An old contender's time has come

IF THE technology boom and bust depended on anything physical, it was silicon. Extracted from humble sand, silicon is the raw material of information technology, as the vast majority of computer chips are made from it. The reason that computer performance has been improving so rapidly is that the density of circuits etched on to silicon wafers has increased. But engineers are now running into the physical limitations of silicon. So there is a renewed interest in replacements.

An announcement from Motorola suggests that one long-awaited replacement may have arrived. Gallium arsenide (GaAs) has seemed a likely alternative to silicon ever since the earliest days of semiconductors. Chips made of GaAs have several desirable properties, because the electrons in them are extremely mobile. They use less power than silicon, and they have much higher switching speeds—nearly 70 GHz (billions of cycles per second), as opposed to silicon's best of 2 GHz. So they can process information faster. And they are very efficient at converting light to electricity, and vice versa, which makes them ideal for the branch of information technology known as optoelectronics.

Unfortunately, GaAs is costly and brittle. That has restricted its applications to things for which there is no alternative. For years, engineers have sought a way to create chips that combine silicon with GaAs, thus reducing both cost (by employing GaAs only in places where it is needed) and brittleness (by backing it with more supple silicon). They have been thwarted because the crystalline structures of silicon and gallium arsenide do not mesh, so the GaAs components will not stick in place. Now Motorola, following up a happy bit of serendipity, seems to have found a way to join the two.

• A bigger pill to swallow
• Seed v seed
• Arsenic and chips
• Silent running

• Integrated circuits
• Engineering
• Circuit engineering
• Science and technology
• Electrical engineering

In 1999, some of the company's researchers developed what was then the world's thinnest transistor. This was made by depositing a microscopic layer of strontium titanate on a bed of silicon. Making it was hard, because of the persistent tendency for a layer of oxygen molecules to interpose itself between the strontium titanate and the silicon. However, while he was studying this problem, Jamal Ramdani, one of the team, noticed that the oxygen also causes the strontium titanate to relax into a crystalline structure that meshes nicely with GaAs. The fit is not perfect, but it provided an opening. Strontium titanate could, in principle, mesh with both silicon and GaAs. It therefore offered a way to stick the two together.

In the intervening years, Motorola has worked out the practical details, produced prototype GaAs-silicon hybrid chips, and has now tested them in mobile phones. That is a big market. A typical mobile-phone handset contains six or seven GaAs chips, which are used as amplifiers and high-speed switches.

Running mobile phones is not, though, the only use to which these chips are likely to be put. More than $35 billion has been spent over the past two years laying some 200m kilometres of optical fibre around the earth. There is a glut of capacity of high-speed, long-distance pipelines, but very few points of access. This is because the optoelectronic technology needed to switch signals from electrons to light, and back again, has been expensive. Until now. Motorola's new chips may, with luck, make the fibre glut useful at last.