Abstract

Silicon Carbide (SiC), as a wide band gap semiconductor, promises superior performance for power devices. Compared to Silicon-based devices, MOSFETs based on SiC can be operated at higher temperature, higher frequency and higher power density. Furthermore, they provide significantly reduced static and dynamic losses, which allows for shrinking passive components and heatsinks at the system level. All of these benefits make full SiC system solutions more efficient, lighter, compact and even less expensive than their silicon based counterparts.

Despite all of the benefits mentioned above, 4H-SiC-MOSFETs still perform far from their theoretical limits. Especially their low channel mobility and increased threshold voltage variations, which differ substantially from silicon based devices in certain aspects, need to be understood and assessed. Therefore, the main focus of this thesis is the investigation of bias temperature instabilities (BTI), which arise after temperature and voltage stress of the gate oxide. Unlike in Si based MOSFETs, where the major part of BTI originates from trap states within the SiO₂-film, 2 independent components are identified on SiC-based MOSFETs.

The first component is especially visible as a gate voltage hysteresis in the subthreshold regime of the transfer characteristics. This hysteresis may reach several volts and originates from hole capture in traps states at the 4H-SiC/SiO₂ interface. The density of these states depends significantly on the crystal plane of the inversion channel and is approximately one order of magnitude higher on devices with the inversion channel along the -axis. Furthermore, the observed hysteresis scales with the charge pumping signal and is nearly independent of temperature. Unlike in classic BTI in silicon based devices, the observed hysteresis is fully recoverable in normal device operation within micro seconds via a gate voltage above the threshold voltage and does not impact the long-term device reliability. Carbon dangling bonds are suggested as the most promising defect candidate for these states.

The second component is similar to what is observed on silicon based devices and most likely originates from border trap states in the SiO₂, which are energetically located close to the conduction band of 4H-SiC. In the first section of this thesis, similarities in BTI of commercially available SiC-power MOSFETs are presented with the conclusion that all devices available on the market today show a nearly identical drift behavior. As opposed to silicon based devices, even an operation close to the threshold voltage causes a voltage shift in the range of hundreds of millivolts. The second section focuses on various voltage shift measurement techniques. It will be demonstrated that most of the observed voltage shift in 4H-SiC based devices originates from fully reversible and stress-independent components. A new drift evaluation technique using device preconditioning is proposed, which allows for a more comprehensive and nearly measurement delay-time-independent extraction of the permanent voltage shift component.

In the last part of this thesis, the impact of various high-temperature processing steps on the charge state of the SiC/SiO₂ interface was investigated. It will be shown that a high thermal budget results in an significant accumulation of positive charges at the SiC/SiO₂ interface. The build-up of positive charges starts after the deposition of the polycrystalline gate contact, and continues in all additional processing steps, in which the sample is exposed to a high thermal budget with temperatures above 500 °C. Furthermore, the presence of an electric field in the oxide, which likely enables diffusion of the positive charges to the SiC/SiO₂ interface during the high temperature processing steps, is of fundamental importance for the observed mechanism. The atomic origin of the charge build-up is still unknown but likely linked to hydrogen, which is incorporated during the poly-silicon (Si) deposition. An energy barrier of approximately 1.3 eV was extracted from the experimental data.