Commercialized in the early 1980s, this revolutionary power transistor has had a tremendous positive impact on the power electronics industry, enabling innovative converter designs, improved system efficiency, and global energy savings. In fact, estimates suggest that IGBTs have helped prevent 75 trillion pounds of CO2 emissions over the past 25 years.
Just as the IGBT technology revolutionized the world of power electronics in the 1980s, today's wideband semiconductor silicon carbide (SiC) is increasingly showing promise for revolutionizing the world of power electronics once again. IGBTs have brought us transistors that can switch off at low on-state (i.e., low on-resistance) losses and well-controlled high-voltage switching. However, these devices are limited in their switching speed, which leads to high switching losses, large and expensive thermal management, and an upper limit on the efficiency of power conversion systems.
The emergence of SiC transistors has virtually eliminated the switching losses of IGBTs under similar conditions of conduction losses (in fact, even lower under light load) and voltage latch-up capability. In addition to reducing the overall weight and size of the system, it has also brought unprecedented efficiency improvements.
However, like most disruptive technologies, the development of commercial SiC power devices has experienced a period of turbulence. This article aims to explain the background and evolution of SiC MOSFET technology, providing a brief history of its advancements and showcasing its current technological advantages and future commercial prospects.
Early silicon carbide
Although research on SiC materials related to devices has been ongoing since the 1970s, the use of SiC in power devices was likely formally proposed by Baliga in 1989. Baliga's quality factor provided aspiring materials and device scientists with additional impetus to continue advancing SiC crystal development and device processing techniques.
In the late 1980s, research institutions around the world made tremendous efforts to improve the quality of SiC substrates and hexagonal silicon carbide epitaxy. These included Kyoto University and the Industrial Technology Institute in Japan, the Iof Institute in Russia, Erlangen and Linköping Universities in Europe, Stony Brook University, Carnegie Mellon University, and Purdue University in the United States. Technological improvements continued for most of the 1990s until Infineon launched the first commercially available device in 2001 in the form of a silicon carbide Schottky diode.
In the years following their product launch, silicon carbide Schottky diodes experienced field failures stemming from material quality and device architecture. Rapid and significant progress was made in improving substrate and epitaxial quality; simultaneously, a diode architecture known as a "barrier Schottky junction (JBS)" was adopted, which could more effectively distribute the peak electric field.
In 2006, the JBS diode evolved into what is now known as the merged pn Schottky (MPS) structure. This structure maintains the optimal field distribution but also achieves enhanced buffering capability by incorporating true minority carrier injection. Today, silicon carbide diodes are so reliable that they exhibit a more favorable FIT rate than silicon power diodes.
MOSFET replacement devices
The first silicon carbide power transistor to hit the market appeared in 2008 as a 1200-volt junction field-effect transistor (JFET). SemiSouth Laboratories followed the JFET approach because at the time, there were obstacles to bipolar junction transistors (BJTs) and MOSFET alternatives that were considered insurmountable.
While BJTs offer impressive current per active region data, these devices have three major drawbacks:
Firstly, the high current required for switching BJT devices is opposed by many designers accustomed to using voltage-controlled devices such as MOSFETs or IGBTs ;
Secondly, the drive current of a BJT is conducted on a base-emitter junction with a huge built-in potential, resulting in huge power loss;
Third, due to the bipolar operation of BJTs, they are particularly susceptible to a device wear phenomenon known as bipolar degradation.
Figure 1: (a) Positive electrode, VGS = +25V, and (b) Negative electrode, VGS = -10V, 2300 hours of high-temperature gate bias (HTGB) stress test at 175°C on 77 devices drawn from three different wafer batches. Negligible deviations were observed at the threshold.
On the other hand, the application of JFETs is hampered by the fact that they are normally-on devices, which scares away many power electronics designers and safety engineers. Designs can certainly be made around this aspect, but simplicity and elegant design are underrated virtues in the engineering world. SemiSouth also has a normally-off JFET device, but it has proven difficult to mass-produce.
Today, USCi offers a normally open SiC JFET device packaged with a low-voltage silicon MOSFET in a cascode configuration, providing a sophisticated solution for many applications. Nevertheless, due to the control similarities between MOSFETs and silicon IGBTs , but with the aforementioned advantages in performance and system efficiency, MOSFETs have remained the 'holy grail' of silicon carbide power devices.
Evolution of Silicon Carbide MOSFETs
SiC MOSFETs have some problems, most of which are directly related to the gate oxide layer. In 1978, researchers at Colorado State University measured a chaotic transition region between pure SiC and grown SiO2, the first observed sign of trouble. Such transition regions were thought to have a high density of interface states and oxide traps that inhibit carrier movement and lead to instability in the threshold voltage; this has since been confirmed by numerous research publications.
In the late 1980s and 1990s, many people in the field of SiC research conducted further studies on the properties of various interface states in the SiC-SiO2 system.
Research in the late 1990s and early 2000s significantly improved our understanding of the sources of interfacial states (density, abbreviated as Dit) and how to reduce these sources and mitigate their negative effects. Among the notable findings, studies observed that oxidation in humid environments (i.e., using water as the oxidant instead of dry oxygen) reduced Dit by two to three orders of magnitude.
Furthermore, the study found that using an off-axis substrate reduced Dit by at least an order of magnitude. Finally, and very importantly, the effect of post-oxidative annealing in carbon monoxide (a method commonly known as nitriding), first discovered by Li and colleagues in 1997, can reduce Dit to very low levels. This finding was subsequently confirmed by six or seven other groups, and a paper by Pantelides provides an excellent summary of this series of research.
Of course, it would be a gross oversight not to emphasize the significant contributions made by the single crystal growth and wafer research community. Previously, we only had pure Leybold wafers, but they brought us 150 mm wafers with microtubes that have almost no equipment damage.
Research progress on SiC MOSFETs slowed in the following years as promising suppliers focused on commercialization. However, preparations for final improvements were made to further enhance clamp voltage stability, process enhancements, and screening to ensure reliable gate oxide and device qualification. Essentially, the SiC research community is getting closer to discovering the holy grail.
MOSFET quality today
Over the past two years, commercially available 1200V SiC MOSFETs have come a long way in terms of quality. Channel mobility has improved to an appropriate level; oxide lifetime in most mainstream industrial designs has reached acceptable levels; and threshold voltage has become increasingly stable.
Equally important from a business perspective is that several suppliers have already reached these milestones, the significance of which will be discussed in the next section. Here, we will demonstrate the quality requirements for SiCMOSFETs today, including long-term reliability, parameter stability, and device durability.
Using accelerated time-dependent dielectric breakdown (TDDB) technology, NIST researchers predict that Monolith Semiconductor's MOS technology will have an oxide lifetime of over 100 years, even at junction temperatures above 200 degrees Celsius.
The NIST research used lifetime acceleration factors with an applied electric field (greater than 9 MV/cm) and junction temperature (up to 300°C) on the oxide; for reference, the oxide electric field in practical applications is about 4 MV/cm (equivalent to VGS=20V), and the junction temperature in operation is typically below 175 degrees Celsius.
It is noteworthy that while temperature-dependent acceleration factors are common in silicon MOS, NIST had not observed this in SiCMOS prior to the study using devices from Monolith Semiconductor. Threshold voltage stability was also convincingly demonstrated, as shown in Figure 1. High-Temperature Gate Bias Testing (HTGB) was performed at a junction temperature of 175°C and below negative (VGS = -10V) and positive (VGS = 25V) gate voltages. Seventy-seven devices from three different wafer batches were tested according to JEDEC standards, and no significant changes were observed.
Another parameter demonstrating long-term stability is the MOSFET's blocking voltage and off-state leakage current. Figure 2 shows the high-temperature reverse bias (HTRB) test data.
Under conditions of VDS=960V and Tj=175C, more than eighty samples were subjected to 1000 hours of stress, and post-stress measurements showed no change in drain leakage current and blocking voltage. Regarding device durability, preliminary measurements shown in Figures 3 and 4 indicate a short-circuit withstand time of at least 5 microseconds and an avalanche energy of 1 joule.
Figure 2: High-temperature reverse bias test data of 82 samples after 1000 hours of stress under conditions of VDS=960V and Tj=175°C, showing that there is no change in drain leakage at (a) VDS=1200V and (b) blocking voltage at ID=250μA.
While we cannot verify the long-term reliability or durability of other manufacturers' products, we can say that, based on our evaluation of commercially available SiC MOSFETs, there appear to be multiple suppliers capable of providing SiC MOSFETs in production volumes. These devices appear to offer acceptable reliability and parameter stability, which will undoubtedly incentivize mainstream commercial applications.
Figure 3: Short-circuit test of 1200V, 80mΩ SiCMOSFET under 600V DC link and VGS=20V conditions, showing that the withstand time is at least 5μs.
Figure 4: Avalanche durability test of 1200V, 80mΩ SiCMOSFET shows that the device with Ipeak=12.6A and L=20mH can safely absorb 1.4 J of energy.
Business prospects
In addition to improvements in quality, commercialization has made significant progress in recent years. This has not only created a competitive landscape favorable to both suppliers and users, but also alleviated concerns about having a second supplier for SiCMOSFETs. As mentioned earlier, given the long evolution of devices, the fact that multiple SiCMOSFET suppliers possess sufficiently reliable devices is a major step forward.
Figure 5, from Yole Développement’s “Power SiC 2016” report, reproduced with permission, shows the SiC MOSFET activity of various suppliers as of July 2016.
Wolfspeed, ROHM, STMicroelectronics, and Microsemi have all launched commercially available components; products from Littelfuse and Infineon will soon be available in the industry. Multichip power modules are also a hot topic among customers and suppliers in the SiC field.
Figure 6, also adapted from Yole's Développement 2016 report, shows the status of SiC module development activity. We believe there are still significant opportunities for discrete-packaged SiC MOSFETs, as optimal layout practices for control and power circuitry can easily extend the applicability of discrete solutions to tens of kilowatts. Higher power levels and the motivation to simplify system design will drive SiC module development, but the importance of optimizing parasitic inductance from the package, control circuitry, and surrounding power components cannot be overstated.
When discussing the commercial prospects of SiCMOSFETs, the last unavoidable issue is price. Our view on price erosion is favorable, primarily due to two aspects of our approach: firstly, our devices are manufactured in an automotive-grade silicon CMOS fab; secondly, this process utilizes 150mm wafers. We explain this in more detail in another research work; however, it can be simply stated that the core advantages of leveraging existing silicon CMOS fabs are the lack of capital expenditure and optimized operating costs (both of which are passed on to the end customer).
Furthermore, manufacturing devices using 150mm wafers yields twice as many as using 100mm wafers, significantly impacting the cost per die. Figure 7 provides some indication of pricing based on a survey of commercially available SiC MOSFETs conducted by Digi-Key.
For example, since its initial announcement at Digi-Key six years ago, the price of a 1200V, 80mΩ device in a TO-247 package has dropped by more than 80 percent, even though SiC MOSFETs are still two to three times more expensive than similar silicon IGBTs. At today's price levels, designers have seen significant system-level price benefits from using SiC MOSFETs compared to silicon IGBTs, and we expect the price of SiC MOSFETs to continue to decline as economies of scale are achieved with 150mm wafers.
Figure 5: Status of SiCMOSFET development activities from different suppliers.
Figure 6: Status of SiC power module development activities. Blue circles represent modules with only SiC devices, while orange circles represent modules using silicon transistors and SiC diodes.
Figure 7: Price survey of commercially available SiC MOSFETs seen at Digi-Key.
in conclusion
In the 1980s, silicon IGBTs had a tremendous positive impact on the power electronics industry, and have remained a mainstay of the sector ever since. The next revolutionary technology will be SiC MOSFETs. The current state of SiC MOSFET development points to solutions to key commercial hurdles, including price, reliability, durability, and supplier diversification.
Despite its higher price premium compared to silicon IGBTs, SiC MOSFETs have achieved success due to system-level benefits that offset costs; as material costs decline, the market share of this technology will increase significantly in the coming years. After more than 40 years of development, SiC MOSFETs finally appear poised for widespread commercial success and to play a vital role in the green energy movement.
