Electric vehicles drive up market for silicon carbide power semiconductors, but cost remains an issue.
The silicon carbide (SiC) power semiconductor market is experiencing a sudden surge in demand amid growth for electric vehicles and other systems.
But the demand also is causing a tight supply of SiC-based devices in the market, prompting some vendors to add fab capacity in the midst of a tricky wafer-size transition. Some SiC device makers are transitioning from 4- to 6-inch wafers in the fab.
SiC is a compound semiconductor material based on silicon and carbon. In the production flow, a specialized SiC substrate is developed and then processed in a fab, resulting in a SiC-based power semiconductor. Many SiC-based power semis and rival technologies are specialized transistors, which switch current in a device at high voltages. They are used in the field of power electronics, where the devices convert and control the electricity in systems.
SiC stands out because it’s a wide band-gap technology. Compared to conventional silicon-based devices, SiC has 10 times the breakdown field strength and 3 times the thermal conductivity, making it ideal for high-voltage applications, such as power supplies, solar inverters, trains and wind turbines. In another application, SiC is used to make LEDs.
The big growth opportunity is in automotive—especially electric vehicles. SiC-based power semiconductors are used in the on-board charging units in electric cars, while the technology is making inroads in a key part of the system—the traction inverter. The traction inverter provides traction to the motor to propel the vehicle.
For this application, Tesla is using SiC power devices in some models, while other electric car makers are evaluating the technology. “The automotive market is definitely the center of focus when people discuss SiC power devices. SiC activities from the pioneers, such as Toyota and Tesla, have brought a lot of excitement and noise to the market,” said Hong Lin, an analyst at Yole Développement. “SiC MOSFETs have potential within the automotive market. But there are several challenges, such as cost, long-term reliability and module design.”
Propelled by the automotive and other markets, the SiC power device business reached $302 million in 2017, up 22% from $248 million in 2016, according to Yole. “We expect a jump in 2018, driven by the automotive industry with the ramp up of Tesla’s Model 3, which already adopted SiC MOSFET modules,” Lin said.
The SiC power semiconductor market is projected to reach $1.5 billion by 2023, according to Yole. Suppliers of SiC devices include Fuji, Infineon, Littelfuse, Mitsubishi, On Semiconductor, STMicroelectronics, Rohm, Toshiba and Wolfspeed. Wolfspeed is part of Cree. X-Fab is the lone foundry vendor in SiC.
Power electronics play a key role in the world’s electrical infrastructure. The technology is used in industrial (motor drives), transportation (cars, trains), computing (power supplies) and renewable energy (solar, wind). Power electronics convert or transform alternating and direct current (AC and DC) in systems.
For these applications, the industry uses various power semiconductors. Some power semis are specialized transistors, which operate as a switch in systems. They allow the power to flow in the “on” state and stop it in the “off” state.
Power semis are fabricated at mature nodes. These devices are designed to boost the efficiencies and minimize the energy losses in systems. Typically, they are rated by voltages and other specs, not by process geometries.
For years, the dominant power semi technology has been (and still is) based on silicon, namely the power MOSFET and the insulated-gate bipolar transistor (IGBT). Considered the least expensive and most popular device, power MOSFETs are used in adapters, power supplies and other products. They are used in applications up to 900 volts.
In conventional MOSFET devices, the source and drain are on top of the device. In comparison, power MOSFETs feature a vertical structure, where the source and drain are located on opposite sides of the device. A vertical structure enables the device to handle higher voltages.
The leading midrange power semiconductor device is the IGBT, which combines the characteristics of MOSFETs and bipolar transistors. IGBTs are used for 400-volt to 10-kilovolt applications.
The problem is that power MOSFETs and IGBTs are reaching their theoretic limits and suffer from unwanted energy losses. A device may experience energy losses for two reasons—conduction and switching. Conduction loss is due to the resistance in the device, while switching losses occur during the on and off states.
“From 5 volts to a few hundred volts, silicon has been a good technology,” said Guy Moxey, senior director of power marketing and applications at Wolfspeed. “When you get from 600 to 900 volts, silicon MOSFETs are good but they start losing some steam. IGBTs are good heavy lifters, but they are not quick or efficient.”
That’s where SiC fits in. Power semis based on gallium-nitride (GaN) are also emerging. Both GaN and SiC are wide bandgap technologies. Silicon has a bandgap of 1.1 eV. In comparison, SiC has a bandgap of 3.3 eV, while GaN is 3.4 eV.
“The electronic bandgap is the energy gap between the top of the valence band and the bottom of the conduction band in solid materials,” according to Landa Culbertson, a contributor for Mouser Electronics in a blog. “It is the bandgap that gives semiconductors the ability to switch currents on and off as desired in order to achieve a given electrical function.”
Wide band-gap devices have several advantages. For example, electric vehicles are propelled by the motor drive, which traditionally have used power MOSFETs or IGBTs. “If you replace that motor drive with SiC, you get 80% lower loss in that drive,” Wolfspeed’s Moxey said. “It means that you can have a smaller battery for the same range. A smaller battery means lower cost.”
SiC-based power semiconductors, meanwhile, are used in 600-volt to 10-kilovolt applications. “600 to 1,700 volts catch most SiC applications. When you go to 3.3- to 10-kilovolts, it’s pretty specialized. It’s wind power and small grids,” he said.
In the power arena, GaN is used in 30- to 600-volt applications. “GaN and SiC are complementary rather than competitive technologies,” he added.
Both GaN and SiC devices are faster than silicon, but they are also more expensive. “At this moment, a single SiC MOSFET device has a cost-per-ampere more than five times higher than a similar IGBT,” said Elena Barbarini, head of department devices at System Plus Consulting, part of Yole.
The first SiC-based device appeared in 2002 with the introduction of SiC diodes, followed by SIC power MOSFETs in 2011. Like power MOSFETs, SiC-based devices are vertical structures.
A SiC power MOSFET is a power switching transistor based on SiC. “A diode is a device that passes electricity in one direction and blocks it in the opposite direction,” explained Mitch Van Ochten, an applications engineer at Rohm.
Regardless, SiC power semis are gaining steam. “Silicon plays a huge role for power devices,” said Mike Rosa, director of strategy and technical marketing for Applied Materials. “But when you talk about higher power and lower weight, manufacturers are looking at materials like silicon carbide.”
SiC-based devices are produced in a fab, where the industry continues to make a wafer size transition. “SiC is available on both 4- or 6-inch,” Rosa said. “The industry is just kicking the tires for 8 inch.”
In fact, Cree has made the transition from 4- (100mm) to 6-inch (150mm) wafers in the fab. Rohm and others are in the middle of the transition. SiC on 200mm wafers won’t happen for some time.
Generally, when moving to a new wafer size, you get 2.2X more die per wafer. A larger wafer size reduces the overall production costs.
In the digital CMOS world, chipmakers made the transition from 4- to 6-inch several years ago. It sounds simple to make the same transition for SiC, but there are some challenges. “Although mass production of SiC power devices on 150mm wafers has been proven for close to five years, the availability and cost of high-performance, low-defect density SiC substrates at 150mm still remains a barrier to adoption,” said David Haynes, senior director of strategic marketing at Lam Research.
“That said, as the transition to 150mm mass production is realized, the associated cost savings will help drive commercial viability in an increasing number of applications,” Haynes said. “Another example is the roadmap for SiC MOSFET technologies. Planar SiC MOSFETs have been proven in commercial applications for some time, but today there is a significant push to the development and commercialization of SiC trench MOSFETs, which can deliver significantly lower specific on-resistance when compared to planar structures.”
In the fab, meanwhile, SiC-based power devices generally follow the same process flow as silicon-based chips. But there are also some differences, such as the development of the SiC substrate.
For silicon-based chips, the first step in the process is the development of a raw silicon wafer. For this, a silicon seed crystal is lowered into a crucible and heated. The resulting body is called an ingot, which is pulled and sliced into silicon wafers at various sizes from 300mm and smaller.
For silicon carbide, though, SiC bulk crystals are lowered into a crucible and then heated. The resulting ingot is pulled and sliced into wafers.
For years, SiC bulk crystals were plagued with defects called micropipes, which are micron-sized holes that run through the crystals. “Micropipe defects and other defects that will kill the device operation are now all but eliminated. Materials suppliers are now offering zero-micropipe products,” said Peter Gammon, an associate professor at the University of Warwick.
Once the SiC wafers have been developed, the next step is to form an SiC substrate. The raw wafers are inserted in a deposition system, where SiC epitaxial layers are grown on the wafer, thereby forming a SiC substrate. Then, the SiC substrate is processed in the fab.
SiC requires a selective annealing process. “For power devices, the current flow is through the wafer. You need a good electrical conduction with the metal interface,” said Hans-Ulrich Zühlke, product manager at 3D-Micromac. “We only heat up the backside of the wafer to anneal the Ohmic contact to make a metal silicon carbide interface.”
Finally, the wafer is inspected for defects using an inspection system. SiC devices are prone to defects, especially as vendors move to larger wafer sizes. “There is a lot of defectivity,” said Lena Nicolaides, vice president and general manager at the LS-SWIFT division at KLA-Tencor. “Our inspection systems are used at a lower wavelength (for SiC). They are able to find discontinuities in the substrate.”
SiC in EVs
Today, meanwhile, automotive is the fastest growing segment in the overall semiconductor sector. “More and more customers are redefining their product portfolios to accommodate the IoT and/or automotive markets,” said Walter Ng, vice president of business development at UMC. “Our automotive-related revenue has grown significantly this year, and we expect it to continue in that direction for the foreseeable future.”
SiC also is seeing growth in automotive, especially for electric vehicles. In total, electric vehicles, including battery-electric cars and hybrids, represent about 1% of the world’s cars sold today. But driven by China and other nations, the electric vehicle market will grow from 1.6 million vehicles in 2018 to 2 million in 2019, according to Frost & Sullivan. By 2025, the market is expected to reach 25 million units, according to Frost & Sullivan.
“The adoption of electric vehicles and hybrid electric vehicles are certainly becoming a reality,” Lam’s Haynes said. “However, the timing and rate of adoption varies significantly around the world and is closely linked to government policy and the consumers’ access to appropriately priced products and charging infrastructure. Without a doubt, the China market is a key growth engine for electric vehicle adoption.”
Within the electric vehicle itself, the system has several domains, such as the entertainment system, on-board charger, traction inverter and others. The traction inverter converts energy from the battery to the traction motor, which then propels the vehicle.
SiC is making inroads in the on-board charger, DC-to-DC converter and traction inverter. The on-board charger charges the vehicle from a power grid.
Fig. 1: Power Electronics for Electric Vehicles Source: STMicroelectronics
A DC-to-DC converter takes the battery voltage and then steps it down to a lower voltage. This is used to control the windows, heater and other functions.
The big battle among device makers is taking place in the traction inverter, especially for pure battery-electric vehicles. Generally, hybrids are moving towards a 48-volt battery. For the power inventor, SiC is generally too expensive for hybrids, although there are exceptions to the rule.
Like the hybrid, pure battery-electric vehicles consist of a traction inverter. A high-voltage bus connects the inverter to the battery and motor. The battery provides the energy for the car. The motor, which propels the vehicle, has three connections or wires.
The three connections extend to the traction inverter, which are then networked to six switches within the inverter module.
Each switch is actually a power semiconductor, which serves as an electric switch in the system. For the switch, the incumbent technology is the IGBT. So the traction inverter may consist of six IGBTs, which are rated at 1,200 volts.
“In reality, they are electric switches. We have a choice of technologies for these electric switches, which enable and disable the various motor windings and effectively cause the motor to rotate,” Rohm’s Van Ochten said. “The most popular electrical semiconductor switch for this function is called an IGBT. They are used by more than 90% of the vehicle manufacturers out there. They are the cheapest way to switch that battery current into that motor as needed.”
There are some tradeoffs using IGBTs, however. “They are probably a third of the price of the latest technology,” he said, “but they are slow.”
That’s where the industry is targeting SiC MOSFETs, which have faster switching speeds than IGBTs. “(SiC MOSFETs) also reduce the switching loss, along with lower conduction loss at low and medium power levels,” said Maurizio Ferrara, director of the Wide Bandgap and Power RF business unit at STMicroelectronics. “They can operate at four times the frequency of an IGBT at the same efficiency, resulting in weight, size, and cost reduction due to smaller passives and fewer external components. As a result, SiC MOSFETs can enhance efficiency up to 90% compared to silicon solutions.”
So for the traction inverter, it makes sense to move from an IGBT to a SiC MOSFET. But it isn’t quite that simple because cost plays a big role in the equation.
Tesla has taken the plunge, however. The company is using SiC MOSFETs from STMicroelectronics in its Model 3, according to Yole, which added that Tesla also is using other suppliers. Other car makers are exploring the technology, as well, although most OEMs are not jumping on the bandwagon due to cost considerations.
Still, there are several ways to make the switch from IGBTs to SiC MOSFETs. According to Rohm, there are a couple of options:
• Keep the IGBT in the system, but replace the silicon diode with a SiC diode.
• Replace both the silicon-based IGBT and diode with SiC-based MOSFETs and diodes.
In the inverter, there are six IGBTs, and each one is accompanied by a separate silicon-based diode. The diode is used for several reasons. “IGBTs do not like it when you put the polarity on or voltage across them,” Rohm’s Van Ochten said. “So a diode needs to be added across each IGBT to prevent destroying it when you go to shut off the switches.”
One way to make the system more efficient is to replace the silicon diodes. “The first step in improving the efficiency of the traction inverter is to leave the IGBTs in there. But then, you put in silicon carbide diodes instead of ordinary silicon diodes,” he said. “Silicon carbide diodes have better properties. That will give you a few percent more efficiency.”
The ultimate solution is to replace both the IGBTs and silicon diodes with SiC-based diodes and MOSFETs. “SiC, because of the price of the material, is more expensive than silicon. However, if you are switching four or five times faster, you can reduce the cost of the magnetics and capacitors,” Wolfspeed’s Moxey said.
Where is this all heading? “When we look at the different applications, we expect the charging station and the on-board charger to be the first applications to adopt SiC technologies,” said Shawn Slusser, vice president of automotive at Infineon Technologies Americas.
“As to automotive applications, we expect IGBT will dominate the market for the next decade. SiC has advantages in terms of higher efficiency and higher power density, but SiC components cost more. This means the system benefits of reduced size and a smaller battery capacity for the same range need to compensate for the higher cost,” Slusser said. “That’s why we believe SiC will be first used in on-board chargers, because SiC efficiency at higher switching frequency with smaller passive components compensates for higher SiC device cost. SiC will be introduced in the broad market for main inverter applications for EVs with large batteries, as long as battery cost savings is higher than increased SiC device cost. For EVs with an 800-volt system, there are additional benefits, such as lower charging time, higher inverter efficiency and reduced cable cost.”
To be sure, SiC is heating up, and so are electric vehicles. SiC power semis appear to be in the driver’s seat—if vendors can drive down their costs. But that’s easier said than done.