Chips at Light Speed
New discoveries push optical computing closer to reality.
Fiber optics swells bandwidth capacity wherever it replaces copper or aluminum wires. Without optical communications, the modern Internet and phone network could not exist. But computers and servers that depend on those networks are still tied down by metal-based technology. Now, however, a solution to that problem appears to be in sight-if not right around the corner.
In the past six months, scientists have made a string of breakthroughs that could promise a fundamental change in the way computer chips are built and tied together. Today in computers, data move on pathways made of very thin strands of metal. In the not-too-distant future, the most critical of those pathways could be replaced with fiber circuits carrying tiny pulses of laser light. It may even be possible to dispense with some of the fibers and wires altogether and move laser light through open conduits within the chip.
This change could be critical. The speed of computer chips could reach an absolute barrier in the next decade. Part of the problem: the physical properties of metal. Today’s chips are like miniature networks tied together by an intricate web of copper or aluminum wires, with signals zipping through circuits of millions of transistors millions of times a second. But like an ancient cobblestone road too rough to allow high-speed traffic, metal interconnects can only allow data to move so fast. When trillion-bits-per-second throughput becomes necessary in chips and computers, the metal wires used today won’t be up to the task, many researchers say. Optical interconnects may then have to take the place of metal ones wherever bottlenecks may occur. In effect, researchers want to add optical express lanes to conventional chips.
Laser on a chip
Electricity moves slower through a metal wire than light moves through the air or an optical fiber. That’s because electrical properties such as resistance limit the throughput of metal wires and also produce a lot of heat. Even over the very short distances within a computer or on a chip, there is no real contest-transmitting data by light is a lot more efficient.
Light has another key advantage: Many different frequencies of light can be sent down the same fiber. Such multiplexing, which is done in telecommunications all the time, could allow several metal wires to be replaced by just one fiber that can transmit just as much data.
In optical systems, laser light pulses carry data. Researchers want to develop a silicon laser that can be integrated within the chip and be flickered on and off to produce these pulses. Building that laser is essential to the development of an optical chip.
This debate sheds more light than heat.
- A Stanford University professor argues for optical interconnects.
- An IBM researcher voices skepticism.
- Pavesi’s and Homewood’s papers are available at Nature.
Not surprisingly, placing a laser on a chip is an enormous challenge. Silicon, the fundamental building block of all chips, can barely emit even a glimmer of light, much less an intense laser. While lasers can be made of other materials, integrating those materials with silicon is too costly and difficult to be economical in most chips. But researchers in Europe and the United States have recently discovered techniques to get silicon to amplify light and then emit it with some efficiency, key steps toward a silicon laser.
The first breakthrough came around Thanksgiving when Lorenzo Pavesi and other researchers at the University of Trento in Italy announced that extremely tiny silicon crystals can amplify light when zapped with a laser. A month later, a group of researchers at the University of Illinois said they too had amplified light with silicon crystals. Then in March, British scientists at the University of Surrey led by Kevin Homewood announced that they had developed a silicon light-emitting diode that operates with encouraging, although still somewhat limited, efficiency. None of the discoveries are themselves enough to solve the problem of developing optical interconnections for chips, but all point to an eventual solution. If one can be found, chip makers will have a way to speed up chips and computers without having to abandon silicon-based chip-making-a process in which they have invested hundreds of billions of dollars.
Just silicon and mirrors?
But they are not there yet. Amplifying light in a crystal or LED is one thing; producing a focused laser beam is quite another. To accomplish this, researchers figure they will have to sandwich the nanocrystals or LED between mirrors so that the light bounces back and forth. That bouncing produces a stimulated emission among the silicon atoms that in turn produces a laser beam. Building chips containing properly aligned mirrors around every crystal or LED will require a host of manufacturing innovations, says Michael Tan, a project manager in the communications and optics research lab at Agilent Technologies.
And that’s just part of the challenge. Silicon lasers must be powered by electricity coming from other parts of the chip, not by other lasers as they were in Pavesi’s lab. Pavesi says he has some projects on the drawing board to stimulate his silicon crystals electronically, but he acknowledges that it is his biggest challenge. Also unsolved: a way to route data within an optical chip, Tan says. In traditional chips, the flow of current around the circuits is governed by the transistors, which are tiny switches. Making analogous optical devices on the tiny scale demanded by chip-making is a breakthrough yet to be achieved.
How long will it take to overcome these challenges? Estimates range from three years to never. But other scientists say they can see the goal shimmering before them. “If we are able to demonstrate a silicon laser, the problem of optical interconnects can be solved,” Pavesi says. “The only limit then will be our imagination.”