Laser Cooling

Definition

Laser cooling is the use of light to remove kinetic energy from atoms. The workhorse version, Doppler cooling, surrounds a cloud of atoms with counter-propagating laser beams tuned slightly below the atomic resonance. Because of the Doppler effect, a moving atom always scatters more photons from the beam it is heading into. Each absorbed photon delivers a momentum kick that opposes the motion, so the light behaves like a thick, viscous fluid — hence the nickname optical molasses.

The result is a velocity-dependent friction force. Slow the atoms down enough and their temperature — which is just a measure of average kinetic energy — drops from room temperature (~300 K) to the microkelvin scale, a few millionths of a degree above absolute zero.

How it works

Three ingredients do all the work.

1. A photon carries momentum. A photon of wavelength λ carries momentum p = ħk = h/λ, where the wavenumber k = 2π/λ. When an atom absorbs it, the atom recoils by the recoil velocity:

v_r = ħk / m

For sodium absorbing its 589 nm D-line photon, v_r ≈ 3 cm/s. That is laughably small next to a thermal speed of ~500 m/s — but the atom doesn't stop at one photon.

2. The Doppler effect makes absorption velocity-dependent. An atom rushing toward a laser beam sees its frequency blueshifted upward, toward the atomic resonance; the beam chasing it from behind is shifted down, away from resonance. If you deliberately tune both beams a little below resonance ("red-detuned", typically by one linewidth Γ), then a moving atom is always closer to resonance with the beam it is running into. It therefore absorbs more photons from the head-on beam than from the tail beam.

3. Absorption kicks add up; emission kicks cancel. Every absorbed photon nudges the atom in the direction the photon was traveling — i.e. against the atom's motion. The atom then re-emits a photon by spontaneous emission, recoiling again, but in a random direction. Over thousands of cycles those re-emission kicks average to zero, while the absorption kicks all point the same way. The leftover, after averaging, is a net drag:

F ≈ −β·v      (for small v — a viscous friction)

Add two more pairs of beams along the other axes and you get the canonical six-beam optical molasses: friction in every direction, no matter which way an atom darts.

Worked example — stopping a sodium atom

Take sodium (²³Na), the original 1985 demonstration atom.

• Mass: m = 3.82×10⁻²⁶ kg
• Cooling transition: 589 nm (D₂ line)
• Natural linewidth: Γ = 2π × 9.79 MHz ≈ 6.15×10⁷ s⁻¹
• Photon wavenumber: k = 2π/λ ≈ 1.066×10⁷ m⁻¹

Recoil velocity per photon:

v_r = ħk/m = (1.055e-34 × 1.066e7) / 3.82e-26 ≈ 0.029 m/s ≈ 2.9 cm/s

How many photons to stop a thermal atom? A sodium atom from a hot oven moves at roughly v ≈ 500 m/s. Number of kicks needed:

N = v / v_r = 500 / 0.029 ≈ 17,000 photons

How long does that take? The maximum scattering rate is Γ/2 (the excited state saturates at half-population), so:

R_max = Γ/2 ≈ 3.1×10⁷ photons/s
t = N / R_max ≈ 17,000 / 3.1e7 ≈ 0.5 ms

Half a millisecond to stop a supersonic atom. The implied deceleration is enormous:

a = v_r × R_max = 0.029 × 3.1e7 ≈ 9×10⁵ m/s²  ≈ 92,000 g

The Doppler limit. Cooling cannot continue forever, because the kicks are discrete and the re-emission is random — the atom diffuses in momentum space (photon shot noise heats it). Balancing this heating against the Doppler drag gives the floor temperature:

T_D = ħΓ / 2k_B
T_D(Na) = (1.055e-34 × 6.15e7) / (2 × 1.381e-23) ≈ 235 µK

So plain Doppler cooling of sodium bottoms out around 235 µK — a couple of hundred millionths of a degree above absolute zero, consistent with the textbook "~100 µK" order of magnitude.

Variants and regimes

Technique	Mechanism	Typical floor	Notes
Doppler / optical molasses	Red-detuned beams + Doppler-dependent scattering	~100–240 µK	Cools but does not trap
Magneto-optical trap (MOT)	Molasses + quadrupole B-field for position-dependent force	~100 µK	Cools and traps ~10⁹ atoms
Zeeman slower	Tapered B-field keeps a beam on resonance as it decelerates	n/a (slows a beam)	Front end that feeds a MOT
Polarization-gradient (Sisyphus)	Atoms climb light-shift hills, get pumped to valleys	~1–10 µK	Beat the Doppler limit (1988 surprise)
Recoil-limited (e.g. VSCPT, Raman)	Dark states / velocity-selective pumping	~0.1–1 µK (≈ T_recoil)	Approaches the single-photon recoil floor
Evaporative cooling	Remove the hottest atoms from a trap (no light)	~10–100 nK	Final stage to reach BEC

The single-photon recoil limit, T_recoil = (ħk)²/(m·k_B), is set by the energy of one recoil kick and sits around 1 µK for alkalis — the deepest temperature a single scattering event allows. Below that you must trade away atoms (evaporation) rather than scatter photons.

Common pitfalls and misconceptions

• "The lasers are cold." The light is not cold — it is a perfectly ordinary, even powerful beam. It cools by carrying momentum away in a biased direction, not by being low-temperature.
• "Cooling = trapping." Optical molasses has friction but no restoring force toward a point. Atoms still random-walk out of the beam region in seconds. Trapping requires the extra magnetic field of a MOT.
• "Blue-detuned would work too." No — blue detuning (above resonance) reverses the sign of the force and heats. Red detuning is essential for Doppler cooling.
• "Laser cooling alone makes a BEC." It stalls in the microkelvin range, ~1000× too hot for condensation. You still need evaporative cooling for the last stretch.
• "More laser intensity always cools faster." The scattering rate saturates at Γ/2; past saturation, extra intensity mostly broadens the line and adds heating, not cooling.
• "You can reach absolute zero." No method can — the third law of thermodynamics forbids it. Laser cooling gets close, never to 0 K.
• "Spontaneous emission is just wasted." The random re-emission is exactly what removes entropy from the atomic motion and dumps it into the scattered light field — it is the part that actually cools.

JavaScript — a 1D optical-molasses simulation

// 1D Doppler cooling: one atom between two counter-propagating beams.
// Force model: F = ħk·(Γ/2)·[ s/(1+s+(2δ_-/Γ)²) − s/(1+s+(2δ_+/Γ)²) ]
// where δ_∓ = δ ∓ k·v are the Doppler-shifted detunings.

const HBAR = 1.0546e-34;
const KB   = 1.381e-23;

// Sodium D2 line
const m     = 3.82e-26;       // kg
const lambda = 589e-9;        // m
const k      = 2 * Math.PI / lambda;
const Gamma  = 2 * Math.PI * 9.79e6; // s^-1
const delta  = -Gamma;        // red-detuned by one linewidth
const s      = 2;             // saturation parameter (I/I_sat)

function scatterRate(detuning) {
  return (Gamma / 2) * s / (1 + s + Math.pow(2 * detuning / Gamma, 2));
}

// Net molasses force on an atom moving at velocity v
function molassesForce(v) {
  const rPlus  = scatterRate(delta - k * v); // beam atom moves into
  const rMinus = scatterRate(delta + k * v); // beam from behind
  return HBAR * k * (rPlus - rMinus);         // sign opposes v near 0
}

// Integrate the equation of motion
let v = 30;            // start at 30 m/s
const dt = 1e-6;       // 1 µs steps
for (let step = 0; step < 4000; step++) {
  const F = molassesForce(v);
  v += (F / m) * dt;
}
console.log('velocity after 4 ms:', v.toFixed(4), 'm/s'); // → ~ a few cm/s

// Doppler limit temperature
const T_D = (HBAR * Gamma) / (2 * KB);
console.log('Doppler limit:', (T_D * 1e6).toFixed(0), 'µK'); // → ~235 µK

// Recoil velocity and recoil-limit temperature
const v_recoil = (HBAR * k) / m;
const T_recoil = (HBAR * k) ** 2 / (m * KB);
console.log('recoil velocity:', (v_recoil * 100).toFixed(2), 'cm/s'); // ~2.9 cm/s
console.log('recoil limit:', (T_recoil * 1e6).toFixed(2), 'µK');

The linearized small-velocity friction coefficient β can be read straight off the force law, and the damping time is m/β. For typical alkali parameters that is well under a millisecond — exactly the timescale our worked example produced.

Performance analysis — where the limit comes from

Expand the two-beam force for small velocity (k·v ≪ Γ). The difference of the two Lorentzian scattering rates is linear in v:

F ≈ −β·v,     β = −8 ħk² · s · (δ/Γ) / (1 + s + (2δ/Γ)²)²

β is positive (a real drag) precisely when δ < 0 — red detuning. Maximum damping occurs near δ ≈ −Γ/2. This drag removes energy at rate β·v². But cooling is grainy: each scattering event is a random momentum kick of size ħk, so the atom's momentum executes a random walk — a heating power that scales with the total scatter rate × (ħk)². Setting cooling rate = heating rate in steady state gives the equilibrium energy, hence:

k_B T_D = ħΓ/2  ⇒  T_D = ħΓ/(2 k_B),  minimized at δ = −Γ/2

The limit depends only on the linewidth Γ — narrow transitions cool colder. That is why narrow-line cooling on weak transitions (e.g. strontium's intercombination line) can reach far below the broad-line Doppler limit, and why optical clocks favor such narrow lines.

Three temperature scales bound the whole field, in order of decreasing temperature: the Doppler limit ħΓ/2k_B (~100 µK), the recoil limit (ħk)²/(m·k_B) (~1 µK), and the condensation temperature reached only by evaporation (~100 nK). Laser cooling owns the top two.

Applications

• Atomic clocks. Cold atoms drift slowly, so spectroscopic lines lose Doppler broadening and become razor-sharp. Optical-lattice clocks now keep time to roughly 1 second in 30 billion years.
• Bose–Einstein condensates. Laser cooling is the mandatory first stage (10⁹ atoms to a few µK) before evaporation finishes the job at ~100 nK — the 2001 Nobel built directly on the 1997 one.
• Atom interferometry. Slow, coherent atoms make exquisite gravimeters, gyroscopes, and inertial sensors — and precision tests of the equivalence principle.
• Quantum computing and simulation. Trapped ions and neutral-atom arrays must be laser-cooled to their motional ground state before high-fidelity gates are possible.
• Fundamental physics. Measurements of the fine-structure constant, searches for an electron electric dipole moment, and tests of relativity all exploit motionless, laser-cooled atoms.

A note on history

The idea was proposed by Hänsch and Schawlow (and, for ions, by Wineland and Dehmelt) in 1975. Steven Chu's group demonstrated optical molasses in 1985. The 1988 surprise — temperatures below the Doppler limit — forced the theory to be rebuilt around Sisyphus cooling by Cohen-Tannoudji and colleagues, with William Phillips's group pinning down the sub-Doppler measurements. Chu, Cohen-Tannoudji, and Phillips shared the 1997 Nobel Prize in Physics "for development of methods to cool and trap atoms with laser light."