Thermoelastic properties of bridgmanite using Deep Potential Molecular Dynamics

The Earth's lower mantle is dominated by the high-pressure Pbnm-perovskite polymorph of MgSiO₃, known as bridgmanite (Bm). The elucidation of Bm's structural and elastic properties is of significant geophysical importance, yet the extreme conditions within the mantle pose challenges for experimental investigations. To overcome this hurdle, ab initio-based methods have emerged as indispensable tools. In this study, we present deep neural network potential models for Bm, developed using density functional theory (DFT) with various functionals. These models enable extensive molecular dynamics (MD) simulations, comprehensively exploring a wide range of pressure-temperature (P-T) conditions. Our research focuses on the compressional behavior and elastic moduli of Bm at high P-T, shedding light on the remarkable performance of the DP-SCAN functionals in accurately predicting high-temperature equations of state and elastic properties. By merging advanced computational techniques with DFT, we offer a robust solution to the accuracy-efficiency dilemma, facilitating precise investigations of elastic properties for minerals like MgPv at high P-T conditions. Our findings signify a significant leap forward in understanding the Earth's internal state and processes, providing an innovative avenue for further exploration.