Post-Perovskite Phase Transition in MgSiO3

See allHide authors and affiliations

Science  07 May 2004:
Vol. 304, Issue 5672, pp. 855-858
DOI: 10.1126/science.1095932


In situ x-ray diffraction measurements of MgSiO3 were performed at high pressure and temperature similar to the conditions at Earth's core-mantle boundary. Results demonstrate that MgSiO3 perovskite transforms to a new high-pressure form with stacked SiO6-octahedral sheet structure above 125 gigapascals and 2500 kelvin (2700-kilometer depth near the base of the mantle) with an increase in density of 1.0 to 1.2%. The origin of the D″ seismic discontinuity may be attributed to this post-perovskite phase transition. The new phase may have large elastic anisotropy and develop preferred orientation with platy crystal shape in the shear flow that can cause strong seismic anisotropy below the D″ discontinuity.

MgSiO3 perovskite is believed to be a principal mineral, at least in the upper part of the lower mantle, but its stability and possible phase transition at greater depths remain uncertain. Because seismic observations have shown unexplained features in the lowermost mantle (14), solid-solid phase transitions that could occur in this region are a subject of debate. Previous experiments have confirmed the conservation of orthorhombic (Mg, Fe) SiO3 perovskite (space group: Pbnm) up to 127 GPa (57), consistent with first-principles total energy calculations (810). In contrast, an experimental study by Shim et al. (11) suggested a subtle change in the perovskite structure above 83 GPa and 1700 K. The dissociation of MgSiO3 into mixed oxides was also previously found at 70 to 80 GPa (12, 13), but it was possibly due to melting or diffusion caused by a large temperature gradient in these earlier studies (14). Here we report in situ x-ray observation of pure MgSiO3 composition at high pressure and temperature up to 134 GPa and 2600 K corresponding to the conditions at the core-mantle boundary region.

Angle-dispersive x-ray diffraction (XRD) spectra were collected at BL10XU of SPring-8 (15). High-pressure and -temperature conditions were generated in a laser-heated diamond anvil cell (LHDAC) (16). MgSiO3 gel was used as a starting material. It was mixed with platinum powder that served both as an internal pressure standard and a laser absorber. The sample mixture (∼25 μm thick) was loaded into a 60-μm hole drilled in the rhenium gasket together with insulation layers of MgSiO3 gel unmixed with platinum (∼10 μm thick on both sides). They were compressed with 200-μm culet beveled diamond anvils. Heating was achieved by a focused multimode continuous-wave Nd: yttrium-aluminum-garnet laser using the double-side heating technique (17), which minimizes radial and axial temperature gradient in the sample (18). A heating spot was about 50 μm in diameter. Temperature was measured from one side by the spectroradiometric method (19). The uncertainty in temperature within the 20-μm area from which XRD spectra were collected was about ±200 K. Pressure was determined using the equation of state of platinum (20) with both (111) and (200) lines. The uncertainty in pressure was ±1.6 to 3.9 GPa, derived mainly from the uncertainty in temperature in the application of P-V-T equation of state. We conducted two separate sets of experiments. The diffraction patterns of the sample were repeatedly obtained at high temperatures during heating and at room temperature after quenching.

In the first run, an amorphous sample was compressed to 124 GPa at room temperature and then heated to 2250 to 2300 K for 11 min at 105 to 114 GPa. The diffraction peaks of orthorhombic perovskite (Pbnm) appeared within 2 min and did not change with further heating. All peaks were indexed by the Pbnm perovskite and platinum (Fig. 1A). We further compressed this sample to 127 GPa at room temperature and reheated it to 2500 to 2600 K for 70 min at 127 to 134 GPa. Eleven new peaks appeared within 9 min. These new peaks grew, and peaks from perovskite became weak with further heating (Fig. 1B). Two-dimensional diffraction images showed circular Debye rings for these new peaks (fig. S1). The consistency of the unit-cell parameters of platinum calculated respectively from (111) and (200) lines indicated that nonhydrostatic stress in the sample was not large after heating in these experiments. The St values—the multiplication of the elastic anisotropy factor S and the uniaxial stress component t—were 0.0010 to 0.0037 at 121 GPa and 300 K (21). The deviatoric stress should have been reduced because of a decrease in the sample volume when perovskite was first synthesized from the gel starting material (22).

Fig. 1.

XRD patterns at (A) 105 GPa and 2250 K, (B) 121 GPa and 300 K after heating for 70 min at 127 to 134 GPa and 2500 to 2600 K, (C) 97 GPa and 300 K after decompression from 125 GPa, and (D) 72 GPa and 300 K after heating at 89 to 101 GPa and 2000 to 2200 K for 10 min. P, Pbnm perovskite; Pt, platinum; N, new phase. The calculated peak positions of MgO and both α-PbO2–type and CaCl2-type SiO2 are shown by small ticks (16, 29). An extra peak position of MgSiO3 perovskite reported by Shim et al. (11) is indicated by a star. The calculated powder XRD pattern of the post-perovskite phase was corrected for Lorentz, polarization, and multiplicity factors to compare with the observed pattern.

In the second set of experiments, perovskite was first synthesized from the amorphous starting material by heating to 1700 to 1970 K for 15 min at 69 to 73 GPa. We then compressed this sample to 122 GPa at room temperature. With heating to 2200 to 2300 K at 128 to 129 GPa, intermittently for a total of 120 min by opening/closing the laser shutter several times at fixed press load, the new peaks were again observed in the diffraction patterns within 10 min of heating and were the same as those in the first run. This sample was then decompressed to 97 GPa at room temperature. The new peaks were still recognized after decompression, although they had broadened (Fig. 1C). After heating to 2000 to 2200 K for 10 min at 89 to 101 GPa, the new peaks disappeared and the diffraction pattern changed back to that consisting only of Pbnm perovskite and platinum (Fig. 1D).

These new peaks do not correspond to the possible dissociation products of MgO or high-pressure polymorphs of SiO2 (Fig. 1B). Shim et al. (11) observed one additional peak in the diffraction pattern of MgSiO3 perovskite and suggested a minor structural change. This extra peak, however, was not found in this study. These new peaks indicate that MgSiO3 perovskite does not dissociate, but is transformed to a new high-pressure form above 125 GPa and 2500 K (Fig. 2). Phase transition in SiO2 from a CaCl2-type to a α-PbO2–type structure, which occurs at similar pressure and temperature conditions (16), did not induce the dissociation of MgSiO3. Xiong et al. (23) reported the phase transition of CaTiO3 perovskite (Pbnm) to hexagonal and tetragonal structures when the pressure was increased to 10 to 15 GPa. However, we failed to assign the new peaks to these structures. Alternatively, the diffraction peaks of a new MgSiO3 polymorph can be indexed by an orthorhombic cell with lattice parameters a = 2.456(0) Å, b = 8.042(1) Å, and c = 6.093(0) Å.

Fig. 2.

Phase diagram of MgSiO3. Solid squares and open circles indicate the stabilities of Pbnm perovskite and post-perovskite phase, respectively. A broken line shows the transition boundary proposed by Sidorin et al. (4) to explain the D″ discontinuity by a solid-solid phase transition.

In order to determine the crystal structure that possesses these lattice parameters, we conducted a molecular dynamics (MD)–aided crystal structure design. The appropriate number of atoms [8 Mg + 8 Si + 24 O; Z = 8 (number of formula unit) for a double unit cell because of the small a-parameter] were positioned randomly in the cell with the experimentally observed dimensions. Classical MD calculations were carried out with the (NVT) ensemble of this system at high temperature (5000 K), and then the system was quenched to 0 K (24) (table S1). We repeated the calculations, each time checking and correcting the atomic positions until the crystal structure became consistent with the crystal chemistry and the calculated XRD pattern matched the observed one. The result revealed a new crystal structure with space group Cmcm (Fig. 3). The crystal data of this post-perovskite phase are presented in Table 1. The calculated powder XRD pattern reproduces both peak positions and intensities of all the observed new peaks (Fig. 1B). We obtained a room-temperature density of 5.536 g/cm3 for the observed new phase at 121 GPa. It is denser than perovskite coexisting in the diffraction pattern by 1.0 to 1.2% at 300 K. The MD calculations also showed smaller molar volumes for the post-perovskite phase (fig. S2).

Fig. 3.

Crystal structure of the post-perovskite phase projected along [001], [100], and [010] directions, and a stereoscopic view showing the layer-stacking structure. Coordination polyhedra of O atoms around Si atoms are shown as octahedra, and the Mg2+ ions are shown as balls. Bold line indicates the unit cell.

Table 1.

Crystal data of the post-perovskite phase at 121 GPa and 300 K.

Crystal system Orthorhombic
Space group Cmcm
Cell parameters a (Å) 2.456
b (Å) 8.042
c (Å) 6.093
Z 4
V3) 120.39
Atomic coordinates x y z
    Mg 0.000 0.253 0.250
    Si 0.000 0.000 0.000
    O1 0.000 0.923 0.250
    O2 0.000 0.631 0.436
Interatomic distances (Å)
    Si-O 1.64 (×2), 1.66 (×4)
    Mg-O 1.84 (×2), 1.94 (×4), 2.13 (×2)
    Si-Si 2.46 (×2), 3.05 (×2), 3.11 (×2)
    Mg-Mg 2.46 (×2), 3.24 (×4)

This new MgSiO3 polymorph has sixfold Si and eightfold Mg coordination, and the SiO6-octahedra share the edges to make an octahedral chain like that of a rutile-type structure (Fig. 3). These chains run along the a axis and are interconnected by apical O atoms in the direction of the c axis to form edge and apex shared octahedral sheets. The octahedral sheets are stacked along the b axis with interlayer Mg2+ ions. The MD calculations suggest that the b axis is more compressible than are the a and c axes (fig. S2), and this compression behavior differs from that of Pbnm perovskite (25). The crystal structure of the post-perovskite phase is isostructural with UFeS3 (26).

Phase transition of MgSiO3-rich perovskite can cause large seismic heterogeneities in the lowermost mantle. The D″ discontinuity is observed in many regions around the world about 200 to 300 km above the coremantle boundary (119 to 125 GPa) with a velocity increase of ∼3% (3), although its ubiquitous occurrence is still a subject of debate (4, 27). The post-perovskite phase transition occurs at depths matching those of the D″ discontinuity (Fig. 2). Although the effects of minor elements such as FeO, Fe2O3, and Al2O3 at the transition pressure remain to be evaluated, this phase transition may be responsible for the D″ seismic discontinuity. A variation in the depth of the discontinuity could be due to a large effect of temperature on phase transition pressure. Sidorin et al. (4) previously proposed a hypothetical transition boundary with a Clapeyron slope of about 6 MPa/K, assuming that the D″ discontinuity is caused by a solid-solid phase transition. Although the Clapeyron slope is not well constrained from the present results, their supposed boundary is consistent with our data (Fig. 2). Masters and Gubbins (28) recently found the excess density of 0.4% in the bottom 500 km of the lower mantle. An expected density increase of 1.0 to 1.2% for the bottom ∼200- to 300-km layer owing to the post-MgSiO3 perovskite transition is consistent with their observations.

The high compressibility of the b axis compared to the a and c axes of the post-perovskite phase suggests slow longitudinal elastic-wave velocities propagating along the [010] direction. In addition, it is also inferred that the post-perovskite phase forms a platy crystal habit parallel to the (010) plane as a result of the sheet-stacking structure. A strong preferred orientation of such platy crystals may develop under the shear flow. A large S-wave polarization anisotropy (VSH > VSV) observed in the D″ region (1, 2) is possibly caused by the preferred orientation of the post-perovskite phase with the (010) plane being parallel to the horizontal shear flow, which is introduced by the down-welling of slabs and upwelling of plumes. This also provides an explanation for seismic anisotropy that is found only below the D″ discontinuity in the deep lower mantle.

Supporting Online Material

Figs. S1 and S2

Table S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article