The idea of using molecules as electronic components has been around since at least 1974, when Ari Aviram of IBM in New York and Mark Ratner, then at New York University, showed theoretically that a molecule placed between two metal electrodes can act as a rectifier. However, it took more than 20 years before an individual molecule was successfully connected to two nanofabricated electrodes in an experiment. The difficulties lay in the manipulation of single molecules and in the ability to build electrodes separated by only a few nanometres.

But by the mid-1990s a wonderful new material - the carbon nanotube - had appeared on the scene. Carbon nanotubes are rolled up sheets of graphene just a few nanometres in diameter that can behave as either metals or semiconductors depending on their atomic arrangement. Aided by the mechanical robustness and long length of the nanotubes, many groups around the world quickly succeeded in attaching contacts to these macromolecules.

One common way of achieving this involves fabricating "source" and "drain" electrodes on a conducting substrate covered with an oxide layer, and then dropping the nanotube onto them. The substrate - often called the gate electrode - and the nanotube can act like the two plates of a capacitor. Varying the voltage on the gate therefore changes the number of charge carriers on the nanotube.

Measurements of semiconducting nanotubes revealed unexpected electrical characteristics, including an increase in resistance by several orders of magnitude when the gate voltage was raised. The electrical characteristics of the early devices were very similar to those of conventional metal-oxide-silicon field effect transistors (MOSFETs) but with much poorer performance.

Since the first report in 1998 of such a carbon-nanotube field effect transistor by Cees Dekker's group at Delft University in the Netherlands impressive progress has been made in improving device performance - in particular during the last couple of months. These efforts have led to the recent report by Phaedon Avouris and co-workers at IBM in New York of a carbon-nanotube FET that can compete with the leading prototype silicon transistors currently available (S Wind et al. 2002 Appl. Phys. Lett. 80 3817).

The nanotube FET made by Avouris's group has a new layout that resembles a conventional MOSFET structure with the gate above the conduction channel. This arrangement means that the gap between the nanotube and the gate can be made very small, thus making the resistance much more sensitive to variations in the gate voltage. Indeed the coupling between the gate and the nanotube is now strong enough to amplify a signal. In contrast the variations in the output voltage of the first generation of nanotube transistors were far too small to control the input of a second transistor, thus preventing nanotubes from being integrated into circuits.

Semiconducting nanotubes typically operate like p-type semiconductors and so conduct holes rather than electrons. This behaviour is due to molecules from the atmosphere being adsorbed onto the nanotube and charge-transfer doping from the electrodes. These adsorbed molecules are a problem because they affect the reproducibility of the device characteristics. However, the IBM group has also made progress in controlling this molecule adsorption by embedding the nanotubes in a film and by applying thermal-annealing treatments in a vacuum (V Derycke et al. 2002 Appl. Phys. Lett. 80 2773). Avouris and co-workers found that the same techniques could be used to fabricate n-type devices, which conduct electrons, as an alternative to the current method of doping nanotubes with alkali metals.

Probably the most important goal in the past few years has been to make devices that conduct electricity better. Large currents mean faster transistors, which can lead to powerful integrated circuits. The currents in the early nanotube devices were limited by the intrinsic resistance of the nanotube as well as by the contact resistance at the electrodes. However, recent advances in growth techniques - such as chemical vapour deposition - have led to high-quality materials and improved performance. For example, Michael Fuhrer and co-workers at the University of Maryland recently reported producing carbon nanotubes with mobilities as high as 20 000 cm2 V-1 s-1, which is higher than the values for silicon MOSFETs.

Important progress has also been made in reducing the resistance at the nanotube-electrode interface. Paul McEuen of Cornell University and Hongjie Dai at Stanford University have grown large-diameter nanotubes that have smaller band gaps and therefore thinner "Schottky barriers" that allow charge carriers to tunnel more easily between the nanotube and the metal electrode. Another route taken by the IBM group involves thermal annealing with titanium electrodes. Devices with the new gate layout that have been treated this way boast currents as large as 3 µA and are competitive with the best prototype silicon transistors.

All these efforts have enabled the assembly of different nanotubes into basic logic circuits, which is an important step towards nanoelectronics. Manufactured at Delft and IBM, and by Chongwu Zhou's group at the University of Southern California in Los Angeles, these circuits include an inverter, a logic NOR, a static random-access-memory cell and a ring oscillator.

In spite of this success the manufacture of such nanotube circuits remains extremely challenging, and advances in this direction have been much less spectacular. Indeed current techniques fall far short of those needed for mass production. To obtain transistors that can conduct the same amount of current as micron-wide silicon transistors, for example, we need to synthesize 2D arrays of parallel tubes with exactly the right width. Meanwhile it is still impossible to control the electrical properties of synthesized nanotubes. The result is a mixture of metallic and semiconducting tubes. Most problematic, perhaps, is the lack of control when it comes to placing the tubes in predetermined positions during device fabrication.

A great deal of effort is being directed towards these issues. Many university and industrial groups are working hard to improve the performance of carbon-nanotube FETs even further. Considering the amazing progress that has been made in the last four years, do not be surprised to hear of more discoveries in the near future.

This article originally appeared in the August issue of Physics World.