The Molasses

There is a cloud of sodium atoms in a laboratory, pooled where six laser beams cross, and it is colder than anything else for miles around — colder than the dark between galaxies, a few millionths of a degree above absolute zero. It is made that cold by light. Everything else we know about light says the reverse. Light is what the sun does to a stone; it warms the wall it falls on, and a strong enough beam will cut steel. To deposit energy is what light does. And yet this cloud is chilled almost to a standstill by laser light poured onto it from every side. The light is not somehow cold — it is ordinary light. What does the cooling is that the beams are tuned slightly too low.

An atom absorbs light only at particular frequencies, its resonances, and near one of them it will take in a passing photon and spit it back out a moment later. Tune your laser exactly to that frequency and a still atom scatters the light happily. Tune it a little below — red-detune it, in the jargon — and the still atom mostly ignores you; you are singing slightly flat, off its note. Now set the atom moving. An atom rushing toward one of the beams meets its waves crowded closer together — the same Doppler shift that raises the pitch of an approaching siren — and to that atom the light you tuned too low is lifted back up onto the note. So it drinks preferentially from the beam it is running into, and every photon it swallows delivers a small push against its motion. Absorb, and be shoved backward. It re-emits the photon an instant later, with a recoil of its own — but in a random direction each time. Over thousands of absorptions that outgoing recoil points every which way and averages to nothing, while the incoming push always points the same way: against the motion. Ring the atom with six beams, a pair along each axis, and whichever way it tries to bolt it runs into a headwind. It bogs down as though empty space had thickened to syrup. The group that built the first one, Steven Chu's at Bell Labs in 1985, called it optical molasses.

The cold is not in the light. The same beams tuned a little too high instead of too low would push the atoms faster and blow the cloud apart; flip the sign of the detuning and the identical apparatus becomes a heater. Nor is it in any one photon, each of them a dumb, identical shove that carries no word of whether it will warm or chill. What cools is the fit between the frequency you chose and the speed of the atom — and the atom's own velocity is what sets that fit. The faster it moves, the further it Doppler-shifts the oncoming beam into resonance, and the harder it is thrown back. Nothing sorts the atoms by speed but the atoms, moving. And the thing doing the sorting is the Doppler shift, which nearly everywhere else in physics is how you read a speed you cannot otherwise reach: you take a star's velocity from the reddening of its light, a storm's from the shift in returned radar, the flow in an artery from the change in an ultrasound echo. The shift is the dial you look at to learn how fast a thing recedes. Here nobody reads it. The same shift is turned outward and made to act — the velocity you would ordinarily measure from it becomes instead the thing it reaches in and changes. Everywhere else it tells you a speed. Here it takes hold of one.

The cooling has a bottom. Those random recoils cancel on average but not exactly, and their jitter is a faint, ceaseless heating — a floor beneath which the drag can no longer win. Physicists worked it out, the Doppler limit, and for sodium it comes to about two hundred and forty millionths of a degree above absolute zero, and that was understood to be that. Then in 1988 William Phillips's group, at what is now NIST, measured the temperature of their molasses carefully, expecting to confirm the number — and found the atoms at forty-three millionths of a degree, more than five times colder than the floor allowed. The floor was not wrong. It was the true floor for the atom in the equations, an idealized thing with a single clean resonance. Real sodium is more crowded: it carries several closely spaced internal states, and in the crossed and polarized light those extra states had opened a second, subtler way to bleed off motion that no one had thought to put in the model.

Where the beams overlap, the polarization of the light shifts from point to point, and an atom drifting through it climbs a landscape of shallow hills, its energy lifted as it goes. Near the top of each rise — right where it is slowest, having spent its motion on the climb — the light pumps it into a different internal state, one for which the hill just climbed is a valley; and so it finds itself at the bottom again, a fresh hill ahead. It climbs, slows, is set down at the foot of the next slope, and climbs again. Jean Dalibard and Claude Cohen-Tannoudji, who explained it in 1989, named it Sisyphus cooling, for the man the gods sentenced to roll a stone up a mountain and watch it roll back, and roll it up again, without end. But the name catches only the shape — the endless climb, the summit never kept. Sisyphus's curse is that nothing comes of it: the same stone, the same hill, the ledger blank at the end of every push. The atom's climb is the opposite where it counts. Each time it is returned to the valley it arrives a little slower than before, and the motion it spent on the climb does not come back. The very thing that looks like futility — always sent down, never cresting — is the ratchet that empties it. Repeat the pointless-seeming climb enough times, and the atom is barely moving at all.

So the cloud lingers where the beams cross, a few millionths of a degree above absolute zero, drifting slowly apart under light that would warm your face. Nothing exotic is doing it. A shift you would ordinarily only read off a dial, and a climb that only looks like getting nowhere, have been turned on a knot of atoms and set, between them, to draw the motion out of it.