Several somewhat similar processes are also referred to as laser cooling, in which photons are used to pump heat away from a material (normally a solid) and thus cool it. The phenomenon has been demonstrated via anti-Stokes fluorescence, and both electroluminescent upconversion and photoluminescent upconversion have been studied as means to achieve the same effects.
This technique works by tuning the frequency of light slightly below an electronic transition in the atom. Because the light is detuned to the "red" (i.e. at lower frequency) of the transition, the atoms will absorb more photons if they move towards the light source, due to the Doppler effect. Thus if one applies light from two opposite directions, the atoms will always scatter more photons from the laser beam pointing opposite to their direction of motion. In each scattering event the atom loses a momentum equal to the momentum of the photon. If the atom, which is now in the excited state, emits a photon spontaneously, it will be kicked by the same amount of momentum but in a random direction. The result of the absorption and emission process is to reduce the speed of the atom, provided its initial speed is larger than the recoil velocity from scattering a single photon. If the absorption and emission are repeated many times, the mean velocity, and therefore the kinetic energy of the atom will be reduced. Since the temperature of an ensemble of atoms is a measure of the random internal kinetic energy, this is equivalent to cooling the atoms.
In other forms of laser cooling, the laser light is similarly tuned to a frequency with an energy just below the energy of an electronic transition, but the type of transition is not necessarily restricted to a simple atomic transition. In these cases, the heat extraction occurs because the thermal energy of the material makes up the energy difference between the laser photon energy and the transition energy, allowing the transition to take place based on their combined energy. If this excited state then decays radiatively, the heat energy is carried away from the material by the emitted higher-energy photon. Abstractly, this can be seen as equivalent to the process of atomic laser cooling, though the processes are usually described in different terms.
The vast majority of photons that come anywhere near a particular atom are almost completely unaffected by that atom. The atom is almost completely transparent to most frequencies (colors) of photons.
A few photons happen to "resonate" with the atom, in a few very narrow bands of frequencies (a single color rather than a mixture like white light). When one of those photons comes close to the atom, the atom typically absorbs that photon (absorption spectrum) for a brief period of time, then emits an identical photon (emission spectrum) in some random, unpredictable direction. (Other sorts of interactions between atoms and photons exist, but are not relevant to this article.)
The popular idea that one can heat up and vaporize objects with a laser is not exactly true when we are looking at individual atoms. If we have an atom that is practically motionless (a "cold" atom), and control the frequency of the laser we shine at it, we find that most frequencies just pass by the atom — it is invisible at those frequencies. There are only a few points on our frequency control dial that have any effect on that atom. At those frequencies, the photon slams into the atom, the atom starts drifting away from the laser, and later the atom releases the photon. If it happens to shoot the photon towards the laser, it makes the atom drift away from the laser twice as fast. If it happens to shoot the photon directly away from the laser, then the atom stops and becomes motionless again. But usually the photon speeds away in some other direction, giving the atom at least some sideways thrust.
Another way of changing frequencies is to move the laser. Imagine that we have a fixed-frequency (single-color) laser that has a frequency that is a little below one of the "resonant" frequencies of this atom (this laser will go right through our atoms without affecting them.) If we strapped that laser to the front of a train coming towards us (and our pile of cold atoms), then the doppler effect would raise its frequency. At one specific velocity, the frequency would be just right for these atoms to start absorbing those photons.
Yet another way of changing frequencies is to move the atoms. If we bolt that laser to the ground, and put an iceball of cold atoms in the beam, then normally the laser beam would pass right through the transparent iceball. But at one specific speed of the iceball towards the laser, suddenly the iceball will absorb photons slamming into it. That slows down the iceball. The iceball will also emit photons in all directions. That accelerates the atoms in the iceball in all directions, which is pretty much the definition of a higher temperature.
Something very similar happens in a laser cooling apparatus, except we start with a warm cloud of atoms moving in all directions at all different speeds. We start with a laser frequency well below the resonant frequency. Photons from any one laser pass right through the majority of atoms. However, atoms moving rapidly towards a particular laser catch the photons for that laser, slowing those atoms down until they become transparent again. (Atoms rapidly moving away from that laser are transparent to that laser's photons — but they are rapidly moving towards the laser directly opposite it). This utilization of a specific velocity to induce absorption is also seen in Mossbauer spectroscopy.
On a graph of atom velocities (atoms moving rapidly to the right correspond with stationary dots far to the right, atoms moving rapidly to the left correspond with stationary dots far to the left), there is a narrow band on the left edge corresponding to the speed those atoms start absorbing photons from the left laser. Atoms in that band are the only ones that interact with the left laser. When a photon from the left laser slams into one of those atoms, it suddenly slows down an amount corresponding to the momentum of that photon (we redraw that dot some fixed "quantum" distance further to the right). If the atom releases the photon directly to the right, then we redraw the dot that same distance to the left, putting it back in the narrow band of interaction. But usually the atom releases the photon in some other random direction, and we redraw the dot that quantum distance in the opposite direction.
We design the apparatus with many lasers, corresponding to many boundary lines that completely surround that cloud of dots.
As we ramp the laser frequency up, the boundary contracts, pushing all the dots on that graph towards zero velocity. That's the definition of "cold".
The atom performs a random walk in momentum space with steps equal to the photon momentum due to spontaneous emission and photon absorption. This constitutes a heating effect, which counteracts the cooling process and imposes a limit on the amount by which the atom can be cooled. Moreover, the optical transition used for cooling in reality must have a finite frequency width, which limits the velocity discrimination (i.e. the likelihood that an atom will scatter light from the "correct" beam, as described above), and therefore the temperature. This temperature is called the Doppler temperature. Lower temperatures, down to the recoil temperature, may be obtained by sub-Doppler cooling. Beyond that, evaporative cooling is used to further cool the ultracold atoms.
The concentration must be minimal to prevent the absorption of the photons into the gas in the form of heat. This absorption happens when two atoms collide with each other while one of them has an excited electron. There is then a possibility of the excited electron dropping back to the ground state with its extra energy liberated in additional kinetic energy to the colliding atoms — which heats the atoms. This works against the cooling process and therefore limits the maximum concentration of gas that can be cooled using this method.
Only certain atoms and ions have optical transitions amenable to laser cooling, since it is extremely difficult to generate the amounts of laser power needed at wavelengths much shorter than 300 nm. Furthermore, the more hyperfine structure an atom has, the more ways there are for it to emit a photon from the upper state and not return to its original state, putting it in a dark state and removing it from the cooling process. It is possible to use other lasers to optically pump those atoms back into the excited state and try again, but the more complex the hyperfine structure is, the more (narrow-band, frequency locked) lasers are required. Since frequency-locked lasers are both complex and expensive, atoms which need more than one extra repump laser are rarely cooled; the common rubidium Magneto-optical trap, for example, requires one repump laser. This is also the reason why, to date, molecules have not been laser cooled: in addition to hyperfine structure, molecules also have rovibronic couplings and so can also decay into excited rotational or vibrational states.
The following is an incomplete list of atoms and ions which have been laser cooled:
Counter-propagating sets of laser beams in all three Cartesian dimensions may be used to cool all the three degrees of freedom of the atom. Common laser-cooling configurations include optical molasses, the magneto-optical trap, and the Zeeman slower. Ions trapped in an ion trap require cooling in only one dimension, since the external electric field couples motion in the three dimensions.