Market Analysis of Wideband Gap Devices in Car Power Electronics
By, Chris Whaling
Electric Vehicle (EV) growth has compelled us to look for innovative components for the whole system which will provide us a better performance in terms of the different metrics on which it is judged. The performance metrics to be considered are cost, efficiency, size and reliability. Semiconductor devices are the primary components which enable us to transform the available electrical power in an EV to the required mechanical power (to propel the car) and other different forms of electrical power (to run the auxiliary electrical components). Optimized power conversion is necessary in this regard, thus our investigation focuses on the types of semiconductor technology which are or will be used in the EV industry.
To start with, an optimum design of the semiconductor technology will ensure that it is able to operate with 600 V, 300 A electrical parameters, is able to withstand more than 200°C (which automotive systems may reach due to inherent thermal conditions) and is able to work with very high switching frequency without going into de-rated operating conditions.
Conventional Si devices have long been unable to adhere to the aforementioned metrics largely due to its inability to work with very high switching frequencies. This increases the operating temperature of the device and reportedly, Si devices can only withstand temperatures up to 150°C. Also due to their lower band gap voltage, the Si devices can only withstand lower voltages.
Further performance improvements on Si devices have become more and more challenging and cost-intensive to achieve, making disruptive power device technologies based on Wide-Band Gap (WBG) materials like GaN and SiC increasingly attractive because of their ability to operate at higher temperatures, higher power densities, higher voltages and higher frequencies
Wide-band Gap Devices: Technology Benefits & Challenges
Wide-band gap devices (WBG) can operate at approximately twice the voltage of Si-based semiconductors as they have a higher energy band-gap than Si, around 3 times of Si, which means higher energy is needed to excite electrons from the valence band into the conduction band. They also have higher thermal conductivity (heat is more easily transferred from the device and is thus less likely to cause heat-related problems in other devices) and can operate at higher frequencies. These devices benefit from low conduction resistance (i.e. Rds-on) as well as lower capacitive charges. Since wide bandgap technologies have high critical fields, they can handle high voltages much better than Si. These properties lead to WBG devices with higher thermal and radiation durability. The critical junction temperature of Si devices which makes it uncontrollable is around 150°C while it is around 900°C for WBG devices. These semiconductors are also much lighter and smaller than Si-based devices of comparable ratings.
• Two different Wide band gap technologies have been the topic of research now. Those are SiC (Silicon Carbide) and GaN (Gallium Nitride) based devices. As already discussed earlier, there are several benefits over Si-based devices;
• Much smaller switching losses, leading to higher efficiency
• Ability to operate at higher temperatures without much change in electrical properties, leading to better reliability
• Smaller heat sink because of lower losses
• High frequency of operation allows smaller filters, which leads to light and compact packaging
Currently, SiC is considered to have the best trade-off between properties and commercial maturity. GaN can offer better HF and HV performances, but the lack of good quality large area substrates is a disadvantage for vertical devices. GaN presents a lower thermal conductivity than SiC and allows forming 2DEG heterojunctions (InAlGaN alloys) grown on SiC or Si substrates. So it is essentially a race between SiC and GaN in terms of performance and cost.
There are also some potential challenges associated with the wide band gap device of today. Those are:
a. High cost associated with WBG devices
b. Normally-on GaN and SiC devices
c. In order to migrate to wide bandgap technologies, gate driver circuits have to be replaced as well. For commercially available SiC MOSFETs, the recommended on-voltage is 20 V which is higher than the Si technology. Hence, replacing the gate driver can impose extra cost.
d. GaN-on-SiC technologies demonstrate very good figures for efficiency and maximum switching frequency among the available technologies. However, these switches are not cost effective for power converter applications.
e. In low voltage applications, GaN-on-Si provides good efficiency and switching characteristics at a reasonable cost. However, these switches are not commercially available at higher voltage levels (currently 650V).
f. Though SiC MOSFETs are available in medium voltage levels, currently, the economic aspects prevent integration of this technology in commercial transportation electrification.
Available Sic/GaN devices
Lower power SiC diodes are commercially available for long time, but research was constrained by a severe cost penalty of the SiC starting material. Boosted by the widespread use of GaN in optoelectronics – GaN is the starting material for blue and green emitting (laser) diodes – researchers started building devices for radar and RF for military and aerospace applications, but recent advances in the challenging epitaxy of GaN on silicon substrates bear the promise to significantly reduce costs down to silicon levels, triggering investigation on GaN use for high volume markets among which the power conversion.
Currently, commercialization of these technologies are restricted by the availability of the switches as well as economic concerns. Manufacturing low defect (defects due to discoloration and dislocation effects) Silicone-Carbide (SiC) substrates are expensive and not compatible with conventional fabrication equipment. For this reason, SiC technology carries a high price tag. On the other hand, high speed Gallium-Nitride (GaN) switches are manufactured on SiC substrate. These switches are costly and not suitable for power conversion applications. For this reason, GaN on Si switches have been introduced and commercialized for applications in power electronics converters. These switches benefit from fast switching times of GaN technology while maintaining a comparable price with Si technology.
Several manufacturers are manufacturing WBG switches these days. For example, BJT from Fairchild, GeneSiC (SJT), APEI(500 A WBGS), Microsemi, JFET from Infineon and MOSFET from Cree are some of the examples. For the voltage rating, most of the announced devices are 1200V, which meet the requirements for 600V DC bus applications in EVs. Only recently, Fairchild and GeneSiC have announced their high temperature SiC BJTs with maximum junction temperature of 250°C which could be an ideal choice for high temperature applications such as traction power converters. Besides the discrete package transistors, manufactures such as Cree and Rohm, have built up power modules with SiC transistors and SiC Schottky diodes for higher power applications.
The production process for the Wide band gap devices is significantly slower than that of silicon crystals for standard semiconductors. Producing the crystals to fabricate the wafers requires temperatures in excess of 2000° C for SiC. The material does not melt; it sublimates and the crystals must then be condensed from the vapors. The largest wafer size currently commercially available for this material is four inches (100mm). While the capability to produce six-inch wafers exists with Cree, market demand is currently insufficient to warrant production.
Gallium nitride (GaN) is produced by condensing the GaN vapors (usually created by combining the vapors of a gallium compound with nitrogen or ammonia vapors) into crystals or by condensing the GaN in layers directly onto the substrate (typically SiC or sapphire, though success in depositing GaN directly onto Si substrates has recently been reported in R&D projects). According to Cree, these processes are prone to producing crystals with substandard lattice structures. The crystals produced are smaller than either Si or SiC crystals, thus driving up the cost of producing bulk materials.
In terms of higher-voltage applications relevant to drive-trains, the result is that market demand for emerging WBGS modules is low because their manufacturing cost is high, in part due to the unavoidably complex manufacturing processes. Low market demand in turn leads to especially high cost-premiums given the large capital expenditures that are spread over relatively low-volume production runs.
Applicability of existing technology: benefits and challenges
It is believed that the wideband gap devices could have a wide range of applications, if the technology becomes feasible. In the transportation context, they could be used for:
• High frequency chargers for electric or hybrid cars: It could reduce the charger footprint, lower parasitic and switching losses. It could also lead to innovative fast charger designs. For level 1 chargers, it may be possible to use GaN devices
• Electric or hybrid car drive train power electronics: This is one of the major application areas currently being explored. It is most likely that GaN devices may not scale up to the voltage and power levels needed for drive train converter and SiC may be a better choice.
• These devices could also find applications in aircrafts, ships and electric train power electronic converters.
The first obstacle to designs with WBG devices is that the current rating per device is too low compared to their Si counterparts. HEV, PHEV applications use IGBT powered inverters and require switches to continually conduct currents up to 450A. Power modules handle these load level requirements through multiple Si IGBT dies per switch integrated into the mechanical package. “Mild hybrid” inverter systems operate at currents levels >200A where a single Si dies can handle this current rating. While a majority of the available SiC devices are generally rated below 50A with typical applications are for demonstration and testing projects. These single devices are currently not suitable for inverter applications. Paralleling discrete SiC devices is a solution, but is not desirable. Electrical issues such as non-simultaneously switching, places extra requirements on the gate driver designs, system layout, device matching and cooling which may not be a cost effective solution. Therefore, the inverter design engineer would expect the WBG single die current rating to increase (>200A) so that paralleling dies might not be necessary for lower power inverter applications. Higher power inverter applications would require parallel dies and contained within the same package.
Placing the entire power electronics system in a higher ambient temperature environment may also thermally challenge the microprocessor control board, but it could be avoided by thermally isolating the control board from the power stage boards.
To accelerate the WBG technology commercialization in automotive industry, it requires all parties’ involvement. The qualification process cannot be fully understood without sharing of specifications, application information and collaborative testing from the OEMs and Tier 1 inverter system suppliers.
Reliability refers to the device and power module. Most of the major power module vendors (including Infineon, Powerex, Fuji Electric, Mitsubishi Electric, RoHM) have released discrete SiC power devices or hybrid SiC power modules (Si IGBTs with SiC diodes) either into production or as demonstration modules targeted at automotive applications. Industry standards (i.e. AEC-10l) are used for semiconductor devices and package qualification in the harsh automotive operating environment (high temperature, strong vibration, high humidity, and etc.). Having off-the-shelf, automotive qualified WBG semiconductor devices is of great importance to the automotive inverter suppliers and OEMs. GaN technology development has not moved as fast in terms of automotive applications due to inherent voltage limitations. The automotive industry remains optimistic about the SiC technology and reviews GaN development for progress.
Different companies and research organisations are looking at developing and making the technology widely available for the transportation sector. The Power Electronics and Electrical Power Systems Research Center at the Energy and Transportation Science Division at ORNL is researching the application of SiC-based power devices in hybrid vehicles and the fabrication of SiC MOSFETS for use in high-power commercial applications. Cree, the producer of 85-90 percent of the world’s supply of SiC, is currently producing SiC in 4” wafers and building SiC-based inverters for solar power stations and power supplies for servers. Cree announced that it has introduced the industry’s first “commercially available Z-Rec™ 1700-V Junction Barrier Schottky (JBS) diode products intended for high-voltage power-conversion applications in motor-drive, wind-energy and traction systems.” The SiC power device leverages silicon carbide’s advantages vis a vis silicon by virtually eliminating diode switching losses. It reportedly increases efficiency, reliability and longevity of power systems while reducing the overall system size, weight and cost. Cree has a 1700 V SiC device which they believe is going to be the driver in future automotive industry.
The overarching driver for lowering SiC prices is volume production. Reducing costs is largely a matter of increasing the utilization rates of facility space and materials. Assessing the automotive industry alone – which is viewed as least likely to be the early adopter of SiC because of its cost-sensitivity – it is clear that market penetration is required to achieve 20 million MOSFET units per year. Working from Cree’s estimate of an average of 60 MOSFETs per vehicle, sales of 334,000 vehicles with SiC MOSFETs per year would be needed to reach the 20 million MOSFET unit benchmark. This is equivalent to 7.1 percent of the total automobile and light truck sales by the “Detroit 3” in North America in 2009 or 3.2 percent of all light vehicles sold in the US in 2009 (see Table 5 below). While it is difficult to foresee the timing of SiC’s widescale market penetration into the automotive sector, we assess that SiC’s adoption in the automotive sector will depend largely on non-automotive sectors and other factors outside the automakers’ control controlling the development.The key advantage of silicon carbide is its higher break-down voltage compared to silicon. More silicon is needed to achieve the equivalent voltage rating as SiC. Therefore, SiC is a natural fit for higher voltage applications. SiC becomes less attractive as voltage declines. In the 2000V plus range, SiC is clearly superior. Si is unlikely to compete in this range in the near-future.
It is notable that the use of GaN in low-voltage Schottky diode applications is increasing. These applications use sapphire as a substrate. GaN on Si substrate is also used by some manufacturers to produce FET devices. However, no information was found to show that strictly GaN MOSFETS or IGBTS are currently being produced. There are indications that research is underway in this area, however.
The use of GaN for lower voltage applications, in particular for the LED industry, is not new. Compound Semiconductor magazine reported in 2008 that GaN-on-sapphire substrates “promise to be a much cheaper alternative to SiC.” The magazine noted that while this combination of materials is reported to suffer from sapphire’s low conductivity – something that ultimately leads to poor thermal resistances and hot, unreliable devices – Velox Semiconductor (Somerset, NJ) has demonstrated this not to be true with GaN-on-sapphire diodes incorporated in an insulating frame. According to Compound Semiconductor, “the compatibility with an insulating frame is a big advantage over SiC, because it reduces the cooling demands of the heat sinks employed in the SMPSs [switch-mode power supplies].” In short, in the context of the multibillion-dollar LED industry, GaN-on-sapphire modules have been shown to deliver efficiencies comparable to SiC and manufacturing cost-savings over SiC using sapphire substrates. Further gains are realized from GaN’s lower growth temperature (1,000 to 1,100°C versus 1,500 to 1,600°C for SiC) for these applications. Finally, it has been reported – though not independently validated – that reactor parts for SiC growth are very expensive, not particularly reliable, and suffer from a small supply base.
Some of the important market views towards WBG devices:• As there is a widespread view that SiC power modules will not be useful for any automotive applications below 500V because of their inherent high band gap voltage property, different car manufacturers are opining for a higher voltage (1200 V) rated transportation system.
• Japanese automakers are currently using a variety of substrate materials, including SiC, to work on power modules in the 1200V range and may be pushing beyond it. For example, Nissan claimed to have developed the world’s first inverter using SiC diodes for vehicle use in September 2008, and implemented it in the XTRAIL FCV (Fuel Cell Vehicle). Mitsubishi is actively pursuing SiC technology R&D for their HEV/EVs, and Honda/ROHM appears poised to bring a SiC power module to the market. • Japanese automakers are currently using a variety of substrate materials, including SiC, to work on power modules in the 1200V range and may be pushing beyond it. For example, Nissan claimed to have developed the world’s first inverter using SiC diodes for vehicle use in September 2008, and implemented it in the XTRAIL FCV (Fuel Cell Vehicle). Mitsubishi is actively pursuing SiC technology R&D for their HEV/EVs, and Honda/ROHM appears poised to bring a SiC power module to the market.
• Bridgestone Corporation of Tokyo (the world’s largest tire and rubber company), commenced production of silicon carbide wafers in 2010. It is reported that the PureBeta™ SiC Single Crystal Wafer being produced by Bridgestone extends the company’s experience in using polymer technology and nanotechnology in developing tires.
• As currently implemented, Si power modules perform effectively due to a dedicated cooling system. There are no indications that eliminating this cooling loop will increase efficiency or reduce weight of current designs. However, as manufacturers design future automobiles, SiC provides an alternative to employing these dedicated cooling systems and will allow greater weight reduction with more efficient packaging and placement of components.
• While a Si-only solution could provide an approach to handling increased operating temperatures, the penalties of increased size and weight do not make it an optimum solution. In the long-run, alternative technology approaches, including SiC, are anticipated.
• Using a hybrid Si MOSFET/IGBT-SiC diode device may provide an entry point for SiC, allowing increased production which should lead to lower costs. However, there is debate regarding the technical merits of transitioning from Si-SiC devices to an all-SiC device.
• Most indicators point to SiC as the best wide bandgap material for use in hybrid vehicle power electronics. However, Toyota is pursuing GaN as an alternative to Si. It is possible, though no evidence was found to confirm, that Toyota is developing a breakthrough technology which would allow cost-competitive production of GaN devices.
Future trends and barriers of SiC/GaN for car PE
• US HEV sales are not sufficient to drive automotive OEMs to transition to SiC at this time. Thus, it appears that OEMs will need to consider the both HEV and non-HEV vehicle sales to achieve the necessary economies-of-scale to drive WBGS costs to a competitive premium over the costs of Si.
• The current state of demand for SiC MOSFETS is a chicken-and-egg challenge: Dramatic reductions in the cost of SiC MOSFETS for automotive applications depend on large-volume production runs. However, without the high volumes, producers can’t achieve the price-points required by automotive users. Automotive applications are particularly demanding, with a low-to-zero tolerance for price premiums.
• Non-automotive industries (e.g., wind, solar, and civilian/military aerospace applications) will drive SiC production growth over the next several years.
• SiC’s adoption in the automotive sector will also be affected by other factors outside the automakers’ control, such as efficiency and emission mandates.
• While the investments being made to advance the state of the GaN-art by the LED industry are a key factor to consider with regard to potential innovative breakthroughs in GaN, it is unclear if such developments may drive the evolution of higher-voltage GaN modules for automotive applications.
• Pure GaN power modules are likely only for gigawatt and tetrawatt range devices, and not for those rated in megawatts (particularly the 2000-4000V range), which is the highest voltage range addressed in this report as of potential relevance to automotive producers.
• Other industries, including aerospace, defence, and medical, may provide the near-term market demand for higher voltage GaN-based WBGS devices. However, these applications will not create the demand needed to drive cost reductions to a point at which these devices are competitive for automotive applications.
• While the experts disagree on the SiC – Si cost differential at which SiC components will begin to be adopted by the automotive industry, they all agree that the cost of SiC and relatively limited supply of SiC are the prime reasons for the continuing use of Si.
• At current market prices, using SiC is cost-prohibitive in mass-produced vehicles. However, with improvements in manufacturing processes and increased demand leading to volume production, the cost could drop enough within the next 3 – 5 years to make SiC a viable alternative to Si. For this to come to pass, the price of SiC components would need to be reduced to only a small premium (e.g., 20-30%) over Si components.
• At the current rate of development it will be at least five years before SiC MOSFET prices are low enough to interest automakers in replacing Si devices in large volumes.
The future of using GaN and SiC devices in transportation electrification seems to be very promising due to the clear advantages, like lower losses and smaller footprint. However, to realize these advantages in practice, research and development activities need to be carried out to specifically address open and unanswered questions. Some of the crucial ones are:
• Develop advanced processes for growing other WBG materials to be used in power electronic devices
• High-voltage WBG semiconductor device design and development
• Develop alternate power electronic structures and assemblies
• Power electronics packaging and thermal design for use in harsh environments
• Effect of higher temperature operation on power throughput and associated control techniques
• Reliability Testing
• Substrate size and cost:
• Device design and cost:
• Issue of Electromagnetic Interference while using very high switching frequency with WBG devices
• Non-availability of higher current rated switches for transportation industry
• New Safe-operating –area of the power converters due to proliferation of the WBG devices
Christopher L. Whaling is co-founder of Synthesis Partners, LLC, a private business and technology analysis company based in Reston, VA. Mr. Whaling has lead assessment teams in all source research, analysis and delivery of technology and market intelligence decision-support products for 20 years. Recently, Mr. Whaling refined and implemented Natural Language Processing (NLP) technology to produce beta-level, scalable decision-support systems. Mr. Whaling has been invited to present analysis products, forecasting techniques and outcomes at the US National Intelligence Council (NIC), the Society for Competitive Intelligence Professionals (SCIP), the Washington Institute for Operations Research and Management Sciences (INFORMS), and the US Department of Energy’s Oak Ridge National Laboratory. You may reach Christopher email@example.com.
About the Newsletter
The TEC eNewsletter is now being indexed by Google Scholar.