Exploring phonon anharmonicity in harmonic materials with machine-learning based force-fields

While the concept of phonons is defined within the harmonic approximation, many experimentally observable spectral and transport properties involving phonons are driven by higher order anharmonic effects, even if the material itself would be considered highly harmonic. Taking such anharmonic effects into account however, can be challenging for first-principles methods, in particular for cases where the commonly used, phonon-based perturbation theories break down or cannot easily be extended beyond the harmonic approximation. Since all the required information about the lattice dynamics, vibrational properties and thermal conductivity is encoded in the trajectory of the atoms, these quantities can alternatively be obtained from molecular dynamics (MD) using machine-learning force-fields (MLFFs). While sharing the same advantageswith the state-of-the-art ab initio MD, these methods require only a fraction of the computational cost.

By utilizing our dynamically trained MLFFs, in this work we are going to explore the role of anharmonicity in determining the thermal- and phonon properties of harmonic materials i.e., c-BAs and Si. The choice of our force-field, which is based on the well-known lattice dynamics expansion of the total energy, will allow us to directly obtain the higher-order force-constant tensors required for an in-depth analysis.