Semiconducting carbon nanotubes could be the centerpiece of low-power, ultra-fast electronics of the future. The challenge is getting them to work with today's manufacturing processes. Now researchers at Stanford University have made an important advance toward large-scale nanotube electronics. They have created functional transistors using an etching process that can be integrated with the methods used to carve out silicon-based computer chips.

A major roadblock to making carbon-nanotube transistors has been the difficulty of separating semiconducting tubes from a typical batch of nanotubes, in which about a third of the material is metallic. Even a tiny percentage of metallic tubes would short a device, causing it to fail. The established but tricky approach to making transistors involves separating out semiconducting nanotubes and then arranging them into circuits.

Hongjie Dai and his colleagues take a new approach. They grow a mixed bunch of semiconducting and metallic nanotubes on a silicon wafer and have them bridge the source and drain of a transistor. Then they expose the devices to methane plasma at 400 °C. The hot, ionized methane particles eat away the carbon atoms, but only in the metallic nanotubes, converting the tubes into a hydrocarbon gas. (The plasma also etches out nanotubes with diameters smaller than about 1.4 nanometers.) Next, the researchers treat the wafer in a vacuum at a temperature of 600 °C; this treatment gets rid of carbon-hydrogen groups that latch on to the semiconducting nanotubes during the plasma treatment. This leaves behind purely semiconducting nanotubes with a consistent range of diameters stretching across the source and drain.

According to Dai, the process, which the researchers described in Science last week, could be made into a bulk manufacturing process, because it is compatible with silicon-semiconductor processing. In fact, the researchers utilize a furnace that was previously used for silicon chips. The process should not be expensive once the equipment is set up, Dai adds, because "methane is really cheap and the temperature is only a few hundred degrees Celsius."

Separating nanotubes by type--according to electrical properties and diameters--has been one of the hardest and most pursued problems in materials science over the past decade, says James Heath, a chemistry professor at the California Institute of Technology in Pasadena, CA. "This recent paper by Dai's group clearly represents a new benchmark for the field," he says. "It should do quite a bit towards enabling the applications of [carbon nanotubes] towards high-performance [transistors] and other types of nanoelectronic devices."

Combining the new process with traditional separating methods would be very powerful, Dai says. Current sorting methods--growing nanotubes selectively or separating them chemically in a solution--are tedious and nonscalable, and at best they create a mix containing 5 to 10 percent metallic nanotubes. But one could use this high concentration of semiconducting material to make devices and then "add selective etching to get to 100 percent selectivity," Dai says.