Holy Grail: Light from Silicon

Lasers and microprocessors could be made on the same chip

3 min read

Data has become a creature of light. The high-bandwidth transmissions that link computers across the Internet are possible only because lasers can send vast numbers of signals through tiny glass fibers. But silicon, the stuff of memory and microprocessors, has little to do with it.

Instead, for bits to become illuminated, silicon circuits must first connect to separate lasers made of more exotic and expensive compound semiconductors. If silicon emitted a useful amount of light, the microprocessors and memory could export data directly as light, simplifying circuit boards and speeding computation by avoiding the data bottleneck that copper interconnects are becoming. Indeed, silicon lasers might even speed signals, such as the all-important clock pulse, within microprocessors.

Since the early days of the laser in the 1960s, physicists have known they could squeeze a few photons out of silicon. But they accomplished little more than that until 1990, when work by British researcher Leigh Canham provided the first glimmer of hope that the speed of the microprocessor could be married to the huge bandwidth of optics. In a burst of excitement, as many as 100 research groups around the world tried to stimulate silicon to produce more light, but for years they managed to get no more than about 1 percent of the energy they poured in to come out as photons.

Silicon has an indirect bandgap. That means the energy states of the charge carriers--negative electrons and positive electron deficiencies called holes--don't match well, and when they combine, they're much more likely to produce a vibration than a photon, making more heat than light. So optoelectronics are instead made from costly compound semiconductors, such as gallium arsenide. And because the crystalline structure of gallium arsenide is different from that of silicon, they don't fit together on the same chip.

But the dream of silicon-laser-powered computing and communications is closer than ever. Salvatore Coffa, research director of soft computing, silicon optics, and postsilicon technologies for STMicroelectronics NV, in Geneva, says STM will sell the world's first commercial device based on silicon light emission by the middle of this year. An optocoupler the company has in mind would use a silicon light-emitting diode (LED) to convert an electrical signal into light, and then back, to link two circuits with different voltages.

Coffa's research group, in Catania, Italy, has made LEDs that convert 15 percent of the electrons injected into them into photons, an external quantum efficiency in the same range as that of gallium arsenide. They emit 1 mW of light per square millimeter of silicon, which Coffa calls "really bright." With fine-tuning to optimize such characteristics as output and lifetime, these LEDs could also compete in microdisplays or solid-state lighting, he claims.

The silicon in the LEDs consists of nanocrystals, tiny clumps measuring only a few nanometers across. Their size changes the physics, so the silicon is more likely to emit light. Coffa's group dopes these "quantum dots" with ions of rare earths; which rare earth is used determines the color of the light.

Philippe M. Fauchet, chairman of electrical and computer engineering at the University of Rochester, in New York, says Coffa's work seems promising, and a commercial product could really stimulate the field. But an LED doesn't have the power, directionality, and narrow spectrum of a laser. Fauchet believes a silicon laser, if it can be created, would be necessary for optical interconnects on a chip. "In my group, we are working very hard on this," he says. "It is not so easy, of course."

But Coffa sees a light at the end of the tunnel, and it's coming from silicon. "I will be very surprised if you don't see the first prototype of the laser in a year from now," he says. He's not promising, though, that a silicon laser will ever be good enough to provide an optical interconnect.

Cautions another scientist working on silicon lasers, Lorenzo Pavesi of Italy's Università di Trento in Povo: "It could take months, or years, or even forever.

Keep Reading ↓Show less

This article is for IEEE members only. Join IEEE to access our full archive.

Join the world’s largest professional organization devoted to engineering and applied sciences and get access to all of Spectrum’s articles, podcasts, and special reports. Learn more →

If you're already an IEEE member, please sign in to continue reading.

Membership includes:

  • Get unlimited access to IEEE Spectrum content
  • Follow your favorite topics to create a personalized feed of IEEE Spectrum content
  • Save Spectrum articles to read later
  • Network with other technology professionals
  • Establish a professional profile
  • Create a group to share and collaborate on projects
  • Discover IEEE events and activities
  • Join and participate in discussions

3 Ways 3D Chip Tech Is Upending Computing

AMD, Graphcore, and Intel show why the industry’s leading edge is going vertical

8 min read
A stack of 3 images.  One of a chip, another is a group of chips and a single grey chip.
Intel; Graphcore; AMD

A crop of high-performance processors is showing that the new direction for continuing Moore’s Law is all about up. Each generation of processor needs to perform better than the last, and, at its most basic, that means integrating more logic onto the silicon. But there are two problems: One is that our ability to shrink transistors and the logic and memory blocks they make up is slowing down. The other is that chips have reached their size limits. Photolithography tools can pattern only an area of about 850 square millimeters, which is about the size of a top-of-the-line Nvidia GPU.

For a few years now, developers of systems-on-chips have begun to break up their ever-larger designs into smaller chiplets and link them together inside the same package to effectively increase the silicon area, among other advantages. In CPUs, these links have mostly been so-called 2.5D, where the chiplets are set beside each other and connected using short, dense interconnects. Momentum for this type of integration will likely only grow now that most of the major manufacturers have agreed on a 2.5D chiplet-to-chiplet communications standard.

Keep Reading ↓Show less