Session: 17-01-01: Research Posters

Paper Number: 145660

145660 - Thermal Conductivity of Perovskite and Post Perovskite Mgsio3 Using Machine Learning Interatomic Potential 

The Earth's lower mantle (LM), constituting a substantial portion of the planet's volume, is a critical domain for understanding the Earth's thermal evolution. Within this region, MgSiO₃ perovskite (MgPv), commonly referred to as bridgmanite, represents a dominant mineral phase, transitioning to MgSiO3 post-perovskite (MgPPv). Understanding the thermal transport of MgPv and MgPPv at LM conditions is crucial to understand the thermal transport inside the earth mantle. Despite its fundamental importance, the thermal conductivity (κ) of MgSiO₃ at ultra-high conditions remains unclear, leaving strong controversy about Earth's internal dynamics and its evolution. In this study, we investigated the thermal transport mechanisms operative at ultra-high temperatures and pressures within the LM using first-principles calculations. Leveraging moment tensor potential (MTP)-based Machine Learning Interatomic Potentials (MLIPs), trained meticulously using snapshots from Ab Initio Molecular Dynamics (AIMD), we achieve accuracy comparable with traditional Density Functional Theory (DFT) calculations. Subsequently, employing Green-Kubo molecular dynamics (GKMD) simulations, we accurately predict the κ of MgPv and MgPPv across a broad range of pressures and temperatures. The temperature and pressure used in our study encompass all the range of LM conditions. Our findings reveal that the total thermal conductivity at high pressure and temperatures arises from both phonon and diffuson contributions, with diffusons gaining prominence at high temperatures and low pressures. At 25 GPa, κ exhibits a temperature dependence of κ~T^-0.8, deviating from the standard Boltzmann Transport Equation (BTE) prediction or standard phonon thermal conductivity (κ ~T^-1), indicative of significant diffuson contribution at high temperatures. However, at elevated pressures, diffuson significance diminishes, aligning κ scaling closer to the BTE prediction. At low temperatures, where phonons prevail, κ increases linearly with pressure, while at high temperatures, κ deviates from linear temperature dependence due to the counteracting effects of phonons and diffusons. This intricate interplay results in a non-linear κ-temperature relationship, flattening at high temperatures. In summary, this study elucidates the governing role of phonon-diffuson interplay in ultra-high temperature and pressure thermal transport and provides accurate prediction of κ at LM conditions facilitating accurate modeling of Earth's mantle thermal dynamics.

Presenting Author: Janak Tiwari University of Utah

Presenting Author Biography: Janak Tiwari as a Ph.D. Student at the University of Utah. He works on first principles, machine learning, and finite element-based thermal transport simulations.

Authors:

Janak Tiwari University of Utah
Tianli Feng University of Utah