But on a conference call with reporters this morning, Intel representatives were quick to note that this program is still in the research phase, with practical, available-for-sale silicon lasers still years off.
"The silicon photonics research and [its associated] building blocks are still a research program, but we hope to transfer that technology [to real-world use] by the end of the decade," says Mario Paniccia, director of Intel's photonics technology lab.
The initial version of the laser has an output of 8 mW, compared with 10 mW for many commercial lasers. Paniccia notes that Intel's laser is still a rough draft, not optimized for maximum power.
Silicon-based photonics has long been a quest for Intel and others (see Light From Silicon). With the truckloads of research dollars that pour into chip manufacturing, it's much cheaper to build a device in silicon rather than in indium phosphide or gallium arsenide, two of the materials commonly used for photonics. Moreover, chips in silicon can be integrated with relative ease, creating smaller optical modules that can include electronics on-chip.
But silicon doesn't emit light, and historically, silicon chips haven't been fast enough to detect or modulate the optical signals used in telecom. All told, Paniccia notes that six innovations are required to make silicon photonics work:
- The light source (today's announcement)
- Waveguides in the silicon, to guide the light
- A modulator, as Intel announced last year (see Intel's Modest Modulator )
- Detectors, to receive the light in silicon
- Packaging and assembly
- Intelligence (i.e., the chips Intel does already)
Intel has gotten the waveguides done, and demonstrated automated assembly last fall at the Intel Developer's Forum. That leaves the detectors as the only piece Intel hasn't discussed publicly. "We hope in 2005 we will produce some technical work there," Paniccia says.
Researchers at The University of California, Los Angeles (UCLA) demonstrated a silicon laser in the fall (see UCLA Claims First Silicon Laser). But that laser delivered light in a pulse of less than 50 picoseconds, Paniccia notes. Such pulse lasers aren't practical in most applications and are usually a laboratory precursor to a more useful, continuous-wave laser.
As with UCLA's project, Intel's silicon laser uses Raman Amplification -- the same photonic effect used in Raman amplifiers for long-distance transmissions. Here's the plan: When a beam of light creates vibrations in the lattice of silicon atoms, this gives off energy in the form of light at a new wavelength. If that light intersects a second beam of the same wavelength (this would be the beam carrying data), the result is amplification of the second beam. Repeated amplification eventually causes the beam to reach the threshold current, where intense light (the laser beam) emits.
Silicon happens to be a happy home for the Raman effect. "The Raman gain coefficient is 10,000 times stronger in silicon than in amorphous glass fiber," Paniccia says. "We can do in centimeters what's done today in kilometers."
But the Raman method hit a snag. Every now and then, two photons would hit a silicon atom simultaneously, and the resulting energy would kick out an electron. Because electrons get reabsorbed into the material slowly, these free electrons would build up, creating a cloud that disrupted the laser's process.
It's this "two-photon absorption" that limited the UCLA laser to 50-picosecond pulses rather than continuous-wave output, Paniccia says. Intel ran into the same problem. By November, the company had gotten only a pulse-wave version of its laser to work; those results got published in January. Intel had gotten the continuous-wave laser to work by then -- on the day before Christmas, as it turns out (awwww).
Intel overcame two-photon absorption by implanting material in the silicon to create an electrical field that grabs the electrons. "I can suck out the electrons similar to a vacuum cleaner," Paniccia says gracefully. This, er, "sucking" is what made Intel's continuous-wave breakthrough possible.
Silicon photonics could lead to increased reach for lasers. Some fancier applications could be made possible too; Paniccia describes the possibility of using silicon waveguides to create a Wavelength Division Multiplexing (WDM) feed out of one light source, for example.
The silicon laser is nice for telecom, but the more glamorous applications could lie in medical equipment and sensors, where a tunable silicon laser could replace models that cost tens of thousands of dollars. Intel also has its eye on optical interconnects for chips and backplanes, not just for the speed of optics, but because thin optical fibers take up less space than electrical cables, leaving more space to get computers and servers cooled down.
— Craig Matsumoto, Senior Editor, Light Reading For more on this topic, check out:
- The Light Reading Beginner's Guide: Laser Basics