Melting of mantle minerals

During the first few hundred million years following the birth of the solar system more than 4.6 billion years ago, Mercury, Venus, Mars, and Earth were mostly molten due to the energetic and prevalent impacts by meteorites. Although these molten bodies solidified quickly, regional or localized melting persisted for billions of years in some (e.g., Mars, Moon) and is still active in others (e.g., Earth). Specifically for Earth, volcanoes are clear manifestations of ongoing melting, at least near the surface. Melting may also occur at depths 1000s of km beneath the surface, where the melt is trapped within. For example, ultralow-velocity zones (ULVZs) are relatively small patches along the core-mantle boundary (CMB) nearly 2900 km below the surface which have extremely low seismic velocities. ULVZs have been suggested to be induced by the partial melting of rocks at that depth and strongly influence core and mantle dynamics. Therefore, it is of great interest to understand the melting of planetary materials as the melting process plays an important role in the evolution of all planets.

During my time at Yale, I have investigated the melting of two important Earth materials, (Mg,Fe)O ferropericlase and FeO₂H_X iron peroxide, at high pressures (> 25 GPa, corresponding to a depth of ~700 km) in order to decipher the dynamics within Earth’s deep interior.

Despite ferropericlase being the second most abundant mineral in Earth’s lower mantle (depths of ~700-2900 km), its melting behavior at high pressures remained poorly constrained before my PhD work. This is because ferropericlase melts at very high temperatures (> 3000 K) and experiments at simultaneous high pressures and high temperatures is challenging. In order to generate the high pressures and temperatures to melt ferropericlase I used the laser-heated diamond-anvil cell (LHDAC) composed of two opposing gem-quality diamonds which generate high pressures (high enough to simulate to Earth’s core conditions) and an infrared laser focused on the sample.

In situ temperature measurements in LHDACs have been a long-standing challenge. In my first paper (Deng et al., JAP, 2017), we use the radiative transfer equation to analyze the radiation process within samples during laser heating and further evaluate the influences of temperature gradients and wavelength-dependent absorption/emission in temperature measurements. Theoretical modeling combined with experimental measurements demonstrates that wavelength-dependent absorption and temperature gradients can account for the largely aliased apparent temperatures (e.g., ~700 K deviation for an actual 3300 K melting temperature). Additionally, a more rigorous temperature determination method has been proposed to calculate the highest temperature reached within the sample during laser heating. While this method is based on the analysis of ferropericlase, it can be applied more generally to other materials in LHDAC experiments with the knowledge of the temperature profile, sample geometry and absorption properties, and hence is important for LHDAC community as a whole.

With the breakthrough in accurate temperature measurements in LHDACs, we successfully performed 11 sets of melting experiments for (Mg,Fe)O ferropericlase (Deng and Lee, 2017, Nature Comms). Our experimental results revealed a pronounced melting temperature depression of ferropericlase of Earth-relevant compositions at pressures greater than ~40 GPa, corresponding to a depth of ~1000 km, where the electronic spin transition of Fe²⁺ occurs, creating local minima in the solidus and liquidus melting curves. To our knowledge, this is the first time the effects of spin transition on melting have been proposed to behave in this manner. We did a thorough analysis on why the spin transition of Fe²⁺ from high to low spin state could induce the melting temperature depression (Deng and Lee, 2019, AM) and found that this melting depression can be explained within the framework of Lindemann’s law for a Debye-like solid. The spin transition of iron from high to low reduces the molar volume and the bulk modulus of the crystal, leading to a decrease in Debye frequency and consequently lowering the melting temperature. Thermodynamically, the melting depression may derive from a more negative Margules parameter for a liquid mixture of high- and low-spin endmembers than that of a solid mixture. Motivated by the melting depression on ferropericlase, we also assessed the influence of spin transition on the melting of iron-bearing bridgmanite, the most abundant mineral in the Earth, and found qualitative consistency with extant literature data.

In a separate paper, we fitted our experimental results to the regular solution model and constructed the phase diagram for ferropericlase to CMB conditions (136 GPa) (Deng and Lee, 2019a, JGR). Based on this phase diagram, we found that ULVZs are not necessarily patches of melts and instead may be explained by a solid-state mixture of ferropericlase with 35 ≤ Mg# = (100×Mg/(Mg+Fe) by mol) ≤ 87 and coexisting bridgmanite.

The melting behavior of ferropericlase not only informs us of the temperatures at which ferropericlase melts, but also many other physical properties that are intimately related to melting, such as diffusivity (a parameter describes how fast the atoms can diffuse) and viscosity (a parameter governs the ease of plastic deformation such as expected in mantle convection). In (Deng and Lee, 2017, Nature Comms), we calculate the relative viscosity profiles of ferropericlase and find that viscosity increases 10–100 times from ~750km to ~1000–1250km, before decreasing at greater depths. This prediction is important because it provides a single mechanism for observations that have been revealed recently by seismic tomography that slabs sometimes stagnate (e.g., the subducting slab below Tonga) and plumes (i.e., ‘hotspot’ volcanoes which are believed to have deep mantle origins) often deflect at these depths.

In contrast to ferropericlase, the abundance of FeO₂H in Earth is much smaller (<<1%). Nevertheless, recent LHDAC experiments suggest that the pyrite-structured iron peroxide with varying hydrogen concentration (FeO₂H_x) is stable up to 2,600 K and 133 GPa while exhibiting thermoelastic properties consistent with ULVZs. As a result, subducted FeO₂H_x has been invoked to be a plausible cause of ULVZs. However, the temperature near the CMB is expected to reach 4000 K at 136 GPa. Under such extreme conditions, the stability of FeO₂H_x remains unknown. If FeO₂H_x decomposes or melts in the deep Earth, it may release hydrogen and oxygen, providing oxygen and/or water at or above the surface. Therefore, we perform the first-principles molecular dynamics (FPMD) simulations to study the thermodynamic and structural properties of FeO₂H_x solid and melt using the density functional theory. Our results show that FeO₂H_x is not stable at CMB conditions and would probably be melted (Deng et al., 2019b, JGR).

References

Deng et al., JAP, 2017

Deng and Lee, 2017, Nature Comms

Deng and Lee, 2019a, JGR

Deng and Lee, 2019b, JGR

Deng and Lee, 2019, AM