Searching For The Next Power Transistor

For decades, the industry has relied on various power semiconductors to control and convert electrical power in an efficient manner. Power semis are ubiquitous, as they are found in adapters, appliances, cars, elevators, switching power supplies, power grids and other systems.

But today’s silicon-based power semiconductor transistor technologies, such as IGBTs, MOSFETs and thyristors, are edging closer towards their physical limits. So over the years, the industry has developed faster and more efficient power chips based on wide-bandgap technologies, namely silicon carbide (SiC) and gallium nitride (GaN) on silicon.

Still, today’s wide-bandgap devices also suffer from various issues, prompting the need for a new and disruptive technology. In fact, there is a growing wave of R&D taking place in the development of next-generation power transistors. The next-generation candidates include bulk vertical GaN, diamond FETs, newfangled SiC and others.

“If successfully developed, these devices could offer a pathway to functional cost parity with silicon-based power devices at higher power levels,” said Tim Heidel, program director at the Advanced Research Projects Agency-Energy (ARPA-E), part of the U.S. Department of Energy. ARPA-E, which focuses on early-stage technologies, recently announced a program aimed at developing next-generation devices and materials for 1,200 volt and higher applications. The program is called SWITCHES, which is short for “Strategies for Wide-Bandgap, Inexpensive Transistors for Controlling High-Efficiency Systems.”

The current generation of GaN-on-silicon power chips and SiC FETs are shipping today, but it could take years, if not decades, before the next-generation technologies ship in volumes. “GaN had been successful in the LED space, but there are some issues with the substrate quality,” said Adam Brand, senior director of the transistor technology group at Applied Materials. “Diamond has a high thermal conductivity, but it’s also expensive. It’s far away in terms of practicality.”

Power semi mania
Besides the ARPA-E program, many other companies, research institutes and universities are developing next-generation power semiconductor technologies and for good reason. Worldwide demand for electricity is expected to increase by 87% from 2010 to 2040, according to Exxon Mobil. But in the transmission and distribution of electricity, the losses can range from 8% to 15%, according to estimates.

So, there is a pressing need for faster and more efficient systems and chips. Today’s power electronics systems range from a few watts to multiple megawatts. In a system, there are various components and transistor types.

At the low end, the most common silicon-based transistor type is the power MOSFET. A souped-up version, the super-junction MOSFET, is a vertical device that tops out at around 900 volts.

The leading mid-range power semiconductor device is the insulated-gate bipolar transistor (IGBT), which is a three-terminal device that combines the characteristics of MOSFETs and bipolar transistors. And at the high end, the industry uses thyristors, which are wafer-level, solid-state devices.

“The incumbent technologies are mostly silicon. In the market, there are also various voltage nodes. For example, there is the 600 volt node. That voltage node is probably well served by today’s silicon-based super-junction MOSFETs. Your laptop charger or other things you plug into the wall probably uses a super-junction MOSFET,” said Isik Kizilyalli, chief technology officer at Avogy, a startup that is developing bulk vertical GaN transistors. Avogy has obtained funding from Intel and others.

“At 1,200 volts and higher, people use IGBTs. But these are slow,” Kizilyalli said. “SiC devices are beginning to penetrate the market, more at 1,200 volts. So that will be the next wave of high voltage devices. GaN is going to be the next wave.”

In theory, wide-bandgap chips like GaN and SiC are smaller, faster and more efficient than silicon. They operate at higher temperatures, frequencies and voltages, thereby helping to eliminate up to 90% of the power losses in electricity conversion. Wide-bandgap refers to higher voltage electronic band gaps, which are larger than one electronvolt (eV).

SiC vs. GaN vs. diamond
Today, several companies are shipping SiC FETs, which are targeted for 600-, 1,200- and 1,700-volt applications. Based on silicon and carbon, SiC has a bandgap of 3.3 eV. Silicon has a bandgap of 1.1 eV. On the downside, SiC devices are made on 100mm or 150mm substrates, making the wafer costs somewhat expensive. SiC FETs also suffer from low-effective channel mobility.

In R&D, however, Cree has recently devised SiC FETs that could one day replace silicon-based IGBTs and thyristors. The unipolar SiC FETs have blocking voltages up to 15-kV. IGBTs and thyristors have blocking voltages up to 8-kV.

At 15-Kv, unipolar devices hit the wall, prompting the need for a bipolar technology. In R&D, Cree has also devised a combination SiC n-IGBT device with a 27-kV blocking voltage. This is the world’s highest voltage for a semiconductor device reported to date, according to Cree.

Though still in R&D, these devices demonstrate the capabilities of SiC in high-power applications, said John Palmour, chief technology officer for power and RF at Cree. “The unipolar SiC MOSFETs show a nearly 30x reduction in switching losses over the 6.5-kV silicon IGBTs,” he said.

Another technology, GaN, has been touted as the next big thing in power and RF. GaN has a bandgap of 3.4 eV. Today’s GaN-on-silicon devices are lateral structures, meaning the current flows from the source to the drain on the surface.

GaN-on-silicon devices suffer from various issues, however. “It’s not latticed matched,” said Avogy’s Kizilyalli. “But people have been able to make good RF devices on it. A lot of people are also trying to do GaN-on-silicon for power electronics at 200 volts and 600 volts. But there are questions about the reliability. There are questions about scalability.”

In fact, lateral GaN-on-silicon devices could hit the wall at 600 volts, prompting the need for a next-generation technology, namely bulk vertical GaN transistors. In vertical GaN devices, the electrons flow from the top to the bottom. “We don’t suffer the issues of lattice mismatch, because we are growing GaN-on-GaN,” Kizilyalli said. “If you go vertical, you can make devices at 1,200 and 1,700 volts. We can make this as large as 3.7-kV. That’s very hard, and not straightforward, in a lateral configuration.”

Using 2-inch wafers and an MOCVD process, Avogy is sampling a limited amount of devices. “We’ve experimented with 4-inch. There are multiple suppliers of GaN substrates. Those are expensive, but the prices are coming down and the quality is improving over time,” he said. “Overall, vertical GaN will change the power electronics industry. It’s not going to happen next year. A lot of fundamental problems need to be solved.”

Bulk GaN substrates are limited to small sizes and are expensive, meaning GaN-on-GaN will play a limited role over the next decade, according to Lux Research. On the other hand, GaN-on-SiC could get some traction, possibly in the transportation market starting in 2017, according to Lux.

Besides SiC and GaN, there is also interest in perhaps the ultimate power device–diamond. Diamond has a wide bandgap (5.45 eV), a high breakdown field (10MV/cm), and high thermal conductivity (22W/cmK). Diamond is a metastable allotrope of carbon. For electronics applications, the industry mainly uses synthetic diamonds, which are grown via a chemical vapor deposition (CVD) process.

It could take years before diamond FETs reach the mainstream. “It’s hard not to be interested in a material with high conductivity,” said Dave Hemker, senior vice president and chief technology officer at Lam Research. “The issues have always been how do you economically grow diamond in a way that is useful in a semiconductor sense.”

Still, researchers are making some breakthroughs in the arena. “The industry is entering a renaissance in diamond electronics,” said Robert Nemanich, a professor in the Department of Physics at Arizona State University (ASU). ASU itself has obtained a grant from the SWITCHES program to develop diamond semiconductor devices.

“The problem is that it has been difficult to dope diamond. But in the last few years, doping has advanced for the n-type. For the p-type, people have used boron as a dopant for many, many years. To get an n-type material, phosphorus has proven to be the dopant,” Nemanich said. “The other thing that was missing was reproducible and high-quality substrates. But right now, you can purchase a relativity inexpensive diamond substrate for a power device. For a 3mm x 3mm substrate or something like that, it’s $100, or maybe $100 to $500, depending on the quality you want.”

Over the years, a number of entities have produced diamond FETs, at least in the lab. For example, Waseda University in Japan, one of the leaders in the area, recently presented a paper on a diamond FET for 1,000-volt applications.

Instead of a traditional boron-doped p-type diamond FET, Waseda makes use of p-channel FETs, based on an H-terminated (C-H) diamond surface. In simple terms, when the hydrogen-terminated diamond is exposed to air, the C-H diamond FET becomes highly conductive at the surface. “C-H diamond FETs show much superior transistor performance, because dense hole accumulation near the surface is effectively modulated by applying gate bias from surface side,” said Hiroshi Kawarada, professor of the School of Science and Engineering at Waseda University.

To make diamond FETS, Waseda grows an undoped diamond layer on a diamond substrate using microwave plasma assisted CVD with a thickness of 0.5μm. A key to the device is the formation of the dense surface holes (2D hole gas, 2DHG) on the top of the FET. Researchers use an atomic layer deposition (ALD) process to apply Al2O3 at 450°C to reproduce 2DHG without adsorbates.

In the SWITCHES program, meanwhile, ASU is working on diamond-based devices. “ASU is concentrating on growing the phosphorus doped layers. Electrical contacts are a little bit tricky, because the diamond material has such a low work function. We are making good progress on it. We are also getting ready to build p-n junctions. After that, we will build a bipolar junction transistor. It will be a vertical configuration,” ASU’s Nemanich said.

Still, the question is clear. Will diamond FETs displace silicon and the other wide-bandgap technologies in the future? “There are some places where SiC and GaN will be the leaders in the future. But there are places where diamond can play a role, particularly in the highest voltage devices,” Nemanich said. “Silicon-based power electronics is also not a static field. Staying ahead of silicon is very difficult. So, it’s hard to say who is going to win at the end.”