The burgeoning SiC market

Q After many years of waiting, some SiC chipmakers are now making transistors. Who is leading the charge? a Cree is definitely number one in terms of the MOSFET. They are the first company to have developed and commercialised the SiC MOSFET, launching a 1.2 kV, 20 Amp device that will deliver 17 Amps at 100 °C. It’s a discrete design that was also developed for a US company Powerex. They made a field-effect module with a SiC MOSFET, which has the discrete as a die, not as a packaged chip. Who else? If we stay with the MOSFET, the most advanced technology besides Cree is probably that of Mitsubishi Electric. It is making this device for internal purposes only, probably for motor control or inverters for AC systems. Fuji Electric is developing MOSFET devices and involved in a product with AIST in Japan. It is supposed to open a SiC line that will be an open fab. Fuji’s is an interesting MOSFET technology, but it’s not fully developed yet. Northrop Grumman was developing a SiC MOSFET too. The company is very defence orientated and issues very little information. And Rohm and Toshiba are also active in this area. Q What about other types of SiC transistor? a SemiSouth introduced the SiC JFET in the market place three year’s ago in very limited quantities. They were the first company with normally-on and normally-off JFETs. One of the key advantages of these transistors is that they are much easier to manufacture – and probably less expensive as well – because there is no oxide and the geometry is simpler. Yields depend on the micropipe density. So they depend on what type of substrate you start with, and how big your device is. With a MOSFET, die are quite big, so the chances of being on top of a micropipe can be huge. To increase the yield, you have to start with zero micropipe wafers that are much more expensive. You have some of these problems with JFETs, but the current density is much higher. So for the same current range you have much smaller die and less risk of being affected by micropipes. Q That is a big advantage. So why are some manufacturers pursuing the MOFSET? a The MOSFET is a discrete device, offering chip-tochip replacement with an IGBT or silicon MOSFET. The JFET has to be driven a bit differently. But if you can operate it in normally-off mode, the industry is not reluctant to use it. Q Does SemiSouth face competition in the JFET market? a Infineon is trying to enter this business. One of the key added values of its JFET is that they are pushing to it higher current densities to reduce the die size and the cost. Today, 50 percent of the cost is related to the substrate and epi, so if you can save space and size, you can save a lot of cost. As far as we understand, Infineon is looking at normally-on JFETs cascode-mounted, using silicon transistors to turn a normally-on JFET into a normally-off JFET. For high temperature applications, using silicon is probably an issue. Northrop Grumman is also involved in JFET development. Q In addition to the MOSFET and the JFET, some other types of device are under commercial development, aren’t they? a Yes, there is the BJT transistor. Leading its development is TranSiC, the Swedish company; and Shindengen, a Japanese company. Shindengen is an advanced silicon-Schottky-diode developer, so it’s a key competitor of ST Microelectronics. It is really active in SiC, even if it is not that visible. The BJT is not in production yet. Cree is also doing some IGBT development. And there is a small company called United SiC, which are quiet active in developing an IGBT. Q What will all these forms of SiC transistor be used for? a The transistor is now opening the door to the fieldeffect power module. As soon as you put the SiC transistor on the module, you have a field-effect module that can be used at very high temperatures. The key highend applications are wind turbines, PV inverters and motor controls. Of course, the dream is the hybrid electric vehicle (HEV), but now it’s a question of qualification time. Starting from scratch, it takes approximately one to oneand- a-half years to qualify a SiC transistor for PV applications. For HEVs, some say two years, while others say four or even five. QAre SiC transistors viewed by potential customers as strong contenders to the silicon incumbents? a Oh yes, for sure. In terms of on-state resistance, there is no discussion. And the high-temperature functionality of SiC transistors is incredible. It means that you can really save costs at the module and system level by shrinking the size of the passives, getting rid of fans and basic devices and so on. Q How would you describe shipments of SiC transistors? a The device is ready, the manufacturing process is OK, and it’s available on the market – we just have to qualify the device. By the end of 2011 we should see the first mass-volume application in production. However, it’s a very limited market – Cree is the only one you can find on DigiKey, and you are talking about $80-100 a piece, which is huge. We hope the price will come down. But I don’t think we’re going to find that kind of 1.2 kV, 20 A device at less than ten bucks – it would be too close to the manufacturing cost. What price could we hope for? $25 would be reasonable. Q Which companies do you tip to lead sales over the coming years? a To generate high business in the power electronics industry, you have to be an established power electronics manufacturer. Competitors like Infineon, ST Microelectronics, Toshiba and Mitsubishi have been in the power electronics business for a very long time, and I think that they could be very successful if they have a device that hits the specs. I would not be surprised to see Mitsubishi Electric re-think its business model. Currently it does not sell devices, but only modules and systems. Cree is a leader in terms of technology. However, it has not been a power electronics maker for a long time. So I don’t know about its access to the market. And getting access to the market is key. For example, ST Microelectronics has been very successful with its SiC Schottky diode. Its Schottky diodes were not better than its competitors, but it had better access to the market because it was leading the silicon Schottky diode business. Q How will sales of SiC transistors pan out? a Everything can occur if we succeed in reducing the cost. We expect $1 billion device revenue, including diodes, by 2020. We think one-third of this market will be diodes and two-thirds transistors. This is also the split in the silicon world. Q SiC diodes have now been in the market place for 10 years. How would you describe their sales today? a The diode itself is very limited in terms of applications. Most of the time you have to couple it with a silicon device so you don’t get a large part of the value of working with a SiC device – especially superior hightemperature performance. You can sell a lot on efficiency, but not on high-temperature capability. Up until now, it has been a market for power systems for the power supply. But we see more and more companies going towards automotive applications, with companies like Nissian and Toyota thinking about putting SiC diodes only in two kinds of devices: the DC-DC boost converter, and the battery charger for the wall-plug to the car. The SiC diode offers some percentage points in terms of efficiency. QAre chipmakers improving their SiC diodes? a  Manufacturers are pushing up the current density to reduce the cost and make the device more and more attractive. Roughly speaking, today the difference in price of SiC diodes and silicon equivalents is about a factor of five. Q What are the most common uses for these diodes? a There are two businesses: one for 600 V and one for 1.2 kV. 600 V is related to everything that you can plug into the wall. For 1.2 kV, it is related to industry. We are also seeing some development of higher voltage diodes, like 1.7 kV and 2.5 kV, for trains and other forms of transport. Q Do these SiC diodes and transistors account for a significant proportion of SiC substrate consumption. Or does Cree’s LED manufacture account for more material? a I can’t give you a precise number, but the dominant production is for LEDs. Even if you include the GaN RF devices grown on SiC, something like 90-95 percent of all SiC substrate area is being used for LEDs. Q Today, how do shipments of 2-inch, 3-inch, 4-inch and 6-inch SiC compare? a In 2011, in production, we are seeing the introduction of 6-inch, but the price remains really high and it is only available from Cree. We think that only a few percent of SiC devices are manufactured on 6-inch substrates, with 90 percent made on 4-inch. If we project to 2018, 60 percent will be 4 inch and 40 percent will be 6-inch, in terms of percentage of square-millimetres processed. 3-inch will disappear – it’s an in-between diameter that nobody wants. 2-inch will probably stay for R&D. It’s good for the first level of qualification. Q What do you expect SiC substrate sales to be by the end of this decade? a We think that in 2020, along with the $1 billion revenue for devices, there will be probably be $350 million for substrates. Today, 50 percent of the cost of the device is the substrate, and by 2020 it will only accountfor around 30 percent. Q Are substrate costs falling fast? a In 2009, based on the 4-inch wafer, the overall device cost – including substrates plus front-end and back-end processing – was $0.30 per square-millimetre. In 2015, it will be less than $0.18 per square-millimetre, based in 6-inch wafers. And we are probably quite conservative, because we are likely to see the emergence of Dow Corning and II-VI with this material, and that will impact wafer prices. Q What is SiC substrate quality liketoday? a Two or three years ago we were in the phase of material improvement, and now we are in the phase of device improvement. Theoretically the wafers are OK – we know how to make zero  micropipe wafers. However, there are still some issues with Basal Planes and other types of dislocation. Q A few years ago, a dozen or so SiC substrate developers were fighting over a relatively small market. Has this led to much consolidation? a No. We don’t see a lot of new entrants, but we don’t see a lot of concentration. The only concentration we have seen is that Rohm has bought SiC Crystal. But I don’t think we need more than five players. That would be an approach like the GaAs business. Q Who are the big players in the SiC substrate market? a Cree still dominates, and it is increasing its revenues. However, its relative market share is decreasing – in 2010 it probably had 40 percent of the business, on both semi-insulating and n-type. It has strong competition on semi-insulating from II-VI, and with n-type substrates, we are about to see Dow Corning moving up. SiC Crystal is now in the hands of Rohm, and it’s very hard to say what it is going to do with that. Up until now Rohm has decided to keep the brand – you can still buy SiC Crystal wafers.   Figure 1. Wide bandgap semiconductors such as SiC can deliver significant improvements in all forms of electrical conversion over silicon incumbents Figure 2.Today, manufacture of SiC power electronics products is mainly performed on 4-inch substrates. During this decade, production on 6-inch will become more common Yole Développement is releasing it next SiC market report this July. © 2011 Angel Business Communications. Permission required.