29 August 2007—Scientists say one way to build a really fast computer is to use light rather than electricity to perform calculations. Now researchers from Mikhail Lukin’s group at Harvard and the Danish National Research Foundation Center for Quantum Optics and the Niels Bohr Institute at the University of Copenhagen have taken a big step toward this goal with the first feasible plan for making a transistor that uses photons of light instead of electricity. Details were recently published online by the journal Nature Physics.

Unlike other schemes, this new optical transistor could be controlled with just one photon, making it very efficient. And it would work for a broad range of frequencies of light, instead of just one (as with some previous proposals), making it easy to use.

A single photon transistor is “the holy grail of optical computation, both classical and quantum,” says John Howell, a professor of quantum optics at the University of Rochester, who was not involved in the research.

If the Harvard researchers can put their theory into practice and build a photon transistor—and they’ve already started initial experiments, IEEE Spectrum has learned—the result would pave the way to an all-optical computer that could potentially do much faster calculations than even some supercomputers.

The digital computers we are familiar with today calculate by channeling electrons through a network of transistors. But even the fastest computer processors available are limited by the speed of their electrons, which travel at a fraction of the speed of light. Substitute photons for electrons and computing could happen at light speed. That’s the idea, anyway. The reality is, to this point, no one has been able to make such an optical computer’s most basic component—a transistor driven by photons.

An electronic transistor is basically a semiconductor device that amplifies a voltage signal. There are three electrodes in a transistor—one of which is used to control the current flowing between the other two.

A photon transistor would work in a similar fashion. Two beams of light would enter the device. One, which could be as weak as a single photon, carries the signal to be amplified. The other would carry the light to be modulated by the signal. An amplified light signal would then exit the device.

Amplifying a light signal is inherently much more difficult than amplifying an electrical one. Unlike electrons, which have negative charge and interact easily with matter and each other, photons in a beam of light are electrically neutral and do not interact much with matter. A photon is also typically not affected by another beam coming into its path. 

A promising way to get light to interact with matter and itself is to create what physicists call a surface plasmon. Surface plasmons form at the junction between a nonconducting material, or dielectric, and a metal. Metals have lots of free electrons, which oscillate when light shines on them. But when a dielectric borders a metal, the movement of the electrons is curtailed, because electrons cannot enter the dielectric. That forces the jiggling electrons to move in waves of density—like sound—along the junction.

In theory, surface plasmons offer a way to create a transistor by using one beam of light falling on the metal to modulate another beam through electromagnetic interaction. But in practice it has proved to be quite difficult, because surface plasmons do not interact well with optical fibers, which are typically used to feed light to the metal and dielectric junction. Most other sources of light fare no better.

Last year, a group at Queen’s University in Belfast, Northern Ireland, led by Anatoly Zayats came close, with a surface plasmon device made up of a gold disk pocked with 360 holes and coated with a polymer. There have been a number of other demonstrations as well, but to this point amplification works only over a very narrow range of wavelengths of light. It would be better to have an approach where a large range of different wavelengths would be able to make the same device work, because then the lasers involved would not have be to as sensitively tuned to each other.

Lukin’s group at Harvard realized that a possible solution lay in using conducting metal nanowires. Nanowires can guide light just as optical fibers do, but they also produce plasmons on their surface—better plasmons than you’d find in other schemes physicists have tried, says Peter Bermel of MIT, an expert in quantum optics. Because the nanowires are thinner than optical fiber, the electromagnetic fields produced in them by light tend to be more intense.

When the surface plasmons “become confined to very small regions like a nanowire,” they acquire unique physical properties, says Darrick Chang, a Ph.D. student at Harvard and the lead author of the Nature Physics paper. “It means any photons that fall on them have to interact with them. This allows the transistor action.”

Researchers in the quantum optics field that Spectrum polled are all uniformly impressed with the Harvard scheme.“[Our] experiments used brute force,” says Queen’s University’s Zayats, comparing his group’s earlier work with the Harvard work. “The proposal of Lukin’s group to build a single-photon transistor by coupling surface plasmons is much more subtle and practical.” 

The Harvard scheme “seems like one of the best proposals I have seen,” says the University of Rochester’s Howell. The first experiments based on the suggested technique are under way, says Chang, but they are keeping the details a secret for now.