Network: Processors that are faster than the speed of light?

Nanolaser technology is set to revolutionise computer chip-making.
Small may be beautiful, but when it comes to lasers, small is also extraordinarily efficient. Current research is driving towards devices smaller than the wavelength of the light they emit - so-called nanolasers - and on the commercial front interest is growing in applications from microprocessing to disease-detection. Indeed, microlasers are being seen by some as the key to keeping Moore's Law, which states that computer processing power will double every 18 months, on track for the next 10 years.

In a laser, photons are bounced between two mirrors. This encourages them to become so orderly that a point is reached when their collective power bursts through one mirror forming the characteristic ray of laser light. Just as it is much easier to entice rather than to force people to walk in line, laser light is stimulated into mass action by the material in the cavity between the mirrors, in a chain reaction that yields radiation with a very high amplitude (laser stands for light amplification by stimulated emission of radiation). However, large lasers are hugely inefficient, requiring substantial amounts of electricity to power up, since, typically, only one in 10,000 photons will march in time.

This situation applies even to semiconductor lasers, such as found in CD players. But in smaller devices, where the cavity is of the order of the wavelength of the light itself, such conspicuous consumption disappears. On this scale, photons are forced to behave in determined ways, because their options for doing otherwise are systematically removed. They are trapped in a quantum well. Indeed, scientists in Japan have proposed that nanolasers, something less than one-hundredth of the thickness of a human hair, might approach 100 per cent efficiency in a process known as thresholdless operation. Researchers at the Massachusetts Institute of Technology have recently built a device into which individual photons can be fed and scientists elsewhere are developing techniques which could reach this ultimate goal.

For now, though, it is the microlaser cousins that are stealing the show in a range of applications likely to be commercially viable in a matter of years. Optical computing is one possibility, in which electrons are replaced by photons. However, such machines, though often talked about, are still some time off. What is less theoretical and more exciting are initiatives such as one led by Professor Gareth Parry of Imperial College, London.

As part of a major European programme, Advanced Research Initiatives in Microelectronics, with sponsorship from industry giants such as GEC Marconi, Parry's team is working on technology that seeks not so much to replace electronic circuits as tackle the problems which are inevitably going to arise in the next 10 years.

The hurdle Professor Parry is trying to leap appears in the race to miniaturise silicon transistors. At the moment, a feature on a chip will be no smaller than 0.35 microns. But according to current predictions, by 2010 this will have reduced to 0.07 microns, with all the increases in processing power Moore's law implies. In order to shift the information in and out of these chips fast enough, the metalwork surrounding the processor will have to increase dramatically. Today's chips are surrounded by approximately 900 input and output paths, of copper wires more than 100 meters long. To support the advances of the next 10 years, peripheral gates will have to increase to nearly 5,000 in number, with wires stretching to over 4km. Such scaling rapidly becomes unrealistic for fabrication. However, microlasers could save the day, as well as the already stretched pockets of chip industry.

Chips could in effect be stacked on top of one another with lasers passing information between them, rather than wires. Microlasers would work in this environment because the current they demand is of the same order as the transistors they would support, as a result of their new efficiencies.

The swift returns these developments suggest arise because of savings during manufacture. Silicon chips are fabricated essentially in two stages. The first produces the wafer. The second etches the electronics on to its surface. Microlasers are grown in one process, layer by layer. It is cutting out this additional phase that suggests attractive economies. Hewlett Packard and Motorola are two more companies that have recently shown interest in Professor Parry's work.

The potential of microlasers also comes from directions other than the electronics industry. Paul Gourley, a principal investigator with Sandia National Laboratories, along with scientists at the University of St Andrews, has developed devices in which biological material can be passed into the cavity of the laser. Doing this alters the spectrum of the light emitted in ways which can be useful. For example, healthy and diseased blood produce characteristic light which can be compared much as the same note played on a piccolo and a flute can be differentiated by the overtones they produce. Such devices have been tested for diagnosing sickle cell anemia and could be extended to detect HIV and even reading DNA. Usual diagnostic methods involve staining blood cells for comparison with the human eye. The microlaser route, in theory, offers a more objective means of assessment.