By Jeff Bairstow
It's quite amazing how "old" technologies seem to get a new lease on life when someone takes a second look. Take for example, silicon germanium (SiGe), a technology that's been around almost since semiconductors were discovered. When IBM scientist Bernard Meyerson dropped a piece of silicon almost two decades ago, little did he know he was eventually going to bring SiGe very much into back into the mainstream of semiconductor development.
While cleaning that dropped piece of silicon, Meyerson noticed it reacted much differently than it should, according to descriptions in the standard textbooks. His subsequent experiments proved that assumptions about silicon held for more than 30 years were incorrect. Meyerson's resulting work in SiGe technology has given new life to an "old" chip technology.
Meyerson's research in growing crystal silicon layers at very low temperatures opened a new door in the quest for faster and more efficient chips. By shortening the chip development process and replacing more expensive and exotic materials with SiGe, Meyerson's team of IBM researchers created smaller and faster chips at a lower price.
In a heavily attended session at the 2002 Compound Semiconductor Outlook Conference in San Mateo, CA, Meyerson, now vice president of the IBM Communications Research and Development Center, described his team's work in developing what he termed "the world's fastest" semiconductor circuits using SiGe. Actually, IBM had disclosed its SiGe research a year earlier with the announcement of "the world's fastest transistor" that ran at the hair-raising speed of 210 GHz. That's right, over 200 GHz. And you thought a 2-GHz personal computer was fast. Such 200-GHz devices are more than fast enough for 40-Gbit/s networks.
Not only are these SiGe devices ultrafast, but they have the great virtue of being manufactured by the widely used and well-understood CMOS (complementary metal oxide semiconductor) process. The CMOS process is far less expensive and produces higher yields than the conventional methods of fabricating the more exotic compound semiconductors that can be tricky to make.
Of course, the world's fastest transistor isn't of much practical value unless it can be commercially fabricated into working integrated circuits. So Meyerson's group built a demonstration ring oscillator that operates at speeds over 110 GHz, thoroughly trouncing the competing compound semiconductors such as indium phosphide and gallium arsenide. "A ring oscillator is the 'gold standard' circuit used to demonstrate performance," said Meyerson.
Noted Meyerson, "People have continually predicted the death of silicon technology—that it will somehow magically run out of speed. However, in recent years, the entire semiconductor industry has gone from 'no, it won't work,' to 'no, we don't need it,' to 'we better do it yesterday.'" IBM has the benefit of many years of research into the properties of SiGe.
IBM is unlikely to have the SiGe field to itself, of course. Conexant Systems, of Newport Beach, CA, has been hard at work and claims to have devices that have exceeded 150 GHz. Japan's Hitachi is also working hard on SiGe devices and processes.
And commercial devices are already in the pipeline. IBM has indicated that the first chips based on the Meyerson group's work may be available later this year. IBM has said that it is working with "early access customers," such as Sierra Monolithics of Redondo Beach, CA, to develop wireless communications applications.
So the next time you hear someone describe a technology as "worked out" or "finished," remind the doomsayer of the story of silicon germanium. Just how many times have you heard doleful predictions of the demise of optical lithography? Don't bet on it.
Jeff Bairstow
ATD Online Editorial Director
[email protected]