SiC suppliers shoot for emerging 100 mm market

Users of SiC substrates, which are currently available in 2 inch and 3 inch diameters, will soon have a choice of vendors for 4 inch - or more precisely 100 mm - material. During September s International Conference on SiC and Related Materials (ICSCRM), Cree announced that it was providing sample quantities of 100 mm SiC substrates to customers and planned to launch commercial production in 2006. However, although Cree is the market leader in the industry, with about a 75% share of the SiC wafer market, the company does not have a clear field. Intrinsic Semiconductor, II-VI, Caracal, Norstel and Dow Corning all intend to offer 100 mm wafers, and some have plans for ramping up to production quantities within the year.

"The key application driver for 4 inch SiC is RF transistors," explained market analyst Richard Dixon of Germany-based Wicht Technologie Consulting, who co-authored a market report on SiC last summer. Because these transistors have a relatively large footprint, larger wafers are better suited to manufacture as they produce less wasted edge-space and greater economies of scale. Smaller but already commercialized devices such as GaN LEDs and Schottky barrier diodes will benefit as well.

However, production costs, which are strongly influenced by yield, are crucial for these devices because they compete with silicon - which has inferior technical specifications but is entrenched and inexpensive. For large-area devices such as RF and power transistors, defect density is a critical factor. But, as Cree s Adrian Powell pointed out at ICSCRM, even if larger wafer sizes have a slightly higher defect density, they can still deliver a higher yield due to more efficient use of the substrate area.

Cree - and most of its competitors - produces SiC from powders using some variation of physical vapor transport (PVT), with growth at 2100-2500 °C. Cree s major advantage is that as well as manufacturing most of the SiC wafers sold today, it uses a huge number of these wafers internally for its highly profitable LED chip production. "Our biggest advantage is really the volume that we run," explained Cree s director of advanced materials John Palmour. Dixon agrees: "Cree benefits from being vertically integrated, which allows the company preferential access to high-quality wafers, and to feed device performance data directly back to its materials development program."

US-based Intrinsic Semiconductor, like Cree, is vertically integrated and can fabricate prototype devices for customers. The company has already started sampling 100 mm substrates and plans to offer production material in the current quarter. It is aiming for full production in the latter half of the year. The company s current claim to fame, however, is producing substrates with a very low micropipe density.

A few years ago micropipes were the major defect worrying device manufacturers, but efforts to reduce their incidence have resulted in lower densities across the board. In late 2005 Intrinsic announced that it had produced 3 inch wafers completely free of any micropipes. Other defects are also important, however. Dixon points out that basal plane dislocations (which are critical for PIN diodes), as well as inclusions of different SiC polytypes, and chips and scratches accrued during the dicing process, can all degrade device performance.

Intrinsic CEO Cengiz Balkas says that his company is focused on providing high-quality materials, with an intense effort on reducing basal plane dislocations. He agrees that controlling polytype uniformity is also important. Because power electronic devices require the 4H form of SiC almost exclusively, the inclusion of 6H material is a problem. Intrinsic has demonstrated both types of material and can produce 4H 100 mm wafers with micropipe densities of 1/ cm-2.

Intrinsic now also has the capacity to deposit epitaxial layers on 100 mm substrates, after working with a tool maker to build a reactor that is capable of accommodating the larger wafers. The company has increased its customer base by half in the past quarter, with the majority ordering 2 and 3 inch SiC epiwafers for power electronics.

Like Cree, Intrinsic uses PVT to produce its substrates. According to Balkas this method is mature and inexpensive. "What makes it very expensive for others," he says, "is if you re only getting a few usable wafers out of a batch."

US-based II-VI, which has SiC production facilities in Pine Brook, NJ, and Saxonburg, PA, differs from Cree and Intrinsic, claiming that it has an advantage because it is not vertically integrated and so doesn t have to compete with its customers. The company produces only substrates and intends to offer sample quantities of 100 mm material by mid-2006. II-VI is currently increasing its production capacity in New Jersey and has plans to expand in a new facility in Mississippi, near to SiC epiwafer supplier Semisouth Laboratories.

Andrew Souzis, II-VI s manager of technology and programs, says that micropipes are no longer the issue they were five years ago. According to him, micropipe densities are now simply a yield-loss issue rather than a show-stopper. He is more concerned about dislocations that may not be immediately obvious but can affect device reliability. He added that the issue of device reliability is being addressed in the three DARPA programs related to GaN that were announced in early 2005. As part of this funding, II-VI received a $7.5 million contract to scale up its growth method (advanced PVT, a variant of sublimation) to 100 mm sizes.

In contrast to Cree, Intrinsic and II-VI, start-ups Caracal and Norstel are using gas-based methods to grow SiC. US-based Caracal is aiming to introduce 4 inch wafers to the market in 2006. The company is focusing on the production of 4H conducting material for the power electronics market, but it can also grow 6H and semi-insulating wafers.

Jerry Connelly, Caracal s chief marketing officer, claims that methods like seeded sublimation face challenges with respect to quality and growth rate, which result in the high cost of today s wafers. "The advantages of gas-based growth include the inherent purity of the gases, greater control over the sublimation process, and the ability to cover pre-existing cracks in material, leading to higher-quality wafers," he explained. Similarly, Norstel CEO Asko Vehanen claims that his company s process has advantages over seed sublimation. These improvements include better control of purity, a more stable chemical composition of gases at the crystalline interface, a more stable temperature distribution and a better control of growth geometry during the single-crystal growth process.

Norstel uses a high-temperature CVD growth method developed at Linköping University in Sweden. In 2006 its new plant in Norrköping will become operational, producing 3 inch wafers initially. The 100 mm wafers, says Vehanen, will be available in small volumes during 2007, and in large volumes the following year. The company produces both substrates and epitaxial wafers for customers worldwide.

The giant US chemicals company Dow Corning recently won a $3.6 million contract from the US Navy to develop SiC, saying that it would use this funding to develop a manufacturing capability for producing device-quality SiC substrates up to 100 mm in diameter. Dow s global director of new business development programs Robert Johns declined to specify when it might enter the 100 mm market, but claimed that Dow s infrastructure and its ability to provide both epiwafers and substrates gives it a competitive advantage.

Given the size of the market, there is currently an unsustainable number of SiC suppliers. In the US the government is providing contracts and support to Cree, Dow Corning, II-VI and Intrinsic, which improves their chances of success, while Norstel also enjoys support from the Swedish government, but in all these cases the funding will not continue for ever.

"We count 15 actual and potential companies on the open market, and others such as Toshiba that have implemented their own production lines for captive use," said Dixon. However, many of these companies are not capable of producing substrates in sizes as large as 3 inches or 100 mm. Dixon added: "Although we expect that increased competition will contribute to a lowering of wafer prices, in the current absence of a process to cost-effectively manufacture large SiC wafers, this [reduction in costs] would be marginal."