Someday, computer chips will be grown, not made.
The concept of nanotechnology — that is, the manufacture of preposterously small objects — is at least familiar to most, although the scales involved continue to boggle the mind (a pinhead is about a million nanometers wide). It's easy to see why such extreme miniaturisation interests semiconductor makers — after all, feeding the beast called Moore's Law grows more difficult with every generation of chips.
A number of companies are betting that the best way to operate in this nanoscale world is via "molecular self-assembly," in which circuits literally grow themselves. IBM, Texas Instruments, Fujitsu and Hewlett-Packard are focusing on incrementally self-assembled components that can be integrated with conventional silicon-based chips. Meanwhile, start-ups such as ZettaCore and Cambrios Technologies aim to eliminate silicon completely by building entire semiconductors from molecules.
But, experts caution, the race is not a sprint but a marathon whose finish line is at least 20 years away.
Researchers have already shown that it's possible to integrate self-assembly with conventional semiconductor-manufacturing techniques — meaning chips that are at least partially self-assembled may be found in commercially available computers in five to seven years, says Jack Uldrich, president of The NanoVeritas Group and co-author of The Next Big Thing Is Really Small .
Self-assembly — the tendency of certain structures to fall naturally into patterns — is one of nature's most common occurrences. On a grand scale, for example, wind direction, temperature and moisture in the air result in predictable types of storms.
Now think smaller — much smaller. Certain molecules combine without guidance in predictable ways. "Some molecules recognise each other and find natural low-energy states," says W. Grant McGimpsey, a biology professor and director of the Bioengineering Institute at Worcester Polytechnic Institute in Massachusetts.
A common example — and one that's expected to play a prominent role in chip making — is the SAM, or self-assembling monolayer. When a substrate and molecules with long carbon chains are combined under the right conditions, SAMs self-assemble.
"The neat thing about SAMs is they're very well ordered," McGimpsey says. A field of these SAMs protrudes from the substrate at a well-defined angle — like a small patch of thick, well-tended grass — and can perform several duties, such as improving conductivity or increasing surface area. Such order, McGimpsey says, "means predictability of structure, and thus of properties."
So far, the management of self-assembled molecules that can be applied to semiconductors is limited to a few basic structures. However, researchers believe that's a benefit, not a drawback.
Because of the high cost of tooling up, process change in the semiconductor industry is slow. Thus, self-assembly is sure to make its way into integrated circuits only gradually. Early applications will be simple and unglamorous.
For example, IBM Research has used self-assembly to boost by 400% the performance of the high-capacity decoupling capacitor, an integrated-circuit component that helps maintain a steady, spike-free power supply.
"Self-assembled materials form very simple patterns," says Chuck Black, a researcher at IBM Research. That means structures can be made that are far smaller than those resulting from lithography, the conventional chip-making technique.
HP has recently spoken with boldness about nanotechnology's role in its future. "We believe we have a practical, comprehensive strategy for moving computing beyond silicon to the world of molecular-scale electronics," Stan Williams, a director at HP Labs, said in March. HP is betting on crossbar arrays — a way to replace traditional transistors with devices created by trapping a switchable layer only a few atoms thick between crossed wires. HP acknowledges that it must answer many questions before manufacturing crossbar circuits, but one possibility is a self-assembly technique in which silicon nanowires would be "grown" between a pair of electrodes.
The concept of a mass-produced structure with dimensions measured in atoms helps explain why researchers are turning to nanotechnology as the next great hope for Moore's Law — the observation, credited to Intel founder Gordon Moore, that the density of transistors on a chip doubles every 18-24 months.
Today's most advanced transistors feature gate lengths of 90nm; Intel says it will roll out 45nm transistors in 2007. That's less than a 500th of the width of a pinhead —- and yet it's an absolute chasm compared with a molecule, which is about 1nm in width.
As a result, McGimpsey says, the potential to shrink chips is vast: "Replace all the gates on [today's semiconductors] with atoms, and you get a 10,000-fold decrease in size and thus increase in speed."
One sign that limited applications of self-assembly are more likely years than decades away is the effort that's been made to integrate the technique with conventional lithography. For example, the recent IBM breakthrough that most excites Black is a process improvement making it possible to "register," or align self-assembled structures with those created through lithography. Conventional chips feature about 30 lithographed layers, and aligning them precisely is one of the prerequisites of production. It's a difficult process that will only get harder as circuits shrink. Thus, IBM's newly developed ability to align self-assembled components "is a big breakthrough," Black says. "This allows us to truly think about building [hybrid chips]."
NanoVeritas' Uldrich says the complexity of manufacturing ever-smaller silicon chips and the billions of dollars invested worldwide in nanotechnology research will hasten the arrival of full-fledged self-assembled chips. "This stuff is coming," he says, "and it's coming a lot sooner than many people believe."