The detector found two thin groups inside a broad beam
In 1988, Alain Aspect and colleagues sent a beam of metastable helium atoms through opposing laser beams. Metastable means that the atoms occupied a long-lived excited internal state. A downstream detector resolved their sideways motion and found two narrow peaks on a broader background.
The peaks corresponded to a temperature scale near two millionths of a kelvin above absolute zero, or two microkelvin. That was below the usual one-dimensional Doppler value of about 23 microkelvin for this transition, and below the four-microkelvin scale associated with a single photon recoil, using the paper’s one-dimensional kinetic-temperature convention. The result concerned a narrow component of transverse motion. The beam kept traveling forward, and the wider background remained part of what was measured.
Read the evidence: A. Aspect et al. (1988), Laser Cooling below the One-Photon Recoil Energy by Velocity-Selective Coherent Population Trapping, PRL 61, 826–829
Two ways to absorb a photon can cancel
The light connected two different lower internal states to the same upper state. One beam supplied a photon moving one way; the opposing beam supplied the opposite momentum. Quantum mechanics allows a single atom to occupy a superposition of the two lower possibilities, with a definite relative phase between them.
That phase can make the two excitation amplitudes cancel, like two waves meeting crest to trough. In the resulting dark state the atom hardly absorbs the applied light. The cancellation concerns the two amplitudes for reaching the upper state; it requires the two lower components to remain in step.
Here the internal state and the momentum belong together. In the ideally centered dark family, one component carries one photon recoil of momentum to one side, and the other carries the same momentum to the opposite side. Their kinetic energies match, helping their relative phase remain fixed.
Read the evidence: A. Aspect et al. (1988), Laser Cooling below the One-Photon Recoil Energy by Velocity-Selective Coherent Population Trapping, PRL 61, 826–829
Atoms outside the favored velocity class keep scattering
Shift both momentum components away from this symmetric arrangement and their kinetic energies differ. Their relative phase then changes during flight. The cancellation unravels, the atom becomes able to absorb again, and spontaneous emission can send it into another momentum class.
Repeated scattering keeps redistributing the atoms. Those that reach a sufficiently dark class stop scattering for longer and accumulate there. This selects a narrow class by protecting it from the light. Longer coherent interaction can sharpen the selection; magnetic splittings, unstable optical phase and finite transit time limit how well that protection lasts.
Read the evidence: A. Aspect et al. (1988), Laser Cooling below the One-Photon Recoil Energy by Velocity-Selective Coherent Population Trapping, PRL 61, 826–829J. Dalibard and C. Cohen-Tannoudji, Laser cooling and trapping of neutral atoms — author-hosted review, Sections 3–5
A narrow peak and the whole distribution have different widths
The two peak centers remain separated by two photon recoils of momentum. Even if each peak became extremely narrow, measurements on the ideal equal superposition would still give one recoil in either direction. Its average momentum is zero, but that separation carries kinetic energy.
For a numerical illustration, give each of two equally weighted peaks a standard deviation of one tenth of a recoil momentum. The full momentum variance is then 1.01 times the recoil momentum squared: one from the separation, and 0.01 from the individual widths. This invented example explains why the quoted narrow-component temperature and the full kinetic variance answer different questions.
Read the evidence: A. Aspect et al. (1988), Laser Cooling below the One-Photon Recoil Energy by Velocity-Selective Coherent Population Trapping, PRL 61, 826–829
Changing the light tested the proposed mechanism
The experiment compared polarization configurations and disturbed the overlap of the beams. The two-peak pattern changed. These interventions connected the observed velocity selection to the optical arrangement that was supposed to produce it.
A momentum histogram gives the populations at different momenta. By itself it does not measure the phase relation between them: a coherent superposition and an incoherent mixture can have the same populations. A phase-sensitive recombination would ask that further question. The 1988 cooling result showed how a narrow velocity component could be selected by a dark-state mechanism, while the full measured profile still recorded the atoms outside it.
Read the evidence: A. Aspect et al. (1988), Laser Cooling below the One-Photon Recoil Energy by Velocity-Selective Coherent Population Trapping, PRL 61, 826–829