In April 2009, an international research consortium demonstrated all-optical signal processing on a silicon-based device at speeds of more than 100Gb per second. Not only did the researchers prove that ultra-fast communication was possible on such a device, they also demonstrated that all-optical data communication was practical.
As Pieter Dumon of Belgium-based microelectronics research organization IMEC wrote in Nature Photonics at the time, "We have presented the first experimental proof of ultra-fast communication signal processing using silicon-based devices for transmission speeds above 100Gb/s. This is, to the best of our knowledge, the fastest silicon photonic optical signal processing demonstrated."1
Research into silicon photonics -- using silicon chips to send and receive data-carrying light signals -- is flourishing. Electronics giants such as IBM and Intel have already invested vast research funds into producing such devices, including hybrid lasers, electro-optic modulators and nanophotonic switches.
Likewise, U.S. startups such as Luxtera and Kotura are busy delivering commercial products for the optical communications industry.
Luxtera was founded in 2001 by researchers from Professor Axel Scherer's laboratory group at the California Institute of Technology. Scherer is known for fabricating the world's first semiconducting vertical-cavity laser at Bell Laboratories. Professor Eli Yablonovitch is another co-founder. The fabless semiconductor company is using silicon photonics to build relatively complex electro-optical systems in a production silicon CMOS process. In 2007, it unveiled a 40Gb/s optical cable, making it the first business to market a product that monolithically incorporated active optics for data communications manufactured with low-cost silicon-based chip processing. The cable provides electrical input and output at both ends but transmits data optically.
Kotura formed in 2004, following the merger of U.S. components startups Lightcross and Arroyo Optics. The company has been mass-producing silicon photonics products for communications, computing, sensing and detection applications, for more than three years. In September 2008, the company signed a five-year, $14 million contract with Sun Microsystems to develop integrated optics devices for the research arm of the U.S. Department of Defense, DARPA, as part of its Ultraperformance Nanophotonic Intrachip Communications program. As part of DARPA's Photonic Analog Signal Processing Engine with Reconfigurability program, the business also partnered with Telcordia in June 2009 to design programmable filters for radio-frequency applications.
Research in universities and other organizations around the world is also flourishing.
Why the drive to develop silicon photonic devices? Historically the linewidths on traditional silicon chips have been reduced according to Moore's Law; as a result chips will soon contain one billion transistors running at many gigahertz.
Unfortunately this near-limitless ability to process data within a chip is increasingly handicapped by the transmission bandwidth capabilities of electronics connections, such as the copper connections between neighboring components. However, integrating high-performance optics devices directly with silicon electronics on a chip could overcome this problem.
In theory, a range of all-optical components could be connected together using silicon-based waveguides. These are fabricated by etching channels into a silicon substrate in order to guide light, in this case between the various components. The resulting circuit could be used to establish very fast optical communication -- at frequencies of tens of terahertz -- between circuit boards, chips on a board or even within a chip.
The propagation of light through silicon devices is governed by a range of nonlinear optical phenomena. This non-linearity is crucial as it enables light to interact with light, permitting functions such as wavelength conversion and all-optical signal routing, in addition to the passive transmission of light.
But despite these materials properties, maximum data rates demonstrated in bare silicon waveguides to date are around 10Gb/s. Silicon alone lacks the strong non-linear properties required to fabricate a device that can deliver the terahertz processing speeds promised by all-optical communication. To rectify this situation, additional organic non-linear materials can be integrated with silicon to produce a device that offers the best of both worlds; the high non-linearities of organic materials with the very high mode field concentrations of silicon waveguides.
Figure 1. Passive Si photonic devices fabricated by IMEC. Clockwise from top left. Arrayed waveguide grating, spiral delay line, photonic crystal bend, grating coupler, ring resonator. (Photonics research group, Ghent University - IMEC)
The silicon-based waveguide fabricated by Pieter Dumon of IMEC and colleagues is a cutting-edge example of this. The researchers used optical lithography -- widely employed by the semiconductor industry around the world -- to fabricate a 4mm-long silicon slot waveguide. Slot waveguides, formed by depositing two strips of silicon onto a substrate, were first demonstrated by a team at Cornell University in 2004 and are very effective at confining and guiding light.
Dumon and his colleagues then coated the slot waveguide with organic DDMEBT molecules using molecular beam deposition. This well-established technique promotes the growth of a uniform layer to prevent light scattering and optical losses.
"[To optimize optical properties] it is of paramount importance to homogeneously fill the narrow slot of the waveguide with the organic material, without forming any voids. The DDMEBT molecule has a geometry that promotes the formation of such an assembly," explains Dumon. "Other research groups have used organic polymers, which have relatively long molecules, but as far as I know we are the only group that is using these shorter organic molecules."
Using the waveguide, the researchers went on to perform all-optical demultiplexing of a 170.8Gb/s signal to 42.7Gb/s signal using four-wave mixing, the fastest silicon photonic signal processing demonstrated to date. They believe this demonstration is a key step towards developing complex silicon-based photonic integrated circuits for all-optical communications.
But what about manufacturing? Are silicon-organic hybrid devices, such as this waveguide, ready to be integrated with silicon chips in high volume CMOS production processes?
"There are no major hurdles," asserts Dumon. "We are working on integration and it will also take time to get through the telecommunication qualification procedures, but [these devices] can be made manufacturable for high volume without any major issues."
Professor Michael Hochberg of the University of Washington agrees. Like Dumon and colleagues, Hochberg has been working on silicon photonics and has developed a range of polymer silicon hybrid devices. Hochberg was also a founder of Luxtera.
"It should be possible to integrate devices that use silicon waveguides into CMOS electronics flows," he says. "Yes, practical challenges exist but as a founder of Luxtera, we have done this in the past, we've run electronics and optics through a CMOS fab."
Industry state of play
For years, many in the industry have asserted that silicon isn't good for photonics. Today, most photonics devices, such as lasers, modulators and detectors, are made of more expensive non-silicon materials such as indium phosphide and lithium niobate.
However, thanks to aggressive development from the likes of Intel, IBM and IMEC, silicon photonics is now a rapidly advancing field that is predicted to displace large portions of the existing industry that rely on discrete assemblies of electronic and photonic devices.
In March 2008, IBM researchers unveiled what they described as the world's tiniest nanophotonic switch for routing light on a silicon chip, capable of switching up to nine 40Gb/s optical channels. Components such as this will be used in the more complex silicon-based photonic circuits of tomorrow. Intel has also developed several key silicon photonics devices, including the first continuous-wave silicon laser and the first gigabit speed silicon modulator.
Figure 2. A fiber I/O coupler. (Photonics research group, Ghent University - IMEC)
Both Dumon and Hochberg believe that we could see this displacement of existing industry happen over the next few years.
As Dumon highlights, high-speed signal processing with hybrid waveguides is only one aspect of IMEC's research into silicon photonics. "We've made a lot of progress in developing on-chip sources, on-chip detectors, devices such as wavelength demultiplexers and so on," he says. "We still need to get everyone convinced that we can integrate organics on a silicon chip in a manufacturable way. However, in about three years from now we will have a nice demonstration of a telecommunications system where you need very fast wavelength conversion."
Meanwhile, Hochberg asserts that the next few years will not only see a proliferation of silicon-based systems, but new companies as well.
"It won't surprise me to see another generation of companies in the next year or two as there are so many opportunities for this work," he forecasts. "Luxtera, for example, has certainly broken a lot of new ground, but it's only going after one of many potential applications for this sort of technology. I think we will see a number of players emerging in the next few years."
Rebecca Pool is a freelance technology writer based in the UK.
1. C. Koos, P. Vorreau, T. Vallaitis, P. Dumon, et al., All-optical high-speed signal processing with silicon-organic hybrid slot waveguides. Nature Photonics 3, no. 4, 784189, 2009. doi:10.1038/NPHOTON.2009.25