Thermal Conductivity of MgSiO₃‐H₂O System Determined by Machine Learning Potentials

Abstract

Thermal conductivity plays a pivotal role in understanding the dynamics and evolution of Earth's interior. The Earth's lower mantle is dominated by MgSiO3 polymorphs which may incorporate trace amounts of water. However, the thermal conductivity of MgSiO3‐H2O binary system remains poorly understood. Here, we calculate the thermal conductivity of water‐free and water‐bearing bridgmanite, post‐perovskite, and MgSiO3 melt, using a combination of Green‐Kubo method with molecular dynamics simulations based on a machine learning potential of ab initio quality. The thermal conductivities of water‐free bridgmanite and post‐perovskite overall agree well with previous theoretical and experimental studies. The presence of water mildly reduces the thermal conductivity of the host minerals, significantly weakens the temperature dependence of the thermal conductivity, and reduces the thermal anisotropy of post‐perovskite. Overall, water reduces the thermal conductivity difference between bridgmanite and post‐perovskite, and thus may attenuate lateral heterogeneities of the core‐mantle boundary heat flux.

Hydrogen Diffusion in the Lower Mantle Revealed by Machine Learning Potentials

“…We note that our training set includes a significant number of frames that sample various defects and interfaces, thanks to the multithermal-multibaric technique used (Deng et al, 2023a;Piaggi & Parrinello, 2019). As such, though not the focus of this study, our MLP can also be useful for studying the phase transition of the (hydrous) MgSiO 3 system, as well as other transport properties including thermal conductivity (Peng & Deng, 2024).…”