We were able to take information encoded on an optical beam and create an ultra-high-frequency sound wave that contained the same information. At a later, controllable time, we could transfer the information back to an optical beam.
We send a stream of optical data (pulses of light represent bits of information) into one end of a short length of optical fiber, and send a short-duration "write" pulse into the other end of the fiber. When the write pulse passes through the data pulses, interference takes place, giving rise to regions of high and low intensity. This change in intensity causes the density of the glass to increase slightly in some locations and decrease slightly in others through a physical effect known as electrostriction. This change in density corresponds to the acoustic wave, and also causes a small change in the refractive index of the glass. The optical pulses leave the fiber and the acoustic wave stays behind (the speed of sound in the glass is around 5,000 meters/second whereas the speed of light in the fiber is around 20,0000,000 meters/second, so the acoustic wave is essentially stationary over the duration of our experiment).
At a later controllable time, we send a short-duration "read" pulse into the fiber along the same direction as the write pulse. When the read pulse encounters the acoustic wave, it scatters from this wave and puts light into the same direction as the original data pulses. During this process, all of the energy in the acoustic wave is depleted.
What is so cool about this process is that the original data stream is recreated with reasonable fidelity.
In the interaction of the light waves and the acoustic wave, one data photon (νd~193 THz) is converted to a write photon (νb~9.6 GHz) and an acoustic phonon. Therefore, only a tiny fraction (νb/νd~5X10-5) of the data pulse energy is contained in the induced sound wave, and most of the data pulse energy is transferred to the write pulse.
In its current form, we hope that these results are enabling in that they will get other people to think of new methods for storing optical information. The existing methods store the information in a so-called spin coherence in a gas of atoms (or in ions doped into a crystal). Our method shows that a different type of material excitation — an acoustic wave — can also be used to store information. We are certain that there are other material excitations that will also work.
To be practical for a wide range of applications, we need to increase the storage time and reduce the peak power needed in the write and read pulses.
The light pulse itself is truly not stored — it is the information that was encoded on the "data beam" that is transferred to the material. We then transfer that information back to the optical domain during the read-out process.
In more complicated terms, we are storing a state of the system. Specifically, it is possible to write down a new state of the system that is a linear combination of "light-like" states and "matter-like" (or acoustic-like) states. As we store the state, it turns from being essentially light-like to nearly fully matter-like. In the read out process, the state is converted from matter-like back to light-like. So, in actuality, we are storing that combined matter-field state.
There is a growing consensus that high speed optical networks need to keep data in the optical domain — with no optical-electronic-optical conversion — the conversion generates too much heat at the very highest speeds people envision. Thus, we need to be able to buffer (or store) bits of information in the optical domain — which take the form of pulses of light. Slow light is one way to do this, stored light is another. Also, there are some optical information processing applications of our technique. If we were to use a write pulse that had a complex waveform (rather than a short pulse), the acoustic wave would be the convolution of the write waveform and the data waveform. This could find application in recognizing a particular optical waveform (e.g., recognizing the header information on a packet of information). The buffer could be used to improve the performance of optical routers. A router has multiple input lines and multiple output lines, but only one switch. Thus, if two packets arrive simultaneously at the input, one can be buffered while the other is being processed by the switch.
There have been many other experiments to achieve stored light. Most of these experiments encode the information on a so-called spin coherence of a gas of atoms (spin coherence is a quantum mechanical feature of most atoms). In these experiments, the data and write/read pulses have to have their frequencies set precisely to an atomic resonance frequency. If there is a slight error, the effect does not work.
In our experiment, we use an acoustic excitation rather than a spin-coherence excitation. Thus, we have shown that the stored-light effect is more general than thought previously, which we hope will prompt researchers to look at other types of material excitations.
Furthermore, the frequencies of our light beam can be set to any value where the material is transparent — essentially the entire near-infrared portion of the spectrum. (What is required in our method is that the frequency DIFFERENCE of the write and data pulses take on a precise value, but the absolute frequency does not matter.)
We also use a room-temperature, standard optical fiber as the storage material so that our method could easily integrate with existing technologies.
Another set of recent experiments by Michal Lipson's group at Cornell demonstrated that you could store a single pulse in a microring resonator and release it a later time. (This was published just a few months ago.) Her work is very promising. In comparison to our work, we stored three pulses for a longer time, but I consider these details — both approaches are promising and need further research to assess potential limitations/tradeoffs, etc.
It all depends on the application requirements. Current technology that uses cryogenic temperatures works very well with regards to bandwidth and storage times. The down side is that the method only works at the resonance of an ion doped in a crystal, and at near liquid helium temperatures.
On the other hand, there may be some niche applications in optical pulse buffering or real-time all-optical data processing that could use our system as is.
We are currently investigating other materials that have a stronger light-sound coupling strength (i.e., larger Brillouin scattering coefficient) and longer acoustic lifetimes. We mention a few candidates in our paper. Soft glasses in the chalcogenide family are very interesting, as well as hollow-core photonic crystal fibers filled with liquids such as carbon disulfide or with high pressure gases such as xenon.
There is a very realistic chance we can achieve at least a 100 fold reduction using materials that are currently available in laboratories. It will just take time and a little money.
I tried to do this in Question #2. Let me also say that our technique is similar to a form of real-time holography. There is an interference between data and write beams that creates a change in refractive index in the material (like a hologram). The read pulse scatters of this change in refractive index to recreate the data pulses (similar to the way the reference beam diffracts of the hologram to recreate the image scene).
Our current setup used a binary waveform and we stored three bits. But there is no reason why we could not have used a more complex waveform so that more bits could have been stored under similar conditions.
There are other methods for storing light, but these often require that the frequency of the light being stored needs to match precisely a material resonance (i.e., energy levels of an atom/ion/ molecule), or that require cryogenic temperatures.