Recently I intended a day-long IBM/UCB seminar entitled "Is small the next big thing?" Amongst discussions on areas like quantum computing, organic electronics, and nanoscale MEMS were three presentations on carbon nanotubes. IBM is one of a few integrated circuit (IC) companies researching carbon nanotubes as possible replacements to silicon MOSFET's. This begs the questions "Why might silicon CMOS be replaced?" and "Why might carbon nanotubes replace it?" I will try to provide answers to those questions.
Why might silicon CMOS be replaced?
Of course the answer is that another technology will be more profitable. So why might another technology be more profitable? Assuming that computer scientists develop new, exciting "killer applications" that require higher processing speeds and more RAM*, transistors will need to get smaller and faster.
* No offense to all the CS guys on here, but I'm becoming less and less convinced of the validity of this assumption!
Making silicon MOSFET's significantly smaller than they already are is both physically challenging and expensive. Producing smaller MOSFET's is physically challenging because of so-called "short-channel effects." As the size of MOSFET's decrease, they behave less ideally. They tend to consume far too much power and remain conductive in the "off" state. The major IC companies are, of course, well aware of the difficulties of scaling-down CMOS. They devote most of their research to modfications of the traditional silicon MOSFET structure. Examples of these modifications are using silicon on insulator wafers (there's an excellent writeup on this) or MOSFET's with double gates. In both cases, especially the latter, fabrication of the circuits becomes more expensive.
Furthermore, producing smaller MOSFET's requires new lithography techniques. Photolithography has reached a point at which further reduction in feature sizes is nearly impossible and extremely costly. Other lithography techniques such as electron beam lithography and extreme ultraviolet lithography are incredibly expensive. E-beam lithography is currently considered unfeasible for mass-production because of the time it takes. Extreme ultraviolet lithography is a new technology, and it is unclear what its technological fate will be.
Why might carbon nanotubes replace silicon CMOS?
Carbon nanotubes have vastly superior electrical properties to any other material of which I am aware. Rather than introduce complicated physics, let me give an example from IBM. A 300nm carbon nanotube field-effect-transistor with 20nm gate oxide provides a higher on-current (defined as current per gate width at Vg = Vt + 1V) than a double-gate, 25nm MOSFET with 2nm gate oxide. In laymen's terms, there is reason to believe that carbon nanotube transistors with dimensions that can be produced by optical lithography could outperform silicon MOSFET's that require new fabrication techniques.
So what are the odds of carbon nanotube FET's replacing silicon MOSFET's? Nobody is sure, but the odds would be much higher if fabrication of carbon nanotubes could be better controlled. The thing holding carbon nanotubes back from widespread use is an inability to control length, direction, and type of nanotubes produced. Modern integrated circuit technology is thin-film technology. Thin films of materials are deposited over silicon wafers, lithographically patterned, and etched. Carbon nanotubes are discrete objects. To make circuits out of them will require new ideas about how to position and contact them. Scientists will need to find a way to selectively grow either metallic or semiconducting tubes (currently it's a coin flip).
It seems certain that nanotechnology will be a reality in the future. When and in what forms remain to be seen, but carbon nanotubes look like strong contenders for widespread technological usage. Perhaps carbon nanotubes will never replace silicon CMOS, but scientists have suggested many other potential applications, such as gas sensors, accelerometers, high-frequency mechanical resonators for wireless communications, and field-emitters for flat panel displays. It will be interesting to see the technological fate of these remarkable carbon macromolecules.