Published online 24 June 1999 | Nature | doi:10.1038/news990624-1


The end of the road for silicon?

The writing is on the wall for silicon chips, according to research published in the 24 June issue of Nature. By the year 2012, miniaturization of microelectronics will have reached its limit, and computer power will be unable to increase any further without drastically rethinking the whole basis of silicon electronics.

The information revolution that has taken place since the invention of the transistor halfway through this century has been governed by a principle known as Moore's Law. Seemingly as inexorable as a fundamental law of nature, Moore's Law was identified in 1965 by Gordon Moore, cofounder of Intel. It is formulated in several ways, of which perhaps the most general is that computer power doubles every eighteen months. This increase is due mostly to the increase in the number of electronic components that can be packed into a given area on a silicon chip. In other words, Moore's Law is at root all about miniaturization: as devices have got smaller, so computer power has risen.

Today, a single silicon chip just half a centimetre square contains millions of tiny circuit components such as transistors. Each has dimensions of less than a micrometre, which is about a hundredth the width of a human hair. These components are carved primarily from silicon, which conducts electricity, and silicon dioxide, which is an insulator. In the world of microelectronics, silicon is the cable and silicon dioxide is the plastic sheath.

Already the technology needed to make transistors this small is being strained to its limits to keep pace with Moore's Law. But technologists are optimistic about adapting the fabrication processes so that even smaller devices can be made. They are determined to keep to Moore's implacable curve, for the simple reason that their competitors will have the same intention.

But if we project Moore's Law into the next century, we can see a serious roadblock ahead. By 2012, it says, the size of the components will be comparable to the size of molecules. In particular, the smallest feature in a silicon-based transistor - one of the insulating films of silicon dioxide, called the 'gate' oxide - will be just four or five atoms across. Will a layer as thin as this still function as an insulator, or will it start to become leaky and let a current through?

Acknowledging the need to be wise before the event, physicists David Miller and colleagues at Lucent Technologies in Murray Hill, New Jersey, have used the latest in fabrication technologies to make a five-atom-thick film of silicon dioxide sandwiched between two layers of silicon, and tested it to see how well it stands up as a gate oxide. It is possible to make devices with dimensions this small in the laboratory, even though this is not feasible for mass-produced commercial chips, which currently have gate oxide layers about 25 atoms thick.

Miller and colleagues found that their ultra-thin gate oxide was no longer able to guarantee insulation between the silicon layers on either side. They found that the mobile electrons carrying current in the silicon oozed out into the gate oxide film like bright light penetrating through murky water.

The researchers estimate that a gate oxide layer thinner than four atoms across will be so leaky as to be useless. In addition, because of the limitations in making perfectly smooth films, gate oxides made by present-day methods would start to break down at a little under twice this thickness.

So is there any way that the computer industry can avoid seeing Moore's Law coming unstuck around 2012? Perhaps a better insulating material will be found - although silicon dioxide is currently the best known. Or maybe circuit components will need a complete redesign, dispensing with the traditional transistor. Some researchers think that, at these tiny scales, we will have to start making active use of the effects of quantum mechanics in electronic engineering, rather than relying on the 'classical' electronics that applies to the 'soldering iron' circuits driving your computer. Either way, we might soon see the end of the Silicon Age - and the dawn of something new.