Modelling of Defects in Semiconductors

Doctoral thesis, 2024

Defects, minute imperfections within crystals, greatly affect the properties of many materials. Understanding their influence is crucial for enabling innovation in the field of materials science. Computer modelling allows for the study of isolated defects, providing us with an important tool in this endeavour.

A substantial portion of this thesis focuses on defects in bismuth vanadate, a promising semiconductor material for photoelectrochemical oxygen evolution. We review the fundamentals of water splitting as well as the theoretical foundations of the applied computational methods. The native defects of the material are investigated with a particular emphasis on the oxygen vacancy. Our results highlight the structural complexity of these defects and the important role of charge localization in the defect chemistry of the material. Additionally, we show that oxygen vacancies distort the host lattice and make phase identification through X-ray diffraction challenging. Furthermore, we find that these vacancies reduce the overpotential required for oxygen evolution if present at the surface of the material and that cobalt doping influences their stability.

The thesis also covers the application of machine learning interatomic potentials to the study of defects. The fundamentals of machine learning, with a particular emphasis on neural networks, are reviewed. We develop a new approach to machine learning interatomic potential design, enabling the construction of such potentials capable of simultaneously describing defects in multiple excitation states, and apply it to a divacancy defect in silicon carbide. This defect can act as a single-photon emitter, making the optical properties of the defect itself highly interesting. We conduct large-scale machine learning-accelerated molecular dynamics simulations and extract the emission lineshape of the defect from the generated trajectories.