Getting silicon to emit light is no easy feat, but it remains the dream of many photonics engineers, who almost reflexively refer to it as “the holy grail.” Now a team of materials scientists at the University of Pennsylvania say they might have managed it.
“This is the first demonstration of bulk silicon emitting light in the visible range,” says Ritesh Agarwal, head of the Nanoscale Phase-Change and Photonics group at UPenn.
A silicon light emitter would be a big boon to the silicon photonics systems being developed by IBM, Intel, Luxtera, and others. Such systems could route data optically between processors at high speeds, overcoming a bandwidth bottleneck imposed by the physical limitations of copper interconnects. And unlike photonics based on III-V semiconductors, such as gallium arsenide or indium phosphide, silicon photonics would rely on the same CMOS manufacturing processes used to make other chip components, which would hold costs down while avoiding the incompatibilities of different materials.
The only problem is that silicon really doesn’t want to emit light. That’s because, unlike gallium arsenide and indium phosphide, it has an indirect bandgap, which means excited electrons and holes are more likely to emit heat than light when they recombine.
To overcome that difficulty, Agarwal and his team turned to the plasmonic effect. Light striking the surface of a metal in contact with a dielectric, such as silicon dioxide, creates a surface plasmon, an electromagnetic oscillation that is strongly confined to the metal-dielectric interface. Agarwal coated silicon nanowires with a 5-nanometer layer of silicon dioxide and laid them down on a glass substrate. He then applied a 100-nm layer of silver atop the wires. The silver covered all of the wire, except the part that was in contact with the substrate, forming a shape like the Greek letter omega. The interface between the silver and the silicon dioxide created a small region where light could be amplified.
The team then shone a blue laser with a wavelength of 458 nm on the nanowires. Afterward, they measured light in the 470- to 700-nm range coming out of the silicon. The plasmonic cavity had forced the excited electrons to combine with the holes and emit a photon before they could drop to their natural heat-producing, lower-energy state.
The work is at an early stage and would need a lot of effort to be made practical, Agarwal says. He’d like to improve the quantum efficiency of the system—the ratio of photons generated to photons taken in—to 5 or 10 percent, up from the 1 percent he has demonstrated. In this initial experiment he achieved diffuse light, but he wants to build an actual laser. And as is generally the case in laser development, he must replace energy input from another laser with an electric current.
Mario Paniccia, Intel fellow and general manager of silicon photonics operations at Intel, calls the UPenn work “an interesting new approach.” But the holy grail of a silicon laser is still a long way off, he says. “People have been working for decades to make an efficient light emitter.”
One drawback of Agarwal’s technology, Paniccia says, is that it emits visible light. Silicon is opaque at visible wavelengths, and Intel’s work uses infrared light, at 1310 nm. Agarwal counters that shorter-wavelength visible light makes for waveguides much closer in size to other components on a computer chip. He adds that computer networks aren’t the only aim of the experimental silicon emitters, which could shine in a variety of applications that need visible light.
In his previous position, as head of Intel’s Photonics Technology Lab, Paniccia developed an integrated silicon photonics system that included a hybrid silicon/indium phosphide laser, a silicon modulator, and a germanium photodetector. Intel announced in January that it will soon be commercializing an optical link, based on that system, that transmits data between racks of servers at a rate of 100 gigabits per second. Engineering samples have already been distributed to companies that will use the technology, although no sales date has yet been announced.
Others have trodden the silicon path before. Back in 2005, semiconductor company STMicroelectronics, of Geneva, was touting silicon light emitters that were doped with rare earth elements to produce light emissions. Company spokesman Michael Markowitz says the technology turned out to be not as promising as hoped. “Many years ago, after a review of our market and technology priorities, we decided to stop this activity,” he says.
Still, Paniccia says, silicon photonics is on its way. “I think we’re at the beginning of the revolution of this technology being commercialized,” he says. He calls Intel’s hybrid laser “a good compromise” that doesn’t achieve holy grail status. “I would love to have a silicon laser. I encourage the research in this area,” he says. “I think it’s a very hard problem.” And it’s exactly what Agarwal’s group is working on.