With the scaling of semiconductor devices down to the nanometer regime, the probability of functional failure due to single point defects or parameter fluctuations steadily increases. In this respect the simulation of semiconductor devices provides an important ingredient for the optimization of device designs and for the assessment of device lifetimes before production. In this context, this work investigates the influence of the charge carrier transport model on the accuracy of bias temperature instability and hot-carrier degradation models in MOS devices.
For this purpose, a four-state defect model based on a non-radiative multi phonon (NMP) theory is implemented to study the bias temperature instability. However, the doping concentrations typically used in nano-scale devices correspond to only a small number of dopants in the channel, leading to fluctuations of the electrostatic potential. Thus, the granularity of the doping cannot be ignored in these devices. To study the bias temperature instability in the presence of fluctuations of the electrostatic potential, the advanced drift diffusion device simulator Minimos-NT is employed. In a first effort to understand the bias temperature instability in p-channel MOSFETs at elevated temperatures, data from direct-current-current-voltage measurements is successfully reproduced using a four-state defect model. Differences between the four-state defect model and the commonly employed trapping model from Shockley, Read and Hall (SRH) have been investigated showing that the SRH model is incapable of reproducing the measurement data. This is in good agreement with the literature, where it has been extensively shown that a model based on SRH theory cannot reproduce the characteristic time constants found in BTI recovery traces. Upon inspection of recorded recovery traces after bias temperature stress in n-channel MOSFETs it is found that the gate current is strongly correlated with the drain current (recovery trace). Using a random discrete dopant model and non-equilibrium greens functions it is shown that direct tunnelling cannot explain the magnitude of the gate current reduction. Instead it is found that trap-assisted tunnelling, modelled using NMP theory, is the cause of this correlation. This shows that an NMP-based theory of the bias temperature instability can both explain characteristic time constants experimentally found in the drain and the gate current after bias temperature stress as well as the overall threshold voltage shift. These findings imply that for an accurate lifetime prediction an NMP-based theory is a good choice. However, in order to obtain an accurate lifetime prediction information on the threshold voltage shift caused by a single discrete trap created during bias temperature stress needs to be investigated. To this end small area MOSFETs have been investigated on a statistical basis using random discrete doping in order to determine the cumulative distribution function (CFD) of threshold voltage shifts caused by random discrete charged traps as well as their characteristic capture and emission times. It is found that the experimentally observed CFDs of the threshold voltage shifts caused by single charged traps cannot be reproduced using Minimos-NT by considering potential fluctuations alone. Thus further investigations into this subject are needed.
Since the study of hot-carrier degradation requires exact information on the energy distribution of charge carriers, a solution of the Boltzmann transport equation is necessary. For detailed investigations into hot-carrier degradation, ViennaSHE, a device simulator based on a spherical harmonics expansion (SHE) of the Boltzmann transport equation, has been extended in the course of this thesis. To compare SHE to moment-based transport models, quantum correction models, variability caused by random discrete dopants, the classical SRH trapping theory as well as a four state degradation model based on non-radiative multi-phonon theory are incorporated into the simulator. These additions to ViennaSHE allow to evaluate the device characteristics of virgin as well as degraded devices under hot-carrier or bias temperature stress or both. Additionally, ViennaSHE is extended by the extended Vecchi full-band model in order to accurately model the charge carrier transport in the presence of high electric fields. For the simulation of hot-carrier degradation in MOSFETs, a new hot-carrier model is developed and implemented into ViennaSHE. This hot-carrier model is successfully validated for multiple stress conditions against measurement using a unique set of model parameters. In the discussion of the new model the importance of the various ingredients for hot-carrier modelling are investigated and discussed. Additionally, it is shown that electron-electron scattering is paramount for a successful reproduction of the measurement data for short-channel devices. In this context it is also found that electron-electron scattering may only be neglected in long-channel devices. These results contradict recent findings in the literature, where it was suggested that electron-electron scattering in the context of hot-carrier degradation can be neglected.