Structural dynamics of hybrid lead halide perovskites with machine learning
Author(s)
Liang, Xia
Type
Thesis
Abstract
In this thesis, I present a unified computational strategy for investigating the structure and dynamics of lead halide perovskites, which remain at the forefront of research for next-generation solar cells and optoelectronic devices. The methods combine on-the-fly force field training connected to machine learning (ML) architectures such as the Gaussian Approximation Potential, Allegro, and MACE, with an automated Python workflow. My code PDynA was built to efficiently quantify perovskite structural descriptors (e.g., octahedral tilting, distortion modes, and molecular orientations). This integrated approach is applied to both simple, pure compositions and more intricate mixed perovskites, where varying A-site cations and halide ratios drive multiple phase transformations and dynamic effects.
Version
Open Access
Date Issued
2025-03-28
Date Awarded
01/06/2025
License URL
Advisor
Walsh, Aron
Publisher Department
Department of Materials
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)