Abstract

The bias temperature instability (BTI) is one of the most serious reliability concerns in state-of-the-art metal-oxide-semiconductor (MOS) field-effect transistors. It is observed when a large voltage is applied to the gate contact of the MOS transistor at elevated temperatures while all other terminals are grounded. The strongest effect is seen when a negative bias voltage is applied to a p-channel MOS transistor, the corresponding degradation is commonly termed negative bias temperature instability (NBTI). Bias-temperature degradation is generally ascribed to electrochemical reactions at point-defects in the oxide and at the semiconductor-oxide interface. Concerning the physical details of the degradation process, however, no general consensus has been found yet and a lively debate has been going on for several years.

For almost four decades, the reaction diffusion model has been the standard interpretation for NBTI. This model assumes the observed degradation to be primarily due to the depassivation of silicon dangling bonds at the Si-SiO2 interface. These dangling bonds originate from the inherent lattice mismatch between the silicon substrate and the oxide layer and are passivated during the manufacturing process using hydrogen. The reaction diffusion model assumes that during NBT stress the hydrogen atoms leave the dangling bonds and diffuse into the oxide and that this diffusion process determines the transient behavior of the degradation.

As the reaction diffusion model is unable to give the experimentally observed recovery behavior, an alternative description has been developed by Grasser and coworkers. This multi-state multi-phonon model ascribes the BT degradation primarily to the capture and emission of holes at point defects inside the oxide. In this model, each point-defect can undergo a charge transition, which is understood as a non-radiative multi-phonon transition, or a structural reconfiguration, which is understood as a barrier hopping process. The multi-state multi-phonon model is able to explain complex BT experiments with very high accuracy and also links BTI to other oxide-defect-related effects. Unfortunately, the model has a large set of parameters that need to be calibrated and the atomic nature of the defect is still unknown.

The present work applies atomistic modeling techniques to both BTI models. The reaction-diffusion mechanism is studied at the stochastic chemistry level. Our results clearly show that the commonly employed mathematical description using rate equations is inappropriate for the reaction-diffusion mechanism. However, the physically more reasonable model developed in the present work fails to give both the experimentally observed degradation and recovery. This suggests that the reaction-diffusion model requires a revision at the microscopic level and that especially the diffusion-limitation of NBTI degradation is not a reasonable assumption.

For the multi-state multi-phonon model it is shown how the number of free parameters can be reduced using an atomistic representation of a point defect and an electronic structure method. For this purpose, the broad literature of multi-phonon transition theory is briefly reviewed. Methods are developed to extract line shape functions, which describe the non-radiative multi-phonon capture and emission of charge carriers, at different levels of physical detail from an atomistic model of a point defect. The oxygen vacancy and the hydrogen bridge in crystalline silicon dioxide are studied for their BTI behavior both as examples and as references for future defect studies.