Nanotoxicology: the asbestos analogy revisited.

378 nature nanotechnology | VOL 3 | JULY 2008 | www.nature.com/naturenanotechnology oxygen vacancies. In the latter case, the actual carriers are lattice ions, which are much heavier than electrons. The scaling characteristics of these metal–oxide–metal devices are not discussed by the HP team, but they should, in principle, be quite good because the oxygen vacancies will probably have to drift only a few angstroms under the applied electric field. The suggestion that heavier particles might be preferred for nanoscale devices may seem counterintuitive because lighter particles can be moved about much faster than heavier particles. However, it is difficult to confine electrons in a very small space because they can tunnel out quite easily. Moreover, response times for atoms can be rapid if they only have to move a short distance. ‘Nanoionic’ devices have already attracted the attention of several research groups3–5. In the ‘atomic relay’, for instance, a nanoscale gap is opened and closed by the movement of a small number of silver atoms, and the switching time in such devices is expected to be about 1 ns (ref. 4). A number of memory concepts based on ion-migration effects in solids are currently being explored and they show potential to overcome the scaling limits associated with traditional electron-based memories5. A common characteristic of all devices whose operation relies on the movement of both ions and electrons is that they require materials that are not typically used in semiconductor devices. This suggests that innovation in nanoelectronic devices is strongly dependent on materials research. As a final note, although one may argue in favour of two-terminal devices such as those developed by the HP team6, the electronic circuit community might be interested in extending this device concept to three-terminal devices with gate electrodes (Fig. 1a). In principle, one could devise a three-terminal device in which the barrier height was regulated by charge, as in conventional devices, but in which information was carried by different and heavier charged particles. Such an approach could limit the ‘OFF’ state currents that arise in devices with feature sizes of ~1 nm because the heavier particle would be much less prone to tunnelling and over-the-barrier transitions than electrons.