Ab initio studies based on density functional theory (DFT) of Earth forming phases have become pervasive in mineral physics. This grant supported the development and application of novel materials simulation approaches to investigate structural, vibrational, thermodynamic, thermoelastic, and heat conductive properties of lower mantle phases and beyond. It also supported the exploration of consequences of these properties for mantle convection. Novel geodynamic simulations methods and applications taking advantage of ab initio results have also been explored.

We have documented the successful performance of the "quasiparticle" method, a fully anharmonic method to compute temperature dependent vibrational properties, lattice thermal conductivity (K_L), and near melting phase transitions. Particularly relevant was the calculation of K_L for MgSiO₃ perovskite, bridgmanite, by this fully anharmonic ab initio method and by perturbation theory (PrTh). We were able to address controversial experimental data in pure Mg-bridgmanite. We concluded that sample size, an effect we could simulate computationally, critically affects the measured values for micron-size samples. This effect is caused by the limited propagation of acoustic phonons in small samples, which are responsible for the majority of heat transport in periodic crystals. The quasiparticle method also proved capable of predicting a pre-melting hcp-bcc phase transition in beryllium. Therefore, this method will be able to address the possibility of this phenomenon at the extreme conditions of terrestrial exoplanetary cores as well. In summary, we developed great confidence in the performance of this method and a code to compute quasiparticle properties is being released.

For over a decade we have been investigating potential post-post-perovskite phase transitions in MgSiO₃. These studies have culminated in the prediction of a series of dissociation and recombination reactions in aggregates of MgSiO₃, SiO₂, and MgO with variable Mg/Si ratios under pressure. We concluded that the sequence of phase transitions depends on the Mg/Si ratio and that they all terminate in the full dissociation of these aggregates into pure oxides, SiO₂ and MgO, at ~ 3 TPa at low temperatures. The thermodynamic properties of all phases involved were used to model convection in super-Earth and heavier terrestrial planets with masses up to 20 M_Earth. Penetrative convection was demonstrated in a planet similar to super-Earth GJ 876 d, with ~ 7 M_Earth using 2D-axisymmetric and 3D-spherical compressible models. Another 2D-cartesian study using the full sequence of phases, their ab initio thermodynamic properties, and ab initio based rheology demonstrated that the internal dynamics of these planets is mass dependent. The number of internal layers depends on the mass because the number of phase transitions increases up to ~3 TPa. Contrary to expectations, the full dissociation into oxides can occur in planets with masses > 13 M_Earth.  Three different (Raleigh-Benard) convective regimes were identified in planets with up to 20 M_Earth: 1) smaller planets (M < 4 M_Earth), showing vigorous convection, 2) intermediate cases (4 M_Earth < M < 12 M_Earth), with sluggish penetrative convection concentrated in a single shallow mantle zone with higher flow velocity, and 3) large planets, (M > 12 M_Earth), showing vigorous convection in two zones near the top and bottom, separated by a high viscosity mid-mantle region with sluggish convection. These are the most advanced simulations of the internal structure of terrestrial exoplanets performed to date.

On the geodynamics front, the density anomalies caused by the spin transition in ferropericlase and possibly in Fe⁺³-bearing bridgmanite have been used to investigate this effect in mantle convection. A viscosity model based on the elastic strain energy method was developed to include also the effect of pressure and temperature dependent viscosity variations in a 2D axy-symmetric convection model. The major effects observed were: 1) the stagnation of cold subducted slabs in the deep mantle caused by the sharp increase in the bulk modulus and viscosity after the midpoint of the spin transition in bridgmanite, i.e., at depth > ~1,600 km depths; 2) a similar effect on the upwelling of plumes producing large plume heads with over 1,500 km in diameter and episodic upward penetration into the upper mantle. This flow mode delivers large amounts of heat to the base of the lithosphere with implications for large volcanic events. These are probably the most complex simulations investigating the effect of the spin transition in lower mantle phases on mantle dynamics.

Finally, novel methods in geodynamics simulations have been proposed: 1) mantle convection based on the lattice Boltzmann method, and 2) application of machine learning algorithms to trace back flow patterns produced by simulations to the physical property parameters that produced them. Both methods hold great promise for the future of this field.

Last Modified: 12/18/2018
Modified by: Andre Mkhoyan