Theis outlined the many nanotechnology efforts currently under way and described how carbon nanotubes and various self-assembling, human-guided molecular structures will likely replace the silicon transistor starting around 10 years from now.

The Dennard scaling rule for transistors, established in 1976, is becoming "less and less of a guideline as transistors scale down," creating a need for a new evolutionary step, Theis said.

Theis showed a baby step toward this evolution from IBM's labs: a double-gate transistor with a 60-nanometer gate length (estimated 30-nm L-effective) and a 0.92-picosecond gate delay. IBM will show the transistor at the International Electron Devices meeting next month in Washington.

"It still needs to go further, but it is an interesting use of drain electrodes to reduce the parasitic resistance in this device," said Theis, who predicted such devices could become common by 2010.

But the real step toward replacing the transistor, he said, will likely come from carbon nanotubes and variations of self-assembly. "Is that the smallest logical switch nature will allow?" Theis asked. "I don't see any reason why it should be. IBM is looking for alternatives."

Researchers are excited about carbon nanotubes as the first replacement for silicon transistors, the keynoter said. "Expose hydrocarbon gas to the right temperatures with the right catalyst and you will form these structures," said Theis, pointing to a small structure that looked like a bail of chicken wire 4 nm wide. "If you give that chicken wire a twist and join the ends, you have a semiconductor."

Theis showed a photograph of IBM researchers using carbon nanotubes to create the "first logic circuit fabricated from a molecular-scale device."

But carbon nanotubes are merely one of many areas of nanotechnology research that show promise for replacing the silicon transistor, Theis said. Other research is examining self-assembly, basically tailoring or guiding atoms to grow into structures that can be used in semiconductors.

Theis reviewed several forms of self-assembly currently being researched, including self-assembling nanocrystals, a self-assembling cubic superlattice, self-assembling silicon germanium quantum dots and a self-assembling one-layer-thick pentacene snowflake.

Though these areas are in their infancy, they show promise because they prove that atom structures can be encouraged to grow in particular configurations, which could be the replacement for the silicon transistor, Theis said.

Over the next decade, increased research in self-assembly will see such structures eventually show up in niche applications, Theis said.

He predicted that in 10 years chemically synthesized nano-building blocks will replace semiconductor logic and memory devices, resulting in an increased need to emphasize redundancy, test and repair and self-repair.

And in 20 to 50 years, he said, we should see pervasive use of self-assembly. "At that point all bets are off, we have no clue what we will need to design and test such devices," said Theis, pointing out that what the semiconductor industry has accomplished today would have been unthinkable in 1960.