The 6th International Conference on Matter and Radiation at Extremes (ICMRE2023)

113 / 2023-04-13 22:32:58

Thermal Conductivity of MgSiO3 under Lower Mantle Conditions Calculated by Machine Learning Potential

Thermal conductivity,Lower mantle,MgSiO3,Machine learning potential

High Pressure Physics and Materials Science > High-pressure geoscience-II

Thermal conduction of mantle minerals determines the magnitude of heat flux across the core-mantle boundary and is related to the thermal dynamics and evolution of the core and mantle, the formation and stability of the mantle plumes, and the generation of the magnetic field [1-3]. Therefore, the thermal conductivity κ of MgSiO₃ (the most abundant mineral in the Earth’s lower mantle) is an important parameter in determining the heat budget of the Earth. Experimental measurement of κ under extreme conditions is still controversial and challenging [4,5]. Theoretical calculation based on the phonon gas model (PGM), in which a phonon is considered a quasiparticle, may underestimate the κ due to the failure of the phonon quasiparticle picture. In low-κ materials, a number of phonons are referred to as ill-defined phonons, for example, whose mean free paths are shorter than the Ioffe–Regel limit [6]. In this work [7], to deal with the ill-defined phonons, molecular dynamics (MD) simulations are used to obtain the κ under lower mantle conditions. We train a machine learning deep potential (DP) model [8] based on density functional theory datasets to accurately capture the interatomic interactions over a wide P-T range. Compared with previous PGM results [9], the κ predicted by DPMD is larger at low P and high T and closer to the experimental value [4]. The discrepancy is attributed that the contribution of ill-defined phonons is underestimated in the PGM but included in DPMD simulations. A larger κ of MgSiO₃ predicted by DPMD would imply a thicker boundary layer and more stable mantle plumes. Our results emphasize the significance of ill-defined phonons and would provide thermophysical data for evaluating the thermal dynamics and improving the Earth model.