Iron Partitioning and Density Changes of Pyrolite in Earth’s Lower Mantle

See allHide authors and affiliations

Science  08 Jan 2010:
Vol. 327, Issue 5962, pp. 193-195
DOI: 10.1126/science.1181443


Phase transitions and the chemical composition of minerals in Earth’s interior influence geophysical interpretations of its deep structure and dynamics. A pressure-induced spin transition in olivine has been suggested to influence iron partitioning and depletion, resulting in a distinct layered structure in Earth’s lower mantle. For a more realistic mantle composition (pyrolite), we observed a considerable change in the iron-magnesium partition coefficient at about 40 gigapascals that is explained by a spin transition at much lower pressures. However, only a small depletion of iron is observed in the major high-pressure phase (magnesium silicate perovskite), which may be explained by preferential retention of the iron ion Fe3+. Changes in mineral proportions or density are not associated with the change in partition coefficient. The observed density profile agrees well with seismological models, which suggests that pyrolite is a good model composition for the upper to middle parts of the lower mantle.

The evolution of the structure and dynamics of Earth’s interior is influenced primarily by its composition. Based on geophysical interpretations, the lower mantle is typically considered relatively homogenous; however, recent seismological studies have demonstrated that there are some minor discontinuous changes in seismic velocities in the upper to middle parts of the lower mantle (13). Some of these changes are attributed to the presence of subducted slabs (1, 2) or to unresolved phase transitions in mantle materials, including electron spin transitions in Fe-bearing minerals (1, 4). To resolve the origin of these seismic discontinuities, precise observations of phase transitions and associated changes in chemical compositions, densities, and elastic wave velocities of mantle materials are needed.

Pyrolite is a hypothetical representative bulk composition for the mantle. Experimental studies of phase transitions and composition changes in this material have been made at high pressures and temperatures using both the multianvil apparatus (58) and the laser-heated diamond anvil cell (LHDAC) (9, 10). The major phases in the pyrolite model for the lower mantle are orthorhombic Mg-silicate perovskite (Mg-Pv), magnesiowüstite (Mw, also known as ferropericlase), and cubic Ca-silicate perovskite (Ca-Pv). Majorite garnet (5) and post-perovskite (11) are known to be present as major high-pressure phases in the uppermost and lowermost portions of the lower mantle, respectively.

Understanding the nature of Fe partitioning between the two major phases, Mg-Pv and Mw, under lower mantle conditions may provide important clues for interpreting seismic discontinuities. Fe2+ has been known to preferentially partition into Mw relative to Mg-Pv in the simple MgO-FeO-SiO2 system (12). Recent multianvil experiments on more complex mixtures of pyrolite and peridotite—the dominant rock type of the upper mantle—demonstrated that the Fe-Mg partition coefficient between Mg-Pv and Mw [KD = (Fe/Mg)Mg-Pv/(Fe/Mg)Mw] increases considerably with pressure and approaches unity as a result of Fe enrichment in Mg-Pv (5, 6, 8, 13). This is attributed to a coupled substitution of Mg2+ and Si4+ by Fe3+ and Al3+ in Mg-Pv (6, 13, 14) due to the progressive transformation of majorite garnet to the perovskite structure under the pressures and temperatures of the uppermost lower mantle (5, 6). In contrast, LHDAC studies showed substantially smaller values of KD = ~0.4 to 0.5 under the pressure and temperature conditions of the entire lower mantle (9, 10).

To determine the phase and density changes in pyrolite, we conducted in situ synchrotron-based x-ray diffraction measurements at pressures up to 47 GPa and at temperatures of 1873 to 2073 K along a typical adiabatic geotherm (15) using a multianvil apparatus with sintered diamond anvils (16). The chemical compositions of the coexisting phases were determined by electron microprobe analyses on the recovered samples (fig. S1). Mössbauer and electron energy-loss spectroscopy (EELS) measurements were also made on some run products to evaluate the Fe valence states in Mg-Pv and Mw (16).

The Fe and Al contents of Mg-Pv increase with increasing pressure to 30 GPa, whereas the Mg and Si contents of Mg-Pv decrease concomitantly (Fig. 1 and table S1). Both Mg and Fe in Mw were found to decrease with increasing pressure in this range, whereas other cations increase (e.g., Al, Cr, Si, Na, and Ni). The chemical compositions of both Mg-Pv and Mw remain virtually constant at pressures between 30 and 40 GPa. At higher pressures, however, the compositional changes trend in the opposite direction in Mg-Pv (i.e., Mg and Si increase while Fe and Al decrease). A considerable enrichment of Fe in Mw is also observed at pressures above 40 GPa.

Fig. 1

Variations of chemical compositions of Mg-Pv (A) and Mw (B) in pyrolite as a function of pressure combined with results from an earlier study using the same starting material (5). The number of oxygen atoms is fixed to 3 and 1 for Mg-Pv and Mw, respectively. Filled circles are results of the present study; open circles are from (5).

Based on this chemical data, KD increases with increasing pressure up to about 28 GPa (Fig. 2), which is interpreted in terms of the coupled substitution mechanism in Mg-Pv. The KD value stays almost constant (~0.85) at pressures between 28 and 40 GPa. In this pressure range, virtually no changes in either phase assemblage or chemical compositions of the individual phases are observed. However, KD notably decreases with higher pressures down to ~0.5, approaching the values observed in earlier LHDAC experiments on similar compositions (9, 10). Thus, the high KD values observed in multianvil experiments below 30 GPa can be reconciled with the relatively low values obtained in LHDAC studies because of the substantial decrease of KD at pressures above ~40 GPa.

Fig. 2

Variations of the Fe-Mg KD between Mg-Pv and Mw in pyrolite or peridotite composition (red symbols) as a function of pressure (5, 6, 9, 10). The solid line is based on data from multianvil studies; the line is dashed when only LHDAC data are available. Values determined or estimated for San Carlos olivine compositions (green symbols) are shown for comparison (2529). Circles denote the results based on multianvil experiments; squares are those obtained using LHDAC.

Recent experimental and theoretical studies (4, 1720) have demonstrated that an electronic high-spin to low-spin transition of Fe2+ occurs in Mw over a wide pressure range between ~35 GPa and ~80 GPa in the lower mantle. Because the effective ionic radius of Fe2+ in the low-spin state is smaller than that of Mg2+, an enrichment of Fe in Mw relative to Mg-Pv is expected to accompany the spin transition in Mw. It is unclear whether such transitions occur in Mg-Pv in the corresponding pressure range, which would presumably influence this enrichment (19, 2124). The notable increase in the Fe content of Mw at the expense of Fe in Mg-Pv (Fig. 1), and hence the resulting decrease in KD (Fig. 2), may thus be attributed to the commencement of the spin transition in Mw at pressures near 40 GPa. This could be effectively completed by pressures of ~70 to 80 GPa, when the KD values constrained by LHDAC experiments and some theoretical predictions are taken into account (9, 10, 1820).

Some LHDAC studies demonstrated a decrease in KD in San Carlos olivine with a composition of (Mg0.9Fe0.1)2SiO4 at pressures ~70 GPa (25, 26) (Fig. 2), which was interpreted as the commencement of the spin transition. However, considering the uncertainties associated with the KD values (~0.02 to 0.1) determined using LHDAC (2527), it is difficult to see a clear trend in the variation of KD attributable to the spin transition (Fig. 2). In fact, a recent LHDAC study with careful examination of chemical heterogeneity in the sample actually shows an increase of KD (28). Thus, it is not clear whether the observed decrease in KD at ~70 GPa (25, 26) is related to the spin transition in Mw, which may start at considerably lower pressures near 40 GPa, as suggested in our study. Alternatively, the discrepancy in the possible commencement of the spin transition could be due to a compositional effect, because pyrolite has a more complex chemical composition.

The valence state of Fe has also been shown to influence KD, as is the case for the increase from 23 to 28 GPa, where the enrichment of Fe3+ occurs in Mg-Pv (14). To examine the possible changes in Fe2+/Fe3+ ratios in Mw and Mg-Pv with increasing pressure, EELS and Mössbauer measurements were made on two samples recovered from the runs at 28.7 GPa and 44.2 GPa. The EELS measurements show that Fe in Mw in both of these samples is almost entirely Fe2+, with the Fe3+/(Fe2+ + Fe3+) value being less than ~0.02 (±0.05) (Fig. 3, A and D). Thus, we propose that the increase of Fe in Mw (Fig. 1) is due to the increase of Fe2+ in this phase with increasing pressure above ~40 GPa. The Fe3+/(Fe2+ + Fe3+) values for Mg-Pv at 28.7 GPa and 44.2 GPa are 0.66 (±0.06) and 0.67 (±0.08) (Fig. 3, B and E), respectively, whereas those based on Mössbauer measurements are 0.52 (±0.10) and 0.52 (±0.16) (Fig. 3, C and F). Although there was only a small sample volume available for the Mössbauer measurements, these values agree reasonably well with those estimated from literature values (14). Thus, Mg-Pv has a nearly constant Fe3+/(Fe2+ + Fe3+) value between 0.52 and 0.67 at these pressures, suggesting that both Fe2+ and Fe3+ were removed from Mg-Pv in similar proportions at pressures above ~40 GPa. It has been shown that Fe2+ preferentially partitions into Mw in this pressure range, but Fe3+ may also have been removed from Mg-Pv in association with the slightly decreasing Al3+ content of this phase with increasing pressure (Fig. 1). The slight decrease of Al3+ in Mg-Pv may be attributed to the formation of a small amount of an Al-rich phase or its possible enrichment in fine-grained Ca-Pv, although we did not see any evidence for the presence of additional phases in either in situ x-ray diffraction measurements or scanning electron microscopy (fig. S1) observations.

Fig. 3

Fe L2,3-edge energy-loss near-edge structure spectra based on EELS measurements of a run product at 28.7 GPa for Mw (A) and Mg-Pv (B), and the Mössbauer spectrum for the same sample (C). The corresponding spectra for 44.2 GPa are shown in (D), (E), and (F), respectively. For the Mössbauer spectra, the doublets corresponding to Fe3+ in Mg-Pv and Fe2+ in Mw are shaded light gray and dark gray, respectively; the remaining doublets correspond to Fe2+ in Mg-Pv.

It has been shown that Mg-Pv becomes essentially Fe-free when all Fe in Mw is in the low-spin state at pressures greater than ~80 GPa (4, 25), but these values are only valid for Fe2+ in the simple MgO-FeO-SiO2 system. A certain amount of Fe will be retained in Mg-Pv in pyrolite even when the spin transition causes a depletion of Fe2+ in Mg-Pv because of the presence of a considerable amount of Al2O3 (4 to 5 weight percent) in this phase (table S1). The effect of the spin transition in Mw on the viscosity of Mg-Pv should also be smaller than expected (4, 25) because of the relatively small depletion of total Fe in Mg-Pv in the deeper part of the lower mantle.

Based on observed chemical compositions and unit-cell volumes of the coexisting phases, the volume proportions of Mg-Pv, Mw, and Ca-Pv do not change with pressure (Fig. 4A). Consistent with an earlier study using the quench method at pressures up to 28 GPa (5), the volume proportions remain at ~75%, 18%, and 7%, respectively. The density of pyrolite along a typical adiabatic geotherm, calculated using directly determined unit-cell volumes of individual phases and chemical composition data (16), is in excellent agreement with the seismological model (Fig. 4B). Thus, pyrolite is a reasonable model composition for the lower mantle.

Fig. 4

Proportion of mineral phases in pyrolite with pressure (A). In addition to the lower mantle phases (Mw, Mg-Pv, and Ca-Pv), majorite garnet (Mj) and ringwoodite (Rw) are shown. Filled circles were computed using data from this study; open circles are from (5). Density changes in pyrolite, based on unit-cell volumes and chemical compositions of the coexisting phases (B), are compared with that of the Preliminary Reference Earth Model (PREM) (30).

Supporting Online Material

Materials and Methods

Fig. S1

Table S1


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank K. Funakoshi, Y. Tange, N. Nishiyama, T. Sanehira, and T. Tsuchiya for technical assistance at SPring-8 and for helpful discussions. T.I. thanks the Alexander von Humboldt Foundation and Bayerisches Geoinstitut for support during this study. Parts of this work were supported by a Grant-in-Aid for Scientific Research and Joint Research Project to T.I. from the Japan Society for the Promotion of Science (JSPS).
View Abstract

Stay Connected to Science

Navigate This Article