7. Future Information Technologies

Much of the impetus for nanotechnology comes from the fact that the scale of engineering is shrinking on many fronts. Entire laboratories may soon be constructed on a single silicon chip, in which microscopic amounts of chemical reagents are driven down channels, mixed together and analyzed. Lithographic methods can already carve out gears and motors too small for the eye to see, which might power flying craft no bigger than insects-to be used in vast swarms for space exploration or, one has to recognize, to be abused for military purposes. But this relentless miniaturization has always been most keenly felt in information technology, which is sure to be one of the most socially transforming technologies of the coming century.

In 1965, Gordon Moore, one of the co-founders of Intel, pointed out that the speed of computers had roughly doubled every 18 months. This rate of change, colloquially dubbed Moore's Law, has since been followed with remarkable fidelity. The increase in speed is largely concomitant with a decrease in scale of the electronic components, so that more processing power can be packed onto a single silicon chip. By 1998, more components could be packed onto a silicon wafer 8 inches in diameter than there are people in the world.

This has so far been a revolution written in silicon. But there is no guarantee that silicon will continue as the bedrock of information technology in the twenty-first century. If Moore's law persists, the smallest dimensions in microelectronic devices such as transistors will reach the size of small molecules by around 2012. At that stage, their traditional function can no longer be sustained. For example, the layer of silicon dioxide that insulates the "gate electrode" of a conventional transistor (a metal-oxide-semiconductor field-effect transistor or MOSFET) will be just five or so atoms thick. At this point, it is no longer a perfect insulator-the device becomes leaky.

Alternative materials, such as diamond doped to a semiconducting state, might possibly stave off the crisis a little longer (although it looms too imminently for the industry to be likely to effect so rapid a change to a new materials basis). But sooner rather than later, a completely new approach to computer hardware will be required if the information industry is to sustain Moore's Law. Some suggest that this might come about by using new types of device-replacing the workhorse of the MOSFET with some other kind of electronic switching device, perhaps exploiting the quantum phenomena that appear at very small dimensions. Others feel that the answer might be algorithmic: using quantum algorithms instead of classical ones to vastly expand the parallelism of computers. But some very dramatic leaps in technical capability will be needed to transform the present modest laboratory demonstrations of few-bit quantum computation into anything useful.

Another alternative is to abandon reliance on electronics altogether, and use a different medium to carry, store and process information: light. To a limited extent, light-based information technology is here already: long- and medium-distance telecommunications are now generally conducted by photonic means by passing pulsed light signals along optical fibers. The capacity of a fiber-optic cable is far greater than that of a copper wire of the same width; and the full potential of optical transmission, barely realized as yet, is awesome. Moreover, information is now commonly stored and read out by optical means from compact disk (CD) systems or, less commonly, from magneto-optic memories. Conceptually, it would seem to make sense to do everything with light-processing as well as transmission and storage-instead of laboriously converting the signal from electronic to photonic form and back again.

But the all-optical computer is still an uncertain prospect-the technologies (including the materials) to realize it are still lacking, and it is fair to say that there is no consensus as to whether the payoffs would justify the effort. It seems clear that for the immediate future, a hybrid of electronic and photonic technologies-optoelectronics-will be pursued with more vigor.

The two aspects tend to be kept separate, meantime: the devices for translating electrical pulses to light pulses and for the converse are housed separately from the electronic components. The light sources are generally miniaturized laser diodes, relatively lumen devices as big as a chip themselves-several hundreds of micrometers in length. They are layered structures in which the light-emitting (lasing) medium is sandwiched between semiconducting materials that inject charge carriers. When these recombine, a photon is emitted; and the bouncing of photons back and forth between the reflective ends of the cavity induces the stimulated emission characteristic of laser action. Lasers used for optical telecommunications and CD players have generally been made from so-called III-V semiconductors such as gallium arsenide and gallium aluminum arsenide, mixtures of elements from groups III and V of the Periodic Table. They emit light in the infrared and red part of the spectrum: near-infrared radiation is used for signal transmission in current fiber-optic networks.

But there are several good reasons to extend the wavelength range of these semiconductor laser diodes to shorter wavelengths. A wider range allows for wavelength multiplexing: using light of different colors to carry many signals simultaneously down a fiber, just as many distinct radio signals can be broadcast simultaneously at different frequencies. And because the size of a focused laser beam depends on its wavelength, shorter-wavelength laser light could be used to read and write smaller bit sizes into optical storage media, permitting a greater density of information. A CD read with blue light could have about four times the capacity of one read with the current generation of near-infrared lasers.

This need to extend the wavelength range of laser diodes has led to much interest in materials that alter the frequency of light transmitted through them. A material that transmits light to a degree that is not simply proportional to the intensity of the illumination is said to have nonlinear optical properties; frequency-doubling and -tripling materials are examples of these. The first commercialized blue-light laser diode used the well-known frequency doubler potassium niobate to "upgrade" near-infrared light. But such devices are now rendered redundant by laser diodes that genuinely emit blue light.

The emission wavelength from a semiconductor laser is set by the bandgap of the material-crudely speaking, the energy difference between a mobile (conduction) and bound (valence) electron. This determines how much energy is given up (as a photon) when a mobile electron falls back to the valence band. The larger the bandgap, the larger the photon energy and so the smaller the wavelength. The search for large-bandgap materials led at first to so-called II-VI compounds such as zinc selenide, but they were never very efficient. In 1996, however, a bright blue/ultraviolet laser was reported in which the light-emitting medium was a "new" III-V material: gallium nitride.

Figure 15. A Blue-light Laser Fabricated from Gallium Nitride

Gallium nitride was long known as a large-bandgap material, but its use in a laser diode seemed unlikely because of the difficulty of growing gallium nitride films on a silicon substrate. The problem faced by all semiconductors for use in optoelectronic technology is that its dependence on silicon devices means that silicon is still the universal substrate. Everything has to be grown on silicon. Yet for most crystalline semiconductors the lattice spacing of the component atoms is very different from that between two silicon atoms. This means that, if a thin film of the material is deposited on silicon, the atoms must either move away from their equilibrium positions, or suffer the occasional discontinuity of regular order if they are to marry up with the silicon atoms to which they must bond. In other words, at least the first few layers of the film are strained, and this can give rise to imperfections that are propagated into the growing film like a kind of crack. Defects like this can severely disrupt the electrical conductivity of the material, hampering the progress of mobile charge carriers. Gallium arsenide has a lattice spacing not far from that of silicon, which is one reason why it has been so favored as a photonic material.

Gallium nitride, however, has a very different lattice spacing, and so it seemed most unlikely that sufficiently regular films could be deposited on silicon to be operative in semiconductor laser diodes. The technical problems were solved, however, by the persistent effort of researchers at Nichia Chemical Industries in Japan, who created the first gallium nitride laser (Figure 15). Since then, several other companies worldwide have introduced their own versions of these blue and violet lasers. Future CD systems are sure to take advantage of the benefits in data storage capacity offered by these laser systems.

next page...