Letter | Published:

The Earth's ‘missing’ niobium may be in the core

Nature volume 409, pages 7578 (04 January 2001) | Download Citation



As the Earth's metallic core segregated from the silicate mantle, some of the moderately siderophile (‘iron-loving’) elements such as vanadium and chromium1,2 are thought to have entered the metal phase, thus causing the observed depletions of these elements in the silicate part of the Earth. In contrast, refractory ‘lithophile’ elements such as calcium, scandium and the rare-earth elements are known to be present in the same proportions in the silicate portion of the Earth as in the chondritic meteorites—thought to represent primitive planetary material1,3. Hence these lithophile elements apparently did not enter the core. Niobium has always been considered to be lithophile and refractory yet it has been observed to be depleted relative to other elements of the same type in the crust and upper mantle4,5. This observation has been used to infer the existence of hidden niobium-rich reservoirs in the Earth's deep mantle5. Here we show, however, that niobium and vanadium partition in virtually identical fashion between liquid metal and liquid silicate at high pressure. Thus, if a significant fraction of the Earth's vanadium entered the core (as is thought), then so has a similar fraction of its niobium, and no hidden reservoir need be sought in the Earth's deep mantle.


Estimates of the bulk composition of the Earth1,2 are generally based on the compositions of the chondritic meteorites with which the Earth has strong chemical affinities. As the relative abundances of refractory lithophile elements are constant across the classes of these meteorites, their ratios (for example, Nb/Ta, Nb/La) in the bulk Earth can be estimated with confidence (Table 1). Although the relationship of these elements to refractory siderophile elements such as Fe, Ni and W is less certain, the mass of the Earth's core and its probable Fe/Ni ratio suggest that the bulk Earth has siderophile/lithophile ratios between those of CI and CM carbonaceous chondrites1,2 (Table 1). Core formation depleted the silicate mantle in the siderophile elements, but left non-siderophile elements in the same ratios one to another as in the bulk Earth. Later separation of the continental crust, which is highly enriched in many lithophile trace elements (the rare-earth elements, K, U, Th for example) left a large depleted reservoir that approximates the upper-mantle source of mid-ocean ridge basalts4. Table 1 shows some ratios of siderophile/lithophile and lithophile/lithophile elements in the bulk Earth, depleted upper mantle and the continental crust. The imprint of core formation can be seen as a marked reduction of siderophile/lithophile element ratios in upper mantle and crust relative to the bulk Earth. No detailed mass balance is required in order to demonstrate that the sum of continental crust plus upper mantle cannot produce the Fe/Al and Ni/Al ratios of the bulk Earth. Most of the Earth's Fe and Ni have clearly entered the core. V and Cr are also somewhat depleted in the silicate part of the Earth, implying that the core contains significant proportions of the Earth's inventories of these two elements. The alternative hypothesis, that V and Cr are depleted in the mantle because they are relatively volatile and did not accrete to the Earth6, seems less likely. Available data on volatility during solar condensation7 indicate that V is less volatile than Fe and Ni, two elements that are undepleted in the Earth.

Table 1: Element ratios in terrestrial reservoirs

Turning to the refractory lithophile elements, Nb, Ta and La, it is clear that no mixture of crust and depleted upper mantle can generate chondritic ratios of Nb/Ta and Nb/La. There must be another reservoir in the Earth with superchondritic Nb/Ta and Nb/La (ref. 5). In contrast, Ta—the ‘geochemical twin’ of Nb—appears to have an essentially chondritic ratio to La in the silicate Earth, implying that no additional reservoir is required for this element. Rudnick et al.5 have argued that the Nb-rich reservoir is subducted oceanic crust transformed into refractory rutile-bearing eclogite. The main difficulties with this hypothesis are first, that large volumes (up to 6% of the mantle) of such eclogite are required for mass balance, and second, that its superchondritic Nb/Ta is not seen in the ocean-island basalts (HIMU basalts) which appear to contain subducted oceanic crust in their source regions8,9. Here we explore an alternative possibility, that Nb is a weakly siderophile element slightly depleted in the silicate Earth due to partial dissolution in the core.

Free energies of reactions between metals and metal oxides10 suggest that Nb should be as siderophile as V and substantially more siderophile than Si, an element considered by some to constitute nearly 8% of the core2. The implication is that Nb should have partitioned into the core in a manner similar to V. The data refer to pure metals and oxides at atmospheric pressure, however, rather than to very high pressure equilibration between silicate in the mantle and Fe alloy in the core11,12. Observed concentrations of Ni and Co in the mantle1,2 are too high to be explained by low-pressure (0–2 GPa) equilibrium between metal and silicate containing about 6% Fe (ref. 11). But high-pressure experiments11,12,13,14 show that, with increasing pressure, these two elements become less siderophile with respect to Fe, and imply that single-stage core–mantle equilibrium would be possible at pressures of 25–30 GPa (refs 11, 13), plausibly at the base of a 700-km-deep magma ocean. Clearly, therefore, investigation of the effect of high pressure on the silicate–metal partitioning behaviour of weakly siderophile elements such as V, Cr and Nb is necessary in order to understand their depletions in the mantle.

We have performed liquid-metal/liquid-silicate partitioning experiments at low (2.5 GPa) and high (25 GPa) pressure for Nb and other lithophile and siderophile elements under conditions where the metal contains between 0.07% and 16.4% Si. Varying the Si content of the metal enabled us to study the relative siderophile behaviour of this element, as well as providing a means to vary the oxygen fugacity (fO2) of the experiment. Experimental starting materials were equal proportions of synthetic basalt, reduced at 1,000 °C in a CO/CO2 atmosphere at 2 logfO2 units below the fayalite–magnetite–quartz buffer, and a metallic component containing appropriate proportions of Fe, FeS, Fe83Si17 and FeSi. These were intimately ground under acetone with the trace-element component consisting of metals (Ni, Mo, Os, W) and oxides (of V, Nb, Ga, Ta, Cr, Ti, Mn, Co). At 2.5 GPa the samples were contained in MgO capsules and the experiments performed in a piston-cylinder apparatus at 1,750 °C for 30 min. The 25-GPa experiments were performed using the multi-anvil apparatus at the University of Bayreuth. Samples were contained in MgO capsules and held at 2,300 °C for approximately 90 s. Experiment duration was limited by the rapid reaction of the silicate with the capsule, generating inviscid ultrabasic melts (Table 2) that occasionally escaped from the capsule. Temperature control in all experiments was via a W/Re thermocouple, placed directly above the capsule; quenching was achieved by turning off the power to the furnace. Analysis was performed using the JEOL 8600 electron microprobe at the University of Bristol using a range of metal, oxide and silicate standards and conditions of 20 kV accelerating voltage and 15 nA beam current. Microprobe data were corrected using standard procedures. Results are given in Table 2 and shown in Figs 1–2.

Table 2: Experimental results
Figure 1: Low-pressure data.
Figure 1

The figure shows partitioning between Fe-rich metallic liquid and silicate melt at 1,750 °C and 2.5 GPa. Solid symbols, data from this study; open symbols, data from refs 15 and 16. Partition coefficients D (metal/silicate) are computed on a weight basis, and oxygen fugacity (fO2) is expressed relative to the Fe–FeO buffer (IW; see text). Error bars are approximately the same size as the symbols. Lines through data indicate that V and Cr are present in the 3+ oxidation state in the silicate, while Si is 4+. Shaded region, range of fO2 appropriate for core–mantle equilibrium, calculated by assuming metal of 90% Fe coexisting with lower-mantle (Mg,Fe)O of composition (Fe/Fe+Mg) = 0.1–0.2 (ref. 19). Horizontal line at D = 100, conditions at which elements change from preferring silicate (lithophile) to metal (siderophile).

Figure 1 shows the partition coefficients Dmetal/sil (=[wt% in metal]/[wt% in silicate]) plotted as a function of oxygen fugacity at 2.5 GPa. For each experiment the latter was calculated relative to the iron–wustite (IW) equilibrium, Fe + 0.5O2 = FeO, by assuming that the activity of Fe (aFe) in the metal was equal to its mole fraction, and that the activity of FeO in the silicate was equal to Fe/(Fe+Mg). In our earlier work15,16, aFeO was obtained from the FeO mole fraction of (Mg,Fe)O solid solution coexisting with metal. We found that the latter correlates well with the Fe/(Fe+Mg) ratio of the melt. Therefore, in the present study we used Fe/(Fe+Mg) of melt to estimate aFeO. Figure 1 shows that V, Cr and Nb have similar partitioning behaviour at 2.5 GPa, and that they partition into the metal much more strongly than does Si. At the oxygen fugacity appropriate for core–mantle equilibrium, partition coefficients of 0.01–0.1 would be expected for V and Cr and values close to 10-4 for Si. Ta (not shown) has similar partitioning behaviour to Si under these conditions15. The main implications of the figure are that if V and Cr are partially dissolved in the core, then some Nb should also be present but little or no Ta or Si.

Results at 25 GPa (Fig. 2) are appropriate for investigating the hypothesis of core segregation at the bottom of a 700-km-deep magma ocean11,12,13. Nb and V remain more siderophile than Si and, as at 2.5 GPa, have virtually identical metal–silicate partition coefficients to one another. In the oxygen fugacity range appropriate for core–mantle equilibrium, partition coefficients of 0.06–0.6 are found for these elements. Partition coefficients increase from 2.5 to 25 GPa and they probably also depend on temperature, so we cannot be sure that we have reproduced the partition coefficient which actually applied during core formation. However a D value of 0.6 would result in 23% of the element having entered the core, sufficient to produce observably subchondritic ratios of Nb/Ta and Nb/La in the silicate Earth. More robustly, the similar behaviour of V and Nb at both low and high pressure means that, if a significant proportion of the Earth's V is in the core, it must be accompanied by a similar fraction of its Nb. At the appropriate oxygen fugacity for single-stage core formation in the Earth, Cr is slightly more siderophile than Nb while Ta is much less siderophile than Nb (Fig. 2). These results are all consistent with significant dissolution of V, Cr and Nb in the core and the completely lithophile behaviour of Ta.

Figure 2: High-pressure data.
Figure 2

The figure shows data from this study on partitioning between Fe-rich metallic liquid and silicate melt at 2,300 °C and 25 GPa. Partition coefficients and oxygen fugacities are calculated as in Fig. 1. Nb and V have almost identical tendencies to partition into the core, while Cr is slightly more siderophile and Ta much more lithophile. Nb data are consistent with 5+ oxidation state in the silicate. Apparent convergence of the behaviour of Ta, Nb, Cr and V at lowest fO2 is probably an artefact due to changing Henry's law constants of the elements as the metal becomes very rich in Si (16.4%). Some fO2 error bars have been omitted for clarity. At each fO2, error bars are same on all points.

Partition coefficients for Si and Ga (Fig. 2, Table 2) suggest that Si is not a strong enough siderophile, even at 25 GPa, for the core to contain 8% Si and that Ga is a stronger siderophile than is required to explain its depletion in the silicate Earth1. Temperature effects on Dmetal/sil can be large, however17, so the limited temperature range of our experiments preclude definitive conclusions in these cases.

The two competing hypotheses for Nb depletion in the silicate Earth depend only on the nature of the hidden reservoir. Either the core contains significant fractions of the Earth's V, Cr and Nb, or the depletions of V and Cr in the mantle are solely due to incomplete accretion to the Earth. In the latter case the Nb depletion must be due to a hidden silicate reservoir such as subducted refractory eclogite5. The second hypothesis would be supported if some HIMU basalts were found to have superchondritic Nb/Ta and Nb/La (ref. 5), or if V were found to be more volatile in the solar nebula than Fe. Current data suggest that neither of these conditions are met7,9.


  1. 1.

    & .-s. The composition of the Earth. Chem. Geol. 120, 223–253 (1995).

  2. 2.

    , , & The chemical composition of the Earth. Earth Planet. Sci. Lett. 134, 515–526 (1995).

  3. 3.

    in Global Earth Physics (ed. Ahrens., T. J.) 159–189 (American Geophysical Union Reference Shelf 1, Washington DC, 1995).

  4. 4.

    Chemical differentiation of the Earth: the relationship between mantle, continental crust and oceanic crust. Earth Planet. Sci. Lett. 90, 297–314 (1988).

  5. 5.

    , , & Rutile-bearing refractory eclogites: missing link between continents and depleted mantle. Science 287, 278–281 (2000).

  6. 6.

    , & V, Cr and Mn in the Earth, Moon, EPB and SPB and the origin of the Moon: Experimental studies. Geochim. Cosmochim. Acta 53, 2101–2111 (1989).

  7. 7.

    Meteorites: Their Record of Early Solar-system History (Freeman & Co., New York, 1995).

  8. 8.

    Mantle geochemistry: the message from oceanic volcanism. Nature 385, 219–228 (1997).

  9. 9.

    & Source characteristics derived from very incompatible trace elements in Mauna Loa and Mauna Kea basalts, Hawaii Scientific Drilling Project. J. Geophys. Res. 101, 11831–11839 (1996).

  10. 10.

    , & Thermodynamic properties of minerals and related substances at 298.15K and 1 bar (105 Pascals) pressure and at higher temperatures. US Geol. Surv. Bull. 1452 (1978).

  11. 11.

    & Geochemistry of mantle-core differentiation at high pressure. Nature 381, 686–689 (1996).

  12. 12.

    , & Prediction of siderophile element metal-silicate partition coefficients to 20 GPa and 2800 degrees C: The effects of pressure, temperature, oxygen fugacity, and silicate and metallic melt compositions. Phys. Earth Planet. Int. 100, 115–134 (1997).

  13. 13.

    & Effect of water on metal-silicate partitioning of siderophile elements: a high pressure and temperature terrestrial magma ocean and core formation. Earth Planet. Sci. Lett. 171, 383–399 (1999).

  14. 14.

    & The influence of pressure and temperature on the metal-silicate partition-coefficients of nickel and cobalt in a model-c1 chondrite and implications for metal segregation in a deep magma ocean. Geochim. Cosmochim. Acta 59, 991–1002 (1995).

  15. 15.

    & Metal-silicate partitioning and the incompatibility of S and Si during core formation. Earth Planet. Sci. Lett. 152, 139–148 (1997).

  16. 16.

    Geochemical Constraints on the Formation of the Earth's Core. Thesis, Univ. Bristol (1999).

  17. 17.

    , , & Solubility of silicon in liquid metal at high pressure: implications for the composition of the Earth's core. Earth Planet. Sci. Lett. (in the press).

  18. 18.

    Making continental crust. Nature 378, 571–578 (1995).

  19. 19.

    Phase transformations and partitioning relations in peridotite under lower mantle conditions. Earth Planet. Sci. Lett. 174, 341–354 (2000).

Download references


This work was supported by the NERC. Experiments at Bayreuth were performed with assistance from the EU Large Scale Facility programme. B.J.W. acknowledges a Max Planck research award.

Author information


  1. Department of Earth Sciences, University of Bristol, Bristol BS8 1RJ, UK

    • J. Wade
    •  & B. J. Wood


  1. Search for J. Wade in:

  2. Search for B. J. Wood in:

Corresponding author

Correspondence to B. J. Wood.

About this article

Publication history






Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.