Features

The microprocessor has given us computers that sit on our desk rather than fill an entire room. But have you ever wondered why semiconductor manufacturers are still so keen to reduce a chip's feature size? The processor accounts for a small fraction of the space inside your PC's case. Would it be such a problem if the feature size was 90 nanometres, 90nm (90 billionths of a millimetre) or 65nm? Not really. At the moment, shaving off a few nanometers doesn't make a difference to how the processor fits in the system case. But that is not the main reason why processors need to get ever smaller.

Moore's Law, coined in 1965 by Intel pioneer Gordon Moore, states that the number of transistors on a chip doubles every 18 months (although this was later amended to every two years). The result has been a continual improvement in microprocessor performance. Today those extra transistors are being used to give us processors with two or even four cores. If this trend continues, we'll have 128-core chips within a decade and, without a corresponding reduction in the feature size, that would be a seriously large chip.

The laws of physics state that a fast chip has to be a small chip. Experts believe that the traditional methods of manufacturing chips can't be scaled down much further. But as one technology starts to run out of steam, some researchers believe they have a replacement. Using nanotechnology, which deals with materials at a molecular and atomic level, they've come up with a radical new way of making electronic circuits.

Semiconductor manufacturing

Since the first commercial integrated circuits of the early 1960s,

ADVERTISEMENT

all chips have been made using a technique called photolithography. This is a multi-stage process in which the circuit is built up in layers. Layers include conductive metal to provide the interconnections, silicon oxide to act as an insulator, and polycrystalline silicon, which is doped to give the semiconductor properties needed to form transistors. Each of these layers has to be formed into an intricate pattern to create the required electronic circuit.

To do that, when each layer is first applied, it is given a coating of so-called photo-resist, on to which the required pattern is projected optically. The photo-resist is then developed, like a photographic film, so only those portions of the photo-resist that correspond to the projected pattern remain. The chip is immersed in a liquid called an etchant, which dissolves those areas that were not protected by the photo-resist. Finally, the remaining photo-resist is removed.

At first sight, it seems as if this method could be used for ever-smaller features by projecting the patterns at an ever-smaller scale. But that's forgetting about the properties of light. As the feature size approaches the wavelength of the light being used, it's no longer possible to focus the pattern sharply on the photo-resist. To date this limitation has been overcome by using light with an ever-shorter wavelength, but since today's state-of-the-art technology uses extreme ultraviolet the technique is running out of steam. Next in line after extreme ultraviolet are electron beams, but to say that this technology is in its infancy would be a gross understatement. Some experts question whether or not it will ever lend itself to large-scale manufacturing. A better method is needed.

The process of photolithography is like trying to carve a pocket watch out of a solid block of steel. A watchmaker would never do that. Instead, the watch is built up from its component parts. Many researchers believe that this is how the chips of the future are going to be made. This up-and-coming field of nanotechnology, and the technique of building a chip from its constituent parts, is referred to as a bottom-up approach. This is in contrast to the top-down method of today's semiconductor manufacturing.