2006 Wood et al., Accretion of the Earth and segregation of its core (1119308), страница 5
Текст из файла (страница 5)
Oxygen fugacityincreases by about three log units and sulphur-rich liquids are theonly metallic phases which can coexist with the putative magma ocean.The giant Moon-forming impact should have completely meltedthe Earth60, dramatically increasing the solubility of sulphur andchalcophile elements in the liquid silicate9. As the Earth cooled,however, an FeS-rich liquid would have ‘rained out’ of the liquidmantle, extracting chalcophile elements. Importantly, Pb is chalcophile, partitioning strongly into iron sulphide liquid64. Segregation of,1% of late sulphide liquid, the ‘Hadean Matte’8, would thereforehave fractionated U from Pb, re-setting the U–Pb age withoutaffecting the Hf–W age because the 182Hf parent was already extinct9.Thus the U–Pb age of the Earth (65–85 Myr after the beginning of theSolar System) plausibly represents the very last stages of coresegregation at the end of accretion, while the W age represents themajor phase of core segregation before the Moon-forming impact.The ‘light’ element in the coreAlthough the Earth’s core is, in Birch’s65 words, an ‘uncertain mixtureof all the elements’, we are able through geochemical and geophysicalarguments to constrain its composition quite well.
The chondriticreference model (Fig. 1), when combined with the compositions ofcarbonaceous chondrites66 and the primitive mantle3,26, leads to acore which is about 85% Fe and 5% Ni by weight. All of the otherrefractory elements discussed here should be present at very lowconcentration in the core, with the exceptions of Cr (#0.9%) and CoFigure 7 | Calculated effect of perovskite crystallization on the Fe31content and hence oxygen fugacity of a magma ocean of peridotitecomposition. The calculation began by assuming peridotitic mantle ofcurrent composition, containing 8% FeO. Perovskite crystallized from amelt of this composition contains about 3.5% ‘FeO’ but rather than beingentirely ferrous, half of the iron is actually ferric7,77.
The Fe3þ content of themolten mantle increases as perovskite crystallizes, then dissolves, releasingFe3þ, and finally recrystallizes at the base of the magma ocean. This‘processing’ of the mantle was converted from Fe3þ content to oxygenfugacity using experimental data on silicate melts78. Oxygen fugacities areexpressed relative to the iron-wüstite (Fe–FeO) buffer.© 2006 Nature Publishing GroupREVIEWSNATURE|Vol 441|15 June 2006(0.25%). This means that the core contains about 8 wt% of nonrefractory elements and this 8% also corresponds reasonably well tothe density deficit of the core compared to pure Fe–Ni alloy undercore conditions45,67.Because only elements of low atomic number can significantlylower the density of an Fe-rich core, the ‘missing’ 8% of core mass isgenerally referred to as the ‘light’ element.
The most likely candidatesare low-atomic-number elements that are soluble in Fe and that arecosmically abundant. This leads us to H, O, C, S, P and Si as possible‘light’ substituents in the core68. Of these, the clearest case can bemade for sulphur, which is a strongly siderophile element. Sulphur ismoderately volatile (Fig. 1), but is almost two orders of magnitudemore depleted in the silicate Earth than an element of similarvolatility, zinc (Zn) (Fig.
1). Making the reasonable assumptionthat this depletion of S relative to Zn in the silicate Earth is due tocore formation gives us a core concentration of 1.9 wt% S (refs 26, 29,69). The same procedure applied to phosphorus leads to about 0.2%of this element in the core29. Hydrogen and carbon are two cosmicallyabundant siderophile elements which are also highly volatile and hencestrongly depleted in the Earth. A rough estimate of their core concentrations based on an extrapolation of the volatility trend of Fig.
1 leadsto 0.1% and 0.2%, respectively29. This leaves the most controversial ofthe candidates, oxygen and silicon to make up the balance.The depletion of Si in the silicate Earth relative to the chondriticreference (Fig. 1) is due to an unknown combination of volatility(incomplete accretion) and dissolution in the core. If dissolution inthe core were the principal reason then the latter would contain6–7 wt% Si (refs 26 and 29) and Si would be the major ‘light’component. Low-pressure core segregation cannot achieve suchhigh Si concentrations without conditions being much too reducingfor the current FeO content of the mantle70. Si solubility in liquid Feincreases with both pressure and temperature6,71, however, such thatsignificant amounts would dissolve at pressures above 20 GPa, evenat the current FeO content of the mantle.
If we assume that, asdiscussed earlier (Fig. 4), most of the core formed under slightlymore reducing conditions, then the continuous growth and segregation models (Fig. 4) lead to 4–5 wt% Si in the core, which impliesthat, without making any explicit assumptions about volatility, Si isan important constituent. The oxygen content of the core is moredifficult to constrain because the available experimental data conflictwith one another. At low pressures and temperatures, oxygensolubility in liquid iron is negligible72.
All studies show that the effectof temperature is to increase solubility dramatically, however72,73, sothat at low pressures and 2,800 K iron liquid in equilibrium withliquid mantle would contain about 4 wt% oxygen73.The main controversy concerns the effect of pressure. Ohtaniet al.74 suggested that the effect of pressure is to increase oxygensolubility in Fe, whereas most recent studies73,75 have found theopposite effect. If we take the combined effects of temperature andpressure as found by Rubie et al.73, and apply them to metal–silicateequilibration on the peridotite liquidus, we find oxygen concentrations of up to about 1 wt% in the metal.
Lowering of oxygenfugacity during most of accretion decreases the oxygen content of themetal. Thus, if we accept the data indicating that pressure reducesthe solubility of oxygen in metallic iron, then the oxygen content ofthe core should be ,1%. This is an area of vigorous current debate,however, and a recent study in the diamond anvil cell76 found theopposite effect of pressure on oxygen solubility. Furthermore, atapproximately the appropriate oxygen fugacity for core–mantleequilibrium, Takafuji et al.76 found oxygen concentrations in themetal to be about 1.5 times the Si concentrations.
Clearly Si is likelyto be present at moderate concentrations in the core and oxygen mayalso be a significant substituent.physical models of planetary development lead to a coherent view ofthe early history of the Earth. The energetics of accretion45 implycontinuous core segregation from a partially molten mantle underconditions that generated the core–mantle partition coefficients ofTable 1. The siderophile elements Ni and Co have strongly pressuredependent partitioning and point towards a mean pressure of coresegregation greater than 30 GPa. Partitioning of more highly chargedweak siderophiles such as V, Cr, Si and Nb provide better constraintson mean temperature and oxygen fugacity of core formation. Thus,at pressures constrained principally by Ni and Co, these elementspoint to temperatures 650 K above the peridotite liquidus (Fig. 4).
Toreduce the required temperature to the liquidus, consistent with thedeep magma ocean model (Fig. 3), some part of core formation musthave taken place under conditions more reducing than those impliedby the current FeO content of the mantle6 (Fig. 4).
Thus, current bestestimates of metal–silicate partitioning are consistent with animportant component of the ‘heterogeneous accretion’ model, thatoxygen fugacity increased during accretion. The constraints providedby the partitioning data would, however, be stronger if there weremore experimental data at pressures above 15 GPa, particularly forthe weakly siderophile elements.Following the demonstration that lower-mantle perovskite forcesdisproportionation of Fe2þ to Fe3þ plus iron7, it has become clearthat the Earth’s mantle probably ‘self-oxidized’ during core formation6.This means that the increase in oxygen fugacity implied by the metal–silicate partitioning data could be explained as a natural consequence ofthe Earth having reached a size large enough to stabilize substantialamounts of perovskite in the lower mantle (Fig. 5).
The samemechanism could have operated on Venus, to a limited extent onMars and not at all on the Moon, or any of the asteroids.Self-oxidation during accretion has also been invoked as the reasonfor the difference between Hf–W and U–Pb ages of the core9. In thishypothesis (Fig.
6), W was partitioned into the core during the early‘reduced’ phase of metal segregation while Pb largely remainedbehind in the mantle. Progressive oxidation means, it is argued9,that the compositions of liquid metals extracted during the last stagesof core formation were sulphur-rich and Pb should partition stronglyinto such metals.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
