2026 Theses Doctoral

Towards Multiscale Computational Methods for Investigating the Earth’s Interior

Wang, Hongjin

This thesis develops and applies a multiscale computational framework to investigate the thermodynamic and elastic properties of Earth materials under the extreme pressure temperature conditions of the deep mantle.

By combining first principles calculations, statistical thermodynamics, and machine learning based molecular dynamics, this work overcomes key limitations of standard 𝘢𝘣 𝘪𝘯𝘪𝘵𝘪𝘰 approaches.

Major contributions include the development of the pgm package for calculating anharmonic free energies using temperature dependent phonon frequencies, the demonstration that the r2SCAN meta-GGA functional provides an improved description of hydrous minerals through a detailed study of brucite, and the construction of a machine learning interatomic potential for serpentine minerals that enables large scale simulations and clarifies the polymorphism and stability of antigorite.

These methods are integrated to build a predictive and internally consistent thermodynamic and thermoelastic database for Earth's lower mantle. Together, this work provides a coherent link between microscopic mineral physics and macroscopic geophysical observations, improving interpretations of seismic data and models of Earth's deep interior.