Report

Phase Transformations and Metallization of Magnesium Oxide at High Pressure and Temperature

See allHide authors and affiliations

Science  07 Dec 2012:
Vol. 338, Issue 6112, pp. 1330-1333
DOI: 10.1126/science.1229450

Abstract

Magnesium oxide (MgO) is representative of the rocky materials comprising the mantles of terrestrial planets, such that its properties at high temperatures and pressures reflect the nature of planetary interiors. Shock-compression experiments on MgO to pressures of 1.4 terapascals (TPa) reveal a sequence of two phase transformations: from B1 (sodium chloride) to B2 (cesium chloride) crystal structures above 0.36 TPa, and from electrically insulating solid to metallic liquid above 0.60 TPa. The transitions exhibit large latent heats that are likely to affect the structure and evolution of super-Earths. Together with data on other oxide liquids, we conclude that magmas deep inside terrestrial planets can be electrically conductive, enabling magnetic field–producing dynamo action within oxide-rich regions and blurring the distinction between planetary mantles and cores.

Magnesium oxide (MgO) is among the simplest oxides constituting the rocky mantles of terrestrial planets such as Earth and the cores of Jupiter and other giant planets. Present in Earth’s mantle as an end-member component of the mineral (Mg,Fe)O magnesiowüstite, it can be abundant in larger planets due to the theoretically expected dissociation of (Mg,Fe)SiO3 perovskite (1). Among common planetary constituents, MgO is thought to be especially resistant to transformation under pressure and temperature, having a high melting temperature (27), a wide electronic band gap under pressure (1, 8, 9), and a simple phase diagram featuring three phases at the high pressures explored in the present study: the B1 (NaCl)–structured solid found at ambient conditions, the liquid state, and an as-yet-unobserved high-pressure crystalline phase having the B2 (CsCl) structure (15, 9, 10).

To characterize MgO at the elevated pressures and temperatures of planetary interiors, we measured its pressure-volume (P, V) equation of state to 0.6 TPa (6 Mbar) and its temperature (T) and optical reflectivity (R) to beyond 1.4 TPa (14 Mbar) and 50,000 K under shock loading (11). We determined shock temperature and corresponding shock velocity (US) using the method of decaying shocks (1114), in which an optically thick and reflecting shock wave decreases in amplitude as it propagates, revealing—through time-resolved measurement of emission, velocity, and reflectivity—the properties of a continuum of shock states (Fig. 1). Determining the corresponding particle velocity (UP) in separate experiments (Fig. 2 and fig. S1) allows for the derivation of pressure and volume from our measurements (11, 15).

Fig. 1

Schematic of laser-shock experiments (A) showing target configuration with drive laser impinging on an Al buffer plate to which the sample (MgO) or a combination of standard (SiO2) and sample were attached (11). Shock velocity, temperature, and reflectivity were determined using velocity interferometry (B) and streaked optical pyrometry (C) that document intensity (of interference fringes or self-emission, respectively) as a function of distance across the target (vertical axis, ~500 μm full scale) and time (horizontal axis, ~22 ns full scale) (1114). Reflectivity and velocity are determined, respectively, from the observed intensity and fringe shift in (B), with shock velocities extrapolated below US = 17.3 km/s where reflection from the shock was not detectable (11). In the shock-velocity (black dots, dashed line where extrapolated) and emission-intensity (blue dots) records (D), events are labeled e1 to e6: Entry of the shock into the MgO sample (e1) and a period of nearly steady shock propagation are followed by steady decay of velocity and emission until e2, where the rate of emission decay decreases, then increases (e3), while velocity continues to decay steadily; then emission increases (e4 to e5) and decays again before the shock exits the MgO (e6).

Fig. 2

Shock temperature T (A), specific heat CV (B), and shock-front reflectivity R (C) measured for MgO in two experiments [blue and red solid curves, systematic uncertainty bounds (11) shown by dotted curves (mean ± SD)] are shown as a function of shock-wave velocity (bottom scale) and pressure (top scale), determined from a linear US-UP equation-of-state (D) based on a fit (11) to data at low pressure (3033) (off scale, fig. S1) and high pressure (open symbols, table S1) measured (34) relative to SiO2 (35). A Drude-semiconductor model fit to the reflectivity data is shown by the green curve in (C) (11); solid and dashed black lines indicate the band gap (Eg) for models constrained to the zero-pressure value for MgO, or not, respectively. Events, labeled as in Fig. 1, correspond to transitions between anomalous (gray shading) and normal (6 Rgas) values of CV.

We observe that the shock temperature does not change monotonically with pressure, as expected for a single phase (15, 16), but instead shows a large anomaly with a temperature minimum at about 0.45 TPa and 8500 K (Figs. 1 and 2). Two additional regions of anomalous behavior (at about 0.65 TPa and beyond 1 TPa) are most evident when examining the specific heat, CV, of the shock states (Fig. 2B) (1113). In broad regions of our data, CV closely matches the Dulong-Petit limit (CV = 6 Rgas, where Rgas is the gas constant per mole of atoms), consistent with the presence of a pure MgO phase above the Debye temperature (~760 K) (12, 13, 15); in three distinct regions, however, there is substantial deviation from this value. The anomalous values of CV suggest the influence of latent heats associated with structural, bonding or electronic transitions.

Indeed, beginning with the second transition (0.55 TPa) there is an increase in optical reflectivity (Fig. 2C) from ~0.5%—a value consistent with the optical properties of low-pressure, insulating MgO (17) (figs. S2 and S3)—toward ~20%, indicating substantial changes in electronic properties. A simple (Drude-semiconductor) model of the measured reflectivities (1113) is consistent with the energy gap between valence and conduction electronic bands nearly vanishing during the second transition (Fig. 2C), resulting in metallic conductivities >104 S/m above 0.60 to 0.70 TPa (fig. S4) (11).

To interpret these experimental results, we constructed a Mie-Grüneisen-Debye, finite-strain equation-of-state (16, 18) for MgO constrained by the shock data (11). From this model, we assess phase-transition properties, such as the volume and entropy of transformation (ΔVtr and ΔStr, respectively). We find that the low-pressure behavior of MgO (to 0.35 TPa and 9500 K) is well described with this model, taking parameters solely from measurements on B1-MgO near ambient conditions (e.g., pressures below 0.005 TPa) (19). Beyond 0.35 TPa, a sequence of phase transformations is required to explain the anomalies in shock temperature, consistent with our specific heat analysis (Table 1).

Table 1

Comparison between experimental and theoretical properties of phase transitions in MgO. Experimental ranges include thermodynamic equilibrium and nonequilibrium [e.g., superheating (12)] interpretations of the transitions (11). The Clapeyron slope is Embedded Image. Uncertainties are mean ± SD.

View this table:

Several interpretations are possible for the first, most prominent transition, although the observed decrease in shock temperature indicates a positive entropy change (positive latent heat) from low- to high-pressure phases. If the transformation occurs at thermodynamic equilibrium, the data follow a phase boundary having a negative Clapeyron slope and implying a volume change as negative as ΔVtr/V ~ –7% (Table 1) (11, 15). This suggests that the first transition corresponds with solid-solid transformation and the second [consistent with positive volume and entropy change (Table 1)] with melting. Indeed, pressure-temperature conditions of the first and second transitions, and Clapeyron slopes obtained assuming equilibrium transformation, are close to theoretically predicted values for the B1-B2 and melting transitions of MgO (Table 1) (2). Alternatively, the first transition could be due to melting (3), with associated superheating upon shock compression (12). In this case, the first transformation can have negligible volume change (Table 1) (11), and the second corresponds with a transformation in the fluid.

Our preferred interpretation is that the two transitions are due to B1-B2 (0.45 TPa) and B2-melt (0.65 TPa), respectively. The second transformation is unlikely to be anything other than melting; there is no theoretical evidence for a liquid-liquid transition in MgO (8), and known fluid-fluid transitions in other shock-compressed insulators typically occur over a much broader pressure and temperature range (12, 13). Also, the transition from electrically insulating solid to metallic liquid in this interpretation is consistent with theoretical predictions of electronic properties for MgO (1, 8, 9) (fig. S4) and reflects the melting behavior of other nonmetals such as carbon (13), silicon, and sulfur (20).

Continuous changes in the reflectivity of the liquid state suggest that liquid electronic properties are strongly sensitive to temperature and pressure (8). The large entropy of the first transition relative to typical solid-solid transitions remains poorly understood but is consistent with available theory for MgO (Table 1). The broad region of anomalously high specific heat beyond ~1 TPa and ~33,000 K is similar to that observed at extreme temperatures in carbon and SiO2 (12, 13) and is similarly attributed here to the onset of complex fluid bonding or to increased electronic specific heat. With this interpretation, we have a self-consistent picture for the phase diagram of MgO (Fig. 3), involving two crystalline phases (B1 and B2) and one liquid phase (14, 9, 10).

Fig. 3

Phase diagram of MgO constructed from shock temperature data [solid circles, this study, red and blue for two shots; open black circles (36)] with B1-phase temperature model (dot-dashed black line) (36). Our proposed phase diagram (heavy black lines) is a modification of the recent theoretical phase diagram of (2, 4), in which the melt curve is roughly identical but the B1-B2 transition has moved to higher pressure. Zero-temperature B1-B2 transition pressures from theory (excluding outliers) are given by the gray bracket (10). Experimental melting temperatures (6, 7) are open triangles. Also shown are the conditions of planetary interiors for Earth (21), a theoretical Earth-like planet of 5 earth masses (Super Earth) (22), Jupiter (23), and a hot Jupiter-mass planet (average prediction) (37); temperature discontinuities at 0.13 and 0.65 TPa in the terrestrial planets correspond to the base of the oxide mantle.

We conclude that MgO is solid at conditions found in the present Earth (21), in large Earth-like planets (22) and—based on extrapolations of the melting curve—in Jupiter and its core (23) (Fig. 3). The B2 phase of MgO is likely to be stable in the deep mantles of Earth-like planets of more than four Earth masses (22) (Fig. 3), so a B1-B2 transition can be relevant to planetary iron-bearing magnesiowüstite (24). With the corresponding decrease in volume (up to 7%) and large latent heat of transformation (Table 1), a B1-B2 transition could influence internal structure and dynamics (22, 25), orbital evolution (25), and exoplanet mass-radius relationships (22). This is particularly true for terrestrial planets exceeding ~8 Earth masses (22), where dissociation of perovskite (1) could make magnesiowüstite the most abundant mineral of the deep mantle.

Major end-member oxides of deep planetary interiors (1)—SiO2 (12), MgSiO3 (14), and MgO—uniformly exhibit high electrical conductivities (σ) in their fluid states at high pressure and temperature, consistent with metallic or semimetallic behavior: σ ~ 103 to 105 S/m at T ~ 5000 to 15,000 K and P ~ 0.1 to 0.7 TPa. Planetary magmas composed of these rocky constituents should likewise be highly conductive at high pressure, as compared with magmas familiar at Earth’s surface (σ ~ 10−2 to 102 S/m) (26), and would thereby resemble liquid iron alloys making up planetary cores (27).

Our experiments therefore reveal a blurring between traditional definitions of planetary mantle and core material: Liquid oxides can be considered either molten mantle constituents or electrically conducting core components. Indeed, for terrestrial planets with extensive melting (i.e., magma oceans) and sufficient interior pressure, oxides could contribute to the dynamo process sustaining a planetary magnetic field. The magnetic Reynolds number Rm=μ0σvL (28) for such a magma ocean (Rm ~ 103 to 107) exceeds the critical value required for a self-sustaining field (Rm > 101 to 102), considering flow velocities v appropriate for a terrestrial magma ocean (4 to 40 m/s) (29), and distance scales L appropriate for terrestrial mantles (106 to 107 m); μ0 is the permeability of free space. Although Earth may be too small (P < 0.13 TPa in the mantle) for substantially elevated oxide conductivity, even a modest (order of magnitude) increase in σ over typical magma values can shift Rm in an early-Earth magma ocean from too small (Rm ~ 10−2 to 102) to large enough to sustain dynamo action. This suggests that an early, short-lived magnetic field could have existed on Earth, perhaps similar to those inferred for Mars and the Moon on the basis of remnant crustal magnetism.

Supplementary Materials

www.sciencemag.org/cgi/content/full/science.1229450/DC1

Materials and Methods

Figs. S1 to S6

Tables S1 to S3

References (3853)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank N. Gómez-Pérez, R. J. Hemley, D. C. Swift, R. E. Cohen, D. A. Dalton, and W. Unites for helpful discussions. This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory in part under contract W-7405-Eng-48 and in part under contract DE-AC52-07NA27344. We also acknowledge financial support from the U.S. Army Research Office (grant 56122-CH-H), a Krell Institute Department of Energy (DOE) National Nuclear Security Administration (NNSA) graduate fellowship (contract DE-FC52-08NA28752), the DOE/NNSA National Laser User Facility Program, the Miller Institute for Basic Research in Science, and the University of California. Data are available in this manuscript and in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article