Researchers at Intel recently announced a silicon-based light detector that, by all measures, is better than those made of more expensive materials. It can detect flashes of light at a rate of 40 gigabits per second, while most of today's fiber-optic networks operate at 10 gigabits per second. The new detector is also more efficient and produces a cleaner signal than other detectors that operate at the same speed. Since detectors made of silicon have the potential to be manufactured on large silicon wafers, through standard processing techniques, researchers could produce detectors that are hundreds of times less expensive than those used in today's networks, which are made of materials such as indium gallium arsenide.

Already, Intel has demonstrated a silicon-based laser and a silicon modulator--a device that encodes data onto light--that operate at 40 gigabits per second. (See "Silicon Lasers Get Up to Speed" and "Moving Toward a Terascale Computer.") The goal, says Mario Paniccia, director of Intel's silicon-photonics lab, is to combine all three devices on a single silicon chip. That chip would be cheap, since it could be made using manufacturing processes well honed by the microchip industry. If implemented in existing fiber-optic networks, inexpensive photonic chips could drastically reduce the cost of Internet bandwidth. Built into computers, they could move and transmit data at much greater speeds.

Intel's silicon detectors use the same basic principles that many other light detectors do, explains Paniccia. When photons strike a traditional detector, they produce pairs of electrons and "holes." (A hole is the absence of an electron where one would be expected; it can be thought of as a positively charged particle.) A voltage is applied across the detector, pushing the negatively charged electrons one way, and the positively charged holes the other way. The resulting electrical current provides a measure of the amount of light the detector collected.

For detectors made of gallium arsenide and indium gallium arsenide, the process is straightforward: both of those materials easily produce electron-hole pairs when photons with a certain energy pass through them. Silicon, however, doesn't react to light in the same way. So in their new device, Paniccia and his team decided to use silicon as a waveguide, a sort of channel that collects and holds light. On top of the waveguide, the researchers grew layers of germanium, a material that does create electron-hole pairs when struck by photons. It's the germanium that does the actual detecting: as light passes through the silicon waveguide, part of it sneaks into the germanium and produces an electric current.

Some of today's silicon devices actually include small amounts of germanium, so using existing manufacturing processes to deposit the material isn't necessarily difficult. What is difficult is depositing it in uniform layers on top of silicon. The distance between the atoms in a crystal of germanium is different from the distance between the atoms in a crystal of silicon. Combining the two produces strains and cracks, which could cause problems in an electronic device.

The Intel researchers focused on developing a process that minimizes the strain on the materials near the part of the device that detects light. Many of the details are proprietary, but Paniccia explains that his team experimented with a number of variations in the materials' growth conditions. In the end, the researchers found an ideal combination of temperature and other factors that sweep defects out to the edge of the detector, where they don't impede performance. "It took us a long time to get there," Paniccia says. "It's not a completely new design, but it's a lot of engineering."

The team's next major hurdle is to develop processes for integrating the detector and other silicon devices on a single chip. While Paniccia doesn't expect integration to pose any major challenges, he says that it could take a while to complete. He adds that, while all three of his team's silicon photonic devices work well in the lab, when they're subjected to quality-control testing, problems could arise. He estimates that consumers could begin to enjoy the benefits of integrated silicon photonics within about five years.