The importance of silicon carbide (SiC) inverters in the rise of electric vehicles is hard to overstate. SiC technology has played a key role in enabling 800V architectures and the improved charging speeds these offer. Faster switching speeds, higher current capacities and a wider temperature range than traditional silicon semiconductors have also helped to send SiC inverters to the top of every manufacturer’s wish list.
But what if there were something better? Something that could take the advantages of SiC devices and push them further?
That’s precisely what gallium nitride (GaN) semiconductors have the potential to do. Compared with SiC, they promise even faster switching speeds, smaller and lighter devices, and lower conduction losses. In the long run, they could even prove cheaper.
Predictably, there’s a catch. Although GaN’s theoretical benefits have already been realized in lower-power devices such as cell phone chargers and laptop power supplies, the technology currently struggles with the higher voltages and power levels required for automotive traction inverters.
“GaN is simply a much better semiconductor,” says Rupert Baines, CEO of UK-based startup QPT. “It has far lower losses, it’s easier to charge and discharge, it doesn’t have a body diode, so there are no recovery losses. Instead of using conduction, it’s based on an electron gas, so the electrons just float through the device with no resistance, which generates a fraction of the heat. Physicists have known all this for 40 years; they just couldn’t make GaN transistors in any sort of volume.”
One of the challenges is that GaN is a lot more fragile than the more common semiconductor materials. There have been improvements in substrate technology and device topology within the transistors themselves, but GaN is still marginal on voltage capability for automotive applications, and clever system-level design is needed to make inverters work reliably.
“In order to switch very quickly without failure, the delivery of energy into the transistor has to be governed very carefully,” notes Baines. “You also need to make sure that if there’s a short circuit for any reason, it remains safe.”
The second challenge is thermal design. GaN is prone to thermal runaway, where heat causes an increase in resistance, which increases the temperature, leading to a chain reaction. Baines says that with careful thermal design it’s possible to remove heat effectively enough to eliminate this danger, as QPT has succeeded in doing with its mid-power devices.
“We can move into the 10kW class today because we’ve solved those issues,” he says. “The reason we can’t yet move into the 100kW range is mainly because of the voltage limit. Cars are rapidly moving from 400V up to 800V. We could produce a 400V design but nobody really wants those anymore. Once GaN transistors are available that will work at that sort of voltage, we will see GaN traction inverters in EVs.”
High-voltage GaN transistors have been produced in the past. Experimental devices on a silicon carbide substrate have been proved to operate at over 3,000V, but most experts agree that the business case for GaN devices relies on the use of a more affordable silicon base.
Inspirit Ventures CEO Geoff Haynes is a semiconductor industry stalwart with more than 50 years’ experience. He was the co-founder of GaN Systems and worked on the 3,000V GaN transistors developed by Taransys.
“The challenge with using a silicon substrate is that it’s a conductor, and the channel that carries the current in the device breaks down to that substrate at typically 1,100 or 1,200V,” he explains. “By the time you’ve put some safety margin on top of that, it limits the present technology to around 650V. There is a lot of work going on among the semiconductor manufacturers to increase the depth of the AlGaN insulator, so the channel is farther away from that breakdown surface of the substrate. As soon as we can get GaN devices up to a safe operating range of 1,200V, we’ll start to see them in automotive.”
Fast switching
The key to the advantages of GaN devices is the speed at which they can switch. No transistor switches on and off instantaneously. Instead, there’s a brief transition period as the power level ramps up (or down), during which time its power losses increase dramatically. This period lasts for around 30 to 60 nanoseconds on a traditional silicon IGBT. SiC slashes that to somewhere between six and 15 nanoseconds, but the fastest GaN devices can get down to around one nanosecond.
These periods of time might seem infinitesimally small, but when the device is switching 10,000 to 20,000 times a second, they soon add up. As Haynes notes, “A transistor that switches in one-tenth of the time wastes one-tenth of the energy.”
Switching faster wastes less heat, which means that the devices themselves can be smaller and lighter. And as Baines points out, it can also set off a virtuous circle throughout the vehicle. “On a WLTP cycle you might be getting 85% efficiency, between the losses in the motor and the losses in the inverter. Going to the best SiC devices, you might get 90%. That 5% increase means you might be able to run a 5% smaller battery for the same range. That improvement has also reduced your heat rejection by a third, which means a third less heat sink and a smaller cooling system.
“But with GaN, the increase might be from 90-95%, so you’re not reducing your waste heat by one-third, you’re reducing it by two-thirds [compared with silicon]. What could that mean? It has been suggested that we’re not far away from the point where lower-powered city cars could replace liquid cooling with air cooling, like Volkswagen did with the original Beetle. At that point, you’re not wasting energy in the cooling, you’re making everything smaller, you’re making everything lighter and you’re making everything simpler.”
Packaging options
The compact size of GaN inverters could make it easier to package the drive system inside the motor itself, aiding integration. It could even assist with the move toward greater recyclability. “People are starting to ask, ‘Can we use aluminum windings in the motors, and make those flat so they’re easier to wind?’,” Haynes notes. “By moving to aluminum, you have the advantage that you could melt everything down at the end of life. And if you can make them square, you might actually be able to mount the transistors directly onto those as heat sinks. So there’s a tremendous amount of work going on into ways to simplify, reduce the heat production and integrate the whole of the motor and its drive system. Switching to GaN makes the components so much smaller – tiny little coils rather than large inductors – which could help to enable this change.”
There’s no doubt that GaN technology would cause the Tier 1 and Tier 2 suppliers some headaches. The mechanical, thermal and electrical design within the inverter would have to be revised to support the new technology – in contrast to SiC, which is pretty much a drop-in substitute for silicon. For vehicle manufacturers the benefits would be much simpler to unlock, says Baines: “The inputs, the outputs and the control signals will all be much the same; it’s just that the inverter will get a lot smaller and the cooling system requirements will go down.”
Delivering these benefits is a work in progress, but Haynes and Baines both believe it should be realistic within the next couple of years. Perhaps what’s most surprising for a new technology backed by such ambitious claims is that they also believe that it could be cheaper than SiC.
“High-power GaN transistors are more expensive than SiC at the moment, but that’s largely because the market isn’t there for them yet,” says Haynes. “The GaN transistor is actually a very simple structure; there are far more processing steps involved in SiC. If the industry can tip the supply and demand for GaN, it will knock SiC out of the park.”
The key to all of this is context. Haynes emphasizes that it isn’t a question of outright superiority, but rather picking the best material for the requirements, with silicon and SiC still having major roles to play in different applications. For automotive traction inverters, however, the case for GaN appears compelling, providing the challenges around voltage and power capability can be solved.
