In the race to develop faster semiconductors, Melissa Hines thinks the playing field needs to be leveled. So Hines and fellow Cornell University chemists are developing a new manufacturing process for computer chips that would make each microprocessor "perfect," or devoid of surface flaws that downgrade performance.
The Cornell researchers will release their findings about reducing "roughness" on the surface of silicon chips at the upcoming American Chemical Society meeting in Dallas. Surface roughness on the atomic scale greatly decreases the performance of a transistor, and as manufacturers develop smaller devices, roughness becomes a larger problem.
"Flattening the silicon wafers leads to more efficient conductivity and hence more efficient computers," said Hines.
Researchers have been trying to tackle the problem for years. Back in the 1960s, Bell Labs scientists first created a new method of removing dust from the silicon wafers used to produce integrated circuits. The technique, called chemical etching, involved washing the silicon wafers in peroxide baths. But today, the smaller circuitry develops atomic-scale roughness from that very chemical.
Hines, who started her career as a post-doctoral researcher at Bell Labs, found a simple solution to the problem. By changing the acidity and composition of the chemical solution, she was able to make small areas on the silicon surface that were "totally flat," even as far down as the atomic level. The chemicals etch away surface atoms, one atom at a time, in a very precise order. She calls the process "unzipping," because neighboring atoms are etched in a sequential fashion in much the same way that teeth in a zipper are opened in sequence.
The chemical-etching method led to a surface roughness equivalent to one protruding atom out of every 30,000 surface atoms on the silicon wafer.
"Making the surface of a silicon wafer perfectly flat is very important in IC (integrated circuit) technology," said Shri Joshi, a researcher at Marquette University in Milwaukee. "The smallest feature that can be produced on the wafer is a strong function of the surface flatness."
How many years from commercialization is this work?
"This is a harder question to answer," Hines said.
The Cornell research has primarily been aimed at understanding the chemistry of a particular kind of silicon, called Si(111). The "(111)" denotes a specific atomic plane in the silicon. But chipmakers use a slightly different form of silicon for making integrated circuits, so "at least for integrated circuits, our work cannot be directly applied right now," she said. The challenge to Hines and her colleagues is thus twofold: They must first learn how and why the Si(111) chemistry works, and then how to adapt it. "We are very far along on the first step, but we are only beginning to work on the second," Hines said.
The Cornell team is not alone in its efforts to re-etch the surface of silicon. Researchers at Bell Labs, primarily Yves Chabal and Gregg Higashi, have been investigating alternate cleaning solutions. IBM's microelectronic division has researchers who are developing an expertise in "etching technology" as well, says Philip Bergman, a spokesman for the company. And Sematech is working with the University of Wisconsin at Madison to develop chipmaking technologies, said spokesman Brian Mattmiller.
Hines reckons that the technological effects of their research will be incremental, but "the next five years will see tremendous advances in our understanding of surface morphology - both as it applies to etching and to deposition, i.e. thin film growth," she said.
The final hurdle for these new etching techniques will be the cost of incorporating them in chip production plants.
Franco Cerrina, professor of electrical and computer engineering at the University of Wisconsin, estimates that launching any single new technology in chip manufacturing would cost at least $1 billion in research and development.
"Fabricating next-generation transistors with today's processes would be like producing a finely detailed painting with a house-paint brush," Cerrina said. "We need a finer brush for the job."
