SiC Semiconductors

While all-silicon power transistors have an enviable record of performance growth, they may be reaching their limits for high-demand power switching and control applications. A newer semiconductor material, silicon carbide (SiC), promises a range of potential benefits: much smaller, more efficient devices with drastically reduced switching losses, low leakage currents, higher switching frequenc...

By Frank J. Bartos, P.E. Control Engineering Consulting Editor October 1, 2006

At a glance

Power transistors

Silicon carbide devices

SiC-based motor control

Embedded control

While all-silicon power transistors have an enviable record of performance growth, they may be reaching their limits for high-demand power switching and control applications.

A newer semiconductor material, silicon carbide (SiC), promises a range of potential benefits: much smaller, more efficient devices with drastically reduced switching losses, low leakage currents, higher switching frequencies than standard (all-silicon) semiconductors, and the ability to operate well above the standard 125°C junction temperature. Miniaturization possibilities and high-temperature attributes also may provide further freedom to embed these control electronics directly inside motor enclosures.

Any new technology has an uphill battle for acceptance versus a mature, well-established one, and SiC power control is no exception. Standard power switches, such as insulated-gate bipolar transistors (IGBTs), have an extremely large product base and production techniques optimized for yield. SiC advancements, on the other hand, require great financial and R&D investment to solve remaining material issues and to perfect semiconductor fabrication methods. Nevertheless, these power-switching devices, which are externally controlled to “switch” rapidly between conducting high currents in the forward mode and blocking kilovolts in the reverse direction, are worth watching.

Initial success and the primary use of SiC has been in light-emitting diodes (LEDs) for automobile headlights, dashboards and other lighting applications. Other market niches are switch-mode power supplies and Schottky barrier diodes. Upcoming target applications include hybrid vehicles, power converters (to help reduce active front-end filter sizes), and ac/dc motor control. These more-demanding applications are not yet commercialized, as they require higher material quality and high production volumes to lower the cost. However, intense R&D is ongoing at numerous companies, labs, and government facilities around the world to make SiC technology viable. The timeline for commercial, industrial, or military products is seen as two to five years or longer, according to some experts.

Motor control manufacturers are especially interested in SiC developments and, some aid that process through collaboration with researchers and semiconductor manufacturers. But most remain tight-lipped about their involvement.

Facilitators for SiC

Rockwell Automation recognizes the new technology’s potential benefits and considers itself a “facilitator” for SiC, according to Gary Skibinski, Ph.D., consulting engineer for the company’s Standard Drives Div. Rockwell also assesses how SiC might fit into its future business. “It’s crucial for a leading company to understand new technology and be ready for it,” he says.

Developments are proceeding stepwise. For example, Skibinski notes the adding of a SiC power diode to each standard IGBT in a drive circuit (acting as an inverter free-wheel diode) as a logical first step to push up production volumes; power switches would follow later. “An all-SiC drive is still under R&D and prototype investigation,” he says.

In a recent study of such hybrid Si-SiC power modules (Si IGBT plus commercial SiC diode), Rockwell found significant improvements in reducing energy loss and increasing carrier frequency, compared to an all-Si module (Si IGBT plus anti-parallel diode switch). Total power loss of the module is the sum of E on +E rr +E off (see “Energy comparison” diagram). No change of loss was found for E off with any value of R gate tested for either Si or SiC diodes, but marked reduction was obtained in the other two loss components when using SiC diodes. Diode reverse recovery loss E rr was reduced to virtual zero (94 %) for any R gate , while IGBT E on was cut 37% when switching with R gate was =25Ω and more dramatically by 85% when R gate was 8Ω.

Hybrid Si-SiC power modules have the potential to substantially reduce Eon and Err power losses, compared to all-silicon modules, says recent investigation by Rockwell Automation. IGBT En power loss value of 3.3 mJ for all-Si module has been normalized to 1.0 per unit for making power loss comparisons.

This study illustrates the possibility to go to higher switching frequencies, which has historically been limited by reverse recovery loss in the all-Si diode. E rr limits additional improvement in reducing turn-on losses. “Silicon module vendors recommend a gate resistor R gate (such as 25Ω) for optimum balance of IGBT turn-on (E on ) and turn-off (E off ) energy loss,” explains Skibinski. However, with SiC diodes, R gate can be further reduced.

“Total power loss reduction (E on +E rr +E off ) with SiC diodes has potential beneficial features in drives,” he says. First, it enables a four times increase in switching frequency using the same cooling system, allowing better performance and possible size and cost reduction of the front-end magnetic filter. Alternatively, he says, existing switching frequency (and cooling) can be retained, to yield benefits such as higher efficiency and reliability, lower loss, or increased output ratings. Lower total power loss may potentially reduce cooling costs.

Yaskawa Electric is another drive manufacturer with SiC developments on its radar screen. The company summarizes basic SiC benefits as high operating temperature, high switching speed, and low losses both in conduction and switching modes, which makes the drive system more efficient.

Tsuneo J. Kume, Ph.D., an IEEE Fellow at Yaskawa Electric R&D Labs in Kokura, Japan, told Control Engineering , “This low loss nature, together with high operating junction temperature makes both the silicon-carbide device and cooling system much smaller—resulting in realization of higher power density drive systems. In addition, high-frequency switching capability improves response or bandwidth of the control system significantly,” says Kume. Yaskawa has been working closely with leading semiconductor manufacturers such as Mitsubishi Semiconductors to be in position to obtain SiC devices with the most advanced technology whenever available. The technology is still being tested for application and quality and no specific drive product development has been started, according to Kume.

Nimble, innovative

Smaller, innovative companies often provide impetus for advancing technology. One such company in the SiC arena is Arkansas Power Electronics International Inc. APEI specializes in R&D of high-performance power electronics systems that apply SiC devices as the core enabling technology. “In particular, APEI engineers are focused on technology development for extreme environment operation (temperatures up to 500

APEI has developed, fabricated, and tested SiC-based dc and ac motor drives, single- and three-phase inverters (rated 3 kW and 5 kW, respectively), and dc-dc converters. Other developments by the company’s researchers include high-temperature packaging technologies demonstrating single-device power operation to 500-plus0 ºC. “High junction temperatures allow reduction of size of the electronics’ thermal management system and operation at high power densities,” he says.

Another small firm active in SiC power transistor development, founded in 2005 by Swedish SiC research experts, is TranSiC AB, a spin-off from the Royal Institute of Technology (KTH) in Stockholm. Recently, TranSiC AB has successfully demonstrated a prototype of its bipolar junction transistor (BJT) in a standard TO 247 package. The first model, BitSiC 1206, is a 1,200 V, 6 A device.

Bo Hammarlund, CEO of TranSiC, mentions that chip packaging was successful and on/off-switching results were “very good compared to competing silicon devices.” The company buys SiC wafers and epitaxial materials from various sources, but all critical chip processing is done in KTH’s lab.

BitSiC’s industrial packaging was done by an experienced outside company, explains Hammarlund, but, TranSiC can provide short run packaging for pilot customers where process development speed is a trade-off with package price.

TranSiC expects it will take at least two years to fully qualify its BitSiC and make the processing cost-effective. Currently, the device is quite costly per chip size, explains Hammarlund. “We expect prices to drop 30% every six months during the next couple of years when we pay more attention to all cost reductions.”

Next for BitSiC 1206 is wider customer sampling before year-end 2006 and a preliminary datasheet on TranSiC’s Web site. “Our goal is to have a 30 A prototype device by end of 2007 plus a packaged prototype able to handle 225 °C junction temperature,” Hammarlund told Control Engineering . Still higher current devices are on the company’s longer development path.

Divergent view

Not everyone agrees about the future of silicon carbide power control. ABB is a specialist in high-power semiconductors, but it discontinued an ambitious SiC development program at its Corporate Research center in Sweden in 2002. One reason given for this move by the company’s semiconductor R&D department is the bipolar on-state degradation effect, due to what is known as Basel plane dislocations. This reflects a view based solely on high-voltage/high-power devices and applications.

“Silicon carbide is considered suitable in the short term for low-voltage unipolar diodes. Also, it has potential for low-power bipolar transistors and JFETs for discrete high-frequency applications. However, due to the much higher barrier height of PN junctions for SiC versus Si material, bipolar diodes only become feasible from the conduction loss viewpoint for devices with voltage ratings exceeding 4.5 kV,” says Dr. Munaf Rahimo, chief engineer for R&D at ABB Switzerland Semiconductors. “On the other hand, SiC bipolar transistors do not suffer from this drawback, making them attractive for high-voltage applications compared to other switch concepts in the long term,” he says.

As for SiC’s promise of rapid switching capability, ABB Semiconductors affirms such high-frequency operation only if very low stray inductances are present, as is the case for low-power/low-voltage systems. “In high power systems, stray inductances are large, requiring semiconductors to switch more slowly. For SiC, this would mean switching devices slowly or fitting them with snubbers, which would reintroduce losses we were trying to get rid of by using expensive SiC,” continues Rahimo.

In addition, current substrate material prices are cited as 100 times higher (for 3-in. SiC wafer), a “factor that might drop to 10x in the long term,” Rahimo says. While SiC wafer quality has improved to enable manufacture of devices with small die sizes (up to 5 mm2) at acceptable yield, devices with larger dies (typically, 50 A diode on 25 mm2 chips) are plagued by very low yields, he explains. For 4-in. SiC wafers, quality is cited as “still poor.” This compares to monolithic bipolar devices with wafer diameters up to 6 in. and 12-in. Si wafers in production today. “Defect densities may reach sufficiently low levels to permit manufacture of bipolar SiC power devices in about five to 10 years,” he adds. Even a longer timeline is forecast for high-power SiC devices.

Other developers recognize shortcomings of SiC, even as they continue development.

Breakthroughs needed

For APEI Inc.’s applications, breakthroughs needed to spur SiC technology boil down to devices and packaging. “At the device level, fab yield is a primary concern. Low yields mean that devices are slower to come to market,” says Lostetter. APEI reportedly works with numerous global manufacturers to obtain devices at the R&D stage, but if devices are unavailable commercially, systems using those devices likewise will not be commercialized.

Other issues cited by the company include the need to develop normally-off, voltage controlled devices that do not degrade power-density performance, and semiconductor metallization processes that withstand long-term reliability needs for high temperature environments. Many presently available devices are either the normally-on type or current controlled.

Developing power systems with normally-on devices is very tricky, explains Lostetter. Particularly tricky is protecting against control system catastrophic failure, such as all power devices switching on and shorting power to ground.

At the materials level, APEI sees the need for strong “die attaches” with long-term reliability; advances in diffusion barriers and corrosive-resistant metallization stacks on power substrates; and mechanically reliable power substrates able to handle extreme temperature swings without failures due to coefficient of thermal expansion mismatch and stress fractures.

While some prototype products are here, other major SiC drive projects have a longer horizon. APEI mentions two of its more exciting development for U.S. government customers. High power density three-phase motor drives are in the works for the U.S. Army’s Future Combat Systems (FCS) program, geared to fielding fully electric and hybrid-electric combat vehicles and with goals as far out as 2020. “While APEI has demonstrated 3-5 kW SiC motor drives, realistically useful motor drives would be required in 100-1,000 kW range,” states Lostetter. “We believe these goals are achievable for demonstration purposes in the next two to three years and for commercial purposes in the next three to five years.”

A more ambitious APEI project involves extreme-environment dc motor drives being developed for NASA Jet Propulsion Laboratory’s Venus In-Situ Explorer (VISE). That program, still in its planning stages, aims to land robotic explorers on the surface of Venus. VISE is similar to the successful Mars robotic explorers, but the environment of Venus is much more extreme: surface temperatures exceeding 485

NASA’s timeline

“APEI Inc.’s SiC-based motor drives would allow robotic traction, appendage, and gear functionality without costly, heavy electronics shielding and thermal management systems that would otherwise be required,” says Lostetter. “In addition, silicon-based systems would fail within a short time when thermal coolants become exhausted.” APEI estimates that all-SiC motor drives will be flight ready by 2010, which meets NASA’s timeline to launch the VISE lander in 2013.

Meanwhile, TransSiC’s Hammarlund is confident that SiC material defect “difficulties” will be solved within the next three years. “We will be able to do larger area components and be able to introduce at least 50-Amp chips by then,” he concludes.

The status of silicon carbide for power control resembles a “Catch 22“ situation. Users and technology implementers will be there when reliable SiC devices are available in volumes that dramatically cut pricing; on the other hand, commercialization will come when there is adequate user demand.

Online Extra

Developing silicon carbide power control is a work in progress Researchers and developers recognize present shortcomings of silicon carbide (SiC) technology. At the same time, most are confident that deficiencies will be solved as they continue to work on issues of material quality, fabrication processes, and cost reduction.

Actually, Swedish company TranSiC AB was founded in 2005 on the premise that “great progress” is being made in improving SiC wafer materials, according to company CEO Bo Hammarlund. He notes the availability of 4-in. SiC wafers from Cree Inc. since early 2006 and the somewhat earlier introduction of zero micropipe (defect-free) SiC material by Intrinsic Semiconductor. The latter company has since been acquired by Cree in July 2006. Hammarlund also mentions meeting with other new SiC wafer material suppliers, from which he has gained confidence that suppliers “will improve material quality orders of magnitude during the next couple of years.”

A line of prototype SiC bipolar junction transistors (BJTs) is under development at TranSiC, with the first model—BitSiC 1206—rated for 1,200 V, 6 A. Larger devices are in the works, with a 30 A prototype BJT scheduled before the end of 2007 (see main article). Relative to the stepwise increase of current capability for these transistors, Hammarlund states that, “One cannot go directly to high currents (10+ A) today since the chip size versus yield is limited by defect density in the SiC-wafer.”

Rockwell Automation notes three current problem areas of SiC technology: material defects, manufacturing processes needing to scale up wafer sizes, and necessity to reduce production costs. Gary Skibinski, Ph.D., consulting engineer for the Standard Drives Div. at Rockwell, adds that lower SiC pricing has to come from driving up production volumes for SiC devices. Also, he sees the need for packaging breakthroughs for cooling of SiC modules.

Skibinski believes that the micropipe defect problem—which limits die area and yield, hence limiting increase of current capability—is essentially solved for 3-in. SiC wafers. At 4-in. wafer size, micropipe defects affecting yield have reoccurred, but suppliers say the problem will be solved soon, followed by efforts to produce larger wafers, explains Skibinski. “Larger wafers are needed to reduce costs as was the case for standard, all-silicon material, which is running six-inch wafers,” he says.

As for a timeline to reach the commercial SiC diode switch, they’re available in the “few amp sample range” and prototype demonstration stage now. Real power-switching SiC devices are coming in the 2008-09 period, concludes Skibinski.

Arkansas Power Electronics International Inc. (APEI) agrees that presently SiC power control is a high-risk technology. “Most companies don’t want to be the first to sink resources into a new area,” remarks Alexander B. Lostetter, Ph.D., president of APEI. “High risk, of course, also means a potential for a high pay off. Even companies heavily pursuing SiC are still very careful of the market.” He offers the example of SiCED —an Infineon Technologies spin-off company—which has taken a SiC JFET (junction field-effect transistor) to commercialization. However, commercialization means technology transfer to Infineon, which would then have to supply expensive production lines and marketing to take the devices from “engineer samples” at SiCED to a full-blown production phase at Infineon, he explains. “Infineon says there is not a large enough demand for it to commit such resources at this time,” according to Lostetter.

As for SiC wafer quality, APEI doesn’t see it as “nearly the issue it was a few years ago.” While still a problem, wafer size availability rather than wafer quality is considered more important. “SiC wafer costs about 10 times that of silicon, while wafer sizes run only in the 4-5 in. maximum diameter range,” he says.

Due to long development times predicted by some sources for full realization of SiC capabilities, the question of an alternative power-switching technology arises. This is not a concern at APEI. “The only possibility on the horizon would be GaN [gallium nitride], and it doesn’t have the thermal conduction of SiC,” adds Lostetter. “It’s also even further behind the development curve.”

While SiC technology represents a work in progress, perhaps clever developers worldwide can shorten the timeline to commercialization predicted by some forecasters.

Frank J. Bartos, P.E.Consulting editor