Silicon Carbide: A Tug-Of-War

Article By : Lefeng Shao

A shortage of silicon carbide wafers is frustrating the growing demand for SiC power devices, but capacity is being added.

Silicon carbide (SiC) has excellent properties as a semiconductor material, especially for power conversion and control. However, SiC is extremely rare in the natural environment. As a material, it was first discovered in tiny amounts in meteorites, which is why it is also called “semiconductor material that has experienced 4.6 billion years of travel.”

Yole Development’s recently published “Power Silicon Carbide (SiC): Materials, Devices and Applications – 2019 Edition” report predicts that, by 2024, the market for SiC power semiconductors will grow to $2 billion by 2024, at an annual growth of 29%. The automotive market is undoubtedly the foremost driver, with around 50% of total device market share in 2024.

Could GaN Be the ‘Green’ Technology That Drives EV?

Will the wafer shortage continue?
In the past couple of years, wafer supply shortages have been a major bottleneck restricting SiC industry growth. Combined with increasing market demand, many players, including fabs, have recognized the need to expand their investments to fill the supply chain.

Cree announced in May it is investing $1 billion to expand and develop an automated 200 mm SiC fabrication facility. Expected to be complete in 2024, it will support Wolfspeed’s SiC and gallium-nitride on SiC business, enabling up to a 30-fold increase in SiC wafer fabrication capacity and 30-fold increase in SiC materials production, to meet the needs of electric vehicles (EVs) and the 5G market.

Rohm SiC Wafer

6-inch SiC MOSFET wafer. (Source: Rohm)

STMicroelectronics has a strong focus on SiC, having generated $100 million in revenue in 2018 and targeting $200 million in 2019; its target revenue for 2025 is $1 billion and it hopes to account for 30% of the SiC market. To this end, STMicroelectronics signed a multi-year supply agreement for SiC wafers with Cree in January this year; as part of the agreement, Cree will supply STMicroelectronics with $250 million of 150 mm SiC bare wafers and epitaxial wafers. A month later, ST acquired a 55% stake in Norstel AB, a Swedish SiC wafer supplier, with the option to acquire the remaining 45%.

Infineon, also a major SiC manufacturer, doesn’t intend to be left behind. In addition to announcing a long-term supply strategy agreement for SiC wafers with Cree, as early as February 2018, it also acquired German startup Siltectra in November of the same year, giving it access to the latter’s “cold split” technology for splitting SiC wafers.

Japanese manufacturer Rohm has been conducting basic research on SiC MOSFETs since 2000 and acquired SiCrystal, a German SiC wafer material manufacturer in 2009 — SiCrystal has a completely vertically integrated manufacturing process from ingot production, to wafer processing, to package assembly. Milestones include the world’s first SiC SBD (Schottky barrier diode)/MOS and mass production in 2010; mass production of full SiC modules in 2012; mass production of trench SiC MOS in 2015; and mass production of 6-inch SiC SBD in 2017.

As reported by market research firm, Yole Development, Rohm’s share of the global SiC market in 2013 was 12%; this market share grew to 23% in 2018. According to Suwon Dejian, director of the Technical Center of Rohm Semiconductor (Beijing) Co., Ltd., Rohm is expected to invest a total of 85 billion Yen in SiC production through to 2025.

Putting the SiC supply shortage in perspective
Bret Zahn, senior director and general manager of On Semiconductor’s low-voltage and battery-protected MOSFETs and wide-bandgap MOSFETs, has reservations about the widely held argument that the wafer supply shortage is constraining development of the SiC market: “I think the strict process and subsequent certification of the platform import design has been the threshold, but market adoption has continued to increase.”

His take is that the SiC market was previously using 100 mm wafers. Then more SiC suppliers began to enter the market, bringing more intense market competition, and 150 mm wafers began to be favored due to cost advantages. However, 150 mm wafer yields are not comparable to 100 mm wafer yields, so suppliers have been working hard to improve the quality of larger diameter crystals, which has led to a shortage of high yield 150 mm SiC wafers. But as 100 mm wafer suppliers are now offering 150 mm wafers of the same or better die quality, and new wafer suppliers are joining the market, the shortage of SiC wafers has begun to ease.

The SiC trend in automotive applications.

On Semiconductor entered the SiC device supplier market in 2017, with technology coming from Fairchild Semiconductor acquired in late 2016. As a relatively new supplier of SiC devices, On Semiconductor has used 150 mm wafers since its inception. Its core strategy is to certify multiple suppliers and focus on acquiring suppliers that offer the highest die yield crystals. At the same time, it also developed an internal SiC crystal growth plan with the goal of providing at least 50% of its own SiC wafers by the end of 2022. This comprehensive SiC vertical integration guarantees supply (especially for automotive customers).

Automotive, the key to reshaping the SiC market
Initial SiC applications have mainly concentrated in areas such as photovoltaic energy storage inverters, data center server UPS power supplies, and smart grid charging stations, which require high conversion efficiency. Taking a 5 kW LLC DC/DC converter as an example, its power control board weighs 7 kg and has a volume of 8,775 cc when using Si IGBT (silicon insulated gate bipolar transistor). When SiC MOSFET is used, the weight is sharply reduced to 0.9 kg and the volume is reduced to 1,350 cc. This is due to the fact that the chip area of the SiC MOSFET is only 1/4 of that of the Si-IGBT, and its high frequency characteristics enable a 63% loss reduction compared to the Si-IGBT.

People soon discovered that silicon carbide’s electrical (lower impedance/higher frequency), mechanical (smaller size), and thermal properties (higher temperature operation) are also highly suitable for manufacturing high-power automotive electronics, such as car charging, buck converters, and main drive inverters. In particular, after Tesla adopted SiC devices in its Model 3 main-drive inverters, the effect was rapidly amplified.

SiC technology is also used in the famous Formula-E. Starting in the third season of 2016, Rohm began sponsoring the Venturi team and replacing the IGBT+ Si FRD solution in the traditional 200 kW inverter with an IGBT+SiC SBD combination in the car. Maintaining the same power, the transformer weight was reduced by 2 kg and its size reduced by 19%. When the SiC MOS+SiC SBD was adopted in the 2017 season, the weight was reduced by 6 kg, the size was reduced by 43%, and the inverter power was increased to 220 kW from the previous 200 kW.

At present, the main drive inverter in xEV cars is still dominated by IGBT+Si FRD; but when considering the need for longer mileage range in future electric vehicles coupled with shorter charging time and higher battery capacity, SiC MOS devices are used. In addition, automotive OBC and DCDC applications underwent major innovations in 2017/2018 — incorporating SiC. Meanwhile, wireless charging and high-power DCDC are under development.

“The economic benefits of using SiC inverters for electric vehicles are obvious,” said Suwon. He added that SiC can increase inverter efficiency 3%–5% and reduce battery cost/capacity through SiC — which he contends is a great opportunity for high-end cars because of its larger battery capacity.

Zahn reminded that everything in the development chain — including wafer fabrication, packaging/testing, application testing, and final qualification testing — must be reconsidered when developing SiC. For example, the large die, low on-resistance RDSon device that is considered to be the most attractive in the xEV market has been identified as a huge challenge. Due to the different properties of SiC and the much smaller die size, industry needs to rethink many thermo-mechanical stress problems and redesign interconnect techniques to achieve higher current densities and lower inductance.

Discrete device vs. power module
Like IGBTs, for SiC, the industry generally expects modules to play a key role. But what form will the full SiC module take? Although some manufacturers use standard silicon packages, most manufacturers have developed their own SiC modules. For example, Tesla has successfully developed an SiC module design supply chain with independent intellectual property rights through cooperation with ST and Boschman, with devices manufactured by STMicroelectronics.

Zahn said the development path for SiC use in the photovoltaic and xEV markets is interesting. In the past two years, the photovoltaic market has experienced accelerated introduction of IGBT/SiC hybrid boost modules, and in 2019 the market began to shift toward full SiC modules.

But the xEV market is a bit different, since they bypass the hybrid solution and go directly to the full SiC module. There are two reasons for this: first, xEV suppliers have found that using a full SiC module inverter can provide better performance for the xEV market at a lower system cost than an IGBT/SiC hybrid solution; second is the competitive factor — many xEV suppliers have realized that they must follow suit after seeing their peers adopt a full SiC system solution, or they will be eliminated by the market.

Power semiconductor device usage scenario summary. (Source: Rohm)

The fastest way to lower the price
Professionals working with SiC tend to “love and hate” SiC devices. On the one hand, SiC devices have the advantages of high voltage, high frequency, and high efficiency. On the other hand, SiC faces significant technical requirements in terms of manufacturing and application.

Jean-Marc Chery, President and CEO of ST, believes two key short term challenges need to be addressed: one is the supply chain and the other is cost. Raw material suppliers and equipment suppliers need to take measures to promote and prove that the use of SiC in electric vehicles and other fields is energy efficient. At the same time, despite SiC’s advantages in terms of breakdown field strength, band gap, electron saturation speed, melting point, and thermal conductivity, hard materials and complex manufacturing processes have greatly increased costs. Related companies must work hard to shrink devices, increase wafer size, reduce material costs, and optimize module design.

But even so, “the cost of a single SiC device will be higher than traditional Si devices.” However, Chery said ST’s emphasis is on enabling ultimate savings in system costs. For example, in electric vehicles, SiC devices may add up to $300 in upfront costs, but overall, they save $2,000 in system costs due to lower battery costs, EV space, and cooling costs.

ON Semiconductor believes that vertical integration is the fastest way to achieve cost parity for SiC and IGBT. Infineon’s Ma Guowei, director of its industrial power control business unit, said the price of SiC has been an emerging technology; this means SiC naturally has all the common problems of emerging technologies: small output, insufficient stability, and high price. Although everyone hopes that SiC technology can be popularized, the process from emerging technology to general technology is often very long.

“The IGBT has been developed since 1990. After 30 years of technological innovation, the wafer size has increased from 4 inches to 12 inches, the chip thickness has been reduced from 300 μm to 60 μm, and the final cost has dropped to one-fifth.

The following figure is a summary of the scenarios for power semiconductor devices provided by Rohm. If the switching frequency is used as the abscissa and the output power or voltage is taken as the ordinate, then the application of SiC-MOSFET is mainly concentrated in the relatively high frequency and high voltage region, and the Si-IGBT/Si-MOSFET/GaN HEMT corresponds to the high voltage low frequency and high frequency respectively.

Shao Lefeng

Although it is quite optimistic about silicon carbide, ST also emphasizes that SiC does not completely replace silicon-based IGBTs or MOSFETs. These technology products have different switching characteristics, power consumption, and cost. Yu Daihui, vice president of Infineon Greater China, believes that SiC can revolutionize its efficiency in certain industries, such as improving energy efficiency and reducing weight and volume. In conclusion, Si and SiC devices will coexist for a long time and develop together.

Lefeng Shao is Principal Analyst of Aspencore China. This article originally appeared in the September 2019 issue of “International Electronic Business.”

Expo booth:

New products & solutions, whitepaper downloads, reference designs, videos

Conference sessions:
  • Internet of Things (IoT)
  • Supply Chain
  • Automotive Electronics
  • Wave of Wireless
3 Rounds Lucky Draw:

Register, join the conference, and visit the booths for a chance to win great prizes.

Leave a comment