Intel's Modest Modulator
Modulators are a key component that encode data onto a light beam. Usually they are made out of other materials such as lithium niobate or indium phosphide. Intel's silicon modulator has the advantage of being compatible with standard electronics manufacturing, and although it's not as fast as commercially available devices, it is a factor of 50 times faster than any previous silicon-based modulator. Details were published this week in the peer-reviewed journal Nature.
The mainstream press looks to be in love with Intel's announcement, but experts say the technology has plenty of shortfalls and isn't likely to have a commercial impact any time soon. Some even believe there could be negative fallout from Intel's PR machine, which has played up the announcement enormously for the masses.
"It's not a breakthrough worthy of so much attention," says Tom Hausken, director of optical component research at Strategies Unlimited. "I hope that they don't feed a new round of hype, centered this time on silicon photonics".
In fairness to Intel, the modulator device is novel. It's based on a Mach-Zehnder interferometer, a widget that splits the light along two paths and then recombines it. If a phase shift can be introduced into one of the arms, so that it is 180 degrees out of phase, then the two light beams will interfere destructively when they recombine, which gives a "zero" at the output. If there is no phase shift, the beams combine constructively, leading to a "one" at the output. This is a common technique for building modulators.
The clever part is in how Intel induces a phase shift. Rather than using the thermo-optic effect (heat) to change the refractive index of the device, as earlier devices did, the device has an oxide electrode at the top. When a voltage is applied, charge carriers accumulate next to the electrode, which changes the refractive index locally. This effect is fast because no current actually flows in the device.
The modulator was developed at Intel's basic-research organization, the Corporate Technology Group, and involved researchers at the Santa Clara, Calif., headquarters and in Intel's Jerusalem location.
The company praised its own achievement as a key step towards building optical components entirely out of silicon. This is a natural target for a company that's such an expert in silicon electronics, and it's also the target of research efforts the world over.
"This is a significant step toward building optical devices that move data around inside a computer at the speed of light," said Patrick Gelsinger, Intel's senior vice president and CTO in a statement. "It is the kind of breakthrough that ripples across an industry over time enabling other new devices and applications." The company believes its work will pave the way towards the more widespread use of optics inside and between computers.
Those are big claims. But there are plenty of reasons why Intel's modulator isn't going to change the world any time soon.
For starters, the whole idea of moving data at the speed of light is misleading, says Hausken. What really matters is the modulation speed you can achieve, and in Intel's prototype device, this isn't very high.
Intel's modulator has a bandwidth of 1 GHz. Traditional encoding techniques assign one bit to each modulation cycle, which means a 1GHz modulator transmits data at 1 Gbit/s. This may be just about fast enough for some applications in datacom, but many telecom applications would require rates of 10 GHz.
Second, by Intel's own admission, this is a very early-stage device. There is plenty of work to be done before its performance approaches that of commercially available modulators. Parameters such as optical loss must also be improved.
"It is now easier to foresee the fabrication of even faster silicon based modulators, but the technical challenge is still substantial," Graham Reed, from the University of Surrey in the U.K., wrote in an article that accompanied the Nature paper. Reed is from the research team that spun out Si-Light Technologies Ltd., a company working on silicon photonics.
Third, Intel has only built one piece of the silicon puzzle, and that's no good on its own. The showstopper for silicon optics has been the fact that silicon doesn't emit light easily, so it's been difficult to make a silicon laser. Although a couple of research teams have made progress towards this goal, right now they still seem to be some way off from having commercial products (see Silicon Starts to Shine).
"Making a modulator is not a problem," says Andy Carter, VP of research and development for Bookham Technology plc (Nasdaq: BKHM; London: BHM). "But you've got to have a laser. At the moment, you can't make a laser out of silicon, so you've got to have two materials and co-package them. At the low end, that's too expensive to do."
Will Intel be making a silicon laser? Apparently not. "Right now the activities are around doing everything else in silicon," says Mario Paniccia, director of Intel's photonics research lab. Intel's approach will be to attach a III-V laser -- one made of materials such as gallium arsenide or indium phosphide, he says. Ultimately Intel would like to wrap CMOS circuitry around the laser as well.
So if you haven't got a silicon laser anyway, why not use the best material for the job? And it isn't silicon!
What may prove the most difficult barrier to overcome for silicon photonics is economics. Although the equipment and many of the manufacturing methods for producing silicon photonics already exist, that's not enough. Electronics rely on chips being produced in huge volumes, something that's not likely to happen with photonics, at least for the foreseeable future. "There isn't enough volume [in photonics], and you can't get economies of scale for low volumes," Hausken notes.
— Pauline Rigby, Senior Editor, Light Reading