Researchers at Intel have developed a method of generating a continuous laser with a silicon device, one of the first steps toward introducing optical interconnects in future processors, servers and PCs, the company said Wednesday.
Exploiting a principle called the Raman effect and using standard chip manufacturing techniques, Intel built a transistor-like device that is capable of producing a continuous beam of light. Those light waves can be used to carry data at faster speeds than copper, the current standard for chip interconnects, said Mario Paniccia, director of Intel's Photonics Technology Lab. Intel's findings were published on the Web site of the scientific journal Nature on Wednesday.
"These building blocks are still a research project, but we hope to transfer the technology by the end of the decade," Paniccia said on a conference call Wednesday.
Intel's research in this area is all part of the chip industry's search for alternative techniques and materials that will allow it to continue to shrink transistors well into the future. Companies such as Intel, IBM and others are working on ways to boost chip performance as chip components become as small as individual atoms, a point at which the decades-long practice of reducing the size of transistors becomes exceedingly difficult.
One way to improve the performance of chips, servers and networking devices is to replace the electrical charges that currently carry data with light particles, or photons. This discipline, known as photonics, is well underway in the networking industry, where fiber-optic materials are replacing older copper wires as the preferred means of transmitting signals over long-haul communications networks, Paniccia said.
But fiber-optic materials are expensive and complex. Using silicon materials to generate light waves would solve many of the cost issues, but silicon does not naturally emit light. However, existing silicon devices can be used as channels for laser beams that can carry data signals, Paniccia said.
In order to build a silicon photonics device, Intel researchers drew upon the properties of the Raman effect. When a strong beam of photons are pumped into a chamber alongside a weaker stream of data, the Raman effect causes vibrating atoms within the chamber to transfer energy from the data stream to photons. This effect is amplified as a second light source is allowed to reflect back and forth across the chamber, producing a stronger data beam at the other end of the fiber.
This effect is extremely pronounced in silicon, as opposed to fiber materials, Paniccia said. Kilometer-long lengths of fiber are needed to produce a laser beam of the same strength generated by just centimeters of silicon, he said.
Intel has been working with the Raman effect and silicon photonics for over a year, but it has now figured out how to bypass a key roadblock that was sapping the strength of the laser beam.
By its nature, silicon is transparent to infrared light waves, and photons passing through silicon usually have no effect. But on occasion, two photons can strike the same silicon atom at the same time and knock electrons out of that atom. This decreases the strength of the laser as electrons build up and start to absorb photons from the amplified laser beam.
In order to get around this problem, Intel built what it calls a PIN device. The PIN device is comprised of a sandwich of positively-charged silicon, intrinsic or neutrally charged silicon, and negatively charged silicon.
The electrical field generated by the opposing charges sweeps the loose electrons out of the waveguide, or the intrinsic silicon layer, Paniccia said. This produces a steady laser beam at the other end that doesn't lose its strength.
The laser technology could be used in anything from chip components to portable medical devices, Paniccia said. The silicon laser presented Wednesday is one of many optical components that must be tested and further developed before the technology can be used in chip-to-chip or motherboard-to-motherboard connections, he said.