Computing at the Speed of Light
With Moore’s law type advances showing signs of stagnation and decline, researchers around the world are hard at work innovating techniques to improve the speed of computing. A research duo from Northeastern University has come up with a breakthrough that could lay the groundwork for a new generation of fast, powerful computing devices.
Assistant professor of physics Swastik Kar and associate professor of mechanical and industrial engineering Yung Joon Jung have created a device that uses optical and electronic signals to perform basic switching operations more efficiently.
At the most essential level, computing is comprised of a series of on-off switches. It takes billions of these operations to carry out even the simplest of computing tasks, so making this switching process even the tiniest bit faster can have a strong net positive effect on overall efficiency and productivity.
“People believe that the best computer would be one in which the processing is done using electrical signals and the signal transfer is done by optics,” Kar said. “It would save precious nanoseconds.”
The partnership began a couple years ago. Kar’s speciality was graphene, an emerging carbon-based material, prized for its strength and conductivity, and Jung’s research centered on carbon nanotubes, nanometer-sized cylinder of carbon atoms.
Early on, the research team made a startling discovery. They found by taking the metal out of traditional nanotube photodiode devices and replacing it with carbon, light-induced electrical currents rose much more sharply. “That sharp rise helps us design devices that can be turned on and off using light,” Kar said.
To better understand the curious phenomenon, the Northeastern team collaborated with Young-Kyun Kwon, a professor from Kyung Hee University, in Seoul, Korea, on the computational modeling. They then got to work building logic circuits that could be manipulated both electrically and optically. The resulting prototype marks the first time that electronic and optical properties have been integrated onto a single electronic chip.
“What we’ve done is built a tiny device where one input can be a voltage and the other input can be light,” Kar told IEEE Spectrum.
The team actually developed three devices: an AND Gate, which requires both an electronic and an optical input to generate an output; and an OR Gate, which can generate an output if either sensor is engaged. The third device works like the front-end of a camera sensor and consists of an array of 250,000 photoactive elements assembled over a centimeter-scale wafer. It functions as a four-bit digital-to-analog converter.
The nanotubes are created in a solution and placed on a patterned silicon/silicon oxide substrate, which should make the technology compatible with existing CMOS processes, according to Jung.
By using light for data movement and some of the logic operations, the technique could pave the way for a new generation of faster computing chips, according to the researchers. Computers process billions of steps each second, so improving their capability begins with the “demonstration of improving just one,” notes Kar.
A paper describing their research appears in a recent edition of journal Nature Photonics.