Gallium Oxide: Power Electronics’ Cool New Flavor

With improved heat conductivity, this semiconductor could leapfrog other challengers to silicon

Source: Yole Développement, Lyon, France
Semiconductors in Comparison: Flosfia’s gallium oxide devices exploit the material’s big bandgap and breakdown field while making up for its poor thermal conductivity. (Values are normalized to silicon’s [blue].)

Ideally, the electronic components that route electricity through power supplies, inverters, and electric motors are cheap, efficient, and capable of handling high voltages. Judged in these terms, gallium oxide could be the best material yet, according to recent work by ­Flosfia, a startup in Kyoto.

That’s because silicon—the incumbent material for making diodes and transistors for the power electronics market—is cheap but not very efficient. And although this weakness is addressed by devices made from silicon carbide and gallium nitride, both have had limited commercial success due to high prices. Flosfia’s diodes are already performing more efficiently than those made from SiC and GaN.

The superiority of these gallium oxide devices stems from the material’s approximately 5–electron-volt bandgap—way higher than that of gallium nitride (about 3.4 eV) or silicon carbide (about 3.3 eV). Bandgap is a measure of the energy required to kick an electron into a conducting state. A bigger bandgap enables a material to withstand a stronger electric field, making it possible to use a thinner device for a given voltage. That’s a big deal because the thinner the device, the lower its resistance, and thus the more efficient it is.

Gallium oxide devices do not excel in all areas. Their Achilles’ heel is poor thermal conductivity. “When you make a high-power device, you need to have a good thermal conductivity to extract the heat out of the device,” explains Hong Lin, senior market and technology analyst at Yole Développement in Lyon, France.

Flosfia’s engineers have improved the device architecture to address that very issue. In particular, they found a way to make the diode chip thinner, according to Naonori Kurokawa, a partner at University of Tokyo Edge Capital, a Flosfia investor. The key is to grow the gallium oxide crystal on a sapphire substrate.

Flosfia chief technology officer Masaya Oda found that “the gallium oxide epilayer can easily lift off the sapphire substrate,” says Kurokawa. Separating the device from its foundation allows the chip to be bonded to a highly conductive, heat-­sucking material, enabling operation at a lower temperature.

Using sapphire makes a lot of sense. “It’s supercheap and already marketed, because of light-emitting-diode manufacture,” says Kurokawa. The only downside is that the crystal quality is not as high as it would be using a gallium oxide substrate, which at the moment is quite expensive.

The process Flosfia uses for making gallium oxide devices was invented by company cofounder and Kyoto University professor Shozuo Fujita. In it, the sapphire substrate is heated and a fine mist of particles is swept into the chamber on a gust of non­reactive “carrier gas.” The mist, which contains metal compounds, decomposes when it hits the hot substrate and forms a film of gallium oxide. The whole process can be cycled through rapidly because, unlike with other methods, the chamber never has to be completely evacuated. And that drives down costs.

Engineers from Flosfia detailed the results of diodes made with this growth process in the February 2016 edition of Applied Physics Express. One device combines a 531-volt breakdown voltage—the potential needed to reverse the flow of current—with an on-resistance of 0.1 milliohm per square centimeter, exceeding the limits of what is possible with silicon carbide.

The highest breakdown voltage Flosfia reported is 855 V. This is not that high for a wide-bandgap diode—devices made from SiC can handle 10 kilovolts or more.

Kurokawa explains that the diodes have modest breakdown voltages because they are bare chips. Introducing insulating layers into this very simple device should lead to significant improvement, he says.

Later this year Flosfia will start to provide samples of its diodes to potential customers. Further ahead, plans include a ramp-up of diode production in 2018 and the development and launch of accompanying transistors.

However, it is not yet clear whether this will ignite a gallium oxide power electronics industry. “Today we are comparing silicon with silicon carbide,” says Lin, who expects gallium oxide devices to undergo a similar evaluation only once they are commercialized and considered as worthy contenders. That is still some way off.