Impurity-related Interface States of 4H-SiC Schottky Diode Observed by Deep Level Transient Spectroscopy

Abstract

In power device technology, 4H-SiC has been expanding its market share among semiconductor materials owing to its high-temperature- and high-voltage-stabilities and high-thermal conductivity. It is expected that the compound annual growth rate of the SiC power semiconductor market and the market size in 2022 would be 40 % and $1 billion, respectively. However, SiC has some problems that are limiting some performances of the SiC devices. Trap level plays important roles in the power devices such as Schottky diode, MOSFET, or IGBT by changing barrier height, subthreshold swing, or carrier lifetime, respectively. Especially, 4H-SiC devices have large amount of trap level at their interface, i.e., interface states, because of an inherent nature of the compound semiconductor. Many researchers have reported that the interface states originate from dangling bonds of interface atoms and they are successfully reduced by nitridation or other deposition techniques. With respect to 4H-SiC Schottky diodes, which are important-rectifying devices, have been well developed and are now commercially available. Even though their good performances, the relations of interface states and the rectifying performances are not completely understood. Therefore, in this dissertation, observing the interface states of 4H-SiC Schottky diode, and finding their origins by changing pre-treatment of 4H-SiC, and the relation of interface states and the rectifying performances are confirmed and discussed. The interface states were observed by Deep level transient spectroscopy (DLTS), which is known as the most sensitive and accurate method to measure the deep levels in semiconductors. Furthermore, some related issues about the dependency of intensity of DLTS signal on pulse width and about the effect of weighting- and trap-parameters on spectral shape of DLTS are also figured out and discussed. Spectral fitting of DLTS is important, because of inherent broad-peaks that is easily overlapped. Although it is important to know the spectral shape for appropriate fitting, there are few of information and data about the effects of shape parameters, like weighting functions and trap properties. In the part 4.1, it is examined that how shape parameters determine the width, position, and asymmetry of spectra. Furthermore, based on this study, spectral fitting is applied to the spectra of the part 4.3, which makes the analysis more precise. The pulse width is one of the important condition for measurement of DLTS. It relates to the time required to fill the electron or hole trap. Despite of its importance, there have been lack of research on the pulse width of 4H-SiC. In the part 4.2, the saturation of trap-fillings, abnormal-negative shift and abnormal-negative peak of the spectra is observed. Furthermore, in order to reveal the abnormalities that are emerged by minority carrier signals, four-different Schottky metals were used to modulate work function and the barrier height of the minority carriers. As the pulse width increases, the positive signal was intensified and saturated by pulse width longer than 1 ms. For the pulse width longer than 103 ms, the signal started to be negatively intensified. When the pulse width reached 107 ms, the negative intensification was saturated. The saturation of the negative signal could relate to the minority carriers introduced by changes of quasi-Fermi level. In the part 4.3, background, peak-2, -3, and -4 of DLTS spectra related to impurity-related interface states of 4H-SiC Schottky diodes. The spectral fitting developed in the part 4.1 was adopted to resolve the overlapped background and peaks. The oxide at the interfaces increased the background and peak-2. Photoresist (PR) residues at the interface increased the background and the peak-3 and -4. The peak-2, -3, and -4 located at 1.28, 1.7, 0.7 eV below the conduction band minimum, respectively. The nitridation increased the background intensity abruptly. The continuous background could imply disorder of defect structure at the interface, and the discrete peaks could associated with short-range order of the defect structure. The relation of interface states and current flow, and the side effect of DLTS measurement on the sample are also discussed.