Silicon microchips, the thumbnail-sized microprocessors that constitute the brains of a PC, are heading for a disaster created by their own remarkable success. As chips get faster, the electrons that carry messages through the tiny metal wires within the integrated circuit are having a hard time keeping up.
One place where this looming problem is particularly acute is in the ultrafast clocks used to pace computation. Roughly speaking, faster clocks mean faster computing; microprocessors now run at clock rates over one gigahertz (a billion pulses per second) and are getting faster all the time. Soon, says Lionel Kimerling, director of MIT’s Microphotonics Center, electrons moving through metal wires will simply be too slow to keep pace. “Assume that somewhere in the future is a 10-gigahertz clock. It’s impossible to distribute that kind of signal electrically,” he explains. The solution, says Kimerling, is tiny pulsed lasers that can distribute the clock signals through the processor chip. “Intel thinks that three gigahertz is a big problem,” says Kimerling, “and that is about two years away.”
Over a dozen research groups are racing to develop miniature optical devices capable of being integrated right into the silicon chip. It would be a kind of optical network to ferry data around the microprocessor, boosting its capabilities in the same way fiber optics have transformed telecommunications. But there’s a problem: silicon is a lousy light emitter.
Silicon’s curse is that it is, in the jargon of physicists, an “indirect-bandgap” material. Other semiconductor materials are good light emitters because when their electrons are kicked up to a higher energy by a current, the electrons can drop right down again and fire off a photon in the process. Pump a lot of electrons rapidly into a higher energy state, and you can make a laser. This is how the semiconductor laser used in a DVD player works, for example. But the laws of physics say that the electrons in silicon cannot travel directly back to a lower state. As a result, the electron usually gives up its energy as heat rather than as light.
There are two strategies for overcoming silicon’s “light” problem. Some researchers, including colleagues of Kimerling’s at MIT’s Microphotonics Center, are developing light emitters and detectors made from silicon’s siblings-semiconductors like gallium arsenide-that can be grafted directly onto silicon chips. Other groups have found ways to get silicon itself to emit the desired light.
In 1996, Philippe Fauchet and his colleagues at the University of Rochester reported a light-emitting diode made from silicon. The device had an important characteristic: an electric current rather than another laser or light source could be used to trigger the light emission. But, says Fauchet, the efficiency of the device in emitting light is too low to interest chip makers. “In these light-emitting devices, the power efficiency is around 0.1 percent,” he explains. “But the minimum acceptable standard in the industry is one percent before they will talk to you.”
Silicon’s light-emitting powers got a boost last November when Lorenzo Pavesi at the University of Trento in Italy found that silicon nanoparticles could amplify light. What made this exciting is that amplification is the first step toward making a silicon laser. “With a laser, it’s a whole new ballgame,” says Fauchet. “Some of the efficiency issues go away.” The nanocrystals, however, have to be stimulated by a laser rather than electric current.
Then in March a group led by Kevin Homewood at England’s University of Surrey discovered another way to get silicon to glow on its own. “Our approach uses absolutely standard silicon technology,” says Homewood. These silicon-based light-emitting diodes are not optimized for efficiency, Homewood acknowledges, but he says they are only a factor of three away from conventional light-emitting diodes. Homewood’s next step is to try to get laser action. “I certainly don’t think the physics is against us,” he says.
Despite these tantalizing hints of success, Fauchet says research in light-emitting silicon faces some tough challenges. “The trouble with all these devices, including ours, is the low efficiency,” he says. “As research, it’s very interesting, but Intel is not jumping yet.”
Still, the future of silicon microphotonics is bright. Whether they’re silicon lasers or light emitters made from some other semiconductor, Kimerling says the integration of optical devices into silicon chips “is the next big step” in photonics. For a multibillion-dollar chip industry built around silicon, the clock is rapidly ticking toward a way to take that step.
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
Data analytics reveal real business value
Sophisticated analytics tools mine insights from data, optimizing operational processes across the enterprise.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.