11 Dec 2007
IBM's silicon optical modulator is one hundred times smaller than previous devices, which could enable high-density optical integration in silicon.
Putting all optical functions onto a silicon chip using standard CMOS technology has been the target of research teams the world over, because it might eventually result in the mass production of cheap silicon-based optical components for telecoms.
Now, IBM has made an optical modulator – a device used to encode information on a light beam – that is 100 times smaller than previous devices. This could enable a high level of optical integration that would make it possible to put lasers, modulators, detectors and electronics all on the same chip. The work was published in the journal Optics Express.
IBM is not the first to build a fast optical modulator out of silicon. That accolade goes to Intel, which announced a 1 Gbit/s optical modulator back in 2004. What's more, Intel researchers have since optimized their device to handle 40 Gbit/s data rates – this was reported in the summer – while IBM's device runs at a more modest 10 Gbit/s.
However, Intel's device was around 3 mm long, which means that you can't pack very many onto a silicon chip. In contrast, the key structure on IBM's device is just 100 to 200 µm long, making it similar in size to other optical components such as lasers.
"This [IBM's announcement] is quite significant, as a major criticism of silicon photonic devices has been their relatively large footprint," says Kevin Homewood from the University of Surrey in the UK, and founder of Si-Light, a company developing silicon photonics.
Like conventional optical modulators, IBM's device is based on a so-called Mach-Zender interferometer. This works accoby splitting light along two paths, altering the refractive index in one of the paths to delay light by just the right amount, and then recombining it. If both paths are the same, the light combines constructively, leading to a "1" at the output. But when a 180° phase shift is introduced into one of the arms, then the two light beams will interfere destructively to give a "0" at the output.
Conventional modulators need long arms because the interaction in the phase shifting section is weak. IBM researchers have managed to increase the strength of the interaction, and thus shorten the distance required for the device to work. They do this by using "nanophotonic" waveguides – in other words, waveguides that are really small!
Nanophotonics in practice
Each waveguide measures approximately 400 nm across and just 220 nm high, giving it dimensions that are about the same as the wavelength of 1.5 μm light inside the silicon (which has a refractive index of 3.5). "Despite the extremely small cross-section of our 'silicon photonic wire' waveguide, over 90% of light of the waveguide mode is still confined within its physical dimensions," explains Yuri Vlasov, who leads the team at IBM’s T J Watson Research Center in New York. "Strong overlap between the injected charges and optical mode is achieved. This is the main reason for the extremely high efficiency of the phase shifter in our modulator."
However, there must be more to the story than this explanation suggests, because both IBM's and Intel's device seem to make use of tiny waveguides and have a similar figure of merit for the phase shifter: 3.6 V mm in IBM's case and about 4 V mm for Intel.
According to Megan Langer, a spokesperson for Intel, there is a clear difference is in application focus. IBM's primary interest in this technology appears to be for on-chip interconnects that would replace copper sometime in the next 10 to 15 years. Intel, on the other hand, believes that chip-to-chip and board-to-board interconnects will be a more immediate application, perhaps in the next three to five years.
"This is definitely an exciting time for silicon photonics. We are just at the beginning of the whole new era and it seems that breakthroughs are coming more and more often," Langer comments.