TOKYO -- When silicon transistor technology hits a brick wall in terms of speed and density, carbon nanotubes could be ready to take over, if Japan's NEC Corp. (Nasdaq: NIPNY) has its way.

In its typical, understated fashion, last month the Japanese giant announced significant advances in the performance of carbon nanotube transistor technology -- in Japanese only.

But before explaining the details of NEC's work, let's go back to basics. What's so special about carbon nanotubes? In short, they are tiny cylinders of carbon atoms, like a sheet of graphite that's been folded around to make a tube, which have a range of novel physical and electronic properties.

Last year, the carbon nanotube community got a boost when IBM Corp. (NYSE: IBM) researchers discovered a way of forming thousands of semiconducting carbon nanotubes on a chip, thus creating the first carbon nanotube integrated circuit. In May 2002, IBM went on to produce carbon nanotube FETs (field-effect transistors) that it claimed could outperform even the most advanced silicon transistor prototypes.

Now it appears that NEC, which discovered carbon nanotubes in 1991, may have stolen back the lead from IBM.

NEC's carbon nanotube FETs are nearly twice as fast as those from the nearest competitors, claims Fumiyuki Nihey, principal researcher at NEC Laboratories’ Nanotechnology Group.

Nihey and his colleagues have developed carbon nanotube FETs that have a transconductance -- a measure of their current-carrying capacity -- of 320 nanoSiemens (nS) at a 100 mV drain current. The devices announced by IBM in May had a transconductance of 190 nS.

"This is a very high transconductance, and this makes for potentially very fast switching speeds and fast large-scale integrated circuits," says Toshio Baba, senior manager of the Nanotechnology Group.

NEC also claims to have optimized the structure of its transistor. The carbon nanotube itself forms what's called the "channel," through which the main current flows. On top of the nanotube is an electrode, or "gate," separated from the channel by a thin layer of insulating oxide, which controls how much current flows through the channel. Traditionally, transistor performance improves as both oxide thickness and channel length decrease.

By using titanium oxide as the insulator, NEC has managed to reduce the gate insulator thickness down to 2-3 nm. By comparison, IBM's device uses a silicon-dioxide insulating layer that's 15 to 20 nm thick.

NEC's transistor is smaller, too. The channel length is 210 nm, compared to IBM’s 260 nm.

Although IBM Research is not NEC's sole rival, clearly there is a bit of a battle brewing. "Our purpose is to compare results with IBM’s to find out what's useful for making devices," says Nihey. IBM did not immediately respond to requests for comment.

Other institutes working on carbon nanotube transistors include California's Stanford University, and Delft University of Technology in the Netherlands.

The need for new electronics technology may be closer than people think. Intel Corp.'s (Nasdaq: INTC) roadmap for silicon comes to an end in 2009 -- a mere six years from now -- with a 35nm process technology, codenamed P1268. If an alternative isn't found it could prove disastrous for industries like computing and telecommunications, whose progress depends on denser and faster silicon integrated circuits.

NEC has its own roadmap, which shows carbon nanotube transistors operating at speeds up to 50 GHz in 2015. Baba acknowledges, however, that this target isn't going to be easy to reach.

Scientists still have a lot to learn about how carbon nanotubes work, he says. Although IBM, NEC, and others have measured transconductance and a few other characteristics, there is plenty they don't know. For example, NEC still doesn't know how fast electrons move inside carbon nanotubes, although it believes it to be between two and eight times as fast as electrons in silicon.

A key challenge will be to develop a simple, and therefore cheap, method for manufacturing these devices. NEC's present design comprises individual carbon nanotubes sitting on little islands. Ironically, while this structure makes it easy to take good measurements, it also makes the circuit very hard to manufacture -- even if NEC had a way of producing uniform nanotubes, rather than random bundles of pipes. NEC is at least three to four years away from controlling the growth of carbon nanotubes, says Nihey."We still have to learn exactly what carbon nanotube's potential is for making FETs and how to make nanotubes that can be placed in fixed positions on substrates," he notes.

— Paul Kallender, special to Light Reading

Additional reporting by Pauline Rigby, Senior Editor, Light Reading