U11A-0001
A Tightly Constrained Composition Model of Earth's Core
We measure compressional sound velocities in light-element alloys of iron (FeO, FeSi, FeS, and FeS2) at high pressure in the diamond anvil cell. Combining this data set with pressure-density (equation of state) systematics, and constraining them with radial seismic models, we propose an average composition model of Earth's inner core that matches both the seismically observed densities and compressional sound speeds. We show that sulphur cannot be the only light alloying element in the core, because it cannot satisfy all at once the needed density, sound velocity, and abundance (from cosmochemical models). On the other hand, the incorporation of silicon or oxygen is compatible with geophysical observations and geochemical abundances. We therefore propose a preferred model, where the inner core contains 2.3 wt% silicon and traces of oxygen, whereas the outer core contains 5.3 wt% oxygen and 2.8 wt% silicon. Our model is in excellent agreement with recent high-pressure silicon solubility data in molten iron. Our model's oxygen-content in the outer core is constrained by high-pressure oxygen solubility data in molten iron. Our compositional model is constrained by seismology, petrology, and mineral physics. Using existing solubility data, we can infer a range of equilibration parameters that provide as many constraints to the pressure- temperature conditions as well as the depth at the base of the magma ocean during core formation. References N. Takafuji, K. Hirose, M. Mitome, Y. Bando, Solubilities of O and Si in liquid iron in equilibrium with (Mg,Fe)SiO3 perovskite and the light elements in the core, Geophys. Res. Lett. 32 (2005). T. Sakai, T. Kondo, E. Ohtani, H. Terasaki, N. Endo, T. Kuba, T. Suzuki, T. Kikegawa, Interaction between iron and post- perovskite at core-mantle boundary and core signature in plume source region, Geophys. Res. Lett. 33 (2006). J. Badro, G. Fiquet, F. Guyot, E. Gregoryanz, F. Occelli, D. Antonangeli, M. d'Astuto, Effect of light elements on the sound velocities in solid iron: Implications for the composition of Earth's core. Earth Planet. Sci. Lett. 254, 233 (2007).
U11A-0002
Effects of Oxygen Fugacity and Temperature on Partitioning of Ni and Co Between Liquid Metal, Magnesium Silicate Perovskite and Magnesiowustite
In the core formation stage, a magma ocean might have extended to the lower mantle depth, and Ni and Co might have been distributed throughout the magma ocean, solid lower mantle, and the Earthfs core. Although it is important to determine effects of oxygen fugacity (fO2) and temperature on partition behavior of these elements between liquid iron and lower mantle minerals, it has not been well constrained because of its experimental difficulty. Partitioning of Ni and Co between liquid iron-rich metal and lower mantle minerals, such as magnesium silicate perovskite and magnesiowustite, was investigated at 26-27 GPa, 2680-3020 K and 1.15- 1.86 log units below the iron-wustite fO2 buffer. The experiments have been achieved at such conditions using a Kawai-type apparatus and a cell assembly optimized for high-temperature generation. Our experiments revealed that partition coefficients of Ni and Co between liquid iron-rich metal and the lower mantle minerals decrease with increasing fO2, whereas they have very week temperature dependence in the present temperature range. The present result makes it possible to evaluate Ni and Co abundance throughout the mantle for a plausible fO2 and temperature. The mantle abundance of Ni and Co may be explained by the core formation in the deep magma ocean with a bottom pressure of around 65 GPa (corresponding to 1500 km depth).
U11A-0003
Laboratory Fluid Experiments on Metal-Silicate Plumes and Core Formation
Differentiation and formation of the Earth's iron core from proto-planetary material is suggested to have occured rapidly (< 30 My) under very energetic and violent accretionary conditions. Such high impact bombardments and evidence from mantle siderophile abundances suggest the presence of a deep magma ocean ({≥ 400} km) in the early Earth that facilitated metal-silicate equilibration. High temperatures of formation are also implied by the ancient dynamo, indicating a superheated core in the past. Conventional core formation models, however, have difficulty satisfying the simultaneous requirement of rapid formation with the residence time needed for metal-silicate equilibration. A favored model for core formation envisions impact-induced metal drops that equilibrate with silicates in the magma ocean, concentrate at the magma ocean base, and then descend to the core as large diapirs or instabilities. We report results from fluid experiments of liquid gallium in stratified sucrose solutions. To model core formation processes, we consider three cases of liquid metal instabilities through the stratified fluids: 1) descending single, small diameter metal drops; 2) metal ponding, instabililty, and descent of large diameter metal diapirs; and 3) ponding of tiny metal droplets in an emulsion, instability, and descent of emulsion diapirs. The experimental parameters represent viscosity ratios ({μsucrose/μga}) up to {106}, Reynolds numbers from {10-6} to {10-4}, and Bond numbers from 2.5 {10-1} - {101}. The most interesting observation in all experiments is the formation of trailing conduits behind sinking metal drops which fill with the upper layer, low viscosity material. These conduits provide extended residence time for chemical exchange and equilibrium between metal and silicate material during descent, after the metal pond goes unstable. We suggest the presence and long life of trailing, fluid-filled conduits may be a mechanism for reconciliation of models which require rapid core differentiation and formation and simultaneous metal-silicate equilibrium. Our experiments also indicate that metal-silicate mantle plumes may evolve into buoyant thermal plumes, connecting core formation to ancient hotspot activity on terrestrial planets.
U11A-0004
Control of the volumetric and viscosity ratios of iron-silicate emulsion on the core formation process
In a vigorously convecting magma ocean, shearing motion would cause an iron-silicate emulsion to form. Iron or silicate droplets would eventually settle or rise to form the core and the mantle. Volumetric and viscosity ratios of the droplet to continuous phase can vary but how these control the phase separation process is not understood. Here we report on a series of experiments using an oil droplet in water or hydroxyethylcellulose solution (all Newtonian) to study how these ratios govern the gravitational phase separation process. First, we changed the volumetric fraction of oil. From tracking the boundary between the layer of solution and emulsion, we find that the initial phase separation rate depends strongly on oil fraction, and can be modelled by a buoyancy driven permeable flow using the Blake-Kozeny-Carman permeability formula. Next, we changed the droplet to continuous phase viscosity ratio (λ) and find that there are two distinct regimes with different styles of phase separation. Cases with λ < 100 are characterized by a sharp lower boundary and a vertically homogeneous mixture layer. On the other hand, cases with λ > 100 are characterized by a diffuse lower boundary and a large vertical gradient of composition resulting from efficient droplet coalescence. As a result, polyhedral foam structure develops at the top of the mixture layer which is slow to rupture and to transform into a uniform oil layer. We interpret these differences to arise from a faster coalescence rate relative to the separation rate at large λ, where the droplet deformation which inhibits coalescence becomes very small. We simultaneously measured electrical resistivity in order to monitor the temporal change of the mean composition in the mixture layer and found that the measurements were consistent with the visual observation. To summarize, we find that the separation rate is controlled by the permeable flow velocity, whereas the vertical compositional structure within the emulsion layer is controlled by the viscosity ratio. If the above viscosity ratio criterion can be applied to silicate-iron emulsion, the case where iron percolates through silicate droplets (λ >> 1) would yield a strongly stratified mantle, compared to the case where iron droplets sink (λ << 1). Future separation experiment using iron-silicate emulsion is needed to confirm this. Sato, M. and Sumita, I., Experiments on gravitational phase separation of binary immiscible fluids, J. Fluid Mech., (in press)
U11A-0005
Earth's Core Formation Aided by Flow Channelling Induced by Rayleigh-Taylor Instabilities
The core formation process remains poorly known. Isotopic constraints by Hf/W systematics indicate a fast process which was largely completed within 33 Ma for the Earth. An unstable gravitational configuration of a dense molten metallic layer overlying a cold chondritic protocore is predicted by most studies for the time a planetary embryo reaches Mars-size. This leads to the formation of a Rayleigh-Taylor instability. We propose the application of Stevenson's (1989) stress-induced melt channelling mechanism in the region surrounding an incipient iron diapir. We therefore perform numerical experiments solving the two-phase, two compositions flow equations within a 2D rectangular box or 3D cuboid with symmetrical boundary conditions. We apply the Compaction Boussinesq Approximation (CBA) and include a depth-dependent gravity. For simplicity we use a constant viscosity for the solid phase and melt-fraction dependent rheology for the partially molten region around the diapir. A systematic investigation of the physical conditions under which the melt channels can form is being performed in 2D and 3D, and results are being compared to the isotopic time scale of core formation and applied to the early Earth. As a result, for sufficiently small retention numbers iron-rich melt channels develop within a region of approximately twice the diapir's size. This could lead to effective draining of the surrounding region and might initiate cascading daughter diapirs. The region of the protocore drained by this cascading mechanism is expected to significantly increase with depth, and thus proposes an effective mechanism to extract iron melt also from deeper parts of the initially chondritic protocore. This mechanism could effectively enhance melt accumulation in the Earth's protocore, accelerate the process of core formation and affect the metal-silicate equilibration in the deep planetary interior prior the Moon-forming giant impact. Therefore the channelling mechanism could also be interesting for planets like Mars, which never experienced complete melting.
U11A-0006
Numerical Simulation Of The Segregation Process Of Metal In Early Magma Ocean
Earth's core is considered to have been formed by separation of iron from the silicate magma ocean. This metal- silicate separation is important because it controls the initial states of the core and mantle, in particular concerning the distribution of heat and chemical species between them. In this process, the size of metal droplet is critical, which is estimated as an order of 1 cm by a simple balance of the surface tension and the shear stress (Rubie et al., 2003). Here we report quantitative estimate how the metal size evolves during the metal-silicate separation based on numerical simulation of two-phase flow under gravity. The gravitational energy that is gained, is transformed in heat by viscous heating and we want to quantify the resulting thermal structure. In the numerical method that simulates the process, tracking of the metal-silicate boundary is crucial. In this study, we adopt the moving particle semi-implicit (MPS) method based on the Lagrangian particle method. This method avoids numerical diffusion in the surface tracking of metal-silicate phase boundary because of its Lagrangian nature. Incorporation of the surface tension between metal and silicate phase is another critical factor in this simulation. In the 2-D simulation, we adopted the model by Nomura et al., 2001 implemented in MPS. In 3-D, we employed a new model that we invented. In the simulations, several interesting features have been clarified: irrespective of the initial size, size of the metal droplets evolve to a stable size of the order of 1 cm, which is mainly controlled by the surface tension and the viscosity of silicate melt. We discuss the size and shape of droplets, their falling speed, the interaction between droplets, and the resulting temperature distribution.
U11A-0007
Mercury's core fraction and ancient crustal composition: Predictions from planetary formation under extremely reducing conditions
Several hypotheses have been suggested to explain the paradox of Mercury's large core, which is on the order of sixty percent of the mass of the planet and recently demonstrated to be at least partially molten. Here we suggest that extremely reducing conditions in the earliest stages of planetary accretion nearest to the Sun may have produced the unusual metallic iron fraction by reducing iron otherwise bound into silicates. We demonstrate the formation conditions necessary for various meteoritic bulk compositions to produce the core/mantle ratio of Mercury. During this hypothetical core formation, we assume the remaining silicate fraction of Mercury (now largely lacking iron) has been heated to produce a magma ocean. The resulting cumulate mantle composition is calculated in a Matlab simulation of magma ocean solidification using a CMAS system adapted for Mercury. Plagioclase flotation, frequently cited as the necessary signature of a magma ocean, is highly dependent upon initial bulk composition. We demonstrate the initial silicate iron content of the magma ocean necessary to make plagioclase buoyant and thus produce a plagioclase flotation crust as seen on the Moon. In addition, over a range of bulk compositions the solidified mantle cumulates are unstable to gravitational overturn. During overturn hot cumulates rise from depth and may cross their solidi and melt, producing an earliest planetary crust. This crust may still exist on Mercury. With the first flyby results of the MESSENGER mission coming this winter, predictions from these models can be compared with initial ground measurements.
U11A-0008
Earth's Basal Magma Ocean and the Delayed onset of the Geodynamo
Partial melt has been proposed to explain the presence of thin ultralow-velocity zones in the lowermost mantle. The cooling of the core necessary to have maintained the geodynamo for at least 3.2 Gyr implies a greater amount melt in the past and a possibly large magma ocean at the base of the mantle following Earth's formation. The gravitational stability of such a melt layer can be explained by the small volume change upon melting at high pressure (Stixrude and Karki, 2005) and a larger amount of FeO in the melt than in solids at equilibrium. The crystallization of the magma ocean leads to progressive enrichment in FeO in both the melt and the solid, leading to formation of dense piles at the bottom of the mantle. Solving the thermal-chemical evolution of the deep Earth involving a basal magma ocean, we show that, to first order, the mass of melt decreases exponentially with time. Many parameters involved in this evolution are poorly known but can be constrained using geochemical observations. Indeed, the basal magma ocean is the perfect "hidden" reservoir that can store 20 to 30% of the Earth inventory in incompatible elements, provided its initial mass is around 1024 kg. In addition, the systematic difference in 142Nd between terrestrial samples and chondrites (Boyet and Carlson, 2005) can be used to constrain the time scale of crystallization. Because of the large amount of radiogenic and latent heat to extract from the basal magma ocean early in Earth's history, the heat extracted from the core can initially be lower than that conducted along an adiabat. Models most readily satisfying geochemical constraints predict that the geodynamo would not start before 3.4-4 Gyr ago, in agreement with lunar (Ozima et al, 2005) and terrestrial data (Tarduno et al, 2007). http://perso.ens- lyon.fr/stephane.labrosse
U11A-0009
Time Variability in Cenozoic Reconstructions of Mantle Heat Flow: Plate Tectonic Cycles and Implications for Earth's Thermal Evolution
The thermal evolution of Earth is governed by the rate of secular cooling and the amount of radiogenic heating. If mantle heat sources are known, surface heat flow at different times may be used to deduce the efficiency of convective cooling and ultimately the temporal character of plate tectonics. We estimate global heat flow from 65 Ma to the present using seafloor age reconstructions and a modified half-space cooling model, and we find that heat flow has decreased by ~ 0.15% every million years during the Cenozoic. By examining geometric trends in plate reconstructions since 120 Ma, we show that the reduction in heat flow is due to a decrease in the area of ridge-proximal oceanic crust. Even accounting for uncertainties in plate reconstructions, the rate of heat flow decrease is an order of magnitude faster than estimates based on smooth, parameterized cooling models. This implies that heat flow experiences short-term fluctuations associated with plate tectonic cyclicity. Continental separation does not appear to directly control convective wavelengths, but rather indirectly affects how oceanic plate systems adjust to accommodate global heat transport. Given that today's heat flow may be unusually low, secular cooling rates estimated from present-day values will tend to underestimate the average cooling rate. Thus, a mechanism that causes less efficient tectonic heat transport at higher temperatures may be required to prevent an unreasonably hot mantle in the recent past.
U11A-0010
Cratons Formation and the Earth Episodic Thermal History
U-Pb ages of zircons in granites and large rivers record a three-stage evolution of the continental crust. Plate tectonics operated in the first stage, from ~4.4 to 2.7 Ga. Huge peaks of crustal growth separated by long troughs dominated the second stage, from 2.7 to 1.8 Ga. Semi-continuous growth punctuated by large peaks characterized the last stage, from 1.8 to 0 Ga. Individual peaks in the second stage open with massive mafic- ultramafic volcanism and climax 30 Ma later with intrusion of voluminous granitoids: each peak opened with enhanced mantle plume activity, climaxed with accelerated plate tectonic activity, and was followed by a long quiet period when little crust formed. We develop a fluid-mechanics model to explain the three-stage evolution and the pronounced peak-and-trough pattern of the second stage. Low temperatures in the upper mantle during the first and last stages allow the formation of thin, subductable oceanic crust, and lead to a plate-tectonic regime. The crustal-growth peaks of the second regime results from destabilization of a hot denser layer at the bottom of the lower mantle. Domes rising from this layer partially melt to form voluminous mafic magmas. Moreover, when the domes hit and spread under the top surface, they peel off the cold thermal boundary layer there, which triggers a ring of enhanced cold instabilities around each thermochemical dome. This period of enhanced downwellings (i.e. subduction) is however followed by a lull, during which the cold thermal boundary layer is growing again. The onset of the first thermochemical instabilities is predicted to be synchroneous on the whole mantle , and followed by several more disorganized events, in agreement with observations. These experiments closely link continental growth and the thermal history of our planet, which appear strongly episodic. The importance of the 2.7 Ga peak, related to the sudden destabilization of a lower denser layer, suggests that extraction of heat from the core could have changed drastically at this time. This could have triggered the inner core crystallization and a new dynamo regim.
U11A-0011
Visco-elastic Models of Neo-Hadean and Archean Greenstone Belt Formation by Diapiric Density Inversion
Diapirism as a model for the formation of Archean granite-greenstone terrains has been periodically in and out of favour since it was first suggested in 1951. Recently, there has been renewed interest in diapirism to explain Archean terrains, which has re-ignited the debate over the importance of vertical vs. horizontal tectonics in the Archean, or at least, the debate over when the change from one to the other is likely to have occurred. In 1980, Mareschal and West performed numerical calculations which tested the physics of crustal diapirism, and were able to simulate the bulk features of many granite-greenstone sequences. However, due to computational restrictions at the time, their rheologies were over-simplified, and recent information on mantle and crustal heat flow suggest they overestimated Archean thermal inputs. Here we present results modeling diapiric overturn using a moving mesh visco-elastic finite-element solver. We test a range of non-Newtonian rheologies, thermal parameters, and supracrustal volcanic thicknesses. Comparison of our results with field observation constrains the range of plausible models. We find that the formation of dome-and-keel structures occurs easily and rapidly (~ 10 Ma or less), even using conservatively cool estimates for Archean thermal profiles, implying that this process may have been important in the Neo-Hadean and Meso-Archean. Indeed, the ubiquity of dome-and-keel structures in Archean terrains suggests that diapirism may have been a dominant crustal process in hotter geological times. We speculate on what the predominance of such a mechanism implies for early crustal and lithospheric evolution.
U11A-0012
Micron-Scale Correlations Among Ti, P, Ce, and Y in Hadean Jack Hills Zircons
Detrital zircons and the inclusions found therein are our only mineralogical constraints on geologic events that occurred on the Hadean Earth. These zircons are commonly small (ca. <100 μm in the longest dimension) and preserve micron to sub-micron chemical zonations indicative of a dynamic petrological history. Trace elements within zircon are of particular interest because concentrations and ratios of these elements can provide information regarding chemical and physical conditions during zircon growth. In this study, we analyzed Hadean-age detrital zircons from Archean metasediment in the Jack Hills (Australia) using the Caltech Microanalysis Center Cameca NanoSIMS 50L. Trace elements analyzed included Ti, P, Ce, and Y. Ti- thermometry [1,2,3] can potentially constrain growth and/or re-equilibration temperatures of zircons; P, Ce, and Y are known to enter the zircon lattice by the coupled xenotime-type substitution mechanism: (Y, REE)3+ + P5+ = Zr4+ + Si4+ [5]. The 89Y/28Si ratio was observed to correlate with, and was used as a proxy for, cathodoluminescence (CL) banding. Growth features manifested in CL (e.g., sector, oscillatory zoning) were observed in all zircons analyzed. CL zones vary from <1 μm to several microns in width; therefore, the NanoSIMS---with a beam diameter resolved to ca. 250 nm on the sample surface when operating with an O- primary beam---is uniquely suited for this scale of analysis. Regions displaying CL banding were imaged as 20 x 20 μm areas. All elements were normalized to 28Si; 49Ti/28Si ratios were converted to [Ti] via calibration based on analyses of synthetic, high-Ti zircons (provided by B. Watson) that were independently analyzed on Caltech's JEOL JXA-8200 electron microprobe. We observe three types of relationships between trace element distribution and CL banding in the zircons imaged: 1) strong positive correlations between CL banding, P, Ce, and Ti; 2) subtle positive correlations between CL banding, P, Ce, and Ti; 3) no correlation between minor/trace elements and CL banding. Positive correlations between CL banding, 3+ cations, and [Ti] have previously been reported by Holden et al. [4]. In this study, gradients at least as sharp as a factor of ~3 in [Ti] are observed between adjacent CL bands in the strongly correlated images. These images also have the highest absolute concentrations of trace elements and display both sector and oscillatory zoning in CL. The correlations observed may be due to: temperature-dependent equilibrium partitioning of all trace elements during rapid cycles in growth temperature; episodic diffusion-limited enrichment of incompatible trace elements in the boundary layer melt adjacent to growing crystals; and/or kinetically controlled, non- equilibrium crystal-melt partitioning caused by trace element enrichments in the boundary layer melt surrounding fast-growing grains (e.g., [6]). We will discriminate between these alternatives based on quantitative relationships between relative enrichments of [Ti] and other trace elements. [1]Watson, E.B. & Harrison, T.M. (2005) Science 308, 841-844. [2]Watson, E.B., Wark, D.A., & Thomas, J.B. (2006) CMP 151, 413-433. [3]Ferry, J.M. & Watson, E.B. (2007) CMP 154, 429-437. [4] Holden, P. et al. (2005) Eos Trans. AGU 86 (52) Fall Meet. Suppl., Abstract V41F-1539. [5] Speer, J.A. (1982) Zircon. In Rev. Min. 5 (ed. P.H. Ribbe), 67-112. [6] Watson, E.B. (2004) GCA 68, 1473-1488.
U11A-0013
Evolution of Plate Tectonics From Hadean to the Modern Geodynamic Regime According to 3D Spherical Models With Self-Consistently Generated Plates
It has been proposed that the plate tectonic style has changed from the Hadean to the modern geodynamic regime. In the past decade, several studies have documented the effectiveness of plastic yielding in causing a basic approximation of plate tectonic behavior in mantle convection models with strongly temperature dependent viscosity, strong enough to form a rigid lid in the absence of yielding. The vast majority of such research to date has been in either two-dimensional, or three-dimensional cartesian geometry. In our recently-presented study, the planforms of self consistent tectonic plates in three-dimensional spherical geometry were calculated and compared to the outcome of similar calculations in which a three dimensional cartesian geometry was used, as in Tackley (2000). Several diagnostics are used to analyze how successful each model is in producing tectonic plates. In general, the observed structures in spherical geometry are consistent with the structures observed in cartesian geometry, but two previously unobserved planforms are found. At low lithospheric yield stresses one great circle downwelling formed, with fragmented oceanic plates at both hemispheres. At intermediate yield stresses two hemispherical plates formed, separated by a spreading centre and a downwelling. Here, new calculations are presented that study the evolution of the tectonic mode and planforms in time as the Earth cools, by decreasing the internal heating rate with time according to the decay of K, U and Th and by decreasing the temperature at the CMB according to a parameterized core model. The influences of these two effects are separated by comparing to calculations where the heat flux across the CMB is set to zero. We study the proposed transitions in tectonic mode (e.g. changes in plate size, rigid lid convection to tectonic plates, smoothly evolving plates to more episodic, time dependent, tectonics) as a function of yield stress envelope. The results are compared to analytical scalings for boundary regimes as well as scalings for heat flux.
U11A-0014
Recycling of ca 4.35 Ga KREEP-Like Crust in Western Australia 4.1 Gy ago
We measured Hf and Pb isotope compositions for 63 single zircons from Jack Hills (JH), Western Australia, using solution chemistry and, respectively, MC-ICP-MS and ICP-MS techniques. The data were pooled with the solution chemistry data of Harrison et al. (2005) totaling 104 analyses. Bulk 207Pb*/206Pb* ages, including ion microprobe dates at >4.33 and 3.32 Ga, are anormally distributed between 3.8 and 4.2 Ga with a well-defined peak at 4.10±0.03 Ga. Thus, for our purposes, we treat the data as a single population that formed at 4.1 Ga, rather than as a series of events, although ion microprobe dating indicates the presence of a small >4.2 Ga component. Most of the analyzed zircons have Hf isotope compositions significantly less radiogenic than both chondrites and the depleted mantle at 4.1 Ga. The simplest interpretation is that the JH zircon granitoid hosts were melts of an enriched reservoir with sub-chondritic Lu/Hf. Single-stage Hf model ages, calculated by back-tracking the Hf isotopic composition at 4.1 Ga to either the chondritic or the depleted mantle reservoirs for a range of Lu/Hf ratios, allow insights into the age and composition of this pre-existing crust. The modeling shows that for 176Lu/177Hf > 0.015, such as in basaltic crust, the resulting histograms are flat and poorly defined with many unacceptable zircon ages older than 4.56 Ga. In contrast, for extremely low 176Lu/177Hf of 0.005-0.010, the calculated histograms define narrow well-defined peaks with maxima at 4.31 and 4.36 Ga for the chondritic and depleted mantle reservoirs, respectively, with all zircons younger than the age of the Earth. This indicates that the source rock of the JH granites was not the equivalent of modern MORB or oceanic plateau basalts, which have 176Lu/177Hf of 0.026 and 0.029, respectively. Neither was it likely to have been granitic, similar to rocks formed in subduction zones today, because the low inferred Lu/Hf ratio for the source rock of the JH granites is different from granites in their present form. Thus, remelting of either hydrous oceanic crust or granites with modern Lu/Hf characteristics does not seem to be how the JH granites formed. Rather, the early crust giving rise to the JH granites may have originated from either the residual liquids after magma ocean crystallization or from melts of the last magma ocean cumulates, both of which would have been extremely differentiated with very low Lu/Hf due to garnet having crystallized at depth. Interaction with the hydrosphere, known from oxygen isotopes in JH zircons to have existed in the Hadean, would have enhanced the hydrous nature of this crust so as to constitute a suitable reservoir for the JH granites. A Rayleigh fractionation model of the terrestrial magma ocean, assuming chondritic Lu/Hf of the magma and a partition coefficient for Lu of 4 between garnet and the liquid, demonstrates that removal of either 85% of a cumulate with 25% garnet or 99% of a cumulate with 10% garnet would have left a residual liquid with 176Lu/177Hf of 0.005. Once this extremely differentiated hydrous KREEP-like proto-crust was in place, the scene was set for plate tectonics to begin and thus modern-type granites to form. The JH zircon Hf model ages place this transition at a minimum of ~4.35 Ga. The loss ultimately of this early crust, either by impacts or foundering into the deep mantle, provides an explanation for the hidden reservoir required by Hf-Nd and 142Nd isotope systematics.
U11A-0015
Heterogeneous Earth Accretion and Incomplete Metal-Silicate Reequilibration at High Pressure During Core Formation
We present a new model of core formation, based on the partitioning of siderophile elements, that involves accreting the Earth through a series of collisions with smaller bodies that had already differentiated at low pressure. Each impact results in a magma ocean in which the core of the impactor reequilibrates with silicate liquid at high pressure before merging with the Earth's protocore. The oxygen contents of the chondritic compositions of the proto-Earth and impactors can be varied. The compositions of coexisting metal and silicate are determined through mass balance combined with partitioning equations for Ni, FeO, Si and other siderophile elements. The oxygen fugacity is fixed by the partitioning of FeO and is a function of P, T and bulk oxygen content. An important constraint for core formation is that core-mantle partition coefficients for Ni and Co must both converge to values of 23-28. Based on a recent study of the partitioning of Ni and Co over a wide P-T range (Kegler et al., EPSL, submitted) together with other published data, this constraint is not satisfied by a single- stage core formation model at any conditions because the partition coefficients converge at values that are much too low. In the present multi-stage model, the correct values can be reached if only part of each impactor core reequilibrates with silicate liquid in the magma ocean (as proposed by previous models based on Hf-W isotope studies). Physically, this would mean that impactor cores fail to emulsify completely as they sink through the magma ocean. Incorporating other elements (e.g. V and Cr) in the model requires, in addition, that the bulk composition of the impactors changes during accretion from reduced (FeO-poor) to oxidised FeO-rich). Then, with the resulting increase in fO2, incomplete reequilibration of the cores during the final 20-30% of Earth accretion is required to satisfy the Ni-Co constraint. In addition, this model enables the concentrations of O and Si in the core to be estimated.
U11A-0016
Terrestrial Planets Accreted Dry
Plate tectonics shaped the Earth, whereas the Moon is a dry and inactive desert. Mars probably came to rest within the first billion years of its history, and Venus, although internally very active, has a dry inferno for its surface. The strong gravity field of a large planet allows for an enormous amount of gravitational energy to be released, causing the outer part of the planetary body to melt (magma ocean), helps retain water on the planet, and increases the pressure gradient. The weak gravity field and anhydrous conditions prevailing on the Moon stabilized, on top of its magma ocean, a thick buoyant plagioclase lithosphere, which insulated the molten interior. On Earth, the buoyant hydrous phases (serpentines) produced by reactions between the terrestrial magma ocean and the wet impactors received from the outer Solar System isolated the magma and kept it molten for some few tens of million years. The elemental distributions and the range of condensation temperatures show that the planets from the inner Solar System accreted dry. The interior of planets that lost up to 95% of their K cannot contain much water. Foundering of their wet surface material softened the terrestrial mantle and set the scene for the onset of plate tectonics. This very same process may have removed all the water from the surface of Venus 500 My ago and added enough water to its mantle to make its internal dynamics very strong and keep the surface very young. Because of a radius smaller than that of the Earth, not enough water could be drawn into the Martian mantle before it was lost to space and Martian plate tectonics never began. The radius of a planet therefore is the key parameter controlling most of its evolutional features.
U11A-0017
Early Formation of Terrestrial Crust
Early (≥4.5 Ga) Formation of Terrestrial Crust T.M. Harrison1, A.K. Schmitt1, M.T. McCulloch2, and O.M. Lovera1 1Department of Earth and Space Sciences and IGPP, UCLA, Los Angeles, CA 90095, USA; 2Research School of Earth Sciences, Australian National University, Canberra, A.C.T. 2601 AUSTRALIA Large deviations in \varepsilonHf(T) from bulk silicate Earth seen in >4 Ga detrital zircons from Jack Hills, Western Australia, have been interpreted as reflecting a major differentiation of the silicate Earth at ca. 4.4 to 4.5 Ga. We have expanded the characterization of 176Hf/177Hf (Hf) in Hadean zircons by acquiring a further 116 laser ablation Lu-Hf measurements on 87 grains with ion microprobe 207Pb/206Pb ages up to 4.36 Ga. Most measurements employed concurrent Lu-Hf and 207Pb/206Pb analyses, permitting assessment of the use of ion microprobe data to characterize the age of the volumetrically larger domain sampled by laser drilling. Our new results confirm and extend the earlier observation of significant negative deviations in \varepsilonHf(T) throughout the Hadean, although no positive \varepsilonHf(T) values were documented in this study. These data yields an essentially uniform spectrum of single-stage model ages between 4.54 and 4.20 Ga for extraction of the zircons' protoliths from a chondritic reservoir. We derived the full error propagation expression for a parameter, \varepsilono, which measures the difference of a sample from solar system initial (Hf) (Hfo), and from this conclude that data plotting close to (Hfo), are statistically meaningful and consistent with silicate differentiation at 4.540±0.006 Ga. δ18O and Ti thermometry for these Hadean zircons show little obvious correlation with initial (Hf), consistent with their derivation through fusion of a broad suite of crustal rock types under near water-saturated conditions. Together with the inclusion assemblage and other isotopic and trace element data obtained from these ancient zircons, our results indicate essentially continuous derivation of crust from the mantle from 4.5 to 4.2 Ga, concurrent with recycling into the mantle and internal crustal re-working. These results represent further evidence that by 4.35 Ga, portions of the crust had taken on continental characteristics.
U11A-0018
Relationship between the Neoproterozoic snowball Earth and Cambrian explosion
Origin of snowball Earth has been debated in terms of greenhouse gas (e.g., Hoffman and Schrag), obliqueness of Earthfs rotation axis (Williams, 1975), true polar wander (Evans, 2003), Galactic cosmic ray radiation (Shaviv and Veizer, 2003; Svensmark, 2006), or weakened geomagnetism (Maruyama and Yoshihara, 2003). A major difficulty for the greenhouse gas hypothesis is the on-off switch causing decrease and increase of appropriate amounts of CO2 by plume- and plate tectonics, and also in available amount of CO2 in atmosphere to be consistent with the observations. In contrast, the cosmic ray radiation models due to the star burst peaked at 2.5- 2.1 Ga and 1.4-0.8 Ga can explain on-off switch more easily than the greenhouse gas model. Cosmic ray radiations, however, must be modified by the geomagnetic intensity, fluctuating 150% to < 10% of the present-day level through geologic time. Our compilation suggests the idea of extensive glaciation appeared when the intensity decreased below 50% of the present-day value, as typically seen in the Neoproterozoic time. This proposes the idea of extensive cloudiness by increased cosmic rays in the Neoproterozoic to cause the snowball Earth. Time difference between the Neoproterozoic snowball Earth and Cambrian explosion is as large as 250 millions years, and this refuses their direct close-relationship. Role of frequent mass extinctions, i.e., 8 times during 100 m.y. from 585 Ma to 488 Ma, during the Ediacaran and Cambrian, has been proposed (Zhu et al., 2007). This frequency is one order of magnitude higher compared to that in the post-Ordovician time. Yet, the Cambrian explosion cannot be explained by mass extinction which replaced the vacant niches shortly after the mass extinction and never created a new animal with a new body plan. A new model proposed herein is derived from weakened geomagnetism and resultant extensive cosmic radiation to alter gene and genome for a long period over advancement of low magnetic intensity and cosmic radiations (Svensmark, 2006) from 1.2-0.8Ga. As to the new body plans of animals, it took an appreciably long time to prepare all 34 genometypes before the apparent Cambrian explosion. Geochemically extreme conditions and widened shallow marine environment on continental shelf by the return-flow of sweater into mantle in the Late Neoptroterozoic were the associated critical conditions to alter genomes during 1.2-0.52 Ga appreciably before the final consequence called the Cambrian explosion.
U11A-0019
Freezing of a Magma Ocean and Subsequent Mantle Differentiation
It seems likely that a Magma ocean, after separation of iron from silicate, did freeze from the bottom up, due to the increase of pressure with depth. A scenario can thus arise in which hot material at the bottom of the magma ocean is compositionally light and underlies colder but compositionally denser material. A rapid the evolution of chemical heterogeneities of this instable configuration has been proposed by several authors. By means of two- and three-dimensional convection models, we investigated the overturn-scenario and especially the subsequent evolution of the mantle, following the magma ocean period. The numerical models include finite element and finite volume procedures as well as front tracking methods to capture the evolution of chemical heterogeneities Our numerical experiments clearly reveals that an overturning puts the mantle into the diffusive regime, characterized by an unstable thermal, but stable compositional stratification. The formation of layered flow structures is, a typical phenomenon in this regime. In a wide parameter range (thermal/compositional Rayleighnumbers, realistic rheologies and various distribution of internal heat sources) , we observe the overturning followed by a long period of layered convection. Vigorous convection takes place in the upper- and lower mantle, while typically a less vigorous convection layer develops in between. The number of layers and their individual lifetime depends on the particular parameters. However in any case, layered structures develop over a significant time span, such that a profound influence on the chemical evolution seems reasonable to expect.
U11A-0020
Nd Isotope Systematics for the 3.8 Ga Nuvvuagittuq Greenstone Belt: Implications for the Degree of Depletion of Early Earth's Mantle
Rare occurrences of Eoarchean mantle-derived crust provide the only compositional and isotopic constraints on the early crust-mantle differentiation of the Earth. The relatively high positive initial εNd values obtained on Eoarchean rocks (Acasta Gneisses, Slave Province, Canada; Nulliak Assemblage, Labrador, Canada; Amîtsok Gneisses, Akilia and Isua supracrustal assemblages, SW Greenland) are interpreted to indicate derivation from a mantle source already strongly depleted in the Eoarchean, implying that significant volumes of continental crust had formed early in the Earth's evolution. Here we present the Nd isotope data of the newly discovered ca. 3.8 Ga Nuvvuagittuq greenstone belt (Northern Québec, Canada), and discuss their implications for early mantle depletion. The Nuvvuagittuq greenstone belt is a volcano-sedimentary sequence mainly composed of cummingtonite- amphibolites, numerous ultramafic and gabbroic sills, and chemical sedimentary rocks. Most mantle-derived samples display positive initial εNd values ranging from -2 to +4, with an average of +1.2 ± 0.2 for 44 samples. Isotopic data for all the samples that have been analysed yield a Sm-Nd regression age of 3.81 ± 0.13 Ga (MSWD = 8.5), and a subset of samples from the gabbroic sills gives an isochron age of 3.83 ± 0.39 Ga (MSWD = 0.39), with initial εNd values of +1.2 and +2 respectively. Sm-Nd isotopic data for the Nuvvuagittuq mantle-derived rocks support an age of 3.8 Ga, consistent with previous geochronology, and the low MSWD for the Sm-Nd regression lines suggest minimal disturbance of the Sm-Nd isotopic system. The few samples that yield εNd(3.8Ga) values greater than +3 suggest that they may have been derived from a mantle source that had already experienced even more extensive trace element depletion than the present-day MORB source by 3.8 Ga. Such a degree of trace element depletion, however, is not supported by the flat to slightly LREE-depleted profiles of the gabbros. Nuvvuagittuq samples which yield εNd(3.8Ga) values greater than +3 have Sm/Nd ratios which are too low to be consistent with melting of a mantle source having an εNd value of +3 or greater at 3.8 Ga. This implies that the trace element depletion is not consistent with the isotopic depletion seen in these most isotopically depleted samples. Consequently, εNd(3.8Ga) values of ~+2 for the Nuvvuagittuq mantle-derived rocks are most plausible, indicating derivation from mantle already depleted in the Early Archean, but depleted to a lesser extent than that which is seen in the 3.8 Ga SW Greenland rocks.
U11A-0021
Constraints on Core Formation From Systematic Study of Temperature Effect on Metal- Silicate Partitioning
Models of core formation are currently established through metal-silicate partitioning results at high pressure and high temperature. Although a large effect of temperature on metal-silicate equilibrium is expected on thermodynamic grounds, very little experimental work has been dedicated to separate this effect from pressure and provide a systematic study of partitioning coefficients across a wide range of temperatures. Utilizing free energy of pure oxides formation data at atmospheric pressure to predict the temperature effect on metal-silicate partitioning might be a source of large uncertainties for some recent core formation models [1, 2]. The present study is aimed at constraining the temperature dependence of partition coefficients for a large number of elements and extending the existing database to extreme temperatures. Using a new piston-cylinder design assembly [3] allows us to determine a suite of isobaric partitioning experiments at 3 GPa within a temperature range from 1600 to 2700°C. Systematic partitioning behaviors between molten metal and peridotite or basaltic melts of elements normally regarded as moderately siderophile, slightly siderophile and refractory lithophile are presented. These include Ni, Co, W, Mo, Cr, Mn, V, P, Ga as well as elements that are usually poorly integrated with any accretion or core formation models (Ge, Nb, Ta, Te, Zn). Absolute measurements of partitioning coefficients combining EMP and LA-ICPMS analytical methods are provided. The individual effects of oxygen fugacity and pressure have also been studied through piston cylinder experiments (2200°C, 3 GPa) between IW- 1.5 to IW-4 and multi-anvil experiments to 15 GPa. These partitioning results are then combined with literature data to refine our understanding of core formation and place constraints on the highly debated Earth's accretion mechanism issue. [1] Wade and Wood, EPSL, 2005. [2] Corgne et al., GCA, in press. [3] Cottrell and Walker, GCA, 2006.
U11A-0022
Si isotopic composition of the Earth's mantle and meteorites
The superchondritic Mg/Si ratio of the Bulk Silicate Earth (BSE) as compared to chondrites has been a long- standing issue in trying to understand the composition and history of the Earth. It is thought that it must be inherited from the formation or the early differentiation of our planet. The reasons commonly invoked to account for this discrepancy are (i) a relative loss of Si due to its slightly greater volatility relative to Mg; (ii) the production of intra-mantle heterogeneity during magma ocean crystallization, whereby the samples from the upper mantle are actually non-representative of the actual BSE; (iii) lastly, the low Mg/Si of the BSE could be explained by the incorporation of Si in in the Earth's core. We have further investigated this issue by measuring with high precision Si isotopes using the large geometry MC-ICPMS Nu1700 that allows a good resolution of potential isobaric interferences. We have further refined the chemical separation and mass spectrometric technique compared with already published methods. In particular, we have investigated all the steps that could induce isotope fractionation (or produce mass bias) during the fusion technique, chemical separation and mass spectrometry. We have obtained Si isotopic compositions for a series of ordinary and carbonaceous chondrites as well as a suite of peridotites and basalts. Our preliminary results are not identical to the results published by [Georg et al., 2007] and the new implications of our data will be discussed at the meeting. References: R.B. Georg, A.N. Halliday, E.A. Schauble and B.C. Reynolds, "Isotopic evidence for silicon in the Earth's core", Nature, 447, 1102 (2007)