Motorolas new way of bonding light-emitting compounds onto a silicon chip may do more to deliver smaller, cheaper and faster semiconductors than anything announced by chip makers this year, analysts said.
“This could go down in history as a major turning point for the semiconductor industry,” said Steve Cullen, Cahners In-Stat Groups principal semiconductor analyst.
Physicists at Motorolas research lab in Tempe, Ariz., attacked a problem that had confounded chip makers for 30 years: how to grow light-emitting semiconductors on a silicon wafer to combine the best qualities of each on a single chip.
Having to use separate silicon integrated circuits and photon chips in a computers microprocessor slows the speed to about one-tenth of its potential capacity, said Jim Prendergast, director of Motorolas physical sciences research lab. “A problem with any silicon technology is getting data in and out of a chip. Once it leaves the chip, the speed of a typical computer is less than 200 megahertz, even though it might be working with a 2-gigahertz microprocessor,” he said.
Until now, scientists have been unable to bond semiconductors that have light-emitting properties to silicon integrated circuits.
But Jamal Ramdani, a Motorola scientist, thought of creating an intermediate layer with properties that bonded well with both silicon and light emitters. A large team of colleagues helped him perfect the recipe.
The team used strontium titinate to bond silicon with light-emitting gallium arsenide, Prendergast said. Now Motorola is perfecting a different intermediate layer to bond indium phosphide, a compound that has more potential uses for fiber-optic communications.
The technology will accelerate the development of optical fiber to the home, streaming video on cell phones and systems that help automobiles avoid collisions, the company said. But the crowd-friendly breakthroughs are a couple of years away.
Computing and communications have always been separated because silicon is in the guts of a PC, while light-emitting compounds are at the core of optical networks.
“This says that theyre not so separate after all, if you can get everything on one chip,” Cullen said. “The history of anything in electronics teaches that the more stuff you can cram on one chip, the better the performance and the cheaper and more pervasive it becomes.”