TOKYO Building on work to promote millimeter-wave communications in Japan, researchers here have built what they call the world's fastest high electron mobility transistor (HEMT), a device that runs at a current gain cutoff frequency of 472 GHz.
The previous HEMT world record, 398 GHz, was set at the Communications Research Laboratory (CRL) this spring. Now, working with Fujitsu Laboratories and Osaka University's Graduate School of Engineering Science, CRL has souped up the device's performance, mainly by reducing the distance between the gate and channel. The performance gain was accomplished by adopting a thinner supply layer, from 16 to 12 nanometers, said the lab's millimeter-wave devices group researcher Keisuke Shinohara.
Several other important tricks also contributed to the breakthrough. Shinohara said the lab found it could halve parasitic resistance by reducing the process temperature to below 200C. At 300, the source and drain metal tends to sink slightly into the crystal. Meanwhile, indium-phosphide materials tend to release stray fluorine atoms into the crystal, increasing source resistance when annealed at more than 250.
"So we decreased the temperature throughout the whole process and we think we have got the parasitic resistance as small as possible," he said.
Modifying the gate structure was also crucial, said Shinohara. T-gates shorter than 50 nm are tough to make because the gate's foot becomes extremely thin. Shinohara added a process step, now under patent application, to thicken the neck of the gate, making it more robust.
Earlier this year, Fujitsu Labs Fellow Takashi Mimura, who invented the HEMT in 1979, told EE Times he was pessimistic about HEMTs breaking the 500-GHz barrier until gate lengths came down to a difficult-to-manufacture 10 nm.
Very high-frequency HEMTs face two huge performance barriers: electron speed and the distance the electrons must travel. CRL's work on indium gallium arsenide with a 53 percent indium content for the free-electron channel layer structure now has a proven performance track record. To overcome the second barrier, the group has developed a simple liftoff technology using electron-beam lithography to form sub-50-nm gates.
Shinohara's latest record-breaking HEMT actually uses a 30-nm gate, longer than its predecessor, said Mimura. But Mimura said Shinohara's new gate process will make 15-nm gates feasible Combined with Shinohara's other improvements, that could open up the potential to boost performance nearer to 600-GHz frequencies.
"It was a surprise for me, too," said Mimura. "We will now go to shortening the gate lengths and we are now aiming at beating 500 GHz, but we don't know about 600 GHz yet."
Despite vast virgin bandwidth available in the millimeter band, millimeter waves suffer rain attenuation problems and transmitters and receivers need wave guides to handle the very high power, narrow beams.
Device performance has always been a problem, with an indium-phosphide substrate essential to achieve the gain required. A 400-GHz-capable HEMT at 60 GHz, for example, achieves about double the gain of the most powerful 150-GHz gallium-arsenide models. Redeveloped versions could cut four to eight amplification stages, shrinking chip sizes and costs dramatically.
Toshiaki Matsui, leader of the CRL's millimeter devices group, said his research received a boost this May at the International Microwave Symposium when the U.S. Federal Communications Commission proposed the 90-GHz band for future potential consumer use. "This indium-phosphide HEMT is the most suitable for high-performance transmitters and receivers," he said. "I feel that we ought to be able to make some good equipment at this band to demonstrate it working for phone links. We need to show excellence in the prototypes and once we show our high quality equipment, then some sort of technology firing will take place."
Apart from Cruise missile radars and other military applications, the millimeter band is largely restricted to such scientific applications as radio telescopes. The research lab is making an interbuilding, millimeter-band-based LAN, but ambitious plans to develop a 4G national communications infrastructure partially relying on the millimeter band remain stalled as the Japanese government deals with the nation's fourth recession in a decade.
Still, Mimura said, the device improvements being perfected at CRL will end up in scientific applications and then, hopefully, there will be a push to develop other essential system-on-chip and antenna technologies.
"We need to make a marketing study and search out the real applications," said Mimura.
"HEMTs of this sort of power will initially be used only in very small markets, for example in radio telescopes. But, technologically speaking, there is a lot of demand for higher performance," Mimura said.