Silicon has conventionally been the semiconductor of choice for power devices. However, the increase of the on-state resistance with increase of the drift region width, especially in unipolar power devices, has been the main barrier impeding the development of unipolar power devices with high voltage blocking capability. To overcome this and aim for high voltage applications, bipolar power devices were developed. However, these also exhibit a relatively long switching transient, followed by a tail current due to the recombination phase. Hence, to develop a high voltage and fast switching device, the only option available is to move towards the use of unipolar devices employing wider bandgap materials.
Wide bandgap semiconductors come in the form of different elements or compounds of which Silicon Carbide (SiC) and Gallium Nitride (GaN) are among the most developed ones. SiC, which is the dominant commercially-available WBG semiconductor technology, is a compound formed between silicon and carbon with a strong covalent bond and can be formed in different polytypes. Unfortunately most of these polytypes are difficult to
Figure 2.12: (a) Technical requirements and current state of power converters, (b) Comparison of typical properties of silicon and SiC devices; Courtesy of [8].
crystalize in extensive diameters, hence for large wafer diameters, three main polytypes namely 4H-SiC, 6H-SiC and 3C-SiC are chosen, among which 4H-SiC is the dominant due to higher mobility and breakdown field. SiC as the trendiest choice of WBG materials, has many advantages which can be beneficial for the development of power devices. Some of these are shown in Table 2.1 obtained from [9]. As can be seen, SiC has over 8 times higher critical electrical field capability compared with silicon; hence it can deliver a significantly higher blocking voltage for the same width of drift region. This is because of the wider band gap between its bands, hence the electrical field that is required to initiate the impact ionization is significantly higher. SiC can also maintain its characteristics at higher temperatures. This is because firstly, it has a higher melting temperature point, meaning that it can withstand higher temperatures without losing its physical characteristics. More importantly, it maintains its semiconductor properties at high temperatures, since the generation of carriers require significantly higher energy. In silicon, the relatively lower bandgap means that the rate of carrier generation with temperature renders the semiconductor degenerate thereby losing its semiconducting properties.
Table 2.1: Comparison of main material properties of silicon and SiC polytypes [9].
Parameter Unit Silicon 4H-SiC 6H-SiC 3C-SiC Bandgap (Eg) eV 1.12 3.26 3.02 2.36
Electrical Breakdown Field (E) (ND=3×1016 cm−3)
MV cm−3 0.37 2.8 3 1.4 Intrinsic Carriers Concentration cm−3 1.5×1010 5×10-9 1×10-6 0.1 Electron Mobility (µn) cm2 V−1 s−1 1350 1020 450 1000 Hole Mobility (µp) cm2 V−1 s−1 480 120 100 100 Thermal Conductivity (λ) W cm−1 K−1 1.49 3.3-5 3.3-5 3.3-5 Relative Dielectric Constant (εr) (–) 11.7 9.76 9.66 9.72
Also the higher thermal conductivity in SiC means it can remove the heat generated from the junction more quickly, resulting in a lower junction temperature. Table 2.2 of [10] provides an estimated comparison between properties of silicon and SiC devices with the same voltage range. It can be seen that the thickness of drift region of device is significantly lower in SiC while its doping is higher. This simply means a considerable reduction in the on-state resistance and conduction losses compared to the silicon coun- terparts. Implementation of SiC in unipolar devices also means that the fast switching capabilities are maintained thereby switching losses are also low. These advantages as well as a few important challenges of SiC devices will be discussed in subsequent chapters.
It can be seen in Figure 2.12 of [8] that reliability and price are important factors that still require significant improvements. Unfortunately these two factors are significant drawbacks in front of the broad application of newly developed SiC power devices, where the reliability issues hinder performance. The price also is considerably more expensive than silicon devices, although it is reducing rapidly as the utilization of the devices be- comes more widespread. By overcoming these constraints, it is expected that SiC devices will replace silicon power devices in high and medium voltage applications; i.e. in electric vehicle applications where this has already started to take place.
Table 2.2: Doping and thickness of silicon and SiC devices with the same ratings [10].
Breakdown Voltage Silicon Silicon Carbide (4H-SiC) Doping (cm-3) Thickness Doping (cm-3) Thickness 500 V 5× 1014 36 1 × 1017 2.2 1000 V 2× 1014 81 4 × 1016 5.3 2000 V 8× 1013 183 1.6 × 1016 12 3000 V 5× 1013 294 1 × 1016 20 5000 V 2× 1013 530 4.4 × 1015 35 10000 V 9× 1012 1200 1.7 × 1015 80
2.3.1
Latest Devices
The very first SiC devices were unipolar to provide a narrow drift region. Among these, the SiC Schottky Barrier diode was one of the initial devices to be developed, commercial- ized with CREE® and Semisouth® in 2001. Compared to conventional silicon Schottky diodes where the blocking voltage could not exceed 200 volts due to the very high (and unacceptable) on-state resistance, SiC Schottky diodes with 1.2 kV ratings have relatively low on-state resistance. Among the transistors, SiC JFETs were also become commer- cially available in 2005, however due to the normally-on characteristics of these devices, they were not very practical. Development of normally-off SiC JFETs removed these disadvantages in 2008, however by then SiC MOSFETs were the main research topic worldwide. These devices, with good reliability and performance became commercially available by CREE® in 2011 and ROHM® in 2012 which eventually overshadowed the normally-off SiC JFETs. SiC MOSFETs reduced the on-state resistance of silicon power MOSFETs significantly, with the integral body diodes showing a much better switching performance with almost no reverse recovery. This, too, will be investigated in depth later in subsequent chapters. However, SiC MOSFETs have higher internal gate resistance (due to the smaller die size) and require high gate voltage (due to lower transconductance).
SiC bipolar devices, including SiC PiN diodes and SiC IGBTs are also in development in research facilities, although they have not been commercially available yet mainly due to the problems with low carrier lifetime in SiC. Unlike SiC unipolar devices, which aim for applications with medium voltage ratings such as EVs, the SiC bipolar devices aim for extremely high voltages, normally 10 kV and beyond, where the main application is deemed to be in high voltage transmission systems. Devices such as SiC Thyristors are also in development, although their point of application is still vague.