Point a laser at the right crystal, in a vacuum, and it does not warm up. It gets cold — ninety-one kelvin from room temperature, with nothing moving and nothing flowing but light. Each photon it absorbs leaves a little richer than it arrived; the difference is the crystal’s own heat, carried out as light. Proposed in 1929, declared impossible in 1945, first seen in 1995. The wait was not for the physics.
An idea told to wait
In 1929 Peter Pringsheim noticed that a substance which absorbs light of one colour and gives it back slightly bluer is losing energy on the books — so light should be able to cool it. Sergey Vavilov objected in 1945 that this would beat Carnot. Lev Landau answered in 1946, and his answer is the issue in a line: light has entropy, and it depends on how many directions and colours the light is spread across.
Sixty-six years after Pringsheim, at Los Alamos, a sliver of ytterbium-doped fluoride glass under a laser got measurably colder. Three-tenths of a degree.
What each photon does
The working ion is ytterbium, which has one excited state and nowhere higher to climb. The laser is tuned just below the energy the ion normally emits: in a lithium-yttrium-fluoride crystal, 1020 nanometres against a mean fluorescence near 1000. The ion absorbs the photon, finds itself slightly short of its own equilibrium, and makes up the shortfall from the lattice: it takes a phonon, a quantum of the crystal’s vibration, which is to say a quantum of its heat. Then it emits a photon carrying, on average, two per cent more energy than it absorbed.
Repeat that a hundred billion billion times a second and the crystal’s heat is leaving as light. This is anti-Stokes fluorescence, emission bluer than absorption. Nothing conducts the heat away; it is radiated, at a wavelength you cannot see, two per cent bluer than the beam.
Where the bill goes
No. 99 found the second law drawing the ring in your sink; No. 106 found a tube that buys cold with a pressure drop it never recovers. Here the bill is paid in the light itself. A laser is as orderly as light gets: one colour, one direction. The fluorescence leaves in every direction, across a band of colours. That spreading was Landau’s point: it is entropy, and there is more of it in the light that leaves than in the light that came. The crystal gets colder by making the light messier.
Cold as a rate
The 2016 record: a crystal of ten per cent ytterbium, pumped for twelve minutes, settling at 91 kelvin while the copper shell around it sat at 265. Settling is the authors’ word. The shell radiates heat into the crystal; the crystal radiates heat out as fluorescence; the temperature stops falling where the two rates cross. No. 125 found transparency to be a pumping rate held against a leak; this is temperature held the same way. Switch the laser off and there is no cold to keep — only a crystal, warming.
Ninety-eight of every hundred
Now the honest half. Two per cent is the whole margin. To cool at all, more than ninety-eight of every hundred photons the crystal absorbs must come back out as fluorescence, because each one that does not wipes out the cooling of fifty that did. Any impurity that absorbs the laser and simply warms — iron, it turned out, at trace levels — counts against that mark.
The record crystal returns 99.6 per cent, and its stray absorption is so small it could only be bounded: a metre of the stuff would turn at most about one per cent of the light passing through it into heat. Landau finished the physics in 1946. Growing a crystal that clean took until the 2010s.
The colder it gets, the less of it can be cooled
One step further, to the part I did not expect. The laser is absorbed not from the ion’s lowest level but from a slightly higher one, which ions occupy only because they are warm. At room temperature the crystal soaks up over ninety per cent of the beam in four passes. At 90 kelvin, after twenty-two passes, it has taken 39 per cent. The colder it gets, the fewer of its ions are warm enough to accept the light that cools them. Even a flawless crystal has a floor, and the floor is made of this: the cooling spends the very population that makes it possible.
Not finished
The record has moved since I promised you 91: 87 kelvin in 2019, and the calculated floor is now nearer 70, because the stray absorption itself falls with temperature. An infrared sensor has been held at 135 kelvin by one of these crystals working as a refrigerator, a later payload at 125. It lifts a fraction of a watt for tens of watts of laser, and the field polices its own claims hard, because the balance is fine enough for a thermometer to get wrong. Not solved. But the crystal is glowing, slightly bluer than the beam, and the glow is the heat.
≈
One loop I’m watching
Next: a bubble of glowing gas, light-years across, whose edge is drawn where a star’s ultraviolet runs out. Every atom inside is stripped of its electron and re-formed, over and over, and the glow you photograph is the re-forming — a balance made visible, its boundary nothing but a count.