Intel unveils silicon laser

Intel has devised a laser out of silicon, the latest in a series of steps that could take the expense and pain out of optical communication.

The Santa Clara, Calif.-based company has created a chip containing eight continuous Raman lasers by using fairly standard silicon processes rather than the somewhat expensive materials and processes required for making lasers today. The lasers emit a continuous stream of light that can then be modulated, or chopped up, into a stream of impulses that can represent data. Cheap optical parts could not only lead to faster computers but also to less expensive and more accurate medical equipment.

While silicon lasers likely won't enter the market for at least four to five years, the chip should generate enthusiasm and interest in the industry. Although manufacturers love silicon, it's typically a terrible carrier for optical data.

"This is a scientific breakthrough, and a psychological breakthrough, because no one thought you could do it," said Mario Paniccia, director of the photonics technology lab at Intel. "Silicon is not a good optical material" in ordinary circumstances, he added.

The laser represents the latest step in Intel's plans to adopt optical links to connect computers, chips or eventually even subcomponents on the same chip. Last year, the company showed off a silicon modulator that is on its way to running as fast as the exotic modulators of today.

"What Intel is talking about is taking a $2,000 modulator and putting it on a piece of silicon and taking all of the parts you need and putting it into a single package," said Martin Reynolds, an analyst at Gartner. "Clearly, there is a market for it."

It is also part of a larger effort at Intel to employ its factories to make silicon chips that can test blood or perform mechanical tasks rather than just calculate ones and zeros.

Carrying data on light comes with tremendous advantages. Power consumption and heat dissipation have become a huge problem for chip designers. Photons, units of light carried on optical fiber, generate far less heat than electrons, the signal carriers on copper wire. Fiber strands can also handle far more data traffic, thereby cutting down on cabling and the internal volume of computers.

The catch? Optical components are expensive to manufacture and require exotic III-V materials. Assembling the components into complete systems also remains an arduous task.

That Raman lasers could reduce the hassle and expense is a "significant breakthrough," said Jalali Bahram, a professor at the University of California at Los Angeles. Bahram invented the first silicon Raman laser. (Intel's is the first with a continuous beam.)

Current optical equipment requires that the optical fiber serving as the light source be carefully aligned

with the laser, said Victor Krutul, senior manager of silicon photonics strategy at Intel. Two-thirds of the cost of finished optical equipment goes into testing and assembling it, he said.

Mass-produced silicon can ameliorate many of these problems. Silicon allows for passive alignment: A groove can be cut into a chip containing a silicon laser. The fiber can then be slotted in quickly, cheaply and accurately.

A Raman laser, in some ways, is ideally suited for silicon. The Raman Effect, discovered in 1928 by Nobel laureate Chandrasekhara Venkata Raman, roughly works as follows: Light hits a substance, causing the atoms in the substance to vibrate. The collision causes some of the photons to gain or lose energy, resulting in a secondary light of a different wavelength. A Raman laser essentially involves taking this secondary light and then amplifying it (by reflecting it and pumping energy into the system) to emit a functional beam.

Because of its crystalline structure, silicon atoms readily vibrate when hit with light. The Raman Effect, in fact, is 10,000 times stronger in silicon than standard glass, which should make it far easier to amplify.

Unfortunately, it falls flat in the second half. When silicon atoms get struck by two photons at once, the struck atom will disgorge an electron. The loose electrons then form a cloud inside the material that absorbs the resulting light.

To get around this problem, the chip creates an electric field around the silicon chamber. This sweeps away the cloud, which permits the light to be captured, amplified and emitted.

Technically, silicon in the experimental laser does not generate the light beam--a separate beam does--but it serves as a medium for creating and amplifying the secondary light. Silicon is not a good light generator. "This is a different way of solving the light emission problem," Intel's Paniccia said.

The experimental chip includes silicon on insulator, a technology promoted by IBM and one Intel has regularly criticized when used in microprocessors. Otherwise, the chip was produced on standard silicon processes, which could reduce the cost of mass-producing lasers because they can be made in the same factories as flash memory or chipsets.

"Lots of people have been trying a variety of ways to make a silicon laser, and some skepticism has grown up. However, now we see a real silicon laser. There may not be a direct part for part replacement--more an enabling of the entire silicon technology--because this was a missing weapon in the silicon armory," said Graham Reed, a professor at the University of Surrey in England. "Silicon is well-established as a low-cost, high-volume medium. Just look how inexpensive electronic devices have now become."

Intel has also been improving the technology behind its other silicon parts for the optical industry. When it showed off the silicon modulator last year, the part ran at 1 gigabit per second. Three weeks ago, it published a paper showing the chip running at 2.5gbps, and a paper currently being reviewed shows an Intel silicon modulator churning at 4gbps. Commercial modulators today run at 10GHz.

A paper providing more details on the laser will be published on Wednesday in Nature magazine, which also published the original paper detailing the Raman Effect in1928 and the paper on the world's first laser in 1960.