PowerUP EXPO 2021: Fundamentals of GaN and SiC power devices

AspenCore’s PowerUP Expo 2021 kicked off the three-day virtual conference with a deep dive into the fundamentals of wide-bandgap (WBG) devices – namely gallium nitride (GaN) and silicon carbide (SiC), which are driving big changes in the power electronics industry. GaN and SiC are making big inroads in several sectors, including automotive, data centers, renewable energy, aerospace, and motor drives.

Leading the discussion on the Fundamentals of GaN is Alex Lidow, Ph.D, CEO at EPC. Lidow believes power electronics is undergoing a renaissance with GaN and SiC as the successors to the silicon MOSFET and IGBT.

In the technical session, Lidow covered some of the basic thermal and electrical advantages of GaN power devices compared to silicon (Si) counterparts and four main applications – LiDAR, DC/DC converters, motor drives, and satellite electronics – for GaN power devices. He also discussed EPC’s product roadmap for performance improvements and increased levels of integration.

Victor Veliadis, Ph.D., professor in electrical and computer engineering, North Carolina State University, and executive director and chief technology officer at PowerAmerica, led the discussion on the Fundamentals of SiC. He believes Si power devices, although continuing to make progress, are approaching their operational limits primarily due to their relatively low bandgap and critical electric field that result in high conduction and switching losses, and poor high- temperature performance.

In his technical session, Veliadis outlines the material properties of SiC and application opportunities where SiC devices are displacing their incumbent Si counterparts. He also covers material and device fabrication challenges and the design of MOSFETs, which are currently used in the majority of SiC-based power electronics systems.

Fundamentals of GaN

Here are highlights from Lidow’s presentation, starting with a discussion on the advantages of GaN.

Some of first questions asked by designers is ‘when is it appropriate to use GaN and what are the advantages,’ said Lidow. While those questions are appropriate, he said the  biggest question is really ‘why continue with silicon.’

There are many advantages of GaN, said Lidow. “It’s smaller, it’s faster, more efficient, and low cost, which is now reflected in pricing in the marketplace over the last couple of years.”

However, two remarkable benefits of GaN technology is its radiation hardness and integration, said Lidow. “The biggest advantage that will define the future of power conversion more than anything else about GaN is that it is easier to integrate multiple power devices.”

In a materials comparison of basic semiconductor properties – bandgap, critical electric field, and electron mobility, GaN is shown to be a superior material. “With Si the bandgap is a bit over an electron volt and the critical electron field is .23 MV/cm and with GaN the mobility of the electronics and the bandgap is much wider, which means there is a tighter bond between the gallium and the nitrogen atoms in the crystal lattice than there are between the silicon,” said Lidow. “it is fairly similar to SiC, which both have a bandgap of about 3.26, said Lidow.

This reflects in the critical electric field, he added, with GaN having more than an order of magnitude higher critical electric field. This ultimately means that power devices can be made much smaller.

Another benefit of GaN is its higher electron mobility than either Si or SiC, which makes it a fundamentally superior semiconductor for all these reasons, but this mobility is an extra benefit, he said.

Because of the critical electric field advantage, both SiC and GaN are far superior to the theoretical limits of Si, and GaN has a basic advantage over all of them, which comes from two things; a somewhat higher critical electric field and the higher mobility, said Lidow.

Lidow goes into more details about how GaN leverages these properties to achieve a two-dimensional electron gas and why this is an advantage. In addition, he discusses enhancement mode devices, which can be created by growing a GaN crystal that is doped with magnesium, as an example, which creates an acceptor-rich crystal on top.

The advantage of enhancement-mode GAN devices, and GaN in general, is that they will conduct not just in the forward direction, he said. “If you simply reverse the terminals and apply a positive bias to the gate compared to the drain, you will conduct in the reverse direction so it looks like it has a reverse diode much like a MOSFET. That is very useful in many circuits.”

Thermal management is another advantage of GaN. Lidow is asked all the time about how to get the heat out of these tiny GaN devices. “They’re small, but they have much lower on-resistance and much lower switching losses, so it doesn’t generate as much heat but there is a basic advantage.”

GaN and Si are basically limited by the amount of solder joint hitting the PC board and limited by the PC board’s ability to conduct the heat away but looking at the thermal resistance junction-to-case, which is getting the heat out in every other direction, the GaN devices have a 6× advantage over silicon, he added.

“What that means is if you have any kind of topside cooling whether it be blowing air across the device or more sophisticated thermal management on the top, you can get a lot higher power density. As a matter of fact you can get about a 10× lower thermal resistance for the same die size out of a GaN device – from EPC chipscale devices – then you can get out of a silicon device.”

This translates into an order of magnitude better thermal opportunity with a GaN device if it is managed well, he said.

Another big opportunity is integration.

“GaN itself is semi-insulating so devices made on the surface don’t talk to each other unless you connect them electrically, so you can make low-power devices and low-voltage devices on the same chip as a high-power, high-voltage device,” said Lidow. “All you have to do is shrink the dimensions. Shrink that source-to-gate dimension and you get a lower voltage device; shrink all the dimensions and you get a lower power device.”

An additional power device can be added to create a monolithic half bridge, which is extremely difficult to do in silicon above 15-20 volts, according to Lidow. Since GaN is not voltage-sensitive, you can have a completely monolithic power stage with high-side and low-side devices as well as a signal-level device on the high side and a level-shifting device to communicate between the top and the bottom levels, he said.

“You can go from here adding sensors and control and make a complete system-on-chip, which EPC has been doing for several years,” said Lidow. “We introduced our first monolithic half bridges seven years ago and we have devices that have drivers on the same chip as FETs and a fully monolithic power stage which has all sorts of sensors, drivers, level shifters, and logic, which is very popular now in DC/DC converters, robotics, and e-mobility applications.”

He also covers the progress that GaN technology has made since 2011 up to 2021 (see chart below).

GaN is replacing silicon and it’s not going to slow down thanks to improvements in EMI, efficiency, cost, size, and integration, said Lidow.

Next up for EPC will be new power chipsets, which will be introduced in about two weeks. The first devices will be a 65-amp power-stage chipset including crossover protection, sensing, and logic. “These will be our first devices in packaged form,” he said.

Key highlights in the EPC roadmap show Gen 6 devices, which are expected to be released in the fourth quarter of 2022 and multi-channel high-side devices (full system-on-chip devices) in 2024.

Fundamentals of SiC

Here are highlights from Veliadis’ presentation. He covers the benefits of WBG devices as well as the challenges and opportunities in SiC.

Veliadis pointed out that both SiC and GaN devices allow for more efficient and novel power electronics thanks to their higher thermal conductivity, energy gap (eV), and critical electric field.

The large bandgap and critical electric field allow for high-voltage devices with thinner layers, resulting in lower resistance and associated conduction losses, low leakage, and robust high-temperature operation, said Veliadis. In addition, the thinner layers and low specific on-resistance allow for smaller form factors that reduce capacitance, enabling higher frequency operation and smaller passive components, and the large thermal conductivity allows for high-power operation with simplified thermal management, he said.

“Those are big advantages for SiC and GaN and have led to the massive growth that we’re seeing today,” he said.

Veliadis also discussed some of the key applications for SiC, including automotive, information technology, grid infrastructure, electric motor drives, and aerospace.

The number application for SiC is the automotive sector with massive growth in electric vehicles (EVs), he said. “High-voltage SiC will solve one of the biggest problems with EVs, which is the ability to charge the vehicle in the time that it takes to do a similar thing at a gas station.”

In data centers heat management is a big issue, so any technology that provides more efficiency like SiC and GaN will go a long way towards improving this application, he added. Other big applications are renewables and electric motor drives, which consume about 50 to 60 percent of the world’s electricity.

Last but not least is the aerospace sector that is moving towards an all-electric aircraft with the goal of achieving better efficiency as well as lower noise and emissions, he said.

“Today’s power electronics engineers have a number of choices; they can use Si, SiC or GaN in their applications,” said Veliadis. “The question is ‘how do you select what type of technology to use in your application’ and the answer is that you have to take into consideration your voltage requirements, the current levels, frequency requirements, the efficiency needed in your application, temperature requirements, and of course, cost considerations.”

The Infineon chart below shows the areas where the different technologies offer the biggest advantages.

“At lower frequencies and very high power, silicon is the strongest contender and as you increase the frequency that’s where silicon becomes more lossy so it’s no longer a good solution and it’s here where SiC becomes the best solution, and of course, at very high frequencies GaN is an excellent solution,” he said. “Silicon is extremely competitive from 15 up to 650 V and GaN is very competitive from 100 to 650 V. [He also noted that one GaN company has a 900-V device.] SiC is extremely competitive at high voltages like 1200 V and 1700 V and 3.3-kV devices have been demonstrated and are close to being released as product by several vendors along with 6.5 kV and 10 kV.”

But Veliadis said a big battleground market is in the 650-V range where Si, GaN, and SiC all compete. “Silicon is very reliable, very rugged, and it’s cheap and capable of high current, while GaN offers very high efficiency at a very reasonable cost. GaN also is a CMOS compatible device so it takes advantage of the economies of scale in silicon and is manufactured in large fabs. SiC is very efficient and operates at high current and high frequencies.”

Veliadis also dives into the differences between planar and Trench SiC MOSFETs, which he calls the workhorse of power electronics. The discussion touches on mobility, electric field, drift layer, on-resistance, blocking-voltage capability, and breakdown field performance.

He also covers several wafer fabrication issues in detail and the road to mass commercialization, providing tips on how to solve these challenges. The overall challenge is the SiC fabrication processes require investment in select tools and development of non-CMOS compatible processes.

“SiC fabrication much like GaN is very important,” said Veliadis. “In the case of SiC, you need to buy specific tools and develop specific processes that are not silicon CMOS compatible. We’ve taken advantage of all of the mature Si processes, which have  been successfully transferred to SiC to take advantage of the economies of scale.”

However, SiC material properties require specific processes to be developed, he said.

Some of the challenges include etching – wet etch is not practical with SiC because of the high temperatures; substrate thinning for lower resistance; doping – conventional thermal diffusion is also not practical with SiC, requiring new processes for implantation and implantation anneal and flattening the wafers to alleviate the impact of high-temp anneal; achieving a good ohmic contact formation, and selecting CTE-matched metals.

Other challenges covered include gate oxides, transparent wafers, the lack of flatness in SiC wafers, and insulation dielectrics. Ultimately, dealing with these challenges is expected to yield better substrates, higher reliability, fewer defects, improved ruggedness, and lower costs.

Being able to improve these processes is needed especially as the industry moves to 200-mm wafers that are expected to reduce SiC costs by about 20 percent or more, according to Veliadis.

He estimates that it will take an investment of about $10-$15 million to enable a silicon foundry to process SiC wafers.

In addition to the manufacturing challenges, he also points out supply-chain issues. “With respect to the silicon world using a SiC foundry you’re competing both on the process and the design, and the second issue is that the wafer capacity supply chain is a concern for SiC as it is growing very fast and all companies are looking for ways to secure low-cost, high-quality substrate wafer sources.

Please visit PowerUP’s conference schedule and register or log-in to access the on-demand presentations.