The ICCAD panel followed a keynote speech by Thomas Theis, director of physical sciences at IBM Research, on the same topic Monday morning. Theis, also one of the panelists, predicted in his keynote that nanotechnology will replace the silicon transistor and reshape the IC industry.

"We heard about a lot of great technology this morning, but what will it actually mean to us in the IC CAD community?" asked Andreas Kuehlmann, panel moderator and research scientist at Cadence Berkeley Labs. Kuehlmann asked panelists to ponder which applications will make the first use of nanotechnology, how soon nanotechnology will be widely introduced, and how the technology will affect design tools and methodologies.

Searching for a definition, Kuehlmann identified nanotechnology as a "bottom-up" engineering practice in which designers take individual atoms and assemble them into useful structures. "It is multidisciplinary," he said. "Physicists, chemists and biologists are all going to meet in this area."

Theis said that the next 10 years will bring increasing usage of self-assembly, with "organic" electronics — the use of organic molecules as switching devices — established in niche applications. After 10 years, he said, chemically synthesized nanoblocks will replace semiconductor logic and memory devices. And in 20 to 50 years, hierarchical self-organization will allow systems to approach "biological" levels of complexity, he predicted.

Theis said dense, cross-point memories are one likely application for nanotechnology, as are sensors. Further out, "anything we can imagine will be possible," he said. "It's like inventing life all over again."

Philip Keukes, senior computer architect at Hewlett-Packard Labs, confirmed that electronic nanotechnology has come of age. "We have computer architects and chemists working together, and we're starting to build some real stuff," he said.

Keukes showed a photo of silicide nanowires, built in HP's lab, that are six atoms across. The wires are overlaid in a grid pattern with one molecule in between. That molecule, he said, can replace the switching activity of six or seven transistors.

But chemical structures typically have a 3 percent defect level, forcing a new way to think about the design process, Keukes said. He predicted that designers will build a programmable system, run a test program to find defects and then modify the design to avoid the defects. "In effect, we'll be building FPGAs with far more routing and interconnect resources," he said.

Seth Goldstein, assistant professor at Carnegie Mellon University, also emphasized that electronic nanotechnology systems will have to be reprogrammable. He said he expects commercial applications for memories within five years and for logic and sensing with 10 years.

Nanotechnology chips will be homogenous, regular and fine-grained, Goldstein said. They'll be based on matrices of wires with active devices at the intersections. Designers will use reconfigurability to detect and avoid defects and will compile programs directly to silicon, he predicted.

Goldstein invited controversy by stating that nanotechnology systems will comprise diode-resistor logic, using molecular latches. "No transistors. Throw them away," he declared.

Some panelists focused on the difficulties of nanotechnology. "Almost any two dissimilar materials can be put together to make a circuit, but the real problem is putting things together. How do you connect things on a nano-metric scale?" asked Paul Franzon, professor at North Carolina State University.

Chemical self-assembly won't provide the high yields of inorganic crystals, he noted. Interconnect will be tough, and systems won't be particularly fast. From a tool standpoint, Franzon said, performance convergence will be even harder than it is today.

Robert Dutton, director of Stanford University's research center for ICs, noted that nanotechnology could allow densities in the sub-20-nanometer range. But the drive current per device will be low, he said, and I/O problems come along with that.

Getting the heat out will be difficult, Dutton said. But he said that "smart I/Os" are a potential first application for electronic nanotechnology.

Software challenges were the focus of Eric Parker, senior software architect at nanotechnology startup Zyvex Inc. Parker observed that nanotechnology systems will require tremendous amounts of data from such diverse domains as electronics, chemistry and structural engineering. What's needed, he said, is a "common data interchange mechanism" through which the domains can share data at the highest possible level of abstraction.

In response to questions, panelists talked about the likely university curriculum for designers of electronic nanotechnology systems. There was general agreement that some knowledge of electronics, chemistry and biochemistry will be required.

"Engineers think differently from scientists," Franzon noted. "[Nanotechnology] needs an engineering perspective. It's up to us to bridge the gap."