By linking loops of silicon on a microcircuit, researchers have taken a major stride toward using light to shuttle information within computer chips. The new approach may lead to circuitry that can manipulate exceptionally large amounts of data and introduce the delays often required for a chip to coordinate calculations and communications.
Computer manufacturers are currently cramming multiple electronic-computing modules, or microprocessors, onto individual chips of semiconducting material. For instance, the powerful electronic brain of the Sony PlayStation 3 video game controller jams nine microprocessors onto a single chip. As the number of modules per chip continues to multiply in the coming decade, the data traffic is expected to outstrip the information-handling capacity of the electronic microcircuits that carry and route the traffic, says Yurii A. Vlasov of IBM’s T.J. Watson Research Center in Yorktown Heights, N.Y.
Light-manipulating, or photonic, components might rescue those overburdened electronic circuits, Vlasov predicts. By developing photonic components in silicon, chip makers stand to exploit their existing infrastructure and know-how for working with that material, he notes.
Light can travel within thin ridges left untouched between etched channels in a silicon surface. Light conducted along such ridges can transport far more information than electrons within ordinary copper wires do.
Now, Vlasov and his colleagues Fengnian Xia and Lidija Sekaric, also of the IBM center, have unveiled a promising new type of photonic component—chains of up to 100 oblong loops. Made of extraordinarily precise silicon ridges, the arrays transmit pulses of light from end to end. That’s a feat in itself, given that in previous research, pulses typically died out even in short chains, Vlasov says.
The new arrays of chained loops take up less than 0.1 square millimeter, so they’re suitable for chips, he adds.
The chains, described in the January Nature Photonics, delay more data than other chip-scale light-slowing technologies do, Vlasov says. The prototype components achieve delays by forcing pulses to circle each loop more than 50 times before proceeding to the next one.
“This is a beautiful result,” comments photonics specialist Masaya Notomi of NTT Basic Research Laboratories in Atsugi, Japan.
“It shows the first practical steps toward being able to store optical bits of data,” says electrical engineer Keren Bergman of Columbia University. The IBM loops can store up to 10 bits of high-speed data so far, but they’ll need to store 10 to 100 times as much to be useful, she adds.
More than a year ago, another IBM team led by Vlasov reported retarding light in silicon by a different means—passing it through ultrathin slabs of the semiconductor punctuated by arrays of holes (SN: 11/5/05, p. 292: Available to subscribers at Light Pedaling: Photonic brakes are vital for circuits). Such a structure, known as a photonic crystal, remains a potentially useful component for manipulating light on chips, Vlasov says.
Indeed, Notomi, Takasumi Tanabe, and their colleagues at NTT report, also in the January Nature Photonics, that they’ve developed a new photonic crystal that retards light more than 170 times as much as the 2005 IBM crystal did.
That’s “a great achievement,” Vlasov says.
Scientists are continuing to pursue both the photonic-crystal and silicon-loop approaches to find the most advantageous combination of light retardation, information capacity, and ease of integration with chip production.
