Slow down to speed up

Slowing down light could speed up telecommunications networks - E&T investigates.

Light, nature's 300,000km/s speed demon, is being domesticated by a select group of physicists who are slowing it to just a few metres per second - the sort of speed a Sunday cyclist could easily achieve.

Fibre-optic communication is expected to be one of the first applications to benefit from the taming of light, which began in the United States of America just over a decade ago.

How can light be slowed to a crawl? It all starts with the realisation that there are two different parameters for measuring the velocity of a wave: its phase velocity and its group velocity. Phase velocity describes the speed at which the wavefronts move; group velocity describes the speed at which a pulse of light moves through a resonant optical material.

"This conceptual difference has been known for 100 years," says Robert Boyd, the M Parker Givens professor of optics and professor of physics at the Institute of Optics, University of Rochester. "But what has much more recently happened is that people have realised one can find physical situations in which these two velocities are very different."

Boyd explains that the phase velocity is never very different from c, the speed of light in a vacuum.

"The phase velocity is c divided by the refractive index, but the biggest refractive index of anything that occurs in nature is about four. That means we can only slow down the phase velocity by as much as a factor of four."

Group or phase?

It was only when researchers started to experiment with the group velocity that they discovered this other parameter offered them significantly greater control. "There are physical situations that can easily produce a group velocity that is ten million times smaller than c," says Boyd.

The group velocity of light is determined by the change of the refractive index of the material through which it is passing with the frequency of the light. The trick to generating extremely slow light is to find a material system in which the refractive index changes very quickly with frequency.

There are at least half a dozen physical processes that can achieve this. Early experiments at Harvard University and Stanford University used very low-temperature atomic gases, in a technique called electromagnetically induced transparency. Results were impressive, but many doubted the real-world engineering potential of a method based on cumbersome refrigerated gas.

Professor Boyd and his team at the University of Rochester are recognised as pioneers in the use of room-temperature solids for slowing light. Their work is based on two other physical processes: stimulated Brillouin scattering and coherent population oscillations.

"We are using non-linear optics," says Boyd. "Non-linear optics means using one beam of light to control another beam of light. We use one beam of light to change the refractive index of a material, and then the signal that we are trying to slow down 'feels' that modified refractive index."

The scientist believes he has now reached the point where he knows enough about how to make light go slow to start working on practical applications.

Putting the brakes on data packets

For all the speed that fibre-optic networks have brought to the telecoms world, they still suffer from a number of inefficiencies that slow them down and make them expensive to operate.

Such inefficiencies arise from the fact that high-speed fibre-optic lines are mainly used to transport packet-switched traffic, often over very long distances. Unlike circuit-switched communications, which use dedicated, point-to-point channels, packet-switched networks break up their original content into multiple digital data packets, which are then individually routed over the network.

As each packet leaves its source, it knows what its final destination is from the address contained in its packet header, but doesn't know how it will get there. Packets hop from switch to switch, sharing the channel with other data packets. If two packets arrive at a switch or router at once, the device can only process one of them. To date there has been no straightforward way to slow down or store packets of light to avoid them contending for the same switching resource. Instead, optical packets are turned into electronic signals so that they can be stored, switched and routed, and then turned back into optical packets for onward transmission. The problem with this approach is that it cuts the throughput of the optical network, as well as costing energy for every conversion between the optical and electrical domains.

"What one could do with the help of slow light is to put one of the two data packets [arriving simultaneously at a switch] into an optical buffer while the other one clears the switch," Professor Boyd explains. "Controllable slow light is crucial for this. Just making light go slow does not help - you have to be able to control the slowness."

Pei-Cheng Ku, an assistant professor at the University of Michigan's electrical engineering and computer science department, says that the motorway system provides the best analogy for this process and the advantages that slow light would bring.

"Let's imagine that all the cars on a motorway are the data packets being transmitted in a fibre-optic network," he says. "If two cars want to go the same way at the same time but there's only one lane, one of the cars will obviously need to slow down a little bit to let the other car go first, or they will collide.

"A similar process takes place in a fibre-optic network today. Being able to control the speed of light would be the equivalent of giving brakes to the data packets so they can independently navigate through the high-speed network."

As things stand today, each 'car' is constantly forced to pull off the motorway so someone can ask where it is intending to go, check everything is in order, fill up the tank with expensive petrol, give it directions to its next stop and wish it good luck.

There is a second way in which slow light could benefit fibre-optic communications, and that is in the process technically known as 'regeneration'. The transmission of light pulses over very long distances requires the use of amplification. Amplification brings noise, and noise may cause data pulses to go out of sync with their time windows, the slots set by a master clock signal that say when the light should be sampled to determine each data bit.

"The ability to slow down or speed up the velocity of light controllably on a bit-by-bit basis would allow you to take each pulse and [re-]centre it in its time window," says Boyd.

The search for an optical memory

Of these two potential telecoms applications, regeneration is seen as more likely to reach commercial deployment first. This is because the resynchronisation of data packets doesn't normally require a delay longer than just one bit (or a pulse width) in the light carrying the signal.

"Even the devices we have today may already be able to do that," claims Professor Ku. He believes that, pending some minor systems engineering issues that need resolving, a slow-light optical retiming mechanism for data pulses could be commercially available within two or three years.

When it comes to the more complex buffering application, network operators will have to wait a little longer. A key requirement for this application will be the development of an optical memory. But there are a number of technical challenges that will need to be overcome before such a device can be built.

One of them is a variable known as the 'bandwidth/delay product'. This is the nearly constant relationship between how much an optical signal can be delayed and how much bandwidth can be squeezed into an optical memory.

As Ku puts it, "if you want to delay more (in order to create a larger memory), then the amount of optical signal that you can squeeze into the memory is reduced as well. That creates a physical limit for the development of a practical optical memory, which needs to be able to store at least several hundred bits of data with the current architecture."

In the same way that a delay of one pulse width would be enough for the regeneration application, data packets processed by modern routers would require optical memories capable of storing more than 1,000 pulse widths.

While there are efforts by various physicists to achieve this kind of performance, "the amount of delay you can generate today is actually quite limited," Ku admits. "That's why I think [we are still] five to ten years from seeing the development of a practical optical memory."

Boyd is equally cautious in his optimism: "We know that there are industrial laboratories that are working on this. What is certainly known publicly is that IBM is working on slow light. And, if industrial laboratories are interested, they certainly think that there are applications. I can easily imagine that, within five years, some of these ideas are going to show up in commercial devices. But that's just a guess."

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them