Ironing Out A Vexing Chipmaking Problem

WHEN SEMICONDUCTOR MAKERS WANT TO COMBINE TWO MATERIALS TO PRODUCE SOMETHING that can't be made from a single material, such as ultraviolet light-emitting diodes (LEDs), they have a limited menu of choices. That's because of the so-called lattice mismatch--the difference in spacing between the atoms in various chip materials. Even tiny, 1% mismatches rapidly multiply into strains that result in material flaws--and defective LEDs. For many combinations, the mismatch factor climbs to 20% or more.

Now, researchers at Cornell University seem to have found a way to do the impossible. And it's so simple that materials scientists are dumbfounded. "Even we don't understand why it works so well," admits Felix E. Ejeckam, a Ph.D. student whose computer simulations indicated that the "too-easy" idea from Yu-Hwa Lo, a Cornell associate professor of electrical engineering, just might work.

It boils down to this: Take a very thin film of gallium arsenide and place it on a gallium arsenide wafer. Rotate the film about 10 degrees out of lattice alignment, then iron it down with heat and pressure. Now, a badly mismatched material can be grown atop this buffer layer, and lattice-mismatch strains are somehow passed through the buffer into the bulk substrate, where they cause no harm. Chipmakers are excited by the technique because it also promises a new silicon foundation for computer chips.

Before it's here, it's on the Bloomberg Terminal.