SEPT. 1, 1997--As computer makers scramble to marry high-speed optical technologies with conventional chips made of silicon--the cheap, sand-type material incapable of 'shining'--University of Delaware researchers today reported a silicon-based device that converts some light into electricity.
The key, UD researchers explain in the September 1997 issue of IEEE Electron Device Letters, seems to be carbon, the same element responsible for diamond, graphite and coal. Germanium, coupled with a tiny bit of carbon and mounted on a silicon substrate, exhibits surprising optical responses to laser light, says Paul R. Berger, an associate professor of electrical and computer engineering and one of the few promising young investigators nationwide to receive a 1996 National Science Foundation CAREER award.
UD's simple prototype device "demonstrates the feasibility of using a conventional silicon platform, enhanced by a germanium-carbon alloy, to exploit the power of light," Berger explains.
Germanium-carbon alloys "will be vital to the future of silicon integrated circuits," adds Berger's colleague, James Kolodzey, a professor of electrical engineering. "In the near term, our research might lay a foundation for improved radio-frequency wireless communication devices such as cellular telephones, as well as solar energy conversion."
Silicon-based, germanium-carbon devices could "bridge the gap" between today's silicon computer chips and next-generation microprocessors based on gallium-arsenide alloys, Berger says. As silicon chips approach speeds of 2 gigahertz, he explains, their performance wanes. Gallium-arsenide devices promise faster speeds--up to "tens of Gigahertz," according to Berger. But, a single, two-inch gallium-arsenide wafer might cost $100, compared to only $10 for a silicon wafer the size of a dinner plate. Germanium-carbon alloys, offering speeds "in the low gigahertz range," cost less than gallium-arsenide materials, Berger says.
And, he adds, the photo-optic properties of UD's prototype device suggest a way to create faster silicon-germanium-carbon chips--by exploiting optical technologies.
To prove this point, Berger and Kolodzey teamed up with students Xiaoping Shao, Sean L. Rommel and Bradley A. Orner to grow a single epilayer of germanium, infused with a very small amount (0.2 percent) of carbon, on a silicon substrate. The researchers "spray-painted" the alloy onto silicon, using a technique known as molecular beam epitaxy (MBE). Germanium and carbon were loaded into separate "ovens," or effusion cells operating at high temperatures (greater than 600-700° Celsius), where they emitted a vapor that stuck to a heated silicon substrate loaded inside a high-vacuum chamber.
The resulting p-n diode--a device that starts and stops a current moving in one direction--effectively converted 1.4 percent of incoming laser light into electricity, Berger says. (Researchers used laser light operating at 1.3 microns, a wavelength undetectible to silicon, but compatible with current fiber optics.) The diode also "rectified," or controlled the flow of electricity even when subjected to an abrupt 80-volt reverse bias, he notes. "That means it could withstand a lightning-storm type power spike with no interrupution in the flow," Berger explains.
A 1.4 percent light-to-electricity conversion rate "is not stellar," Berger concedes. And, the prototypical diode included molecular flaws. But, "this is the first proof of this principle," he says. "It shows that, even when we cut corners to develop this crude, stripped down device, we still got pretty good light-conversion and rectification numbers--and we did it all on a silicon platform."
Moreover, he adds, the UD device demonstrated a relatively efficient conversion rate, given the fact that light was transformed through direct contact with an ultra-thin (50 angstroms) active region of only 20 atomic layers, under "surface-normal" conditions--and not through a concentrating waveguide.