Microscopic Modeling of NBTI

Following the International Technology Roadmap for Semiconductors (ITRS), modern metal-oxide-semiconductor (MOS) transistors have been shrunk to scales where even a single charge trapped in their dielectric noticeably alters the transistor parameters, such as the threshold voltage. These charge trapping events are often reversible but can also lead to permanent degradation. The fluctuations in the device behavior resulting from such charge capture processes pose serious reliability questions for device design and hence have aroused a great deal of industrial interest. Also, from the scientific side, this issue has been extensively studied for many decades; nonetheless, a convincing explanation that bridges the gap between recorded data sets and the predictions from physics-based models is often still missing. A particular example where charge trapping in the oxide is crucial is the negative bias temperature instability (NBTI). We have recently suggested a two-stage model (TSM) which can successfully explain a large number of experimental data sets. However, although the model is based on defect properties published in literature, like the cyclic charging behavior of E` centers (oxygen vacanies) and the creation of Pb centers (interface states) via E` centers, the actual association of defect properties in the TSM with the properties of E` and Pb centers prompts many questions which we will try to answer in this project. In the TSM defects can be in one of three states with each of them linked to a certain atomic configuration and charge state. The probability for a transition between these states is governed by their defect levels as well as the separating thermal barriers for atomic movement. During the last couple of years, ab initio calculations have been extensively used to find answers to similar questions. Especially density functional theory calculations with the emergence of hybrid functionals have been established as a well suited tool for the assessment of the required defect properties. In literature, possible defect candidates with suitable models of their atomistic configuration have been proposed and related to results obtained by measurement techniques such as electron spin resonance and optical absorption. So far, however, none of these studies have been able to convincingly link certain defects to the rich experimental data sets which are available for NBTI. In this project we will investigate possible defect configurations which are compatible with the experimentally observed NBTI properties. Using extensive ab initio calculations, the TSM will be refined and its parameters will be either calculated or their possible range narrowed down. The refined TSM model will be implemented into a Schrödinger-Poisson solver for the evaluation against the large body of experimental data available. In order to address state-of-the-art device designs, investigations will be carried out for nitrided oxides, modern high-k dielectrics, and the frequently observed silicon dioxide interlayers. Finally, a simulation tool based on the refined TSM will be made available to a wider public using the distribution channels of the Institute for Microelectronics.

Microelectronic devices have become the foundation to many aspects of our modern lives and take an important role in our technology-driven society. The ongoing advancement of those devices is strongly related to their continuous miniaturization. Due to this, device dimensions have entered scales where even a single charge trapped in the dielectric can severely alter the transistor characteristics. This poses a serious reliability issue, referred to as the bias temperature instability (BTI). Unfortunately, its behavior for different gate biases and temperatures, as well as its physical origins have eluded our understanding so far. However, the recently developed time dependent defect spectroscopy (TDDS) allows for the investigation of single defects, leading to a new microscopic model called four-state NMP model. As its name suggests, this model is based on several different defect states, which differ in their configuration, their stability, and the trapped charge. These defect properties can be extracted from a calibration of this model to experimental data but solely enter as empirical values for a hypothetical defect. As such, the actual defect responsible for BTI remains unknown. Hence, we set out on the search for the atomistic structure of the responsible defect, by employing density functional theory (DFT) simulations.Our studies started from defects located in amorphous silicon dioxide (SiO2) as this is the material in which BTI was discovered initially. The number of defects that had to be investigated in detail could be narrowed down to a few candidates, including the oxygen vacancy, the hydrogen bridge, and the hydroxyl E' center, as only they have a metastable in addition to a stable configuration according to our DFT studies. This was a special feature, which has been deduced from our TDDS studies and used as a stringent criterion for defect identification throughout this work. Another restrictive criterion was that the trap level of the defect must lie close to the valence band edge of the silicon substrate, which is not met by the oxygen vacancy. This was found to hold even if the amorphous nature of SiO2 was considered, which gives rise to large variations in the local atomic bonding structure and thus in the trap level of the defects at different sites. Such statistical studies have also been performed for the hydrogen bridge and the hydroxyl E' center, demonstrating that both of them are still good candidates for a BTI defect.Recent long-time TDDS studies have revealed that defects involved in BTI occasionally disappear and reappear in the neutral or positive charge state. This phenomenon termed defect volatility usually occurs on large time scales and is thus related to the device lifetime of transistors, an important aspect for device reliability. These TDDS findings required an extension of the four-state NMP model with additional states. Using DFT simulations, these new states could only be identified for the hydroxyl E' center. As such, the hydroxyl E' center was the only one meeting the TDDS criteria and is therefore regarded as the sought as the most likely defect responsible for the BTI phenomenon, the identification of which being the primary goal of this project.