The study of seismic wave propagation and normal mode oscillation are direct probes of the Earth's interior, sensing sound velocities and density. However, to derive accurate compositional models, these seismological observations need to be combined with experiments constraining the density and elasticity of highly compressed minerals. Looking at velocity vs. density, Birch proposed in the 1950's that the Earth's core is composed of iron alloyed with nickel and some “light element(s)”. More recently, high-pressure measurements on several iron compounds (FeO, FeSi, FeS, FeS2 and Fe3C) suggested an Earth's inner core model of iron alloyed with silicon (2.3 wt.%) and traces of oxygen (0.1 wt.%). However, these results are based on assumptions that can only be confirmed by direct measurements on samples of realistic chemical compositions. Furthermore, a clear limitation of the proposed model is that it is derived only considering the aggregate compressional sound velocity (VP), whereas the largest discrepancy between the seismological observations and mineral-physics results is reported for the shear wave velocity (VS).

We carried out inelastic X-ray scattering measurements at beamline ID28 on a Fe-Ni-Si alloy with 4.3 wt.% of Ni and 3.7 wt.% of Si (Fe0.89Ni 0.04Si0.07). We collected data from the hexagonal close-packed (hcp) phase, at 27, 37 and 47 GPa on quasi-hydrostatically compressed samples, and at 32, 68 and 108 GPa on non-hydrostatically compressed samples. We then derived VP from the aggregate longitudinal acoustic phonon dispersion, and we obtained VS combining our measurements with values of the bulk modulus. The densities (ρ) were directly determined by X-ray diffraction.

In Figure 13, the measured VP is plotted as a function of ρ, together with values for pure Fe and Fe0.78Ni 0.22 alloy. While no systematic offsets can be observed between data for pure Fe and Fe–Ni alloy, velocities for Fe0.89Ni 0.04Si0.07 are about 9% higher at the same density. We also note that our experimental results compare favourably with calculations on Fe0.9375Si0.0625 (Figure 13), further stressing that the increase in VP is solely due to the Si incorporation.

Fig. 13: Aggregate compressional sound velocity vs. density. Circles: Fe0.89Ni0.04Si0.07; squares: Fe; triangles: Fe0.78Ni0.22; open hexagons (calculations): Fe0.9375Si0.0625. Lines are linear regressions to the experimental data (solid – Fe0.89Ni0.04Si0.07; dotted – Fe; dashed – Fe0.78Ni0.22).

Figure 14 shows the extrapolation of our results to core conditions, reported along with the density evolution of hcp-Fe as compared with the seismic velocity profile from the preliminary reference earth model (PREM). VP for pure Fe is lower than the PREM, whereas adding 3.7 wt.% Si yields a velocity that is too high. If we consider simple linear mixing of Fe and Fe0.89Ni0.04Si0.07 (i.e. an ideal solution mixing model), we match the PREM values of VP and ρ for an alloy with 1.2 wt.% of Si (blue dashes in Figure 14). However, both Fe and Fe0.89Ni0.04Si0.07 exhibit values of VS significantly higher than PREM. From the above, it is clear that a simultaneous solution for VP, VS and ρ cannot be obtained by simply varying the amounts of Ni and Si.

At core temperatures (4000–6000 K) anharmonic effects are, however, expected, and we therefore applied temperature corrections (at constant density) to our ambient-temperature results. Starting from computational estimations for pure Fe, we assumed 4% softening of VP and 30% softening of VS at 5000 K and 13000 Kg/m3. The thermal softening of VP requires an increase in the Si content to ≈1.5 wt.% to match the seismic observations. Most importantly, this composition appears to provide a simultaneous solution for both VP and VS, consistent with PREM values (red dots in Figure 14).

Fig. 14: Aggregate compressional (VP) and shear (VS) sound velocities and density extrapolations. Circles: IXS data on Fe0.89Ni0.04Si0.07; diamonds: PREM. Solid lines – Fe0.89Ni0.04Si0.07; dotted lines – pure Fe. Blue dashes are estimated values for Fe0.936Ni0.040Si0.024 (≈1.2 wt.% Si), neglecting temperature corrections. Red dots are estimated values for Fe0.93Ni0.04Si0.03 (≈1.5 wt.% Si) at 13000 Kg/m3 and 5000 K.

Our results suggest an inner core composition containing 4–5 wt.% of Ni and 1–2 wt.% of Si. The exact amount of Si might vary depending upon the temperature corrections or if other elements are also present in the inner core. Our conclusions pertain strictly to solid Fe-alloys and hence the inner core. Elements such as oxygen that are expected to reside mainly in the outer core cannot be adequately constrained here. However, our proposed core composition matches all three primary geophysical observables (VP, VS and ρ) simultaneously for the first time. Taking literature values for the partition coefficient for silicon between the liquid and solid phase of iron, we estimate an entire core composition with Si ranging from 1.2 to 4 wt.%.


Principal publication and authors

D. Antonangeli (a,b), J. Siebert (a,b), J. Badro (a,b), D.L. Farber (b), G. Fiquet (a), G. Morard (a) and F.J. Ryerson (b), Earth Planet. Sci. Lett. 295, 292-296 (2010).
(a) IMPMC, IPGP, Univ. Pierre et Marie Curie, Univ. Paris Diderot (France)
(b) LLNL, Livermore-CA (USA)