For sending information across continents and around the globe, light is the medium of choice. The ability to send multiple wavelengths at high speeds within fibers has transformed communications. But light could do even better, much better, if it weren’t hobbled by the electronic switches, routers and other devices of optical communications technology.
Since they operate by converting optical signals to electronics and back again, these devices considerably reduce the efficiency of current optical networks. Is it possible to create all-optical circuitry — something analogous to the microcircuitry of “chips” but that doesn’t require converting light to electrical current? It’s a challenge many scientists worldwide are addressing.
Shanhui Fan and Fatih Yanik, Stanford University
Using LeMieux, PSC’s terascale system, to simulate how light behaves, applied physicist Shanhui Fan of Stanford and graduate student Mehmet Fatih Yanik have made notable progress. Using all 3,000 LeMieux processors, they showed that it’s possible to stop light and hold it captured — in an optical holding cell — until a subtle shift in optical features releases it. Unlike earlier attempts to capture light, their finding — reported in 2004 — suggests it may be possible to corral complicated light pulses and, moreover, to do it in a way that integrates easily with existing chip technology.
So far, Yanik and Fan’s device exists only in simulation, but they have teamed with a laboratory group at Stanford to build and demonstrate their scheme. Because of the powerful ability of their simulations to accurately predict how light behaves in fascinating materials called “photonic crystals,” the researchers are confident the laboratory work will yield an all-optical device to stop light in its tracks.
It made news in 2001 when researchers brought light to a standstill for the first time. Two groups at Harvard demonstrated a technique that captured light in clouds of gaseous atoms. But these systems of atomic gases are impractical for an all-optical circuit.
Rather than gases, the Stanford team’s approach relies on photonic crystals — layered materials, often silicon or other semiconductors, made with cavities in patterns within the crystal. Because such a device will operate at room temperature and be only microns in length, it could easily integrate with traditional microcircuitry.
By careful design of irregularities in the patterns of the cavities, photonic crystals can allow — or forbid — the passage of certain wavelengths of light. This handy trick makes them attractive filters, with the potential to act as gatekeepers that allow only selected wavelengths to pass through the crystal on prescribed paths. Exactly which wavelength, or band of wavelengths, can travel through or not depends on the properties of the crystal.
Yanik stumbled on the light-stopping mechanism while using LeMieux to simulate the impact of changing one property of a crystal, the index of refraction — the ratio of light’s speed in a vacuum (well established at 186,000 miles per second) to its speed in a medium, where it travels more slowly. His original goal was a tunable switch — a crystal that could be prompted, by small changes in the refractive index, to allow safe passage to different wavelengths of light.
This graphic from simulation shows snapshots of the positive (red) and negative (blue) electric fields as an optical pulse propagates (left to right) through a photonic crystal, shown in three segments at four times (top to bottom). Resonant frequencies of the cavities (black dots) are tuned to stop the pulse during the time interval shown in the second and third snapshots, until the cavities are detuned and the pulse is released.
For one possible design of such a switch, the simulations indicated the effect could be quite strong. Small changes in refractive index allowed a large change in the bandwidth of allowed wavelengths. And that wasn’t all. “I saw an optical signature very similar to the ones observed in atomic media,” says Yanik. “So the question became, could we use the cavities in the crystal to store electromagnetic pulses, just as they were stored in atomic media? If somehow we could get light into this structure, and then change the properties of the entire structure while the light was inside, we could change the properties of light as well and trap it.”
The idea depends on a phenomenon called optical resonance, which is similar to why long and short pipes in an organ produce notes of different frequency. In an organ, each pipe is cut to the length required to amplify sound waves of a desired frequency. The sound energy bounces back and forth inside the pipe and establishes an unmoving wave pattern, or resonance, at the desired frequency. In the Stanford team’s approach, the role of the organ pipe is played by a waveguide — either an empty channel or closely spaced cavities inside the crystal that allow light to propagate.
Prior to this work, many groups had used optical resonators to trap light of a single wavelength. Optical communication, however, uses light pulses to encode and transmit information, with each pulse composed of many wavelengths. Trapping such a multi-wavelength pulse in a single resonator would lose the information carried by the pulse.
Yanik and Fan’s idea, however, goes a crucial step further by tuning all of the wavelengths within a pulse to the same frequency and, at the same time, adjusting the crystal to resonate at that frequency. They do this by adjusting the index of refraction once the pulse has entered the crystal. As all the frequency components are collapsed to a single frequency, the information becomes encoded by the phase and intensity of light along the waveguide.
Changing the resonance of the crystal, Yanik explains, is like adjusting the spacing of stepping stones across a river. Shifting the crystal’s index of refraction is similar to spreading the stones out, so that photons — the tiniest energy chunks of light — of a particular frequency can no longer hop from stone to stone. They have been trapped. When the pulse needs to be released, the index of refraction is shifted back, the stones move closer together, and the photons zip away.
“The entire idea,” says Yanik, “from refractive-index switches to light-trapping devices, was first realized on a supercomputer.” Once he and Fan identified the light-stopping possibility, Yanik adapted software he’d already written to simulate it. Using almost every one of LeMieux’s 3,000 processors, they simulated a series of possibilities until arriving at a 100-micron waveguide with 120 side-cavities.
“A hundred microns,” says Fan, “fits on a chip, a small distance in practice, but a long distance to simulate.” The beauty of photonics simulations, he explains, is the ability to use the full form of Maxwell’s equations. This set of four equations, named for James Clerk Maxwell, a 19th century Scottish physicist, governs most optical and electromagnetic phenomena. Not so long ago, notes Fan, limitations in computing technology required clever approximations to apply these equations.
“With a system like LeMieux,” he says, “we have the ability to solve the entire set exactly.” This means that the computational experiments precisely mimic physical reality and give the researchers high confidence that their predictions can be realized in the laboratory.
To exploit the large-scale parallelism of LeMieux’s 3,000 processors, Yanik’s software parceled separate parts of the crystal waveguide to separate processors. It took 10 simulations to describe the light-trapping behavior, with each simulation of a light pulse entering the wave guide requiring two hours, which Yanik estimates as a year’s worth of computing on a desktop PC.
The simulations showed that shifting the index of refraction around the pulse forces the wavelengths to adopt a single frequency, and traps the pulse in and between cavities. In the 100 micron, 120 side-cavity waveguide, a 1/10,000th shift in the index of refraction is enough to capture the information in commonly used pulses of light.
Another surprising result of the simulations, says Fan, is that if the index of refraction were tuned beyond the point where the light pulse screeches to a halt, the pulse would not merely stop, but reverse in its tracks, backing out of the crystal as though it were a train reversing direction to re-emerge, caboose first, from a tunnel. This time reversal effect, he says, might prove useful in repairing signal degradation.
Efforts to build the device in the lab, in collaboration with Stanford colleagues Martin Fejer and James Harris, are now running parallel to more simulations. “What we’ve done so far is a two-dimensional simulation,” says Fan, “as a proof of principle. We are now extending it to a three-dimensional simulation to arrive at the exact structure the device needs to take.”
For optical networks, a device that can catch and hold light for an arbitrary length of time offers promise to alleviate the congestion that happens when too many pulses arrive simultaneously at a network junction. Beyond that, there’s the promise of quantum computing, the vision of transistors that manipulate single photons rather than electrons. It’s a future, perhaps sooner than we think, in which circuits will be a thousand times smaller and faster. Yanik and Fan’s simulations with LeMieux bring us a step closer.