Carbon nanotubes are rolled sheets of graphite just nanometres in diameter. The electronic properties of a tube - metallic, semiconducting or a combination of the two - depend on whether it is made from one or many layers of graphite, and on the angle at which the sheets are rolled. The tubes are called zigzag- or 'armchair'-shaped depending on the angle of roll.

Lieber and coworkers used scanning tunnelling microscopy to image zigzag and armchair nanotubes - both of which were thought to be metallic - on the atomic scale. But in fact the zigzag nanotubes have 'gaps' in their energy profile, which means that the zigzag nanotubes are not fully metallic but have unexpected semiconductor properties. The team found that the energy gaps are larger for tubes with a smaller radius. Clusters of armchair nanotubes also exhibit a gap-like feature, suggesting that they also have some non-metal character. Single specimens of armchair nanotubes, on the other hand, have no energy gaps, supporting the idea that they are fully metallic. According to the researchers, these findings have important implications for our understanding of the electronic properties and potential applications of carbon nanotubes.

Carbon nanotubes are usually constructed in a process that produces a mixture of tubes. They are rolled at a variety of angles and with different numbers of layers and therefore have a range of electronic properties. This has hindered efforts to exploit the tubes in electronic devices in the past, but now Avouris and colleagues have developed to separate metallic and semiconducting nanotubes.

The team deposited a layer of unsorted nanotubes onto a silicon wafer and superimposed source, gate and drain electrodes using electron beam lithography. Applying a gate voltage temporarily shifts the semiconducting nanotubes into their insulating state. This protects them from the very high current that is subsequently applied to the electrodes, which destroys the metallic nanotubes but leaves the semiconducting nanotubes intact. The team has already created an array of field-effect transistors using the technique, and believes that the principle may also work in other molecular electronics systems.