Speed demon gets hooked on silicon.

Industrial scientists have devised a way to coat wafers of silicon, the stuff of the microelectronics revolution, with a high-performance semiconductor whose wider use could be a boon to many areas of electronics. Mating silicon to gallium arsenide, which currently shows up in special applications, which has been a technological goal for more than 30 years.

If the fabrication advance announced this month by Motorola in Schaumburg, Ill., works on a commercial scale, fast low-power chips may become less expensive and more common, semiconductor specialists say. The same may prove true of chips hosting solid-state lasers and other optical components.

Those changes could benefit consumers by shrinking the size and cost of cell phones, according to the technique's developers. The new process may also make more affordable such technologies as collision-avoidance systems for vehicles and fiber-optic telecommunications to homes.

Wafers of gallium arsenide and other semiconductors that can transmit electrons at high speeds and efficiently emit or detect light are costly and fragile. Previous efforts to graft such compounds onto a cheap, flexible silicon base failed because their crystalline structures don't align with that of silicon. In gallium arsenide, the atoms are 4 percent further apart than atoms in a silicon wafer are. Although researchers have grown gallium arsenide on silicon in the past, the resulting crimping and strain in the compound's crystal lattice made it unsuitable for optical electronic components.

To get around the crystal mismatch, the Motorola researchers say they have squeezed two other materials between the silicon and gallium arsenide. Because atoms shift around in those interlayers, the gallium arsenide bonds to the intermediate layer beneath it without warping. "That's the magic of it," says Jamal Ramdani of Motorola's Physical Sciences Research Laboratories in Tempe, Ariz.

After 2 years of secret development, Motorola has filed more than 270 patent applications based on the technique. Semiconductor specialists outside the company agree that successfully mating the two materials promises to be important. However, until the Motorola work appears in the scientific literature, those specialists also are greeting the announcement with caution. On Tuesday, William Ooms, a research manager at Motorola, described the technique at a materials science workshop in Chattanooga, Tenn.

"If this is real, it's really important," comments Joe C. Campbell of the University of Texas in Austin. "But there have been some premature claims before."

The recent advance arose from an earlier setback. Ramdani and other researchers were working toward tiny transistors free of the current leaks that plague such components (SN: 3/25/00, p. 204). In 1999, the researchers made what the company declared to be "the world's thinnest functional transistor" by growing a scanty layer of the ceramic material strontium titanate (STO) on silicon. Yet when they tried to pare that layer further, oxygen corrupted the underlying silicon, creating a spongy layer of silicon dioxide beneath the STO.

At first, Ramdani considered the new layer an obstacle. But he says that while vacationing on a beach, he realized that the layer might be a godsend. He had noticed that it enabled the STO crystal to relax despite a 2 percent mismatch with the silicon. Ramdani began wondering whether the STO would serve the same function for gallium arsenide.

Initial experiments confirmed his hunch. In the 2 years since, the roughly 100 researchers assigned to the project have surmounted many hurdles, including getting gallium arsenide to bond to the entire surface of an underlying layer of STO.

These investigators have also made working cell-phone amplifiers and an infrared-light-emitting device using the new wafers. What's more, they've made wafers with 300-millimeter diameters--twice that of the largest wafers of gallium arsenide alone. The amplifiers will hit the market within months, the company predicts.

The gallium arsenide feat may be just a beginning. Different interlayers should allow other semiconductors, such as indium phosphide, to grow on silicon. Whereas gallium arsenide transistors are about 10 times as fast as those of silicon, indium phosphide transistors up the pace by another factor of 10.

"If they can do it with gallium arsenide, they can do it with other materials," surmises Vincent P. LaBella of the University of Arkansas in Fayetteville.