This thesis covers the two important topics in the field of high-voltage and smart power devices, which are the simulation of devices of this class and the modeling of reliability issues. Based on this foundation, failure and degradation behavior related to the two phenomena, impact-ionization generation and hot-carrier degradation, were presented and integrated into a TCAD device simulator.
For the evaluation of impact-ionization, the snap-back behavior in a smart power structure was investigated. The obtained results demonstrated good agreements to measurement data. The relatively large devices permit to use the drift-diffusion model. Convergence problems during this simulation have been carefully analyzed in the context of vector discretization schemes. It was demonstrated that the different schemes have only little influence on the convergency behavior. However, differences in terms of accuracy could be shown.
For the hot-carrier degradation modeling, a physics-based model is presented, which relies on the thoroughly evaluated carrier energy distribution function obtained using the Monte Carlo method. Since the model is physics-based, it allows one to captures the main peculiarities of hot-carrier degradation as, e.g., an interplay between a single- and a multiple-particle process of the Si-H bond dissociation. However, more efficient approaches which are less computationally demanding than the Monte Carlo method are required to adapt the model for industrial needs. Therefore, a model based on the drift-diffusion framework was developed. It is demonstrated that the approach delivers a good agreement in comparison to the Monte Carlo results. However, the limitations of the drift-diffusion model have already been reached. First, a row of newly introduced fitting parameters requires a specific calibration for each particular device technology. Second, the interface damage caused by secondarily generated holes cannot be captured using the drift-diffusion model. Therefore, the degradation in devices with a channel length above 0.5 µm, cannot be modeled.
Future work is therefore related to development of methods for an efficient calculation of the carrier energy distribution function. The most promising candidate for this is the deterministic solution of the Boltzmann transport equation based on the spherical harmonics expansion. First results obtained within this technique with the presented hot-carrier degradation model have already been accepted for publication. Additionally, the degradation model needs some extensions. The probably most important missing feature seems to be the consideration of oxide traps.