Low Core-Mantle Boundary Temperature Inferred from the Solidus of Pyrolite

See allHide authors and affiliations

Science  31 Jan 2014:
Vol. 343, Issue 6170, pp. 522-525
DOI: 10.1126/science.1248186

Melting Moments

The boundary between Earth's core and mantle defines where the iron-rich liquid outer core meets the more chemically heterogeneous solid lower mantle and is marked by a sharp thermal gradient of nearly 1500 kelvin. The precise relationship between temperature and melting of the lowermost mantle constrains the structure and heat flow across the core-mantle boundary. In order to identify trace amounts of liquid as melting initiates, Nomura et al. (p. 522, published online 16 January) performed x-ray microtomographic imaging of rocks of a primitive mantle composition that had been subjected to high pressures and temperatures in a diamond anvil cell. The experimentally determined maximum melting point of 3570 kelvin suggests that some phases typically thought to lose stability in the lowermost mantle, such as MgSiO3-rich post-perovskite, may be more widely distributed than expected.


The melting temperature of Earth’s mantle provides key constraints on the thermal structures of both the mantle and the core. Through high-pressure experiments and three-dimensional x-ray microtomographic imaging, we showed that the solidus temperature of a primitive (pyrolitic) mantle is as low as 3570 ± 200 kelvin at pressures expected near the boundary between the mantle and the outer core. Because the lowermost mantle is not globally molten, this provides an upper bound of the temperature at the core-mantle boundary (TCMB). Such remarkably low TCMB implies that the post-perovskite phase is present in wide areas of the lowermost mantle. The low TCMB also requires that the melting temperature of the outer core is depressed largely by impurities such as hydrogen.

The core-mantle boundary (CMB), located at a depth of 2900 km inside Earth, is the interface between molten metal and rock. The temperature jump across the thermal boundary layer (TBL) above the CMB has been believed to be about 1500 K (1), which has important consequences for the dynamics and thermal evolution in the mantle and the core. The temperature at the top of the core should be lower than the solidus temperature of a primitive mantle to avoid global melting above the CMB. Conventionally, the temperature at the CMB (TCMB) has been estimated to be about 4000 K, primarily based on the melting temperature of iron at the inner core boundary (ICB), where solid and liquid cores coexist (13). Such high TCMB implies that MgSiO3-rich post-perovskite, a primary mineral in the lowermost mantle, changes back into perovskite with steeply increasing temperature near the CMB (4), which allows detailed modeling of the thermal structure in the CMB region and the heat flux from the core into the mantle (5). Previous experiments using laser-heated diamond-anvil cell (DAC) techniques showed that the solidus temperature of a primitive mantle is about 4200 K at the CMB, supporting the high TCMB around 4000 K (68). The determination of solidus temperature using the DAC is, however, challenging because detecting a small amount of partial melt is difficult.

We determined the solidus temperature of the lower mantle on the basis of the textural and chemical characterizations of quenched samples after subjecting them to a high-pressure environment like that of the CMB. Our starting material possessed a natural primitive mantle (pyrolite) composition, with about 400 parts per million (ppm) H2O (SM text). It was heated to 2100 to 3900 K at 25 to 169 GPa in a DAC (9), which covers the entire pressure range of the lower mantle (table S1). We subsequently collected x-ray microtomographic images to explore the three-dimensional (3D) internal structure of all samples (9). Additionally we prepared cross sections of heated samples in order to examine more-detailed melting texture and composition in two dimensions.

Comparison between images collected with two different x-ray energies, 7 and 8 keV, showed iron enrichment in the sample, because the K-absorption edge of Fe is 7.11 keV. Some of the samples exhibited an iron-rich pocket at the center, corresponding to the hottest part during laser heating (Fig. 1). Electron microprobe analyses revealed that such iron-rich regions represented quenched partial melt with nonstoichiometric composition, which was enclosed by MgSiO3-rich perovskite (Fig. 2 and fig. S1) (10, 11). We thus bracketed the solidus curve of pyrolite by monitoring the iron-rich pocket in the tomographic images (Fig. 3). With synchrotron x-rays, we successfully detected a melt pool as small as 3 μm, formed by ~3 volume % partial melting at 142 GPa (Fig. 1A).

Fig. 1 Computed tomography (CT) slice image of a pyrolitic material quenched from the deep lower mantle conditions.

The brightness contrast is based on the x-ray linear attenuation coefficient (LAC) of an object. Comparison between the images obtained with 7- and 8-keV energies shows an Fe-rich melt pocket at the hottest part of the sample in (A) but not in (B), bracketing a solidus curve between (A) 142 GPa/3690 K and (B) 151 GPa/3680 K (Fig. 3). Scale bars, 5 μm.

Fig. 2 Melt pocket revealed by both tomographic imaging and microprobe analyses.

(A and B) CT images of the sample quenched from 42 GPa/3200 K for two different slices [the yellow line in (A) corresponds to (B)]. The cross section of this sample was prepared by a focused ion beam and examined by a microprobe (C). The x-ray maps for Si, Mg, Fe, and Ca were obtained for the area marked by white in a backscattered electron image. Pv, perovskite; Fp, ferropericlase.

Fig. 3 Solidus curve of a pyrolitic lower mantle.

Solid and open circles indicate the presence and the absence of partial melt, respectively, based on x-ray tomography. The melting curve is based on the Simon and Glatzel equation. The experimental temperature in this study represents the maximum value, whereas pressure uncertainty is ±10% (9). Light green (7) and purple (8) curves with squares show the solidus temperature of primitive mantle determined by previous in situ measurements in a DAC. The orange band illustrates a possible range of pyrolite solidus reported by (6). The results of earlier multi-anvil experiments are shown by crosses (above solidus, red; subsolidus, blue) (12, 13).

The solidus curve obtained in this study is consistent with the results of multi-anvil experiments performed below 25 GPa (12, 13) but is lower than those of recent laser-heated DAC studies combined with in situ x-ray diffraction measurements (7, 8) (Fig. 3). Although earlier experiments contained little water (68), our experiments included ~400 ppm H2O, which may represent the lower bound of the range of geochemical estimates of water concentrations in the lower mantle (14, 15). The lower solidus temperature found in our study is therefore attributed to both the effect of water and the difference in criteria for the onset of melting. The diffuse scattering used in (7) certainly indicates the presence of melt but is hardly observed for silicate when a melt fraction is minor (8). In contrast, a small amount of partial melt in the DAC sample is best evidenced by the ex situ characterization using microtomography techniques. The solidus curves found in (68) might represent a certain degree of partial melting, which explains their steeper temperature/pressure slopes, because the liquidus temperature increases more rapidly with increasing pressure than the solidus temperature (Fig. 3).

A lower solidus temperature helps explain the present-day thermal and chemical structure of the mantle and the core. The seismic evidence of ultralow velocity zones (ULVZs) may suggest the presence of partial melt above the CMB; however, ULVZs are local, not global, features. Because the CMB temperature is isothermal, such local occurrence indicates that regions where the ULVZ is observed have distinct chemical compositions with lower melting temperatures (16). The upper bound of TCMB should therefore be 3570 ± 200 K from the solidus temperature of a typical lower mantle including 400 ppm H2O (Fig. 3), which is much lower than the conventional estimate of around 4000 K (1). These previous estimates assumed that the liquidus temperature of the core alloy is depressed by 600 to 700 K at the ICB by the effect of impurities (2, 3). Such an impurity effect is, however, strongly variable depending on the unidentified light elements present in the core. On the other hand, the lower limit of TCMB is given by the melting temperature of the Fe-H alloy, which is lower than those of other possible core constituents (17). Earlier experiments demonstrated that pure Fe and FeH [Fe with 1.8 weight % (wt %) H] melt at 4190 and 2600 K, respectively, at the CMB (3, 17), although the latter was estimated by extrapolation and may thus include relatively large uncertainty. Considering that the 10% core density deficit is reconciled with 1.2 wt% H (18), TCMB could be as low as 3100 K (Fig. 4).

Fig. 4 Temperature profile (geotherm) in the lower mantle and the outer core.

Dark blue curve, solidus of pyrolite (this study); light green curves, liquidus and solidus of pyrolite (7). Melting (liquidus) temperatures of pure Fe (3), Fe-O-S alloy (26), and FeH (17) are shown by black, pink, and blue lines, respectively. The dashed curves represent extrapolations of experimental data.

Seismological analysis shows the pairs of positive and negative velocity discontinuities in the lowermost mantle, which have been attributed to the double-crossing of the post-perovskite phase transition boundary by the geotherm (4). Earlier experimental work (7) demonstrated that perovskite is formed above 3500 ± 150 K at the CMB pressure in a natural mantle material. Considering the upper bound of TCMB (3570 ± 200 K) found in this study, the stability of perovskite at the base of the mantle is therefore marginal. Alternatively, the low TCMB suggests that post-perovskite occurs not only in cold subducted slabs but also in wider regions of the lowermost mantle.

The temperature extrapolated along an adiabat from the transition zone is 2500 to 2800 K at the CMB (5), indicating a temperature jump of 300 to 1100 K across the TBL developed near the base of the mantle. The thermal conductivity of the lowermost mantle, a mixture of (Mg,Fe)SiO3 post-perovskite and (Mg,Fe)O ferropericlase, has been estimated to be 7.4 to 11.0 W/m/K at 136 GPa and 3600 K (19, 20). Recent theoretical predictions of high thermal conductivity of the core suggest the adiabatic heat flux of 14 to 20 TW at the top of the core (21, 22), which should be comparable to the heat flux across the CMB unless the topmost core is thermally stratified in a wide depth range (22). In this case, the TBL thickness is limited to less than 130 km (fig. S2). Such thin TBL gives an alternative explanation for the seismic velocity reduction observed ~100 km above the CMB (23).

We also calculated the temperature at the ICB (TICB) from TCMB and the isentropic gradient across the outer core given by (dln T/dln ρ) = γ, where ρ is density and γ is the Grüneisen parameter. With uniform γ = 1.51 in the core (24) and ~3100 K < TCMB < 3570 ± 200 K, we obtain ~4200 K < TICB < 4860 ± 270 K (Fig. 4). The core must be fully molten at both the CMB and the ICB. Recent density and sound velocity measurements proposed Fe90O0.5S9.5 (in weight ratio) as a liquid core composition (25). However, the liquidus temperature of Fe90O0.5S9.5 should be higher than that of Fe85O2S13 (3600 K at the CMB and 5630 K at the ICB) (26) and thus exceeds the outer core temperatures estimated above. On the other hand, geochemical models based on Si isotope data suggest 6 wt % Si in the core (27). Nevertheless, even the eutectic (solidus) temperatures of Fe90.35Si7.35S2.3 and Fe80.07Si10.36S2.57 were reported to be 3750 and 3600 K at the CMB, respectively (28), and their liquidus temperatures should therefore be higher than TCMB.

Recent DAC experiments revised the melting temperature of pure Fe upward to be 4190 and 6230 K at CMB and ICB pressures, respectively (3), which is consistent with both ab initio calculations and shock-wave experiments. The upper bounds of TCMB and TICB found in this study require that the liquidus temperature of the outer core alloy is depressed by >600 K at 136 GPa and >1400 K at 329 GPa from that of pure Fe (Fig. 4). Such a large depression would be impossible without H in the core as discussed above (2, 17). Given that the incorporation of 6 wt % Si contributes to diminishing the liquidus temperature by ~150 K (29) and the density by ~5% (30), Fe93.4H0.6Si6 (in weight ratio) may be compatible with the low TCMB as well as the 10% core density deficit, although a more precise estimate requires knowledge about liquid density and melting behavior in the Fe-H-Si system. A large amount of H may have been incorporated into metals from a hydrous magma ocean at the time of core formation (31).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S7

Tables S1 and S2

References (3255)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank M. Ishikawa, Y. Kudo, and T. Tomomasa for their assistance in the experiments at BL10XU and BL47XU, SPring-8 (proposals no. 2012B0087 and 2012B1706). Discussions with E. Takahashi and Y. Nakajima were helpful. T. Kawamoto and F. Tomiyasu supported the Fourier transform infrared and thermal conversion elemental analyzer measurements, respectively. R.N. was supported by a Japan Society for the Promotion of Science Fellowship for Young Scientists. Data are available in the supplementary materials.
View Abstract

Navigate This Article