12 October 2007--A new theoretical model, showing that metals bend differently than previously thought, suggests that engineers and materials scientists may encounter serious problems as they try to make the wires connecting computer chips ever slighter.
Professor Ferenc Csikor of the department of materials physics at Eötvös Loránd University in Budapest and his colleagues found that metals yield to pressure by a series of random events called dislocation avalanches. In a report published today in the journal Science the group described a model of a microscopic stretch of aluminum wire that shows what is really happening when a wire deforms under pressure. As a wire curls from a straight line into a loop, it yields to the pressure only in certain, random regions.
That randomness could pose a serious challenge for constructing nanometer-scale devices, says James Sethna, a professor of physics at Cornell University, in Ithaca, N.Y., and the author of an accompanying article in this week's Science .
To make the wires connecting transistors on a chip, microprocessor builders deposit atoms of metal to fill holes and channels in a silicon-based wafer. Nothing gets bent. But the bond wires that connect chips to their packages do bend, notes Vladimir Stojanovic, a computer scientist at MIT. According to Stojanovic, bond wires are typically about 100 micrometers thick and probably won't get much smaller. But if they do, avalanche dislocations could become a problem.
Such avalanches occur in many systems--magnets as they respond to fluctuations in temperature, for example, or tectonic plates as they slide past each other. Some are two-dimensional and others three-dimensional. Until now, physicists have largely avoided the problem of avalanches in 3-D objects, because the problem is so complex, Cornell's Sethna says. Simplified descriptions of a bending piece of metal tend to assume that a small layer of atoms on the surface of the metal will yield to incoming pressure smoothly. But Csikor reports a much jerkier process whereby the crystal lattice of a metal deforms throughout the entire width of the wire at points where the strain becomes too intense and the structure fails. The result is an irregular pattern of kinks that tangle up the crystal lattice ”like spaghetti,” Sethna explains.
By studying what happens to the lattice structure of a wire when it bends, engineers might come to depend less on trial and error to develop new materials, he says. If engineers can come to understand the inherent randomness of dislocation avalanches, he adds, they might find a way to control them, instead of blindly experimenting with metals, trying to get the wire to come out strong and pliable.
”I don't know of any examples where people have tried to do anything and failed because of these avalanches,” he says, but ”ignoring the avalanches surely is a terrible idea.”