Accretion of the Earth and segregation of its core

The Earth took 30-40 million years to accrete from smaller 'planetesimals'. Many of these planetesimals had metallic iron cores and during growth of the Earth this metal re-equilibrated with the Earth's silicate mantle, extracting siderophile ('iron-loving') elements into the Earth's iron-rich core. The current composition of the mantle indicates that much of the re-equilibration took place in a deep (> 400 km) molten silicate layer, or 'magma ocean', and that conditions became more oxidizing with time as the Earth grew. The high-pressure nature of the core-forming process led to the Earth's core being richer in low-atomic-number elements, notably silicon and possibly oxygen, than the cores of the smaller planetesimal building blocks.     

The Earth's missing lead may not be in the core

Relative to the CI chondrite class of meteorites (widely thought to be the 'building blocks' of the terrestrial planets), the Earth is depleted in volatile elements. For most elements this depletion is thought to be a solar nebular signature, as chondrites show depletions qualitatively similar to that of the Earth. On the other hand, as lead is a volatile element, some Pb may also have been lost after accretion. The unique (206)Pb/(204)Pb and (207)Pb/(204)Pb ratios of the Earth's mantle suggest that some lead was lost about 50 to 130 Myr after Solar System formation. This has commonly been explained by lead lost via the segregation of a sulphide melt to the Earth's core, which assumes that lead has an affinity towards sulphide. Some models, however, have reconciled the Earth's lead deficit with volatilization. Whichever model is preferred, the broad coincidence of U-Pb model ages with the age of the Moon suggests that lead loss may be related to the Moon-forming impact. Here we report partitioning experiments in metal-sulphide-silicate systems. We show that lead is neither siderophile nor chalcophile enough to explain the high U/Pb ratio of the Earth's mantle as being a result of lead pumping to the core. The Earth may have accreted from initially volatile-depleted material, some lead may have been lost to degassing following the Moon-forming giant impact, or a hidden reservoir exists in the deep mantle with lead isotope compositions complementary to upper-mantle values; it is unlikely though that the missing lead resides in the core.     

Partitioning of oxygen during core formation on the Earth and Mars

Core formation on the Earth and Mars involved the physical separation of metal and silicate, most probably in deep magma oceans. Although core-formation models explain many aspects of mantle geochemistry, they have not accounted for the large differences observed between the compositions of the mantles of the Earth (approximately 8 wt% FeO) and Mars (approximately 18 wt% FeO) or the smaller mass fraction of the martian core. Here we explain these differences as a consequence of the solubility of oxygen in liquid iron-alloy increasing with increasing temperature. We assume that the Earth and Mars both accreted from oxidized chondritic material. In a terrestrial magma ocean, 1,200-2,000 km deep, high temperatures resulted in the extraction of FeO from the silicate magma ocean owing to high solubility of oxygen in the metal. Lower temperatures of a martian magma ocean resulted in little or no extraction of FeO from the mantle, which thus remains FeO-rich. The FeO extracted from the Earth's magma ocean may have contributed to chemical heterogeneities in the lowermost mantle, a FeO-rich D" layer and the light element budget of the core.     

Evidence against a chondritic Earth

The (142)Nd/(144)Nd ratio of the Earth is greater than the solar ratio as inferred from chondritic meteorites, which challenges a fundamental assumption of modern geochemistry--that the composition of the silicate Earth is 'chondritic', meaning that it has refractory element ratios identical to those found in chondrites. The popular explanation for this and other paradoxes of mantle geochemistry, a hidden layer deep in the mantle enriched in incompatible elements, is inconsistent with the heat flux carried by mantle plumes. Either the matter from which the Earth formed was not chondritic, or the Earth has lost matter by collisional erosion in the later stages of planet formation.     

Evidence for a late chondritic veneer in the Earth's mantle from high-pressure partitioning of palladium and platinum

The high-pressure solubility in silicate liquids of moderately siderophile 'iron-loving' elements (such as nickel and cobalt) has been used to suggest that, in the early Earth, an equilibrium between core-forming metals and the silicate mantle was established at the bottom of a magma ocean. But observed concentrations of the highly siderophile elements--such as the platinum-group elements platinum, palladium, rhenium, iridium, ruthenium and osmium--in the Earth's upper mantle can be explained by such a model only if their metal-silicate partition coefficients at high pressure are orders of magnitude lower than those determined experimentally at one atmosphere (refs 3-8). Here we present an experimental determination of the solubility of palladium and platinum in silicate melts as a function of pressure to 16 GPa (corresponding to about 500 km depth in the Earth). We find that both the palladium and platinum metal-silicate partition coefficients, derived from solubility, do not decrease with pressure--that is, palladium and platinum retain a strong preference for the metal phase even at high pressures. Consequently the observed abundances of palladium and platinum in the upper mantle seem to be best explained by a 'late veneer' addition of chondritic material to the upper mantle following the cessation of core formation.     

Non-chondritic distribution of the highly siderophile elements in mantle sulphides

The abundances of highly siderophile (iron-loving) elements (HSEs) in the Earth's mantle provide important constraints on models of the Earth's early evolution. It has long been assumed that the relative abundances of HSEs should reflect the composition of chondritic meteorites--which are thought to represent the primordial material from which the Earth was formed. But the non-chondritic abundance ratios recently found in several types of rock derived from the Earth's mantle have been difficult to reconcile with standard models of the Earth's accretion, and have been interpreted as having arisen from the addition to the primitive mantle of either non-chondritic extraterrestrial material or differentiated material from the Earth's core. Here we report in situ laser-ablation analyses of sulphides in mantle-derived rocks which show that these sulphides do not have chondritic HSE patterns, but that different generations of sulphide within single samples show extreme variability in the relative abundances of HSEs. Sulphides enclosed in silicate phases have high osmium and iridium abundances but low Pd/Ir ratios, whereas pentlandite-dominated interstitial sulphides show low osmium and iridium abundances and high Pd/Ir ratios. We interpret the silicate-enclosed sulphides as the residues of melting processes and interstitial sulphides as the crystallization products of sulphide-bearing (metasomatic) fluids. We suggest that non-chondritic HSE patterns directly reflect processes occurring in the upper mantle--that is, melting and sulphide addition via metasomatism--and are not evidence for the addition of core material or of 'exotic' meteoritic components.     

A diffusion mechanism for core-mantle interaction

Understanding the geochemical behaviour of the siderophile elements--those tending to form alloys with iron in natural environments--is important in the search for a deep-mantle chemical 'fingerprint' in upper mantle rocks, and also in the evaluation of models of large-scale differentiation of the Earth and terrestrial planets. These elements are highly concentrated in the core relative to the silicate mantle, but their concentrations in upper mantle rocks are higher than predicted by most core-formation models. It has been suggested that mixing of outer-core material back into the mantle following core formation may be responsible for the siderophile element ratios observed in upper mantle rocks. Such re-mixing has been attributed to an unspecified metal-silicate interaction in the reactive D' layer just above the core-mantle boundary. The siderophile elements are excellent candidates as indicators of an outer-core contribution to the mantle, but the nature and existence of possible core-mantle interactions is controversial. In light of the recent findings that grain-boundary diffusion of oxygen through a dry intergranular medium may be effective over geologically significant length scales and that grain boundaries can be primary storage sites for incompatible lithophile elements, the question arises as to whether siderophile elements might exhibit similar (or greater) grain-boundary mobility. Here we report experimental results from a study of grain-boundary diffusion of siderophile elements through polycrystalline MgO that were obtained by quantifying the extent of alloy formation between initially pure metals separated by approximately 1 mm of polycrystalline MgO. Grain-boundary diffusion resulted in significant alloying of sink and source particles, enabling calculation of grain-boundary fluxes. Our computed diffusivities were high enough to allow transport of a number of siderophile elements over geologically significant length scales (tens of kilometres) over the age of the Earth. This finding establishes grain-boundary diffusion as a potential fast pathway for chemical communication between the core and mantle.     

Aluminium control of argon solubility in silicate melts under pressure

Understanding of the crystal chemistry of the Earth's deep mantle has evolved rapidly recently with the gradual acceptance of the importance of the effect of minor elements such as aluminium on the properties of major phases such as perovskite. In the early Earth, during its formation and segregation into rocky mantle and iron-rich core, it is likely that silicate liquids played a large part in the transport of volatiles to or from the deep interior. The importance of aluminium on solubility mechanisms at high pressure has so far received little attention, even though aluminium has long been recognized as exerting strong control on liquid structures at ambient conditions. Here we present constraints on the solubility of argon in aluminosilicate melt compositions up to 25 GPa and 3,000 K, using a laser-heated diamond-anvil cell. The argon contents reach a maximum that persists to pressures as high as 17 GPa (up to 500 km deep in an early magma ocean), well above that expected on the basis of Al-free melt experiments. A distinct drop in argon solubility observed over a narrow pressure range correlates well with the expected void loss in the melt structure predicted by recent molecular dynamics simulations. These results provide a process for noble gas sequestration in the mantle at various depths in a cooling magma ocean. The concept of shallow partial melting as a unique process for extracting noble gases from the early Earth, thereby defining the initial atmospheric abundance, may therefore be oversimplified.     

Silicon in the Earth's core

Small isotopic differences between the silicate minerals in planets may have developed as a result of processes associated with core formation, or from evaporative losses during accretion as the planets were built up. Basalts from the Earth and the Moon do indeed appear to have iron isotopic compositions that are slightly heavy relative to those from Mars, Vesta and primitive undifferentiated meteorites (chondrites). Explanations for these differences have included evaporation during the 'giant impact' that created the Moon (when a Mars-sized body collided with the young Earth). However, lithium and magnesium, lighter elements with comparable volatility, reveal no such differences, rendering evaporation unlikely as an explanation. Here we show that the silicon isotopic compositions of basaltic rocks from the Earth and the Moon are also distinctly heavy. A likely cause is that silicon is one of the light elements in the Earth's core. We show that both the direction and magnitude of the silicon isotopic effect are in accord with current theory based on the stiffness of bonding in metal and silicate. The similar isotopic composition of the bulk silicate Earth and the Moon is consistent with the recent proposal that there was large-scale isotopic equilibration during the giant impact. We conclude that Si was already incorporated as a light element in the Earth's core before the Moon formed.     

Density of hydrous silicate melt at the conditions of Earth's deep upper mantle

The chemical evolution of the Earth and the terrestrial planets is largely controlled by the density of silicate melts. If melt density is higher than that of the surrounding solid, incompatible elements dissolved in the melt will be sequestered in the deep mantle. Previous studies on dry (water-free) melts showed that the density of silicate melts can be higher than that of surrounding solids under deep mantle conditions. However, melts formed under deep mantle conditions are also likely to contain some water, which will reduce the melt density. Here we present data constraining the density of hydrous silicate melt at the conditions of approximately 410 km depth. We show that the water in the silicate melt is more compressible than the other components, and therefore the effect of water in reducing melt density is markedly diminished under high-pressure conditions. Our study indicates that there is a range of conditions under which a (hydrous) melt could be trapped at the 410-km boundary and hence incompatible elements could be sequestered in the deep mantle, although these conditions are sensitive to melt composition as well as the composition of the surrounding mantle.     

Chemical interaction of Fe and Al(2)O3 as a source of heterogeneity at the Earth's core-mantle boundary

Seismological studies have revealed that a complex texture or heterogeneity exists in the Earth's inner core and at the boundary between core and mantle. These studies highlight the importance of understanding the properties of iron when modelling the composition and dynamics of the core and the interaction of the core with the lowermost mantle. One of the main problems in inferring the composition of the lowermost mantle is our lack of knowledge of the high-pressure and high-temperature chemical reactions that occur between iron and the complex Mg-Fe-Si-Al-oxides which are thought to form the bulk of the Earth's lower mantle. A number of studies have demonstrated that iron can react with MgSiO3-perovskite at high pressures and high temperatures, and it was proposed that the chemical nature of this process involves the reduction of silicon by the more electropositive iron. Here we present a study of the interaction between iron and corundum (Al(2)O3) in electrically- and laser-heated diamond anvil cells at 2,000-2,200 K and pressures up to 70 GPa, simulating conditions in the Earth's deep interior. We found that at pressures above 60 GPa and temperatures of 2,200 K, iron and corundum react to form iron oxide and an iron-aluminium alloy. Our results demonstrate that iron is able to reduce aluminium out of oxides at core-mantle boundary conditions, which could provide an additional source of light elements in the Earth's core and produce significant heterogeneity at the core-mantle boundary.     

Spin crossover and iron-rich silicate melt in the Earth's deep mantle

A melt has greater volume than a silicate solid of the same composition. But this difference diminishes at high pressure, and the possibility that a melt sufficiently enriched in the heavy element iron might then become more dense than solids at the pressures in the interior of the Earth (and other terrestrial bodies) has long been a source of considerable speculation. The occurrence of such dense silicate melts in the Earth's lowermost mantle would carry important consequences for its physical and chemical evolution and could provide a unifying model for explaining a variety of observed features in the core-mantle boundary region. Recent theoretical calculations combined with estimates of iron partitioning between (Mg,Fe)SiO(3) perovskite and melt at shallower mantle conditions suggest that melt is more dense than solids at pressures in the Earth's deepest mantle, consistent with analysis of shockwave experiments. Here we extend measurements of iron partitioning over the entire mantle pressure range, and find a precipitous change at pressures greater than ～76?GPa, resulting in strong iron enrichment in melts. Additional X-ray emission spectroscopy measurements on (Mg(0.95)Fe(0.05))SiO(3) glass indicate a spin collapse around 70?GPa, suggesting that the observed change in iron partitioning could be explained by a spin crossover of iron (from high-spin to low-spin) in silicate melt. These results imply that (Mg,Fe)SiO(3) liquid becomes more dense than coexisting solid at ～1,800?km depth in the lower mantle. Soon after the Earth's formation, the heat dissipated by accretion and internal differentiation could have produced a dense melt layer up to ～1,000?km in thickness underneath the solid mantle. We also infer that (Mg,Fe)SiO(3) perovskite is on the liquidus at deep mantle conditions, and predict that fractional crystallization of dense magma would have evolved towards an iron-rich and silicon-poor composition, consistent with seismic inferences of structures in the core-mantle boundary region.     

Ancient, highly heterogeneous mantle beneath Gakkel ridge, Arctic Ocean

The Earth's mantle beneath ocean ridges is widely thought to be depleted by previous melt extraction, but well homogenized by convective stirring. This inference of homogeneity has been complicated by the occurrence of portions enriched in incompatible elements. Here we show that some refractory abyssal peridotites from the ultraslow-spreading Gakkel ridge (Arctic Ocean) have very depleted 187Os/188Os ratios with model ages up to 2 billion years, implying the long-term preservation of refractory domains in the asthenospheric mantle rather than their erasure by mantle convection. The refractory domains would not be sampled by mid-ocean-ridge basalts because they contribute little to the genesis of magmas. We thus suggest that the upwelling mantle beneath mid-ocean ridges is highly heterogeneous, which makes it difficult to constrain its composition by mid-ocean-ridge basalts alone. Furthermore, the existence of ancient domains in oceanic mantle suggests that using osmium model ages to constrain the evolution of continental lithosphere should be approached with caution.     

Constraints on the composition of the Earth's core from ab initio calculations

Knowledge of the composition of the Earth's core is important for understanding its melting point and therefore the temperature at the inner-core boundary and the temperature profile of the core and mantle. In addition, the partitioning of light elements between solid and liquid, as the outer core freezes at the inner-core boundary, is believed to drive compositional convection, which in turn generates the Earth's magnetic field. It is generally accepted that the liquid outer core and the solid inner core consist mainly of iron. The outer core, however, is also thought to contain a significant fraction of light elements, because its density--as deduced from seismological data and other measurements--is 6-10 per cent less than that estimated for pure liquid iron. Similar evidence indicates a smaller but still appreciable fraction of light elements in the inner core. The leading candidates for the light elements present in the core are sulphur, oxygen and silicon. Here we obtain a constraint on core composition derived from ab initio calculation of the chemical potentials of light elements dissolved in solid and liquid iron. We present results for the case of sulphur, which provide strong evidence against the proposal that the outer core is close to being a binary iron-sulphur mixture.     

Mixing, volatile loss and compositional change during impact-driven accretion of the Earth

The degree to which efficient mixing of new material or losses of earlier accreted material to space characterize the growth of Earth-like planets is poorly constrained and probably changed with time. These processes can be studied by parallel modelling of data from different radiogenic isotope systems. The tungsten isotope composition of the silicate Earth yields a model timescale for accretion that is faster than current estimates based on terrestrial lead and xenon isotope data and strontium, tungsten and lead data for lunar samples. A probable explanation for this is that impacting core material did not always mix efficiently with the silicate portions of the Earth before being added to the Earth's core. Furthermore, tungsten and strontium isotope compositions of lunar samples provide evidence that the Moon-forming impacting protoplanet Theia was probably more like Mars, with a volatile-rich, oxidized mantle. Impact-driven erosion was probably a significant contributor to the variations in moderately volatile element abundance and oxidation found among the terrestrial planets.     

A crystallizing dense magma ocean at the base of the Earth's mantle

The distribution of geochemical species in the Earth's interior is largely controlled by fractional melting and crystallization processes that are intimately linked to the thermal state and evolution of the mantle. The existence of patches of dense partial melt at the base of the Earth's mantle, together with estimates of melting temperatures for deep mantle phases and the amount of cooling of the underlying core required to maintain a geodynamo throughout much of the Earth's history, suggest that more extensive deep melting occurred in the past. Here we show that a stable layer of dense melt formed at the base of the mantle early in the Earth's history would have undergone slow fractional crystallization, and would be an ideal candidate for an unsampled geochemical reservoir hosting a variety of incompatible species (most notably the missing budget of heat-producing elements) for an initial basal magma ocean thickness of about 1,000 km. Differences in 142Nd/144Nd ratios between chondrites and terrestrial rocks can be explained by fractional crystallization with a decay timescale of the order of 1 Gyr. These combined constraints yield thermal evolution models in which radiogenic heat production and latent heat exchange prevent early cooling of the core and possibly delay the onset of the geodynamo to 3.4-4 Gyr ago.     

Cooling of the Earth and core formation after the giant impact

Kelvin calculated the age of the Earth to be about 24 million years by assuming conductive cooling from being fully molten to its current state. Although simplistic, his result is interesting in the context of the dramatic cooling that took place after the putative Moon-forming giant impact, which contributed the final approximately 10 per cent of the Earth's mass. The rate of accretion and core segregation on Earth as deduced from the U-Pb system is much slower than that obtained from Hf-W systematics, and implies substantial accretion after the Moon-forming impact, which occurred 45 +/- 5 Myr after the beginning of the Solar System. Here we propose an explanation for the two timescales. We suggest that the Hf-W timescale reflects the principal phase of core-formation before the giant impact. Crystallization of silicate perovskite in the lower mantle during this phase produced Fe(3+), which was released during the giant impact, and this oxidation resulted in late segregation of sulphur-rich metal into which Pb dissolved readily, setting the younger U-Pb age of the Earth. Separation of the latter metal then occurred 30 +/- 10 Myr after the Moon-forming impact. Over this time span, in surprising agreement with Kelvin's result, the Earth cooled by about 4,000 K in returning from a fully molten to a partially crystalline state.     

Water content in the transition zone from electrical conductivity of wadsleyite and ringwoodite

The distribution of water in the Earth's interior reflects the way in which the Earth has evolved, and has an important influence on its material properties. Minerals in the transition zone of the Earth's mantle (from approximately 410 to approximately 660 km depth) have large water solubility, and hence it is thought that the transition zone might act as a water reservoir. When the water content of the transition zone exceeds a critical value, upwelling flow might result in partial melting at approximately 410 km, which would affect the distribution of certain elements in the Earth. However, the amount of water in the transition zone has remained unknown. Here we determined the effects of water and temperature on the electrical conductivity of the minerals wadsleyite and ringwoodite to infer the water content of the transition zone. We find that the electrical conductivity of these minerals depends strongly on water content but only weakly on temperature. By comparing these results with geophysically inferred conductivity, we infer that the water content in the mantle transition zone varies regionally, but that its value in the Pacific is estimated to be approximately 0.1-0.2 wt%. These values significantly exceed the estimated critical water content in the upper mantle, suggesting that partial melting may indeed occur at approximately 410 km depth, at least in this region.     

A perovskitic lower mantle inferred from high-pressure, high-temperature sound velocity data

The determination of the chemical composition of Earth's lower mantle is a long-standing challenge in earth science. Accurate knowledge of sound velocities in the lower-mantle minerals under relevant high-pressure, high-temperature conditions is essential in constraining the mineralogy and chemical composition using seismological observations, but previous acoustic measurements were limited to a range of low pressures and temperatures. Here we determine the shear-wave velocities for silicate perovskite and ferropericlase under the pressure and temperature conditions of the deep lower mantle using Brillouin scattering spectroscopy. The mineralogical model that provides the best fit to a global seismic velocity profile indicates that perovskite constitutes more than 93 per cent by volume of the lower mantle, which is a much higher proportion than that predicted by the conventional peridotitic mantle model. It suggests that the lower mantle is enriched in silicon relative to the upper mantle, which is consistent with the chondritic Earth model. Such chemical stratification implies layered-mantle convection with limited mass transport between the upper and the lower mantle.     

Volatile accretion history of the Earth

It has long been thought that the Earth had a protracted and complex history of volatile accretion and loss. Albarède paints a different picture, proposing that the Earth first formed as a dry planet which, like the Moon, was devoid of volatile constituents. He suggests that the Earth's complement of volatile elements was only established later, by the addition of a small veneer of volatile-rich material at ～100 Myr (here and elsewhere, ages are relative to the origin of the Solar System). Here we argue that the Earth's mass balance of moderately volatile elements is inconsistent with Albarède's hypothesis but is well explained by the standard model of accretion from partially volatile-depleted material, accompanied by core formation.