Terrestrial Accretion Under Oxidizing Conditions

See allHide authors and affiliations

Science  08 Mar 2013:
Vol. 339, Issue 6124, pp. 1194-1197
DOI: 10.1126/science.1227923

Earth's Ingredients

What was the composition of the earliest terrestrial starting blocks? The answer lies in understanding how Earth's interior separated into mantle and core components. Siebert et al. (p. 1194, published online 10 January) performed a series of high pressure and temperature experiments to track how chromium and vanadium, which have a slight affinity for iron, partition into metal and silicate fractions. Combined with accretionary models, the data suggest that Earth accreted under the same relatively oxidizing conditions under which the most common types of meteorites formed. Transferring oxygen in the form of FeO from the mantle to the core could have gradually reduced the mantle to its present-day oxidation state.


The abundance of siderophile elements in the mantle preserves the signature of core formation. On the basis of partitioning experiments at high pressure (35 to 74 gigapascals) and high temperature (3100 to 4400 kelvin), we demonstrate that depletions of slightly siderophile elements (vanadium and chromium), as well as moderately siderophile elements (nickel and cobalt), can be produced by core formation under more oxidizing conditions than previously proposed. Enhanced solubility of oxygen in the metal perturbs the metal-silicate partitioning of vanadium and chromium, precluding extrapolation of previous results. We propose that Earth accreted from materials as oxidized as ordinary or carbonaceous chondrites. Transfer of oxygen from the mantle to the core provides a mechanism to reduce the initial magma ocean redox state to that of the present-day mantle, reconciling the observed mantle vanadium and chromium concentrations with geophysical constraints on light elements in the core.

The depletion of siderophile (i.e., "iron-loving") elements in Earth's mantle relative to chondrites can constrain the redox state of accreting materials during terrestrial accretion and core differentiation (14). For example, metal-silicate partitioning experiments at atmospheric pressure indicate that the observed depletion of slightly siderophile elements (SSEs) such as V and Cr can only be produced at conditions more reducing than those required to account for the abundance of moderately siderophile elements (such as Ni, Co, and W) or highly siderophile elements (5). Using metal-silicate partition coefficients obtained at pressures up to 25 GPa, homogeneous accretion models posit that metal-silicate equilibrium took place at the base of a deep terrestrial magma ocean at a single oxygen fugacity (fO2) (68). However, the pressure-temperature (P-T) conditions required to produce the observed depletions for V and Cr (at the present-day fO2) require temperatures that greatly exceed that of the mantle liquidus (2, 4, 9, 10). Such conditions are physically inconsistent with the magma ocean hypothesis, where the P-T conditions at the base of a magma ocean necessarily lie between the mantle solidus and liquidus, thereby creating a rheological boundary that enables the metal to pond and equilibrate with the silicate melt.

To satisfy this rheological constraint and SSE abundance patterns, recent models of core formation constrain metal-silicate equilibration to the P-T conditions of the peridotite liquidus and invoke early accretion of highly reduced materials with a FeO-poor silicate component (2, 4, 10, 11). These initially low fO2 conditions (~IW-4, corresponding to 4 log f O2 units below the iron-wüstite buffer) enhance the siderophile character of the SSEs at the relevant P-T conditions. Subsequent, gradual oxidation of the mantle to ~IW-2 over the course of core formation is required to account for moderately siderophile element abundances and, most important, to reach the current mantle FeO content [8 weight percent (wt %) FeO in silicate]. Under reducing conditions, silicon is likely to be the only light element entering the core in large amounts (1012). This scenario relies on extensive pressure and temperature extrapolation of SSE partitioning data, as existing results are restricted to relatively low-P, low-T conditions (up to 25 GPa and 3000 K) and do not cover a large fraction of the P-T conditions relevant to core-mantle equilibrium in a deep magma ocean (up to 60 GPa and 4000 K) (13, 14). Should the solubility of additional components, such as oxygen, be enhanced in this P-T range, their influence on the thermodynamics of the metallic phase would not be captured by extrapolation.

Here, we present partitioning data for V and Cr at the conditions directly equivalent to the full range of P-T conditions relevant to core formation and metal-silicate equilibration at the base of a deep magma ocean (Fig. 1). The experiments were performed in a laser-heated diamond anvil cell (LHDAC) (15) between 35 and 74 GPa and between 3100 and 4400 K. We found that the mantle concentration of Cr and V can be satisfied in a single high-P, high-T equilibrium (Fig. 1) in relatively oxidizing conditions. In fact, core formation is a continuous process taking place over the course of accretion, with the core and mantle continuously growing and segregating from one another (2, 4). Metal-silicate partition coefficients for SSEs along any possible thermodynamic path (P, T, composition) hypothesized during core formation are required to accurately calculate mantle siderophile trace element concentrations. The P-T conditions of core formation necessarily increase as Earth grows, but there is no robust constraint on the evolution and composition of accretionary material. The composition of these materials does, however, establish the initial oxidation state of Earth's interior—an important variable controlling metal-silicate equilibrium. Hence, siderophile element partitioning provides a diagnostic by which the composition of accretionary materials may be inferred.

Fig. 1

(A and B) Exchange coefficients (KD) for V (A) and Cr (B) plotted as a function of reciprocal temperature: this work (red diamonds with labeled pressure conditions in GPa) and other studies (2, 9, 10, 1619) at different pressure conditions. Where no error bars are shown, the uncertainties do not exceed the symbol size. The solid lines are the result of least-squares regressions of previous lower-pressure partitioning data (2, 9, 10, 1619). The partitioning data from our DAC experiments, well above the low-pressure regression, are attributed to the influence of oxygen solubility in the metal on the activity coefficients of Cr and V in the metal. (C and D) Equilibrium constant K [i.e., KD corrected for activity coefficients (15), K = KD⋅(γMFen/2)] for V (C) and Cr (D) as a function of reciprocal temperature. This is the same data set as in (A) and (B), except that the exchange coefficient has been corrected for the activity coefficients of Fe, Cr, and V. These values depend strongly on oxygen content in the metallic phase. All data (O-bearing and O-depleted) fall on a single linear trend, showing the consistency of our activity model. Note that it is difficult to correct the activity in metals that are too O-rich, like our point at 74 GPa that has the highest O content in the metal (17 wt %), which falls systematically below the trend; however, high oxygen concentration increases the siderophile behavior of V and Cr.

Plausible accretionary composition models coupled with core formation scenarios can be tested against observables to constrain better the nature of the building blocks of Earth. For any core formation model, the primary metrics are (i) that final silicate mantle composition coincides with present-day mantle compositions, and (ii) that the composition of the core is consistent with geophysical observations. In terms of composition, the most important influence on partitioning is fO2, which describes the redox state of the accretionary environment and controls the FeO content of the silicate magma ocean. The effect of core formation on the siderophile trace element concentration in the mantle cannot be determined unless the partition coefficient is known as a function of P, T, and fO2. The Fe-normalized partition coefficient (KD) of an element M can be parameterized asEmbedded Image (1)where n is the valence of element M, γFe and γM are the activity coefficients of iron and cation M in the metal phase, respectively, and a, b, and c are regression constants.

In accord with results from large-volume press (LVP) experiments at lower pressure (2, 9, 10, 1619), we see that the KD values for Cr and V obtained in the LHDAC both increase with temperature between 35 and 74 GPa and are relatively insensitive to pressure (Fig. 1, A and B). However, the results of our experiments show that Cr and V are more siderophile (higher KD) than would be expected from the extrapolation of lower-P, lower-T data and directly match the terrestrial target values (Fig. 1, A and B). Indeed, all of our data lie above the trend extrapolated from lower-P, lower-T LVP experiments. This is due to another important parameter that influences metal-silicate partitioning: the presence of minor constituents in the metallic phase, which in turn affects the activity coefficients of the siderophile elements (the last term in Eq. 1). In the high-P, high-T, high-fO2 (~IW-1) conditions of our experiments (15), large amounts of oxygen (between 3.7 and 17 wt %) are present in the metallic phase (tables S2 and S3). This alters the activities of Cr and V in iron (20, 21) and explains the disparity between our results and those extrapolated from lower-P experiments in which the metallic phases contain negligible oxygen. We used the epsilon formalism (22) to calculate the activity coefficients for Cr and V due to the incorporation of oxygen, correcting both the LHDAC and lower-P data (2, 9, 10, 1619) to obtain an equilibrium constant K = KD⋅(γMFen/2), where γM is the activity coefficient for element M at the relevant oxygen concentration. The low-P and high-P data follow a linear trend (Fig. 1, C and D), supporting our contention that the apparent discrepancy between the low-P and high-P values of KD (Fig. 1, A and B) is entirely due to the change in activity coefficient resulting from enhanced oxygen solubility in our experiments. The regression constants a, b, and c and activity models using LVP data [for experiments with metal phases containing no additional light elements such as C and S (2, 9, 10, 1619)] combined with our data are reported in table S4 (15).

We used the results of our regression in a continuous core formation model (2, 4) to evaluate the influence of oxidation state on the core-mantle evolution of SSEs (15). We explored all possible magma ocean depths (from 0 to 100% of the mantle) and all possible geotherms relevant to the base of that magma ocean (temperatures between the mantle solidus and liquidus) (23, 24). Three accretionary models distinguished by the compositions of accretionary components and oxidation/reduction paths during planetary growth were considered: (i) a reduced model (4, 10), in which the initial 28% of accretion is characterized by the addition of highly reduced material (0.6 wt % FeO in the mantle) followed by continuous oxidation to yield a mantle with the current FeO content of 8 wt %; (ii) a homogeneous accretion model, in which the FeO content of the mantle is constant and equal to its present value over the entire course of accretion; and (iii) an oxidized model, in which Earth accretes from oxidized material, such as carbonaceous chondrites or a mixture (21% FeO in the mantle) of chondrites (25), and is then gradually reduced (through O solubility in the core) until the FeO content of the mantle reaches the present-day value.

The partition coefficients DV and DCr evolve as a function of magma ocean depth (Fig. 2) for the three accretionary scenarios (fig. S2 reports results for DNi and DCo). We find that the redox state of the accreting material only slightly influences the acceptable range of magma ocean depths, 50 to 56 (±7) GPa. In contrast to previous investigations (2, 4, 911, 26), we show that the terrestrial SSE partition coefficients (V and Cr) can be satisfied with any of our three accretion models. On the basis of our data, there is no requirement that accretion must have occurred under highly reducing conditions in order to explain present-day mantle siderophile trace element concentrations. This contradicts previous claims (2, 4, 10, 11) that V and Cr constrain the redox conditions of core formation and terrestrial accretion to allow only the highly reduced model.

Fig. 2

Final core-mantle partition coefficients of Cr and V (DCr and DV, top and middle rows) resulting from continuous core formation, as a function of final magma ocean pressure (i.e., depth). Solid lines show the evolution of the average partition coefficients; dashed lines define the uncertainty envelope calculated by Monte Carlo propagation of all uncertainties listed in table S4. The horizontal shaded bars represent the observed terrestrial target values for Cr and V (9). The bottom row shows the concentrations of O and Si in the core. The calculations are performed along three different oxygen fugacity paths: (A) the highly reduced model, (B) the homogeneous accretion model, and (C) the oxidized model. The relative size of the magma ocean with respect to that of the whole mantle was varied from a shallow magma ocean (final pressure of 0 GPa) to a global magma ocean (final pressure of 135 GPa). The gray vertical shaded bars show the pressure range where Cr and V abundances match the target values, along with additional constraints from Ni and Co partitioning (15). They show that the core must have formed at a final pressure between 40 and 60 GPa.

The fO2 path during terrestrial accretion also strongly affects the concentration of light elements in the core. On the basis of our LHDAC partitioning experiments, we calculate the O and Si concentrations in the core (table S5) and find that the highly reduced model leads to excessively high concentrations of silicon in the core, above 10 wt %, along with less than 1 wt % oxygen (Fig. 2, bottom row). The constant FeO–homogeneous accretion model leads to a core containing 2.5 to 5.5 wt % Si and 2 to 2.5 wt % O, whereas the oxidized path yields an oxygen-rich core with 4.5 to 5.5 wt % O and 1.5 to 2.2 wt % Si. The Si content of the core in the highly reduced model (>10 wt %) is incompatible with cosmochemical and geophysical constraints. Indeed, cosmochemical models (2729) suggest that the upper limit for Si content in the core is 8 wt %. High-pressure measurements of sound velocities in iron and iron alloys indicate that the core must contain between 1 and 4 wt % Si for it to be compatible with seismic radially averaged wave speeds and densities (30, 31).

Terrestrial accretion under relatively oxidizing conditions requires a mechanism for reducing the mantle to yield an oxidation state and FeO concentration compatible with present-day conditions. Because our model requires large concentrations of oxygen in the core, transferring FeO from the silicate mantle to the core solves both problems: reducing FeO in the mantle and increasing O in the core, while at the same time rendering the SSEs more siderophile because of the thermodynamic influence of enhanced oxygen solubility on their activities. This mechanism of FeO solubility in the core has been proposed to explain the difference in mantle redox between Earth and Mars (32).

A mutually compatible solution to the problems of light elements in the core and trace siderophile element concentrations in the mantle is a major discriminant in validating various models of accretion and core formation. On this basis, we show that it is unlikely that Earth formed under highly reduced conditions, such as those found in enstatite chondrites. Conversely, accretion under oxidized conditions similar to those of the most common meteorites (e.g., ordinary and/or carbonaceous chondrites) is consistent with the siderophile trace element concentrations in the mantle and with the geophysical constraints on the composition of the core. Finally, accretion of large objects (planetesimals and planetary embryos) that had redox conditions similar (~IW-2.3) to those of Earth also satisfies the geochemical and geophysical constraints, while at the same time proving the most realistic in terms of dynamical planetary formation models (33).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 and S2

Tables S1 to S5

References (3450)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: The research leading to these results has received funding from the European Research Council (ERC) under the European Community's Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement 207467. This work was performed under the auspices of the Lawrence Livermore National Laboratory, U.S. Department of Energy, under contract DE-AC52 07NA27344 and was supported by the Office of Basic Energy Sciences, Geosciences Research Program (F.J.R.). The Focused Ion Beam (FIB) facility of the IMPMC is supported by Région Ile de France grant SESAME 2006 N°I-07-593/R, INSU-CNRS, Institut de Physique (INP) CNRS, University Pierre et Marie Curie–Paris 6, and French National Research Agency (ANR) grant ANR-07-BLAN-0124-01 (J.B.). We thank I. Esteve for assistance during sample preparation with the FIB, and G. Garbarino for assistance on the high-pressure beamline (ID27) of the European Synchrotron Radiation Facility.
View Abstract

Navigate This Article