During formation of the Earth's core siderophile (metal-seeking) elements were concentrated into core forming metal. As a result, the major fraction of the Earth's inventory of siderophile elements is now in the core. Absolute and relative abundances of the small fraction of metallic elements remaining in the mantle provide information on details of the core forming process and thus on the early history of the Earth. The most important parameters governing the metal extraction process are metal-silicate partition coefficients (Dm/s ). They depend on temperature, pressure, oxygen fugacity and composition of metal and silicate phases. During the last years considerable progress has been made in the experimental determination of metal-silicate partition coefficients, leading to more and better defined constraints for core formation and late accretion of the Earth.
Highly siderophile elements with metal-silicate partition coefficients above 10 000 occur in essentially chondritic abundances in the Earth's mantle at a level of 0.7 % of a nominal CI-component. Their abundances are, by most authors, attributed to the accumulation of a late chondritic component after the end of core formation (e.g. Chou, 1978). Earlier suggestions to explain the comparatively high concentration of metals in the mantle by equilibration at very high temperatures (Murthy, 1991) are not supported by the new experimental data (Borisov et al., 1994; Borisov and Palme, 1995). There are, however, new analyses of upper mantle rocks that seem to indicate, at least locally, deviations from a strictly CI-chondritic pattern (Pattou et al., 1995; Schmidt et al., subm.). The implications of these observations are presently not clear.
The moderately siderophile elements have metal-silicate partition coefficients below 10 000. Their abundances in the Earth's mantle vary from 0.3 (Fe) to 0.04 (Ge) of a nominal CI-component. The comparatively high abundances of Fe, Ni, and Co and the largely unfractionated ratios among them, in particular the nearly chondritic Ni/Co ratio, have long been recognized as an important clue to the accretion and core formation of the Earth. Experiments within the last years have shown that (i) Ni and Co occur exclusively as oxide species during metal segregation (Dingwell et al., 1994; Holzheid et al., 19994). (ii) There is a trend for Ni and Co metal-silicate partition to decrease with increasing pressure and temperature (e.g. Thibault and Walter, 1995; Holzheid et al., subm.). The absolute values of the Dm/s as well as the Ni/Co-partition coefficient ratio can neither produce the absolute level nor the observed chondritic ratio of these two elements. Although experiments have only been performed up to pressures of about 20 GPa and temperatures of 1500°C the Ni and Co partition coefficients appear to converge, with increasing pressure and temperature, to an approximately constant ratio of about four, far from unity required to produce a chondritic Ni/Co-ratio by metal-silicate equilibration. (iii) Ni and Co partition coefficients between metal melt and lower mantle minerals may be significantly lower than those between metal and liquid silicate (Urakawa, 1991).
These findings exclude models where the Ni and Co abundances of the upper and probably whole mantle were established by equilibrium of liquid silicates with core forming metal (solid or liquid). Segregation of metal in a late global magma ocean is, therefore, unlikely to produce the observed high mantle contents of Ni and Co and their chondritic ratio.
The amount of Ni, Co, and Fe in the Earth's mantle corresponds to about 20 % of a CI-chondritic component. If, as is assumed in inhomogeneous accretion models (Wänke et al., 1984; Ringwood, 1984; O'Neill, 1991), these 20 % were added after core formation, earlier episodes of accretion involving a magma ocean cannot be excluded.
The chondritic Ni/Co provides further constraints on the nature and size distribution of late accretional components. Recent models of the accretion of the Earth involve collisions of the proto-Earth with large, up to Mars-size planetesimals. The geochemical constraints imposed by the high Ni and Co and their chondritic ratios, require that these planetesimals were not differentiated into core and mantle. Numerical simulations have shown that the cores of such planetesimals would simply recombine with the Earth's core without equilibrating with Earth mantle silicates (Benz and Cameron, 1990). The abundances of Ni and Co in the mantles of the impactors would be far too low to account for the present inventory of metallic elements in the Earth's mantle.
A second constraint is imposed by the size of the late planetesimals colliding with the Earth. Mars sized objects such as the Moon forming planetesimal would produce a terrestrial magma ocean (e.g. Melosh, 1990). Metal segregation would be the inevitable consequence leading to depletion of Ni and Co in the Earth mantle. Kato et al. (1988) argued against a late magma ocean because fractionation of Mg-perovskite would disturb the observed chondritic ratios of refractory lithophile elements. Tonks and Melosh (1990) however, showed that cooling of a magma ocean is possible without fractional crystallization. These argument apply to crystallization of Mg-perovskite. However, it is doubtful if the much denser metal particles could be kept in suspension during cooling of the magma ocean.
If a giant impact of a Mars-size body cannot account for the origin of the Moon, the formation of the Moon could be achieved by a series of smaller, late-accreting bodies, thus avoiding a global magma ocean (Ringwood, 1989).
Benz, W. & Cameron, A.G.W., In Origin of the Earth (Newsom, H.E. & John, J.H., eds.) 61-68 (Oxford University Press, 1990).
Borisov, A. & Palme, H., Geochim. Cosmochim. Acta 59, 481-485 (1995).
Borisov, A., Palme, H. & Spettel, B., Geochim. Cosmochim. Acta 58, 705-716 (1994).
Chou, C.L., Proc.Lunar Planet.Sci.Conf. 9th, 219-230 (1978).
Dingwell, D.B., O'Neill, H.St.C., Ertel, W. & Spettel, B., Geochim. Cosmochim. Acta 58, 1967-1974 (1994).
Holzheid, A., Borisov, A. & Palme, H., Geochim. Cosmochim. Acta 58, 1975-1981 (1994).
Holzheid, A., Palme, H., Rubie, D.C. & O'Neill, H.St.C., Nature (1995, submitted).
Kato, T., Ringwood, A.E. & Irifune, T., Earth Planet. Sci. Lett. 89, 123-145 (1988).
Melosh, H.J., In Origin of the Earth (Newsom, H.E. & John, J.H., eds.) 69-84 (Oxford University Press, 1990).
Murthy, V.R., Science 253, 303-306 (1991) and corrections, p.1467.
O'Neill, H.St.C., Geochim. Cosmochim. Acta 55, 1159-1172 (1991).
Pattou, L., Lorand, J.P. & Gros, M., Nature (1995, in press).
Ringwood, A.E. Earth Planet. Sci. Lett. 95, 208-214 (1989).
Ringwood, A.E., Proc.R.Soc.London A 395, 1-46 (1984).
Schmidt, G., Palme, H., Kratz, K.L. & Kurat, G., Chemical Geology (1995, submitted).
Thibault, Y. & Walter, M.J., Geochim. Cosmochim. Acta 59, 991-1002 (1995).
Tonks, W.B. & Melosh, H.J., In Origin of the Earth (Newsom, H.E. & John, J.H., eds.) 151-174 (Oxford University Press, 1990).
Urakawa, S., Earth Planet. Sci. Lett. 105, 293-313 (1991).
Wänke, H., Dreibus, G. & Jagoutz, E., In Archean Geochemistry (Kröner, A., ed.) 1-24 (Springer Verlag, Berlin, 1984).