Could You Freeze Light?
It Has Been Done Twice Over, and the Thing That Stops Is Not the Light.
Slow a beam of light down until it stops. Not absorb it, not reflect it, not lose it as heat: hold it still, keep it, and then let it go again and have the same light come out. I want to know whether that has ever been done, and if it has, what is actually sitting there while it is stopped.
It has been done, and the second half of the request is the interesting half. A pulse has been brought down to seventeen meters per second and halted outright, and what comes out afterward is the same light in every sense an experiment can test. But nothing decelerated. At seventeen meters per second the thing crossing the apparatus is already 99.999994 percent atoms and almost none of it is light, and by the time it has stopped there is no light in the sample at all. The pulse was not slowed. It was traded away and traded back.
What Physics Says, Wall by Wall
Granted, and the number is seventeen meters per second
A cloud of sodium atoms is cooled to around 450 nanokelvin, which is cold enough that they are barely moving. On its own the cloud is opaque at the frequency of interest: shine that light in and it is absorbed. Then a second laser, called the control, is turned on at a different transition, and the cloud becomes transparent in a very narrow window of frequency. This is electromagnetically induced transparency, and it was not invented for this purpose.
What matters is not the transparency but its steepness. The refractive index changes enormously across that narrow window, and a pulse whose frequencies span the window travels at a group velocity set by how fast the index is changing rather than by its value. The steeper the window, the slower the pulse. In the first such experiment that came out at seventeen meters per second: about the pace of a car in town, and 17,634,850 times slower than the same light in vacuum.
A pulse two and a half microseconds long is 749 meters long in vacuum, and the front of it enters the cloud while the back of it is still most of a kilometer away. Once it is inside, all of it fits in 42.5 micrometers, which is smaller than the cloud. That compression by a factor of seventeen million is the first sign that the pulse is not doing what a slowed thing does.
And it has been stopped outright, and restarted
Once the pulse is entirely inside the cloud, turn the control laser down. The transparency window narrows, the group velocity falls with it, and if the control is taken smoothly to zero the pulse stops where it is. Turn the control back up and a pulse leaves the far side, with the shape it went in with and the phase it went in with. Two groups did this in the same month, one in a Bose-Einstein condensate and one in hot rubidium vapor, which is worth noting: the effect does not need the cold. The cold buys the extreme slowness, not the storage.
So the request is granted twice over. Light has been slowed to walking pace, and light has been stopped and let go again. Everything after this is about what the words in the request mean.
Nothing decelerated. The light was traded for something else.
The clean description is that the excitation moving through the cloud is neither a photon nor an atomic excitation. It is a fixed mixture of the two, and the control laser sets the mixture. Write the fraction that is light as the square of a cosine, and the fraction that is atomic spin as the square of a sine. Then the group velocity is simply the speed of light multiplied by the light fraction.
That one line does all the work. Seventeen meters per second is not light going slowly. It is a mixture that is one part in seventeen million light, and the rest of it a pattern written across the spins of the atoms. Put the other way round: at the moment everyone calls slow light, the thing in the apparatus is already 99.999994 percent not light. Turning the control down does not brake it. It rotates the mixture further, and the last of the light goes out of it.
The picture below is that mixture. The ruler is the group velocity, eleven and a half powers of ten of it, and the readout says what fraction of the thing is still light at whatever setting you choose. The number to watch is not where the bar is. It is how early the light runs out.
While it is stored there is no light in the apparatus
Take the control laser to zero and the light fraction goes to zero with it. There is no pulse sitting still between the atoms, no photon parked somewhere, nothing that a detector placed inside the cloud would register. What is there is a pattern: each atom carries a phase, and the phases together form a wave written across the cloud in the atoms’ spins rather than in the field.
That pattern is what is stored, and it holds everything the pulse was carrying. When the control comes back the pattern drives the atoms to radiate coherently, and they hand it back as a pulse with the same shape, the same phase and the same polarization. It is the same light in every operational sense: put a single photon in, get a single photon out. It is not the same light in the sense the request had in mind, because for the duration of the storage the light did not exist.
The speed of light never changed, and this is not a technicality
Nothing in these experiments touches the constant. Between the atoms, light travels at exactly the speed it travels at everywhere. What is slow is the envelope of a pulse, and an envelope is not a thing that moves: it is the place where the waves happen to add up, and it can be made to sit still by arranging the adding up correctly. Ordinary glass does the same thing in miniature and nobody finds it remarkable.
The one honest way to say it is that the group velocity is a statement about the medium and the pulse together, and about nothing else. A different medium gives a different number, and there is no number it cannot be given, including zero and including negative, which has also been measured. None of that is a statement about light.
The storage has a clock, and it is the atoms’ clock
What limits how long you can keep it is nothing to do with light. It is how long the atoms can hold a phase relationship before their own motion and their own magnetic surroundings scramble it. The first experiments managed about a millisecond. A crystal laced with praseodymium later held one for a minute. One laced with europium has held the atoms’ phase for six hours, at which point the limit is a materials problem and not an optics one.
Those numbers are worth converting. In a millisecond, light in vacuum covers 300 kilometers. In a minute it covers eighteen million. In six hours it covers six and a half billion kilometers, which is past Neptune. That is the distance the pulse did not travel, and the reason it did not travel it is that for all of that time it was not a pulse.
Those three results are three points on a ruler ten decades long, and the gaps between them are the history of the field rather than anything about light. Put the clock anywhere on it. The distance changes by factors of a thousand; the line underneath it, which says what actually ran out, never changes at all.
The Pushback
The first of these is the objection the walls invite. The second is the one that decides whether the headline was ever fair.
That is too harsh, and the reason is that the test was passed. Send in a state and get the same state back, including its phase, its shape and its polarization. Send in one photon of an entangled pair, store it, retrieve it, and the entanglement with its partner is still there, which has been done. There is no operation you can perform on the output that distinguishes it from the input having simply waited.
What is fair to object to is the mental picture, not the claim. Storing light is an accurate description of what the device does. Frozen light, meaning light that is present and stationary, is a description of something that is not in the apparatus at any moment.
Two things, and the second is the one that matters. The first is that glass slows the phase as well as the envelope, by a factor of about one and a half. Its mechanism is the same everywhere in the band, so it is not sitting on a knife edge of frequency. In the cold cloud almost all of the slowing comes from the steepness of the index across a window a few megahertz wide. Step outside that window and the medium is opaque again.
The second is the mixture. In glass, the excitation is overwhelmingly the field, and the atoms are a small correction to it. At seventeen meters per second the proportions have reversed past anything glass approaches: the atoms are essentially all of it. That is why the control laser can take the last of the light away and leave something behind, and why glass cannot. It is not the same effect made bigger. It is a different arrangement that happens to be quoted in the same units.
Where the Request Was Right
The request asked for something to still be there while it is stopped, and refused to accept absorption as an answer. Both instincts were right, and they are the two that a great many summaries of this work get wrong in the other direction, by saying the light is merely absorbed and re-emitted. It is not. Absorption loses the phase, and this does not; that is the whole difference between a storage device and a piece of soot.
The request was also right to ask what is sitting there. That turned out to be the productive question, because the answer is a specific, writable thing: a phase pattern across the spins of a few million atoms, which can be measured, which decays on a known clock, and which can be moved. A pulse has even been stopped in one cloud of sodium, its pattern carried across a gap by the atoms themselves, and released from a second cloud entirely.
The one word to give up is freeze. Freezing is a thing you do to a substance by taking energy out of it, and the light here is not a substance and does not have its temperature lowered. The cloud is cold; the light is not anything. What was found was not a way to chill light but a way to put a light field and a spin pattern into the same account, and then move the balance from one column to the other and back.
An Opinion, Dated
The walls above are physics. This box is the drafting model’s own bets, and they are scoreable.
Four in five that the practical work stays in doped crystals and warm vapour cells rather than condensates, because the storage time is a materials problem and the condensate buys slowness the applications do not need.
One in three that a quantum repeater link long enough to beat direct fibre over the same route is demonstrated by 2035. The storage is the part that works; the parts that do not are the sources and the detectors either side of it.
And one that is a consequence rather than a wager: no experiment ever detects a photon sitting still, because there is no configuration of this apparatus in which a stationary photon is what is present.
These intuitions were distilled from human physics writing and carry its fashions. Weigh them accordingly.
The Updated Map
The picture to give up is the one where a pulse is a small fast object and the experiment is a brake. There is no object and there is no brake. There is an account with two columns, a field and a set of spins, and a control laser that decides how the balance is split; the group velocity is not a speed anything travels at but a readout of that split. Slow light is a mixture that has already gone almost entirely over to the atoms. Stopped light is the same mixture with the last of the field taken out of it, which is why it can be held for six hours in a crystal and why nothing about that six hours is a fact about light. The remarkable thing was never that light was made slow. It is that a light field and a pattern of spins turn out to be two ways of holding the same thing, and that you can hand it from one to the other and back without losing so much as a phase.
