Atomistic modelling of diffusion mechanisms of carbon defects at the SiO2/4H-SiC interface

Abstract

4H-SiC is an attractive material for high-power, high-temperature electronics because it has a wide band-gap, favourable thermal-conductivity and the beneficial SiO2 native oxide. However, SiC device characteristics are degraded by electrically-active defects with electronic states located deep within the band-gap arising from native defects in the SiC and close to the SiO2 interfaces. In this thesis are presented the results of a computational study to understand the dynamics of selected native defects, with reference to the SiO2/4H-SiC interface. Densityfunctional simulations have been used to explore defect structures, electrical properties and diffusion energetics. The modelling exploits the computational advantages of periodic boundary conditions to represent both bulk and interface cases. Different interfaces corresponding to the internal structure of the SiC have been examined, and many individual diffusion steps have been examined to explore the impact of the system models and role of depth into the SiC for key processes. It is determined that carbon-vacancies (VC), which are known to be key carrier traps, have their diffusion hindered in the vicinity of a SiO2/(0001)-4H-SiC interface, with the overall activation energy 15% higher in the immediate interface than two-to-three bilayers into the SiC where they behave as in bulk SiC. It is also thought that oxidation of SiC introduces carbon interstitials to device channels, and the diffusion Ci in the vicinity of the interface with SiO2 has also been simulated. It is found that the interface stabilises Ci even more than VC, lower its energy by 1 eV relative to bulk. Such stabilisation is expected to inhibit the injection of self-interstitials into SiC. The calculated hindering of diffusion of these native defects is consistent with a relatively high density of traps in this region, in line with experimental findings, but effective approaches to remove vacancy-related traps remains a challenging problem to be solved.