Telecommunications

Could graphene switches lead to 100-times faster internet?

Could graphene switches lead to 100-times faster internet?
Graphene continues to strengthen its reputation as a wonder material, this time in the field of telecommunications (Image: Shutterstock)
Graphene continues to strengthen its reputation as a wonder material, this time in the field of telecommunications (Image: Shutterstock)
View 1 Image
Graphene continues to strengthen its reputation as a wonder material, this time in the field of telecommunications (Image: Shutterstock)
1/1
Graphene continues to strengthen its reputation as a wonder material, this time in the field of telecommunications (Image: Shutterstock)

Researchers from the Universities of Bath and Exeter have shown that a few layers of graphene stacked on top of each other could act as a formidable material for optical switches, delivering speeds up to 100 times faster than current telecommunications technology.

Optical switches are a crucial part of our telecommunications infrastructure and although their performance is already quite impressive, researchers have long suspected that a class of materials named semimetals (which includes graphene) could be used to make them even better.

Graphene and next-gen telecommunications

When light hits a doped semiconductor, the resulting energy makes electrons jump from a lower-energy state (the valence band) through a small gap (the energy gap) and into a higher-energy state (the conduction band) in which electrons can move more freely, conducting electricity. Eventually, when the initial energy has been exhausted, the electrons go back to their original state, through the band gap and back to the valence band.To establish whether a material can be used to process information quickly, researchers need to look closely at how electrons flow up and down the energy gap in that material. The critical factor is the recombination time – the time it takes for an electron to make the trip back from the conduction to the valence band.

In recent years researchers have advanced the hypothesis that graphene could be an excellent material for optoelectronics, but they couldn't test their hypothesis because graphene is a semimetal, or a "zero-gap semiconductor." In other words, in graphene there is no energy gap between the valence and the conduction band, and this means that scientists aren't able to effectively analyze its potential with current techniques.

The Bath researchers found a way to measure recombination time in semimetals such as graphene. They did so by measuring how the electrons moved in the infrared part of the spectrum, transitioning between different quantum states. What this adds up to is that the researchers' suspicions proved to be right: while ordinary optical switches respond at rate of a few picoseconds (around a trillionth of a second), the physicists observed that the recombination time of an optical switch using a few stacked layers of graphene was of the order of only one hundred femtoseconds – nearly one hundred times quicker.

This could open the door to significantly faster telecommunication, but it is also an important step toward the development of graphene-based quantum cascade lasers, which could be used for anything from remote sensing of environmental gases and pollutants, breathalizers, medical diagnostics, lasers, collision avoidance, and even cruise control for your car.

The research is published in the journal Physical Review Letters.

Source: University of Bath

4 comments
4 comments
silentnelite
Very informative and easy to understand article. Thanks for this!
RichDavis
So, does this mean they have to dig up all of the fiber optic lines they have been putting to the homes?
Onihikage
I think the endpoint of network interfaces is a pure optical I/O, where the same photons going through the fiber are also going through the electronics of the network interface chip. Potentially, that chip could then relay those photons to a photonic processing unit, and perform some (if not all) computation on the light directly (ya know, with those photonic transistors that were just invented), without having to convert the signal to electrons first.
Daishi
Photons move through optical fiber at about 2/3 of the speed of light. As the photons travel there is attenuation or signal loss. As the power fades you can pass it through an optical amp to boost the signal on the fiber with fairly low latency like this diagram: https://upload.wikimedia.org/wikipedia/commons/6/6d/Doped_fibre_amplifier.svg
The problem is as it goes through the optical amp aside from boosting just the carrier signal any "noise" is also amplified so depending on the path you can only optically amplify the signal about 10 times before there isn't a large enough difference between the actual channel and the (amplified) noise to distinguish the difference between them. This is reflected as a poor signal to noise ratio (SNR).
To remedy this, you need electrical stations along the path that receive the signal, clean up any errors against a checksum or Forward Error Correction (FEC) method, and retransmit out a clean signal. The problem with these electrical stations is they are both generally more costly and add some latency to the path.
The issue is over 1000 miles of fiber you may have only 1 electrical station so the light propagation delay over the geographic path is generally a much larger factor.
Shooting from the hip: Of the 60 - 75 milliseconds of round trip time it takes to go coast to coast in the US on optical fiber these electronic regen stations are only responsible for maybe a few milliseconds of latency. Even if the graphene systems had only half the total latency it would change coast to coast round trip times from like 67.5 ms to 66 ms.
Over long distance spans I don't see it having that much impact but it could matter in short reach low latency applications like data centers and super computing. Graphene switches combined with the "Fiber cables made of air" that improves light propagation delay to 99.7 % of the speed of light could have a noticeable impact in to latency in supercomputing nodes. The current top supercomputer uses ~400k 8 core Xeons so interconnecting them with minimal latency is a huge part of the challenge.