- © 2014 The Mineralogical Society
Ultrahigh-pressure and -temperature (P-T) experimental techniques have progressed rapidly in recent years. By combining them with X-ray diffraction measurements at synchrotron radiation facilities, it is now possible to examine deep Earth mineralogy in situ at relevant high P-T conditions in a laser-heated diamond anvil cell (DAC). The lowermost part of the mantle, known as the D″ layer, has long been enigmatic because of a number of unexplained seismological features. Nevertheless, the discovery of a phase transition from MgSiO3 perovskite to ‘post-perovskite’ above 120 GPa and 2400 K indicates that post-perovskite is a principal constituent in the lowermost mantle, which is compatible with seismic observations. The ultrahigh P-T conditions of the Earth's core have not been accessible by static experiments, but the structure and phase transition of Fe and Fe-alloys are now being examined up to 400 GPa and 6000 K by laser-heated DAC studies.
Materials existing below 200 km depth are generally inaccessible in Nature. However, the mineralogy of the deep Earth has been examined in the laboratory by high P-T experiments. A laser-heated DAC is now widely used for this purpose (Fig. 1), in which a sample is first compressed to a desired pressure at room temperature by a pair of opposed single-crystal diamonds and subsequently heated to high temperatures by the application of a near-infrared laser. The sample is usually sandwiched by a pressure-transmitting medium, which also works as a thermal insulator, and irradiated from both sides with a laser beam passing through the diamonds. Such a laser-heated DAC can generate the highest static pressure and temperature on a sample for available high-pressure devices. The most recent DAC experiments have been made up to 407 GPa and 5960 K (Tateno et al., 2012a), which is beyond the conditions estimated to exist at the centre of the Earth. A large-volume multianvil press has recently reached 100 GPa at high temperature (Yamazaki et al., 2014). In a laser-heated DAC, the high P-T conditions are generated only in a small area of a sample (~10−15 m3). A relatively homogeneous temperature (<10% fluctuation) is obtained in an area of <10 μm around the hot spot when the sample is heated to ~5000 K at ~300 GPa (Tateno et al., 2012a). Nevertheless, the size of the X-ray beam available at synchrotron sources is small enough (<6 μm half-width of the maximum) to probe the crystal structure of materials under high P-T.
Primary phases in the mantle are Mg silicates, which change from Mg2SiO4 olivine in the upper mantle to Mg2SiO4 wadsleyite/ringwoodite in the transition zone. A MgSiO3 phase with the perovskite structure was first synthesized in 1974 at 30 GPa, corresponding to the conditions at the uppermost part of the lower mantle (Liu, 1974). Since then, MgSiO3-rich perovskite has long been believed to be a primary constituent of the lower mantle from 660 km down to 2900 km. Thus, MgSiO3 perovskite would make up about half of the Earth's volume, making it the most abundant constituent phase in the planet. However, seismology shows a shear-wave velocity jump a few hundred kilometres above the core-mantle boundary (CMB) (Lay and Helmberger, 1983). The depth of this D″ discontinuity varies; it is observed at deeper levels underneath the circum-Pacific region, consistent with a solid–solid phase transition with a positive Clapeyron slope (Sidorin et al., 1999). A corresponding deep-mantle phase change was not known at that time. However, a phase transition from MgSiO3 perovskite to post-perovskite was discovered in laser-heated DAC experiments performed by Murakami et al. (2004), based on a drastic change in the X-ray diffraction (XRD) pattern above 120 GPa and 2400 K, 30 years after the perovskite phase was first synthesized.
The P-T conditions in the Earth's core are >135 GPa and 3600 K (Boehler, 1993; Nomura et al., 2014). It is more difficult to heat a sample in a laser-heated DAC at higher pressures, because the thermal insulation layers between the sample and the diamonds become thinner. With SiO2 glass thermal insulation layers, Tateno et al. (2010a) first succeeded in making XRD measurements of Fe at inner-core conditions up to 377 GPa and 5700 K. More recent experiments have shown stable crystal structures and phase transitions in Fe compounds such as FeO (Ozawa et al., 2011b), Fe-Ni (Tateno et al., 2012a), and Fe-Si alloys (Tateno et al., 2012b; Fischer et al., 2013) under ultrahigh P-T conditions of the core.
Discovery of post-perovskite
The research group to which the present author belongs first observed a number of unknown diffraction peaks in the XRD patterns of a pyrolitic material (representing the natural mantle composition) in 2002, when the sample was heated to high temperatures at pressures >120 GPa. Subsequently the sample was switched to pure MgSiO3 and similar unassigned diffractions peaks were obtained upon heating the perovskite sample (Fig. 2a and b) (Murakami et al., 2004). Importantly, such XRD patterns were reversible; all of these unknown peaks disappeared and the perovskite peaks formed again when the sample was decompressed and then reheated at 89 GPa.
The crystal structure of a new ‘post-perovskite’ polymorph of MgSiO3 was determined by Murakami et al. (2004) on the basis of structural relaxation from random structures by molecular dynamics simulations using the unit-cell dimensions determined by the XRD data (Hirose and Kawamura, 2007). This phase exhibits a layered structure along the b axis (Fig. 3), which enhances electrical and thermal conductivity (see below), and is isostructural with the ambient-pressure phase of CaIrO3 (Rodi and Babel, 1950). The post-perovskite phase has six-fold Si and eight-fold Mg coordination, similar to those in the orthorhombically distorted perovskite structure of MgSiO3. However, the volume of the post-perovskite phase is ~1% smaller at the phase transition; comparison of the structures shows that this is due to the Mg2+ site being remarkably smaller in the post-perovskite phase (Iitaka et al., 2004).
The post-perovskite phase transition was soon confirmed by ab initio calculations (Tsuchiya et al., 2004; Oganov and Ono, 2004; Iitaka et al., 2004). Similar phase transitions are now known to take place in a variety of ABO3 compounds. The distortion from an ideal cubic perovskite structure is more enhanced with increasing pressure in these materials, which induces the phase transition from perovskite to post-perovskite (Tateno et al., 2010b).
The post-perovskite phase transition has also been examined in natural pyrolitic mantle and MORB materials (e.g. Murakami et al., 2005; Ohta et al., 2008a; Grocholski et al., 2012) as well as in the (Fe,Al)-bearing MgSiO3 system (Catalli et al., 2009; Andrault et al., 2010). A wide pressure range for the transition was suggested that would not cause a seismic discontinuity (Catalli et al., 2009; Andrault et al., 2010; Grocholski et al., 2012), but a sharp transition between perovskite and post-perovskite was observed when they coexisted with (Mg,Fe)O ‘ferropericlase’ (Fiquet et al., 2010; Sinmyo et al., 2011).
Unique properties of post-perovskite
An unusually large positive Clapeyron slope for the phase transition boundary between perovskite and post-perovskite was first predicted from theory (+8–10 MPa/K; Tsuchiya et al., 2004; Oganov and Ono, 2004). Experiments have shown an even greater value of +13.5 MPa/K (Tateno et al., 2009) (Fig. 4), four to five times larger than those of other major mantle phase transitions in the upper mantle and the transition zone. The high Clapeyron slope leads to a large fluctuation in the pressure of the phase transition, which is consistent with the fact that the depth of the D″ seismic discontinuity ranges from 2550 to 2700 km. The large Clapeyron slope also suggests that the post-perovskite phase transition can promote both upwellings and down-wellings in the lowermost mantle. The active plume formation transports heat efficiently to the shallower part of the mantle, increasing the mean mantle temperature by a few hundred degrees (Nakagawa and Tackley, 2004).
Further, it is possible that the post-perovskite phase can revert to the perovskite structure at deeper levels, due to a very steep temperature gradient just above the CMB. This ‘double-crossing hypothesis’ may explain a pair of positive and negative jumps in the seismic wave velocity observed in the D″ region (Hernlund et al., 2005). If this is the case, the Clapeyron slope gives the minimum temperature gradient in the bottom thermal boundary layer where heat is transported mainly by conduction. It suggests a low CMB temperature of 3700 K and a high global CMB heat flow of 6.6 TW (Tateno et al., 2009).
The layered crystal structure of post-perovskite suggests it has high electrical conductivity because electrons are more conductive within a sheet. The electrical conductivity of (Mg0.9Fe0.1)SiO3 post-perovskite was measured to be ~102 S m−1, higher by three orders of magnitude than that of the perovskite phase at a similar pressure range (Ohta et al., 2008b). The highly conductive lowermost mantle causes strong electromagnetic coupling, and therefore enhances the exchange of angular momentum between the solid mantle and the liquid outer core. This explains the observed variation of a few milliseconds in the length of a day on decadal time scales (Holme, 1998). The thermal conductivity of post-perovskite has also been found to be much greater than that of perovskite by both theory and experiments, at least at 300 K (Ohta et al., 2012a; Haigis et al., 2012).
Phase relations in Fe and alloys
The P and T conditions corresponding to the Earth's core range from 136 GPa and 3600–4000 K to 364 GPa and 4900–5700 K, respectively, much greater than those of the overlying mantle (e.g. Boehler, 1993; Alfè et al., 2007; Nomura et al., 2014). Nevertheless, recent synchrotron XRD studies have revealed the stable crystal structure, solid-state phase transition, and melting temperature of Fe and Fe alloys under such ultrahigh P-T conditions of the core (Anzellini et al., 2013; Hirose et al., 2013).
Several different crystal structures have been proposed for pure Fe under the P-T conditions of the core. Recent experiments have shown that the hexagonal-close-packed (hcp) structure is the stable form of Fe at least up to 377 GPa and 5700 K (Tateno et al., 2010a) (Fig. 5). Earlier calculations suggested that hcp Fe undergoes a phase transition to the body-centred-cubic (bcc) structure above 5400 K at 330 GPa (Belonoshko et al., 2003), but this is not supported by more recent predictions (Stixrude, 2012; Cui et al., 2013).
The core is not pure Fe but is known to include ~5 wt.% nickel (Allègre et al., 1995) and light elements such as Si and O (Takafuji et al., 2005; Siebert et al., 2013; Campbell et al., 2007; Mookherjee et al., 2011; Fukai and Suzuki, 1986). The identification of the light elements in the core is still highly controversial even though Birch (1952) first pointed out, more than 60 years ago, that the density of the outer core is substantially smaller than that of an Fe-Ni alloy under core conditions. The effect of 10 wt.% Ni in Fe on a stable crystal structure has been examined experimentally under the inner-core conditions. While Ni is known to be a stabilizer of the face-centred-cubic (fcc) structure (Kuwayama et al., 2008), XRD measurements performed by Tateno et al. (2012a) demonstrated that the hcp phase is stable to 340 GPa and 4700 K and obtained no evidence of phase transition to the fcc or the bcc structure (Dubrovinsky et al., 2007).
The effects of Si and S in Fe on the stable crystal structure remain controversial. While Vočadlo et al. (2003) predicted that a small amount (~5 atom%) of Si or S impurity in Fe stabilizes the bcc phase with respect to the hcp phase under inner-core P-T conditions, the more recent calculations by Cui et al. (2013) found that the effects of 6.25 atom% Si/S are not enough to stabilize the bcc phase at 325 GPa and 6000 K. On the other hand, Fischer et al. (2013) reported phase diagrams in the Fe–FeSi system to 200 GPa, demonstrating that the CsCl (B2)-type phase appears in addition to hcp with increasing Si concentration in the system. The XRD measurements performed by Tateno et al. (2012b) to 407 GPa and 5960 K indicate that 9 wt.% (16 atom%) Si is soluble in the hcp phase at 4900 K and 330 GPa. These experiments suggest that the hcp phase is present as a single phase in the inner core, considering that the maximum Si content in the core due to geochemical constraints is ~6 wt.% (Allègre et al., 1995; Shahar et al., 2009). The Fe–S phase diagram has not been examined experimentally at inner-core conditions. Recent experiments reported that Fe3S dissociates into a mixture of the Fe-rich hcp phase and the S-rich B2 phase above 250 GPa (Ozawa et al., 2013).
The FeO is probably not a constituent of the inner core but could be formed at the CMB by exsolution from an O-bearing outer core upon secular cooling (Buffett et al., 2000). The FeO is known to be metallic with the NaCl (B1) structure under the CMB P-T conditions (Fischer et al., 2011; Ohta et al., 2012b). Phase relations in FeO have been examined by XRD measurements up to 324 GPa, showing a phase transition from B1 to B2 structures above 240 GPa at 4000 K (Ozawa et al., 2011b; Fig. 6).
Both C and H have strong affinities with Fe and could therefore be present in the core to a large extent (e.g. Wood, 1993; Okuchi, 1997). The formation of Fe3C, cementite, at inner core P-T conditions has been reported by Tateno et al. (2010a). Recent theoretical and experimental studies suggested that Fe7C3 is a dominant phase in the inner core (Mookherjee et al., 2011; Chen et al., 2012). On the other hand, the relatively low CMB temperature may require substantial amounts of H in the core (Nomura et al., 2014).
Phase transitions in extrasolar planets
High P-T ab initio calculations have been used to predict the phases and phase transitions in extrasolar planets, including rocky (terrestrial) planets with masses 2–10 times that of the Earth. More than 1000 exoplanets have already been confirmed (http://exoplanet.eu/catalog/), and >3000 planet candidates were discovered by NASA's Kepler spacecraft.
Stixrude (2012) calculated the phase diagram of Fe up to 100 TPa and 40,000 K, indicating that the hcp phase would still be stable at the centre of a super-Earth with 5 times the mass of Earth, while pure Fe would adopt the fcc structure at the centre of Jupiter. The occurrence of MgSiO3 post-perovskite is limited to the bottom few hundred km of the mantle inside the Earth, but it should have a wide stability region in the interiors of larger planets. The calculations performed by Umemoto and Wentzcovitch (2011) reported dissociation of MgSiO3 post-perovskite into B2-type MgO and MgSi2O5 around 0.9 TPa in both super-Earths and the gas giants. The MgSi2O5 phase has Mg in nine-fold coordination and Si in both seven- and eight-fold coordination. The structure is a monoclinic distortion (P21/c) of the Pbam structure of post-titanite CaGe2O5 (Németh et al., 2007), which is shared by ambient pressure phases such as YAlGeO5 (Jarchow et al., 1981). MgSi2O5 dissociates further in MgO and SiO2 with the dense hexagonal structure of barringerite (Fe2P) at ~2.1 TPa. They also argued that both dissociations have large negative Clapeyron slopes and such reactions promote mantle layering.
The second abundant phase in the Earth's lower mantle is (Mg,Fe)O ‘ferropericlase’ with the B1 crystal structure. While the B1–B2 phase transition occurs in FeO above 240 GPa and 4000 K (Fig. 6), theory predicts that a similar phase change in MgO takes place at much higher pressures (e.g. Belonoshko et al., 2010). Since this phase transition has a negative Clapeyron slope, it disturbs mantle convection in exoplanets. The shock compression experiments by McWilliams et al. (2012) observed a solid–solid phase transition in MgO above 360 GPa and 9000 K. More recent ramp-compression experiments were combined with XRD measurements, which observed directly the B1–B2 transition in MgO above 600 GPa at relatively low temperatures (Coppari et al., 2013).
The author is grateful to colleagues, in particular Dr Y. Ohishi, at BL10XU, SPring-8. Comments from S. Ghosh and an anonymous reviewer helped to improve the manuscript significantly.
This paper is published as part of a special issue of Mineralogical Magazine, Volume 78(2), 2014, in celebration of the International Year of Crystallography.
- Manuscript received 25 November 2013.
- Manuscript Accepted for publication 12 February 2014.