ × 0.180″ × 1.24″) sleeve between the BaCO3 pressure medium and the furnace, which needed to be appropriately thinned to accommodate the inserted sleeve. The graphite sample capsule also needed to be shrunk, all Al2O3 in the assembly was replaced by UHP MgO, and the top contact of the furnace with the steel sealing plug was bolstered with Re foil. One experiment (#60) at 1 GPa and 2200 °C was run in a 20 mm inside-diameter compound vessel without BaCO3, using grade BTM (Ozark Technical Ceramics) MgO as the insulator and pressure medium. Experiments were typically pressurized cold then heated in slow steps to 800 °C over an hour after which the assembly was left at that temperature, typically overnight, to stabilize and close the porosity in the pressure media and graphite sample capsules. After repressurization the temperature was raised in steps over typically 20 min to the run temperature, repressurized after 20 min, and left to cook for a few minutes to a week after which quenching was accomplished by turning of the electric power. Cooling to less than 400 °C typically took ~5 s. Charges were sent, still in their graphite capsules, and opened for Os analysis at the University of Maryland (UMd). Selected charges were opened, sectioned, and polished for optical and electron microprobe examination at LDEO. Typical charges contained a green silicate glass of composition close to the starting material with a thin dispersion of metal micro-nuggets, some of which were just detectable optically, on polished surfaces sectioned through the charge. The meteoritic metal was recovered in dendritically-quenched liquid form, typically as a subspherical blob, often, but by no means always, near the bottom of the charge.

2. Experimental techniques 2.1. Starting materials In most previous studies, metal–silicate partition coefficients for HSE have been measured by using compositionally compromised starting materials. These include FeO- and S-free silicates and/or HSErich metals with Fe–Ni concentrations that are markedly different from plausible planetary core compositions. To remedy this issue, we used paired samples of well-characterized iron meteorites and basalts with very different Os concentrations and 187Os/188Os ratios (Table 1). The silicate sample 91117 is an Archean (~ 2.45 Ga) komatiitic basalt collected from a lava lake in Lion Hills, central Vetreny Belt, Baltic Shield (Puchtel et al., 1996). The sample is composed of hollow columnar and chain zoned pyroxenes, plumose and chain pyroxenes and euhedral to subhedral hopper olivines embedded in a devitrified matrix glass. The sample KI-75-1-139.3 is a picritic basalt collected from the Kilauea Iki lava lake produced by the 1959 eruption (Helz et al., 1989; Picher et al., in press). The sample has a porphyritic texture which is enriched in olivine crystals. Chinga (USNM#3451) is categorized as an ungrouped iron meteorite (Rasmussen et al., 1984) although it has formerly been classified as an anomalous IVB iron meteorite. The HSE concentrations in Chinga are much lower than the IVB irons with a comparable Ni concentration. Filomena (North Chile) (USNM#1334) is a low-Ni member of the IIAB iron meteorite group (Cook et al., 2004; Wasson et al., 1998).

2.3. Osmium isotope analysis To measure concentrations and isotopic compositions of Os in the silicate portions of the run products, we used isotope dilution coupled with negative thermal ionization mass spectrometry (ID-NTIMS). The drawback to this approach compared to previous in situ analytical methods (e.g., LA-ICP-MS, SIMS) is that we must physically separate silicate from the charge, equilibrate the material with a 190Os spike, then purify the Os from the matrix. Thus, spatial resolution is much more limited. However, the sensitivity of ID-NTIMS is sufficient to simultaneously measure concentrations and isotopic compositions of Os at sub-pg quantities, with analytical uncertainties of only a few %. After each high P–T experiment, the graphite capsule was physically broken into several pieces with a stainless blade in a clean room at the UMd. These pieces were numbered as A, B, C… from the top to the bottom of the capsule where the starting metal was originally located. The quenched glass of the run product was separated from the graphite mantle with ceramic tweezers, and was subsequently ground to a fine powder using a small agate mortar and pestle. Between 2 and 9 mg of powdered sample was carefully weighed and put into a quartz glass tube (5.5 mm i.d. × 20 cm long) together with an 190Os spike and 0.6 mL of acid (HCl:HNO3 = 1:2). The glass was then sealed and heated at 240 °C for

2.2. High P–T experiments The high P–T experiments were conducted at Lamont-Doherty Earth Observatory (LDEO) using piston-cylinder apparatus. Typical experiments at 1450 °C and 1 GPa were performed in 1/2″ inside-diameter Boyd–England compound pressure vessels with cylindrical, Pb-wrapped BaCO3 pressure media (0.487″ × 0.313″ × 1.25″), graphite furnaces (0.312″ × 0.25″ × 1.25″), high-density Al2O3 sample sleeves (0.248″ × 0.181″ × 3/8″), and graphite sample capsules (0.180″ × 0.140″ × 8 mm internal cavity length). These capsules were typically able to encapsulate ~100 mg of basalt powder, in the lower portion of which was embedded a ~10 mg chunk of meteoritic metal sawn and clipped from a ~mm

Table 1 Osmium data and major element concentrations in starting materials Silicates

Os (ppb)



91117 (Archean komatiitic basalt) KI-75-1-139.3 (Hawaiian basalt)

0.064 0.233

2.1626 0.1280












152 33

51.3 48.77

0.662 2.45

12.9 12.37

11.7 11.91

0.201 0.16

10.1 11.64

10.4 10.56

2.52 2.06

0.13 0.52

0.09 0.27


Os (ppb)



Chinga (ungrouped iron meteorite) Filomena (IIAB iron meteorite)

8457 1033

0.1401 0.1693




0.00090 0.00742

16.5 5.63

0.577 0.454

Data are from Puchtel et al. (1996), Helz et al. (1989), Rasmussen et al. (1984), Wasson et al. (1998), Cook et al. (2004), Picher et al. (in press), J.M. Honesto (personal comm.), and this study. Os concentrations are in ppb. Major elements are in wt.%.

T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173

N24 h. After the sample digestion, Os was separated by CCl4 extraction and purified by micro-distillation (Birck et al., 1997). The Os isotope compositions of the spiked samples were measured by N-TIMS using a Thermo-Fisher TRITON or an NBS-designed mass spectrometer, both at UMd. The amount of Os dedicated for a single analysis ranged from 15 pg down to 0.1 pg. Trioxides of Re and Os (TRITON: −m/e = 233, 234, 235, 236, 237, 238 and 240; NBS: −m/e = 233, 235, 236, 238 and 240) were measured in jumping mode using a single ion counting collector equipped with a secondary electron multiplier. In the data reduction determining 187Os/188Os ratios and Os concentrations, the oxide correction was first carried out using 17O/ 16 O = 0.000375 and 18O/16O = 0.002044 (Nier, 1950) then the mass fractionation was corrected by assuming 192Os/188Os = 3.08271 in natural Os (Luck and Allegre, 1983). Uncertainties of Os isotope ratios in individual NTIMS runs depend on the amount of Os loaded on filaments, which ranged from 0.1 to 3% (2σmean).


3. Results 3.1. Petrographic results Selected charges were sectioned for light microscopy and electron probe analysis at LDEO rather than for Os abundance and isotopic composition analysis at UMd. Fig. 1 shows incident and oblique illumination of a ground and polished axial section through experiment #40 (23.5 h at 1 GPa and 1450 °C) that began as a solid chunk of Chinga at the base of a mass of basalt powder (91117) in a graphite capsule. The quenched product of this experiment is green glass with a very dilute dispersion of micro-nuggets of metal, some of which are large enough to see and analyze by electron probe, and a large blob of dendritic metal in the upper region of the green glass. The flotation of the metal blob may have been assisted by the large pontoon of graphite

Fig. 1. Incident and oblique illumination of a section along the axis of experiment #40. Large metal blob has moved upward from its initial position at the capsule bottom, perhaps by flotation on the attached chunk of graphite. Basal silicate is polycrystalline clinopyroxene (cpx) with embedded metal blebs. Bulk of the charge is green glass broken into a stack of decompression discs. Cpx is thought to be consequence of Soret-induced transfer of Mg-rich, Si-poor material to the cold end of a temperature gradient of a few 10s of degrees that may exist in the basal mm of the charge. Growth stratigraphy exists within the cpx and its embedded metal grains suggesting that the metal grains were present before and during the clinopyroxene growth and are not a later product of quench crystallization. This conclusion is in accord with the conclusion from Os that the charge had heterogeneous metal distribution at experimental pressure and temperature. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)


T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173

beneath it. The finding of the major mass of metal at a position in the charge other than the bottom is not atypical in both this study and the one by Cottrell and Walker (2006). At the very base of the charge is a thin, compact layer of polycrystalline clinopyroxene with a few included metal nuggets. Average analyses of phases by Camebax electron microprobe at LDEO are given in Table 2. Typical analytical conditions were 15 kV at 20 nA and counting times on peak and background of 30 s. Metals were used as standards for the metals. Silicates and oxides were used as standards for the silicates. Standard Cameca PAP corrections were applied. Ni and S were analyzed separately in a few places in the silicates as trace elements with 15 kV at 135 nA with counting times of 500 s in a wavelength-dispersive traverse of 5–11 wavelengths about the peak to accurately characterize the peaks and backgrounds. The average of 20 glass analyses taken in raster mode of areas 30 μm on a side spaced along both the axial and radial directions in the charge show a bulk composition very like that expected for 91117 (Puchtel et al., 1996). Standard deviations were close to those expected from analytical precision. There are very modest systematic variations (up to 5% of the amount present for Si) in the silicate glass only in the lowest mm at the base of the charge approaching the compact clinopyroxene layer. Silica decreases, whereas FeO, CaO, and MgO increase in abundance downward. The average of 10 analyses of metal within the large dendritic-structured blob listed in Table 2 shows much larger standard deviations than expected for analytical uncertainty. This heterogeneity reflects the compositional variations between the metal dendrites and their matrix. Small metal nuggets within the silicate glass and beneath the clinopyroxene layer, large enough to find with the probe beam but not to quantitatively probe because the probe beam penetrates into the silicate beneath, have the same Ni ~ 10–30% as the main blob. Metal nuggets within the clinopyroxene layer have much lower Ni ~ 2–3%. Slightly low analytic totals for the large metal blob suggest the presence of a few wt.% C that was not directly analyzed but which is unsurprising in metal equilibrated within a graphite capsule. The average composition of the large metal blob and the average silicate give 2 × log(XSilicate / XMetal ) = −2.05, suggesting the pO2 FeO Fe of the experiment was ~2 log units below IW (Hillgren et al., 1994). 3.2. Accuracy of the Os analysis In order to confirm the accuracy of our approach, we analyzed Os abundances and isotope compositions in multiple pieces of quenched glass of the komatiitic basalt that had been processed at a high P–T condition. Approximately 100 mg of the komatiitic basalt was placed in a graphite capsule without the iron meteorite present, which was processed at 1 GPa, 1450 °C for 1 h. The quenched glass was physically broken and split into three different pieces (#32A, #32B and #32C), which were subsequently powdered and analyzed for Os by using b10 mg of each sample with the same chemical approach described

Table 3 Osmium data for 91117 komatiitic basalt 187

Whole rock powdered samples #1 (2000 mg)a #2 (642 mg) #3 (705 mg) #4 (622 mg) Average (±2σ)


2.352 2.171 2.197 1.932 2.16 ± 0.35

Os (ppb) 0.0669 0.0606 0.0600 0.0669 0.064 ± 0.008

Quenched glasses in high P–T exp. using 100 mg of the basalt with no metal #32A (7.3 mg) 2.243 0.0589 #32B (9.3 mg) 2.313 0.0494 #32C (9.1 mg) 2.350 0.0590 Average (±2σ) 2.30 ± 0.11 0.056 ± 0.011 a

Measured by I.S. Puchtel (personal comm.).

above. Table 3 presents the 187Os/188Os ratios and Os concentrations in these glass pieces, as well as those in larger fractions of the ‘unprocessed’ komatiitic basalt powder (N600 mg) for comparison. The komatiitic basalt shows large variations both in terms of 187Os/ 188 Os ratios (12%, 2σ) and Os concentrations (16%, 2σ), exceeding analytical errors and evidently owing to the heterogeneous distribution of Os in the sample (i.e. nugget effect). In contrast, we obtained reproducible 187Os/188Os ratios (4.8%, 2σ) and Os concentrations (21%, 2σ) in the quenched glass which generally agree with those in the starting basalt, although we analyzed limited amounts (b10 mg) of the glass samples. We interpret this as indicating that Os is heterogeneously distributed in the komatiitic basalt but was homogenized in the charge at above liquidus temperature. This method confirms the accuracy for determining 187Os/188Os ratios and Os concentrations in small pieces of the quenched glass samples processed at high pressures and temperatures. 3.3. Blank corrections Because of the limited amounts of Os used in the NTIMS analyses, uncertainties regarding the 187Os/188Os ratios and Os concentrations in the glass samples determined are ultimately controlled by uncertainties associated with blank correction, rather than from the measurement statistics associated with the NTIMS analysis. Table 4 shows the repeated analyses of total procedural blank in the analytical campaign of this study, in which we obtained 187Os/188Os = 0.18 ± 0.02 and Os = 100 ± 50 fg for the possible blank composition. In the case of Chinga–komatiitic basalt pair, the uncertainty of 187Os/188Os ratio in the blank does not affect the results significantly because of relatively higher 187Os/188Os ratios in the glass samples compared to the blank. Therefore, we fixed the blank 187Os/188Os ratio to be 0.18 but varied the amount of blank Os from 50 to 150 fg. In contrast, the uncertainties of 187 Os/188Os ratio and Os amount in the blank both largely affect the

Table 2 Electron microprobe analyses for experiment #40 Metal blob


1σ (n = 10)



1σ (n = 20)



1σ (n = 3)

Fe Ni Co S Total

77.6 19.8 0.44 0.06 97.93

4.3 8.1 0.19 0.07

SiO2 TiO2 Al2O3 Cr2O3 MgO MnO FeO Ni ppma CaO S ppma K2O Na2O Total

50.45 0.69 13.00 0.13 9.42 0.21 11.82 48 10.31 218 0.12 2.13 98.27

1.23 0.05 0.50 0.03 0.23 0.04 0.18 15 0.26 20 0.02 0.37

SiO2 TiO2 Al2O3 Cr2O3 MgO MnO FeO Ni ppma,b CaO S ppma,b K2O Na2O Total

51.85 0.27 4.15 0.33 21.49 0.27 8.61 188 12.07 497 0.00 0.26 99.29

0.78 0.10 0.79 0.05 1.97 0.02 0.76 40 2.18 20 0.00 0.16

a b

Analytical uncertainties for Ni and S are counting error. Ni and S in first, basal clinopyroxene.

Table 4 Osmium data for total procedural blanks 187

10/25/2007 10/26/2007 11/7/2007 11/7/2007 12/7/2007 12/8/2007 1/25/2008 2/14/2008 2/17/2008 3/22/2008 Average (±1σ)


– – 0.212 0.209 0.187 0.168 0.163 0.142 0.146 0.170 0.18 ± 0.02

Os (fg) 60 97 52 191 79 72 178 146 100 55 103 ± 51

Accurate 187Os/188Os measurement was hindered in some samples because of extremely low 187Os signal intensities.

T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173

result for the Filomena–Kilauea Iki basalt pair. Complete dataset for the 187Os/188Os ratio and Os concentrations in the quenched glass samples with different blank corrections are presented in the supplementary material (see Appendix A). 3.4. Osmium concentrations and isotope compositions in run products 187

Os/188Os ratios and Os concentrations in high P–T run products for experiments involving the iron meteorite Chinga and the komatiitic basalt 91117 are presented in Table 5. All data in this table are corrected for a constant procedural blank contribution (Os = 100 fg, 187 Os/188Os = 0.18). Osmium in the run product silicate glasses showed large variations both with respect to concentration and isotopic composition (Os = 0.008–3.0 ppb, 187Os/188Os = 0.136–2.04). The variation of the 187Os/188Os ratios in the silicate glasses are almost within the range of two starting materials. In contrast, some samples have Os concentrations markedly lower than the starting silicate (Os = 0.064 ppb). This will be discussed later in detail. Similar results are obtained in “isotopic” reversal experiments involving the iron meteorite Filomena and the Kilauea Iki basalt (Table 5), in which the starting silicate has higher 187Os/188Os ratio than that in the iron meteorite. The high P–T run products have the 187Os/ 188 Os ratios that are within the range of the two starting materials, whereas most of the silicate glasses have Os concentrations much lower than the starting silicate.


4. Discussion 4.1. Heterogeneous Os distribution in run products Run product silicates in high P–T experiments have variable 187Os/ Os ratios and Os concentrations. Given that N99.9% of all Os present in each experimental charge was contained in the metal, silicate that isotopically equilibrated with the metal would have the 187Os/188Os ratio of the starting metal (0.140, in the case of Chinga; 0.169, in the case of Filomena). The variable concentrations and isotopic compositions in the run product silicates, therefore, indicate heterogeneous distribution of Os in the charge, along with a lack of isotopic equilibration. The compositional variations of major elements at the base of the silicate glass also support the conclusion that the charge was heterogeneous in its metal distribution at the pressure and temperature of the experiment (see Appendix B). One important outcome of these results is that the Os isotopic heterogeneity in the run product silicates, which does not correlate with recovered position within the charge, provides strong evidence that the concentration variations do not result from the exsolution of metal nuggets during quench of a uniform liquid to a glass, nor from diffusion. This observation is in contrast to a previous, similar study of metal–silicate partitioning of the HSE Pt. Cottrell and Walker (2006) observed no detectable Pt concentration variations from location to location on the scale of quenched charges. This observation reinforced the textural suggestion that the metal micro-nuggets that 188

Table 5 Osmium data of high P–T run products 187


Chinga and 91117: 1 GPa #37 (1450 °C, 2 h) A (top) Replicate B Replicate C Replicate D (bottom) #34 (1450 °C, 7.5 h) A (top) Replicate B Replicate C Replicate D (bottom) #29 (1450 °C, 25 h) A (top) Replicate B C (bottom) #53 (1450 °C, 7 days) A (top) B #60 (2200 °C, 10 min) A (top) Replicate B Replicate C


0.4816 1.327 0.2641 0.3130 0.1671 0.1585 0.1654

33.4 159 12.9 21.6 2.55 3.78 2.57

0.239 0.055 0.601 0.362 3.006 2.026 2.986

2.041 0.8213 0.2388 1.017 0.1942 0.2775 0.2220

220 98.4 23.1 136 16.7 25.6 14.4

0.043 0.084 0.335 0.063 0.460 0.304 0.534

0.7686 0.6174 0.1666 0.1828

78.8 58.5 4.46 6.14

0.105 0.139 1.719 1.252

1.284 0.324

288 50

0.030 0.155

0.3178 0.4035 0.2051 0.1942 0.1432

20.17 29.51 10.14 6.02 8.82

0.388 0.268 0.760 1.277 0.866

Filomena and Hawaiian basalt: 1 GPa #33 (1450 °C, 19.5 h) A (top) 0.1538 Replicate 0.1366 B 0.1693 Replicate 0.1666 C (bottom) 0.1578 Replicate 0.1483

50.9 87.6 109 12.8 51.8 32.8


Os (ppb)

0.150 0.087 0.070 0.600 0.148 0.233

Chinga and 91117: 2 GPa #43 (2000 °C, 0.5 min) A (top) Replicate B C (bottom) #44 (2000 °C, 25 min) A (top) Replicate B D (bottom) #42 (2000 °C, 30 min) A (top) Replicate B C (bottom) #57 (2300 °C, 10 min) A (top) B #58 (2400 °C, 15 min) B



Os (ppb)

0.2296 0.6318 0.2954 0.3005

14.1 103 28.5 29.4

0.548 0.079 0.274 0.265

0.8794 0.1935 0.2308 0.1358

141 36.0 19.8 2.63

0.059 0.214 0.391 2.901

0.8413 0.4820 0.4089 0.3778

226 129 324 288

0.037 0.062 0.024 0.027

0.1516 0.1391

16.6 24.4

0.462 0.313






Filomena and Hawaiian basalt: 2 GPa #47 (2000 °C, 5 min) C (bottom) 0.1543

All data are corrected for a constant procedural blank contribution (Os = 100 fg, 187Os/ 188Os = 0.18).


T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173

line Y is not perpendicular to line X because the total concentration of Os in the silicate glass is a function of 187Os/188Os and 1/188Os. The fractions of the individual experimental run products do not plot along trends consistent with either chemical equilibrium and accompanying isotopic disequilibrium (line Y), or isotopic equilibrium and accompanying chemical disequilibrium (line X). It is possible to envision a scenario where pure silicate glass partially equilibrates with metal both isotopically and chemically. In such a case the silicate glass data from an individual experiment

Fig. 2. Plot of 187Os/188Os versus 1/188Os for run products of high P–T experiments using Chinga (iron meteorite) and 91117 (komatiitic basalt) as starting materials. The line M indicates binary mixing between two starting materials. The horizontal bold line represents isotopic equilibration between starting metal and silicate. Vertical broken lines indicate glass samples in chemical equilibrium with metal defined by uniform Dmetal/silicate values. Data from a given experimental condition define linear correlations Os of which one end converges at Chinga. This suggests metal-nugget contamination in the quenched glass. Line A is a projection for glass samples processed at 1 GPa, 1450 °C for ≤25 h, and the end-member component (filled star). It gives the maximum Dmetal/silicate Os of 2.3 × 105 for that P–T condition. Line B and the open star are data for experiment #53 run at 1 GPa, 1450 °C for 7 days. All Os data plotted are corrected for a constant procedural blank contribution (Os = 100 fg, 187Os/188Os = 0.18), and gray bars represent analytical uncertainties caused by the correlated uncertainties for blank corrections (Os = 50–150 fg, 187Os/188Os = 0.18; see the text in detail).

formed in the Pt experiments were products of quenching from a homogeneous liquid. Quenching evidently rearranged the very local metal/silicate distributions of Pt on the scale of the micro-nugget spacing within the charges. The isotopic disequilibrium observed in the present study between the Os in metal and silicate is not consistent with the observations made on Pt and indicates the need to reexamine Pt partitioning with an isotopic dimension to confirm whether or not Pt isotopic compositions are uniform as expected for products quenched from a uniform liquid to glass plus metal nuggets. 4.2. Osmium concentrations and isotope compositions in pure silicate glasses The Os data for experiments involving the iron meteorite Chinga and the komatiitic basalt 91117 are plotted in the 187Os/188Os versus 1/188Os diagram (Fig. 2). In this diagram, all data depart from a simple mixing line between the two starting materials. Instead, data for run products from a given experimental condition typically form linear trends that extend from the starting iron meteorite composition to compositions with Os concentrations that are lower than the starting silicate, and more elevated 187Os/188Os ratios than the iron. Fig. 3a is a schematic diagram illustrating hypothetical silicate glass compositions that might occur by achievement of chemical equilibrium or isotopic equilibrium. Here we define the term “isotopic equilibrium” for the case where silicate glass and metal have identical Os isotope compositions (points along line X), whereas the term “chemical equilibrium” represents the case where silicate glass and metal have Os concentrations defined by the Dmetal/silicate value at the Os appropriate temperature and pressure (points along line Y). Note that

Fig. 3. (a) Schematic diagram showing possible silicate glass compositions for Os in experiments involving Chinga and 91117. Lines X and Y define possible isotopic and chemical equilibrium for Os between metal and silicate, respectively. The star represents thermodynamic equilibrium for Os concentration and isotope composition. Glass samples in chemical and isotopic disequilibrium would fall in the gray zone. (b) Schematic diagram showing the case when a pure silicate glass is partially equilibrated with metal both isotopically and chemically. The data for the silicate glass lie along a trajectory (broken arrow) between the starting silicate and the final composition (star) that is in isotopic and chemical equilibrium. (c) Schematic diagram showing the linear trend defined by glass samples (with entrained metal nuggets) from a single experimental condition that represent binary mixing between the starting iron and pure silicate glass without isotopic equilibrium. Because the extent of metal-nugget contamination in individual glass samples is unclear, the nugget-free component cannot be uniquely determined. The sample with the highest 1/188Os, therefore, gives the minimum Dmetal/silicate value, while the extrapolation of the trend toward the Os Os isotope composition of the starting silicate yields the maximum possible D value.

T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173 Table 6 Estimated minimum and maximum Dmetal/silicate values Os P and T


Dmin (× 105)

Dmax (× 105)

Chinga and 91117 1 GPa, 1450 °C 1 GPa, 1450 °C 1 GPa, 2200 °C 2 GPa, 2000 °C 2 GPa, N2300 °C

2–25 h 7 days 10 min 0.5–30 min 10–15 min

2.0+− 0.1 0.4 2.8+− 1.7 0.8 0.33+− 0.01 0.01 3.5+− 4.2 1.3 10.0+− 5.3 6.6

2.3 4.5 2.1 9.7 15.0

Filomena and Hawaiian basalt 1–2 GPa, 1450–2000 5 min–19.5 h

0.15+− 0.03 0.02


would lie along a linear trend between the starting silicate and a final melt composition that is in isotopic and chemical equilibrium with the metal, as schematically illustrated in Fig. 3b. Two lines of evidence suggest that this did not occur. First, data for individual experiments do not plot along comparable trends. Second, Ni data suggest that chemical equilibration between metal and silicate was achieved in these experiments. Electron microprobe analyses of the silicate glass for experiment #40 (1450 °C, 1 GPa, 23.5 h) indicate homogeneous Ni concentrations throughout the charge (48 ± 15 ppm; Table 2). This result yields a Dmetal/silicate value (~ 4 × 103) that is only slightly higher Ni than D values for Ni at similar P–T conditions (0.3–1 × 103) obtained by prior studies (Li and Agee, 2001; Righter et al., 1997; Thibault and Walter, 1995). The offset is presumably caused by clinopyroxene crystallization in the base of the silicate glass in our experiments (see Appendix B). Instead, the linear trends in Fig. 2 are interpreted to reflect binary mixing between the starting iron and pure silicate glass that is close to being in chemical equilibrium with the metal nuggets. The trends presumably are defined by and extend through metal nuggets of Chinga, which have been entrained in the silicate glass. The composition of the pure, nugget-free silicate glass for each experiment should plot along the extrapolations of the individual trends toward the direction of lower Os concentrations where chemical equilibrium is achieved, as schematically shown in Fig. 3c. Key to this interpretation is the observation that one of the glass chunks processed at 1 GPa, 1450 °C for 7.5 h (#34A) retained the Os isotope composition of the original basalt (187Os/188Os = 2.04) but had an Os concentration much lower than the starting silicate (1/188Os = 220). Evidently, this glass chunk was least affected by metal-nugget contamination. Its composition suggests that Os abundances in the silicate were substantially drawn down by interaction with metal, without the imposition of the metal isotopic composition on the silicate. This implies extremely low equilibrium concentrations of Os in the silicate melt compared to the metal. The metal nuggets scattered throughout the charge evidently have a much greater capacity to remove Os from the silicate than to exchange Os. As a consequence, the metal–silicate system was most likely in chemical equilibrium for Os concentration, but not in steady-state equilibrium with respect to Os isotopes.


of the starting silicate (Fig. 3c). For example, for the experiments involving Chinga and 91117 processed at 1 GPa and 1450 °C for 2–25 h, the Os concentration in sample #34A gives the minimum Dmetal/silicate Os 5 of 2.0+− 0.1 0.4 × 10 . The projection of the trend to the Os isotopic composition of the starting silicate gives Dmetal/silicate = 2.3 × 105 (filled star; Os Fig. 2). Table 6 summarizes the minimum and maximum Dmetal/silicate Os values obtained for individual experimental conditions. The silicate chunks obtained from an experiment (#53) that was run at 1 GPa, 1450 °C for longer periods of time (7 days) tend to define a shallower slope on this type of plot (line B; Fig. 2). The silicate chunk showing the lowest Os concentration at this condition (#53A) gives 5 Dmetal/silicate = 2.8+− 1.7 Os 0.8 × 10 . It is presumed that this chunk contains negligible metal nuggets. Alternatively, if this chunk was affected by metal-nugget contamination, the shallowing of the slope would mean that the D value is much higher than 2.8 × 105, and that as a consequence of the longer run time, Dmetal/silicate increased from ~ 2 × 105 Os (processed for ≤25 h) to 4.5 × 105 (processed for 7 days) (open star: Fig. 2). For higher P–T conditions (2 GPa, 2000–2400 °C), some glass chunks retained Os concentrations much lower than the starting silicate and also have much lower 187Os/188Os ratios. Minimum calculated Dmetal/silicate Os values are similar to, or slightly higher than those estimated from lower P–T conditions (Table 6). Our preferred interpretation of the collective data is that the isotopic exchange of Os between metal and silicate was very sluggish, but that the experiments were ultimately proceeding towards isotopic equilibrium. This means that the rate of chemical equilibrium for Os is much faster than the isotopic exchange rate. In contrast, if we take the maximum Dmetal/silicate values estimated from the Os extrapolation method, the data indicate further progress of chemical partitioning of Os into metal, with very little isotopic exchange, to overshoot Dmetal/silicate ~ 2 × 105 and approach the higher values of Os metal/silicate DOs = 1.5 × 106. We deduce this latter case is unlikely because it permits no isotopic exchange, even at the highest P–T condition where the diffusion rate for Os should be greatest. Nevertheless, for either interpretation Os solubility remained extremely low and minimum Dmetal/silicate values of N2 × 105 are indicated for all P–T Os conditions examined by this study. In order to verify the observed Os behavior under high P–T conditions a similar set of experiments was investigated using

4.3. Estimation of metal–silicate partition coefficients for Os Our data suggest that Os was removed from the starting silicate into metal that is dispersed throughout the charge as micro-nuggets. If it is presumed that the nugget-free silicate is entirely homogeneous regarding Os and chemically equilibrated with the metal, the pure silicate composition would fall somewhere along the linear trend in Fig. 2 that extends from the starting metal. The pure silicate composition cannot be uniquely determined because the extent of metal-nugget contamination in each glass sample analyzed is unclear. However, the glass sample which has the lowest Os concentration for a single experimental condition would give the minimum Dmetal/silicate Os value, whereas the maximum D can be obtained via extrapolation of the trend to a pure silicate composition that has the 187Os/188Os ratio

Fig. 4. Plot of 187Os/188Os versus 1/188Os for run products of high P–T experiments using Filomena (iron meteorite) and KI-75-1-139.3 (Hawaiian basalt) as the starting materials. The same projection approach as for Fig. 2 gives the maximum Dmetal/silicate Os of 2.9 × 104. All data plotted are corrected for a constant procedural blank contribution 187 188 (Os = 100 fg, Os/ Os = 0.18), and gray zones represent analytical uncertainties caused by variable blank corrections (Os = 50–150 fg, 187Os/188Os = 0.16–0.20; see the text in detail).


T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173

different starting materials. Isotopic “reversal” experiments were conducted using the iron meteorite Filomena and a Kilauea Iki (Hawaii) basalt with 187Os/188Os ratios that were reversed relative to the prior experiments (Table 3). Data for the experiments involving these starting materials define similar trends on the 187Os/188Os and 1/188Os diagram (Fig. 4). The minimum and maximum Dmetal/silicate values are calculated Os by the same approach (1.5–2.9 × 104; Table 6), although they are less well constrained because of the fewer data and the larger error envelopes. These values are lower than the experiments using the Chinga– komatiitic basalt pair and may be due to the different chemical compositions of the starting materials. Exploration of the effects of additional changes in chemical parameters will bear further examination. 4.4. Implication of high P–T Dmetal/silicate Os As described above, we conclude that our data provide Dmetal/silicate N Os 2 × 105 at pressures as high as 2 GPa, and temperatures as high as 2400 °C. In prior 1 atm experiments, it was reported that Os solubility strongly correlates with the increase of oxygen fugacity, which gives Dmetal/silicate ~ 1 × 105 at an fO2 condition of IW-2 (Borisov and Walker, Os 2000; Fortenfant et al., 2006). This is in general agreement with our high P–T Dmetal/silicate values. The retention of a high D value for Os at high P–T Os contrasts with some prior observations that showed dramatic decreases in Dmetal/silicate values for some moderately siderophile elements (P, W, Co, Ni and Mo) (Chabot et al., 2005; Righter, 2003) and HSE (Pt, Au and Pd) (Cottrell and Walker, 2006; Ertel et al., 2006; Holzheid et al., 2000; Righter et al., 2008) at elevated pressures (up to 20 GPa) and temperatures (N1500 °C). The Dmetal/silicate value required for equilibrium core–mantle Os partitioning can be calculated from the following equation metal=silicate DOs

1=f −1 = 1=x−1


Acknowledgement We thank I.S. Puchtel, R,T. Helz and the Smithsonian Institution (Washington DC) for providing the starting materials. Helpful comments by R.W. Carlson and two anonymous reviewers improved the manuscript as did discussions with J. Brenan and W.F. McDonough. This research was supported by NSF CSEDI grants EAR0757808 (to RJW) and EAR0757853 (to DW), and by NASA grant NNX07AM29G (to RJW). These sources of funding are gratefully acknowledged. Appendix A. Supplementary data


where x stands for the mass fraction of silicate Earth (= 0.675) (Anderson and Kovach, 1967) and f represents depletion of Os abundance in the bulk silicate Earth (BSE) relative to CI chondrites that are normalized to Mg abundances to correct for volatile enrichments in CI chondrites: f=

The new data also allow a preliminary evaluation of the kinetics of Os isotope exchange between metal and silicate under high P–T conditions. At high P–T conditions, Os was removed from the starting silicate into micro-nuggets that dispersed throughout the charge. None of the pure silicate glasses in this study showed complete Os isotopic equilibrium with the starting metal despite, in some cases, dramatic changes in concentration. This suggests that for Os, chemical equilibrium and isotopic exchange between metal and silicate proceeded at very different rates, at least within the length-scale and time-scales of our experimental conditions. Isotopic exchange was evidently much slower than the rate of chemical equilibration. This may imply that isotopic exchange of HSE between planetary cores and silicate mantles subsequent to primordial differentiation, without substantial accompanying chemical exchange, may be limited. Our data at present do not permit a quantitative evaluation of the isotopic exchange rate of Os on geological timescales of millions of years. In addition to Os, exploration of isotopic exchange for other elements, using similar methodologies, may prove key in assessing the likelihood of geochemical detection of possible core–mantle interaction (e.g. Hf–W, U–Pb) (Wood et al., 2006).


By using Mg abundances of McDonough and Sun (1995) and Os abundances in CI chondrites (450 ppb) (Horan et al., 2003) and PUM as substitute for BSE (3.9 ppb) (Becker et al., 2006), we obtain a necessary Dmetal/silicate = 5.6 × 102 in the case of equilibrium partitioning. ThereOs fore, our high P–T Dmetal/silicate values are much higher than permitted Os to explain the Os abundance in the silicate Earth via metal–silicate equilibration at high P–T, and alternate mechanisms for generating the relatively high abundances observed may be necessary. 5. Concluding remarks We have devised a new methodology to experimentally determine metal–silicate partition coefficients for Os under high P–T conditions by utilizing iron meteorites and basalts for starting materials. The run product silicates have heterogeneous Os concentrations and isotopic compositions due to micro-nugget contamination. Osmium concentrations projected for nugget-free silicates indicate extremely low Os solubility in silicate (Dmetal/silicate N 2 × 105) at P–T conditions as high as Os 2 GPa, 2400 °C. This indicates that metal–silicate segregation during planetary differentiation at high P–T conditions may leave silicate that is extremely depleted in Os compared to the observed Os abundance in the silicate Earth.

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.epsl.2008.12.046. Appendix B. Additional petrographic information The compositional variation in the base of the silicate glass deserves special comment. One hypothesis is that clinopyroxene growth during the experiment (it was not present at the start) induces changes in the base of the feed liquid. Chemical depletions and enrichments occur in the liquid adjacent to the growing crystal by incorporation and rejection of liquid ingredients into the growing crystal. This hypothesis is difficult to sustain because the crystal and liquid have the same Si abundance away from the interface, yet there are complementary gradients on both sides of the interface. Also Mg and Ca are both enriched in the crystal compared to the liquid and therefore should be depleting the liquid during crystal growth. However the Mg and Ca abundances in liquid increase toward the crystal rather than decrease as they should in the liquid depletion model. An alternate model that we feel better fits the facts is that the base of the charge has a slightly lower temperature than the rest of the charge ~1450 °C. These charges are unusually long, to maximize their product yield. If their base extends a few 10s of degrees below the target temperature, exactly the type and small magnitude of chemical gradient observed in the liquid would be expected to develop by Soret differentiation (Walker and Delong, 1982). Once the Soret gradient begins to develop, clinopyroxene growth at the cold spot begins by burying the Ni-rich nuggets already present at the base of the charge. Clinopyroxene at the base is 4 times richer than the silicate in Ni (Table 2), so clinopyroxene growth rapidly depletes the basal silicate liquid in Ni, with the result that any new metal nugget growth in parallel with the clinopyroxene growth will be Ni-depleted, as observed (see Section 3.1). The clinopyroxene also becomes depleted by a factor of ~ 10 upwards in its growth stratigraphy. We feel that the crystals are a consequence of liquid-state

T. Yokoyama et al. / Earth and Planetary Science Letters 279 (2009) 165–173

compositional changes rather than their cause, because the Soretfollowed-by-clinopyroxene-growth hypothesis makes sense of the gradients and partitioning between the crystals and also of the Ni abundance stratigraphy of metal nuggets trapped within the clinopyroxene layer. An important implication of this reasoning is that Ni-rich metal nuggets were present before clinopyroxene growth and that further metal nuggets were coprecipitated with the clinopyroxene. Neither population was generated by quenching of a homogeneous silicate liquid. This petrography-based conclusion reinforces the Os isotopic-based conclusion that the charge was heterogeneous in its metal distribution at the pressure and temperature of the experiment. References Anderson, D.L., Kovach, R.L., 1967. Composition of terrestrial planets. Earth Planet. Sci. Lett. 3, 19–&. Becker, H., Horan, M.F., Walker, R.J., Gao, S., Lorand, J.P., Rudnick, R.L., 2006. Highly siderophile element composition of the Earth's primitive upper mantle: constraints from new data on peridotite massifs and xenoliths. Geochim. Cosmochim. Acta 70, 4528–4550. Birck, J.L., Barman, M.R., Capmas, F., 1997. Re–Os isotopic measurements at the femtomole level in natural samples. Geostand. Newsl. 21, 19–27. Borisov, A., Palme, H., 1997. Experimental determination of the solubility of platinum in silicate melts. Geochim. Cosmochim. Acta 61, 4349–4357. Borisov, A., Walker, R.J., 2000. Os solubility in silicate melts: new efforts and results. Am. Mineral. 85, 912–917. Brandon, A.D., Walker, R.J., 2005. The debate over core–mantle interaction. Earth Planet. Sci. Lett. 232, 211–225. Chabot, N.L., Agee, C.B., 2003. Core formation in the Earth and Moon: new experimental constraints from V, Cr, and Mn. Geochim. Cosmochim. Acta 67, 2077–2091. Chabot, N.L., Draper, D.S., Agee, C.B., 2005. Conditions of core formation in the Earth: constraints from nickel and cobalt partitioning. Geochim. Cosmochim. Acta 69, 2141–2151. Cook, D.L., Walker, R.J., Horan, M.F., Wasson, J.T., Morgan, J.W., 2004. Pt–Re–Os systematics of group IIAB and IIIAB iron meteorites. Geochim. Cosmochim. Acta 68, 1413–1431. Corgne, A., Keshav, S., Wood, B.J., McDonough, W.F., Fei, Y.W., 2008. Metal–silicate partitioning and constraints on core composition and oxygen fugacity during Earth accretion. Geochim. Cosmochim. Acta 72, 574–589. Cottrell, E., Walker, D., 2006. Constraints on core formation from Pt partitioning in mafic silicate liquids at high temperatures. Geochim. Cosmochim. Acta 70, 1565–1580. Day, J.M.D., Pearson, D.G., Taylor, L.A., 2007. Highly siderophile element constraints on accretion and differentiation of the Earth–Moon system. Science 315, 217–219. Ertel, W., O'Neill, H.S., Sylvester, P.J., Dingwell, D.B., 1999. Solubilities of Pt and Rh in a haplobasaltic silicate melt at 1300 °C. Geochim. Cosmochim. Acta 63, 2439–2449. Ertel, W., O'Neill, H.S., Sylvester, P.J., Dingwell, D.B., Spettel, B., 2001. The solubility of rhenium in silicate melts: implications for the geochemical properties of rhenium at high temperatures. Geochim. Cosmochim. Acta 65, 2161–2170. Ertel, W., Walter, M.J., Drake, M.J., Sylvester, P.J., 2006. Experimental study of platinum solubility in silicate melt to 14 GPa and 2273 K: implications for accretion and core formation in Earth. Geochim. Cosmochim. Acta 70, 2591–2602. Fortenfant, S.S., Dingwell, D.B., Ertel-Ingrisch, W., Capmas, F., Birck, J.L., Dalpe, C., 2006. Oxygen fugacity dependence of Os solubility in haplobasaltic melt. Geochim. Cosmochim. Acta 70, 742–756. Helz, R.T., Kirschenbaum, H., Marinenko, J.W., 1989. Whole-rock analyses of drill core from Kilauea Iki Lava lake, Kilauea volcano, Hawaii. U.S. Geological Survey Open-File Report, pp. 1–30. Hillgren, V.J., Drake, M.J., Rubie, D.C., 1994. High-pressure and high-temperature experiments on core–mantle segregation in the accreting Earth. Science. 264, 1442–1445. Holzheid, A., Sylvester, P., O'Neill, H.S.C., Rubie, D.C., Palme, H., 2000. Evidence for a late chondritic veneer in the Earth's mantle from high-pressure partitioning of palladium and platinum. Nature 406, 396–399.


Horan, M.F., Walker, R.J., Morgan, J.W., Grossman, J.N., Rubin, A.E., 2003. Highly siderophile elements in chondrites. Chem. Geol. 196, 5–20. Jones, J.H., Drake, M.J., 1986. Geochemical constraints on core formation in the Earth. Nature 322, 221–228. Jones, J.H., Neal, C.R., Ely, J.C., 2003. Signatures of the highly siderophile elements in the SNC meteorites and Mars: a review and petrologic synthesis. Chem. Geol. 196, 21–41. Kimura, K., Lewis, R.S., Anders, E., 1974. Distribution of gold and rhenium between nickel–iron and silicate melts: implications for the abundance of siderophile elements on the Earth and Moon. Geochim. Cosmochim. Acta 38, 683–701. Li, J., Agee, C.B., 2001. The effect of pressure, temperature, oxygen fugacity and composition on partitioning of nickel and cobalt between liquid Fe–Ni–S alloy and liquid silicate: Implications for the Earth's core formation. Geochim. Cosmochim. Acta 65, 1821–1832. Luck, J.M., Allegre, C.J., 1983. Re-187–Os-187 systematics in meteorites and cosmochemical consequences. Nature 302, 130–132. Luguet, A., Pearson, D.G., Nowell, G.M., Dreher, S.T., Coggon, J.A., Spetsius, Z.V., Parman, S.W., 2008. Enriched Pt–Re–Os isotope systematics in plume lavas explained by metasomatic sulfides. Science 319, 453–456. McDonough, W.F., Sun, S.-S., 1995. The composition of the Earth. Chem. Geol. 120, 223–253. Newsom, H.E., 1990. Accretion and core formation in the Earth: evidence from siderophile elements. In: Newsom, H.E., Jones, J.H. (Eds.), Origin of the Earth. Oxford Univ. Press, New York, pp. 273–288. Nier, A.O., 1950. A redetermination of the relative abundances of the isotopes of carbon, nitrogen, oxygen, argon, and potassium. Phys. Rev. 77, 789–793. O'Neill, H.S.C., Dingwell, D.B., Borisov, A., Spettel, B., Palme, H., 1995. Experimental petrochemistry of some highly siderophile elements at high-temperatures, and some implications for core formation and the mantles early history. Chem. Geol. 120, 255–273. Picher, L.L., Helz, R.T., Walker, R.J., Piccoli, P.M., in press. Fractionation of the platinumgroup elements and Re during crystallization of basalt in Kilauea Iki lava lake, Hawaii. Chem. Geol. Puchtel, I.S., Hofmann, A.W., Mezger, K., Shchipansky, A.A., Kulikov, V.S., Kulikova, V.V., 1996. Petrology of a 2.41 Ga remarkably fresh komatiitic basalt lava lake in Lion Hills, central Vetreny Belt, Baltic Shield. Contrib. Mineral. Petrol. 124, 273–290. Rasmussen, K.L., Malvin, D.J., Buchwald, V.F., Wasson, J.T., 1984. Compositional trends and cooling rates of group IVB iron meteorites. Geochim. Cosmochim. Acta 48, 805–813. Righter, K., 2003. Metal–silicate partitioning of siderophile elements and core formation in the early Earth. Annu. Rev. Earth Planet. Sci. 31, 135–174. Righter, K., Drake, M.J., Yaxley, G., 1997. Prediction of siderophile element metal–silicate partition coefficients to 20 GPa and 2800 °C: the effects of pressure, temperature, oxygen fugacity, and silicate and metallic melt compositions. Phys. Earth Planet. Inter. 100, 115–134. Righter, K., Humayun, M., Danielson, L., 2008. Partitioning of palladium at high pressures and temperatures during core formation. Nat. Geosci. 1, 321–323. Thibault, Y., Walter, M.J., 1995. The influence of pressure and temperature on the metal– silicate partition-coefficients of nickel and cobalt in a model C1 chondrite and implications for metal segregation in a deep magma ocean. Geochim. Cosmochim. Acta. 59, 991–1002. Walker, D., Delong, S.E., 1982. Soret separation of mid-ocean ridge basalt magma. Contrib. Mineral. Petrol. 79, 231–240. Walter, M.J., Tronnes, R.G., 2004. Early Earth differentiation. Earth Planet. Sci. Lett. 225, 253–269. Walker, R.J., Walker, D., 2005. Does the core leak? EOS 86, 237–242. Walker, R.J., Morgan, J.W., Horan, M.F., 1995. Os-187 enrichment in some plumes — evidence for core–mantle interaction. Science 269, 819–822. Walker, R.J., Horan, M.F., Shearer, C.K., Papike, J.J., 2004. Low abundances of highly siderophile elements in the lunar mantle: evidence for prolonged late accretion. Earth Planet. Sci. Lett. 224, 399–413. Wasson, J.T., Choi, B.G., Jerde, E.A., Ulff-Moller, F., 1998. Chemical classification of iron meteorites: XII. New members of the magmatic groups. Geochim. Cosmochim. Acta 62, 715–724. Wood, B.J., Walter, M.J., Wade, J., 2006. Accretion of the Earth and segregation of its core. Nature 441, 825–833.