Elsevier

Geochimica et Cosmochimica Acta

Volume 64, Issue 22, 15 November 2000, Pages 3887-3895
Geochimica et Cosmochimica Acta

Metal–silicate partitioning of Co, Ga, and W: dependence on silicate melt composition

https://doi.org/10.1016/S0016-7037(00)00489-0Get rights and content

Abstract

This study investigates the effect of silicate melt composition on metal/silicate partitioning for Co2+, Ga3+, and W4+ at 1300°C, 1 atm, and a log fO2 of −12. Five glasses in the system MgO–CaO–Al2O3–TiO2–SiO2 with nbo/t (nonbridging oxygens/tetrahedrally coordinated cations) values ranging from 0.25 to 1.52 were used as starting materials. For W and Co experiments the five glasses were equilibrated with W or Co wire loops, respectively, at the specified run conditions. For Ga experiments the glasses were doped with 2 wt.% Ga2O3 and equilibrated with pure Fe. All phases were analyzed by an electron microprobe. The metal/silicate partition coefficient for W depends strongly on melt basicity, whereas the effect of melt composition on Ga partitioning is less pronounced and for Co it is negligible. DW decreases rapidly with increasing nbo/t, DGa decreases moderately with increasing degree of melt basicity, and DCo remains relatively constant over our compositional range. The findings of this study indicate that the effect of melt composition on trace element solubility is a function of cation oxidation state such that high valency cations like W4+ are more readily dissolved in depolymerized melts which have a higher ratio of nonbridging oxygens, and lower valency cations like Co2+ are relatively independent of the parameter nbo/t. These results confirm that the composition of the primitive mantle is an important factor in constraining how siderophile trace elements distribute themselves between an Fe–metal core and the bulk silicate Earth during an early magma ocean differentiation event.

Introduction

A current hypothesis for the formation of Earth’s core is that a large-scale melting event occurred early in Earth’s history and, as a result, the planet differentiated into an Fe-rich metal core and a silicate mantle. Whether or not Earth’s core formation was an equilibrium process is a topic of current debate. The primary concern lies in developing a working model that can explain mantle depletions of all siderophile elements and the chondritic relative abundances of the highly siderophile elements in the upper mantle (see Jones and Drake 1986, Righter and Drake 1997, Wanke 1981). Therefore, in trying to understand the process of Earth’s core formation, it is necessary to constrain the metal/silicate partitioning behavior of the siderophile elements under the appropriate conditions.

The partitioning of an element between a metal and a silicate phase is dependent on temperature, pressure, oxygen fugacity, the chemical composition of the silicate melt, and the composition of the metal phase. Of these five variables, melt composition is the least investigated. It has been recognized for some time that melt composition has an influence on trace element partitioning. Watson (1976) showed that, in two-liquid systems, many elements have increased solubility in more basic (less polymerized) melts where more nonbridging oxygens are available to aid in the stable coordination of the cation. However, until recently the effect of melt composition on metal/silicate partitioning of the siderophile elements has been considered minor relative to the strong influence of oxygen fugacity. It began to receive attention as an important variable when studies by Walter and Thibault (1995) and Hillgren et al. (1996) independently demonstrated that silicate melt composition plays an important role in siderophile element partitioning at high temperatures and pressures. In a subsequent study, Jana and Walker (1997) found that the effect of melt composition is of greater importance in the high valency siderophile elements than in cations with lower valence states.

In all of the aforementioned studies there has been at least one free variable in addition to composition. In the present study we focus our investigation on the effect of melt composition on W, Ga, and Co partitioning by keeping temperature, pressure, metal composition, and fO2 constant. We find that the valence state of the dissolving species plays an important role in determining the effect of composition on solubility.

Section snippets

Starting materials

Five glasses in the system MgO–CaO–Al2O3–TiO2–SiO2 were prepared as starting materials for our experiments. We choose this system because it is liquid at the conditions of this study. The acidic and basic end-members (referred to as A and B) are approximately the eutectic compositions of the systems silica–anorthite–sphene (Agamawi and White, 1953) and diopside–CaTiAl2O6–akermanite (Yagi and Onuma, 1969), respectively, as indicated in Figure 1. These glasses were synthesized by grinding CaCO3

Analytical methods

All samples were analyzed using a Cameca SX50 electron microprobe with four wavelength dispersive spectrometers and a PAP ZAF correction program. Major elements were analyzed using a beam current of 20 nA on brass, an accelerating voltage of 15 kV, and a 10 second count time with a defocussed beam. The standards used were diopside glass (Mg, Si, and Ca); anorthite (Al); rutile (Ti); gallium arsenide (Ga) and pure metals for Fe, W, and Co. Tungsten was measured on the PET crystal with count

Results

Table 2, Table 3, Table 4 report the applied oxygen fugacity for each run and the average compositions of all Co, Ga, and W samples, respectively, as calculated from the raw electron microprobe data. In the ideal case each run would have a consistent oxygen fugacity of 10−12 atm O2 at the sample. In practice, however, experimental deviations from this value were observed. For Ga runs the deviations were well within the measurement error, so no correction was applied to Ga data. For Co and W

Time series experiments

Results from the time series experiments for W and Ga are plotted in Figure 3. Partition coefficients from the 1, 4, and 7 day runs agree closely with each other. The 2 h experiments are also consistent with this trend at the most basic compositions, but deviate with lower nbo/t. This is interpreted as evidence that although 2 h is not enough time to equilibrate the more acidic compositions, 1 day is adequate time to achieve equilibrium for all compositions. Similarly, Ga experiments were run

Conclusions

Under experimental conditions of 1 atm, 1300°C, and 10−12 atm O2, the Fe-liquid metal/liquid silicate partition coefficients for Co2+, Ga3+, and W4+ respond differently to variations in the silicate melt composition. Cobalt partitioning is seemingly unaffected by changes in the degree of polymerization of the silicate phase, while the W partition coefficient decreases rapidly with increasing melt basicity. The response of Ga partitioning to variations in nbo/t is intermediate to that of DCo and

Acknowledgements

Financial support for this research was provided by NASA Grant NAG59435. The authors thank David Walker for helpful reviews. The authors thank Chris Capobianco for invaluable advice in all stages of this project.

References (26)

  • K. Righter et al.

    Metal–silicate equilibrium in a homogeneously accreting earthNew results for Re

    Earth Planet. Sci. Lett.

    (1997)
  • W. Schmitt et al.

    Experimental determination of metal/silicate partition coefficients for P, Co, Ni, Cu, Ga, Ge, Mo, and W and some implications for the early evolution of the Earth

    Geochim. Cosmochim. Acta

    (1989)
  • Y.M. Agamawi et al.

    The quatermary system CaO-Al2O3-TiO2-SiO2, Part I. The ternary system anorthite-sphene-silica

    Br. Ceram. Soc. Trans.

    (1953)
  • Cited by (44)

    • Geochemical processes between steel projectiles and silica-rich targets in hypervelocity impact experiments

      2014, Geochimica et Cosmochimica Acta
      Citation Excerpt :

      That implies the W enrichment around PD3; could probably occur during cooling. Jaeger and Drake (2000), however, found that the partition coefficients of Co and W respond differently to variations in the silicate melt composition: Cobalt seems almost unaffected by changes in the degree of polymerization whereas, DWM/S decreases with increasing basicity of the silicate melt. The authors emphasize cations with higher oxidation states (W4+) can be better incorporated in basic (higher NBO/T – nonbridging oxygen per tetrahedral-coordinated cation, see Mysen et al., 1982) melts than cations with lower oxidation states (Co2+).

    • Partition Coefficients at High Pressure and Temperature

      2013, Treatise on Geochemistry: Second Edition
    • Structure-property relationships of COHN-saturated silicate melt coexisting with COHN fluid: A review of in-situ, high-temperature, high-pressure experiments

      2013, Chemical Geology
      Citation Excerpt :

      Dissolved COHN volatiles, melt polymerization and element partitioning. The NBO/T of melts is a useful variable against which mineral/melt element partition coefficients are often regressed (e.g., Walter and Thibault, 1995; Jaeger and Drake, 2000; Kushiro and Mysen, 2002; Toplis and Corgne, 2002; Mysen and Shang, 2005). Qualitatively, such relationships exist because most elements, be they major, minor or trace elements, are network-modifiers and form, therefore, bonding with nonbridging oxygen in melts.

    View all citing articles on Scopus
    View full text