Optical Molasses

Optical molasses is the trick that made cold-atom physics an experimental science: six laser beams, tuned a hair below an atomic resonance, that between them behave like a thick fluid for atoms. Drop a rubidium atom into the overlap of three perpendicular counter-propagating pairs and any velocity it has is bled away in tens of microseconds, leaving a glowing cloud at a hundred microkelvin — atoms crawling at centimetres per second instead of the hundreds of metres per second they had in the oven. Steven Chu's group at Bell Labs coined the name in 1985 because the atoms move exactly as a marble would through honey: whichever way they drift, the medium pushes back in proportion to their speed. The engine underneath is Doppler cooling, and it borrows from every article upstream. From the radiation-pressure scattering force it takes $F_{\text{scatt}}=\hbar k R_{\text{scatt}}$ , the recoil an atom feels as it cycles photons; from the two-level line shape it takes the Lorentzian scattering rate $R_{\text{scatt}}(\delta)$ , sharply peaked at resonance. The molasses effect is what happens when that rate depends on velocity: red-detune the light, and a moving atom sees the beam it runs into Doppler-shifted up toward resonance — scattering more — while the beam chasing it is shifted away. The imbalance always opposes the velocity, and expanding to first order drops out a genuine friction, $\vec F=-\alpha\vec v$ . But photons are grainy, and each random recoil kick heats as surely as the average force cools. The truce between the two — a fluctuation-dissipation balance identical in structure to Brownian motion — fixes the lowest temperature Doppler molasses can reach, the celebrated Doppler limit $k_B T_D=\hbar\Gamma/2$ . That molasses routinely beats its own limit was the surprise that opened the door to Sisyphus cooling and, eventually, to Bose-Einstein condensation.

Key takeaways

Optical molasses is three orthogonal pairs of red-detuned counter-propagating beams producing a viscous friction $\vec F=-\alpha\vec v$ in all three dimensions.
The mechanism is the Doppler shift: a moving atom scatters more from the beam it moves into (shifted toward resonance) than from the one behind, so the net force opposes the velocity.
The friction coefficient $\alpha$ peaks near $\delta=-0.29\,\Gamma$ , but the temperature is minimized at $\delta=-\Gamma/2$ , so molasses is run detuned by roughly half a linewidth.
Molasses cools but does not trap: the force depends on velocity, not position — a damper without a spring, so the cloud slowly diffuses out.
Random photon recoils heat the gas; balancing that against the friction gives the Doppler limit $k_B T_D=\hbar\Gamma/2$ — about 235 μK for sodium, 146 μK for rubidium-87.
Key idea: a red-detuned beam scatters preferentially from atoms moving toward it, so six of them make a linear friction $\vec F=-\alpha\vec v$ in three dimensions — cooling without a restoring force, hence cooling without trapping.

The core insight

A red-detuned beam scatters preferentially from atoms moving toward it (the Doppler shift brings it toward resonance); six beams turn that asymmetry into a linear friction $\vec F=-\alpha\vec v$ in three dimensions, but with no dependence on position — a damper without a spring, so molasses cools without trapping.

Where this lives today

Every experiment that reaches for ultracold atoms passes through a molasses stage, usually without ceremony. In a typical rubidium apparatus a magneto-optical trap first collects $10^9$ atoms; the magnetic field is then switched off for a few milliseconds of pure "optical molasses," during which polarization-gradient cooling drives the cloud from the Doppler limit toward 10 μK before loading into a magnetic or optical-dipole trap for evaporation to Bose-Einstein condensation. The same physics runs the world's primary time standards: a caesium fountain clock gathers atoms in molasses, launches the cold ball up through a microwave cavity, and reads the ultra-narrow Ramsey fringe as the atoms rise and fall — the low velocity spread is what makes the fringe narrow enough to define the SI second. Molasses is now so standard that it appears in teaching labs and in chip-scale devices: portable gravimeters, inertial sensors and optical clocks carrying a molasses region a few millimetres across. What began as atoms hanging in crossed beams at Bell Labs is the quiet first step of the entire quantum-gas and quantum-sensing enterprise.

The physical picture

Imagine running forward through driving rain: the drops hitting your face come faster and harder than those catching your back, so the net push is always backward. Optical molasses is that picture made out of photons and made deliberately lopsided. Replace the rain with two laser beams pointing at each other. If they were tuned exactly to the atom's resonance, running toward one and away from the other would shift the two equally and oppositely, the scattering rates would still match to first order, and there would be no net force. The trick is to detune to the red, below resonance. Now the beam the atom runs into is Doppler-shifted up, toward the peak, so it scatters more from that side; the beam behind is shifted further into the red and scatters less. The push is always a headwind.

Figure

Six red-detuned beams meet on one atom. A moving atom sees the beam it runs into shifted toward resonance by the Doppler effect δ → δ plus kv so it scatters more from that side, and the net radiation force opposes the velocity in every direction

Stack three such pairs along $x$ , $y$ and $z$ and the atom is caught from every direction at once — six beams, each opposing whatever component of velocity points into it. But this is not a trap. Nowhere did the atom's position enter; the force reads only velocity. Molasses is pure friction, a damper with no spring, and a damped particle with no restoring force does not sit still — it random-walks. The second figure plots the force against velocity: a steep linear slope through the origin (the friction), and a turnover at a few metres per second, beyond which a fast atom is Doppler-shifted clean out of resonance and slips through uncooled.

Figure

Net molasses force versus velocity for rubidium-87 at δ = −Γ/2 and s = 0.2. Near v = 0 the force is linear, F ≈ −αv, the viscous friction; it peaks around 2.4 m/s and dies beyond the capture velocity v_c ≈ Γ/k ≈ 4.7 m/s where the Doppler shift outruns the linewidth

The temperature is set by a different contest. Every scattered photon lands an average kick that cools, but the kicks are random in direction, so the atom's momentum also random-walks and heats. Cooling scales with the friction, heating with the scattering rate, and where they meet is the equilibrium temperature — precisely Einstein's Brownian-motion balance $k_BT=D_p/\alpha$ , which minimizes to $\hbar\Gamma/2$ . Below that the two-level story runs out, and real atoms, with their multiple ground sublevels and polarization-gradient light shifts, quietly go colder still.

The core equations

Everything begins with the force one beam exerts on a two-level atom at rest. A photon carries momentum $\hbar k$ ; the atom absorbs at the scattering rate $R_{\text{scatt}}$ and re-emits in random directions that average to nothing, so the beam delivers a steady push equal to momentum-per-photon times photons-per-second.

F_{\text{scatt}}=\hbar k\,\frac{\Gamma}{2}\,\frac{s}{1+s+(2\delta/\Gamma)^2} \qquad (1)

where: $F_{\text{scatt}}$ — force from one beam on a stationary atom; $\hbar k$ — photon momentum, $k=2\pi/\lambda$ ; $\Gamma$ — natural linewidth (spontaneous decay rate); $s=I/I_{\text{sat}}$ — saturation parameter per beam; $\delta=\omega_L-\omega_0$ — laser detuning, negative for red. The Lorentzian factor is the scattering rate $R_{\text{scatt}}$ ; it saturates at $\Gamma/2$ on resonance and falls off as $1/\delta^2$ far away. Now put the atom in motion. In its rest frame each beam is Doppler-shifted, and because the two beams of a pair travel oppositely the shifts have opposite sign: the beam the atom moves into has its frequency raised, the one behind lowered.

\delta_\pm=\delta\mp kv \qquad (2)

where: $\delta_-$ — effective detuning of the beam the atom runs into (raised toward resonance for red $\delta$ ); $\delta_+$ — effective detuning of the beam behind it (lowered, further into the red); $kv$ — the Doppler shift at speed $v$ . For red detuning the "into" beam sits nearer resonance and therefore scatters harder. Add the two Doppler-shifted single-beam forces of a counter-propagating pair and Taylor-expand for $kv\ll\Gamma$ . The zeroth-order terms cancel by symmetry; the first-order term survives, linear in $v$ — a bona fide viscous friction.

\vec F=-\alpha\vec v,\qquad \alpha=-8\hbar k^2 s\,\frac{2\delta/\Gamma}{\left[1+s+(2\delta/\Gamma)^2\right]^2} \qquad (3)

where: $\vec F$ — net force from the pair (or, summed, all six beams); $\alpha$ — damping coefficient, positive when $\delta<0$ ; $\vec v$ — atomic velocity. Because $\alpha>0$ only for red detuning, the sign of the detuning is the difference between cooling and blowing the cloud apart. Derivation 1 does the expansion in full. Cooling cannot win outright: the same photons that cool also heat, since each absorption and emission delivers a random $\hbar k$ kick and the momentum diffuses. Balancing diffusive heating against frictional cooling gives an Einstein relation, and minimizing over detuning gives the floor of Doppler cooling.

k_B T=\frac{\hbar\Gamma}{2}\,\frac{1+s+(2\delta/\Gamma)^2}{4|\delta|/\Gamma}\;\xrightarrow{\ \delta=-\Gamma/2,\ s\to0\ }\;k_B T_D=\frac{\hbar\Gamma}{2} \qquad (4)

where: $T$ — steady-state temperature at detuning $\delta$ ; $T_D$ — the Doppler limit, the minimum over $\delta$ at low intensity; $k_B$ — Boltzmann's constant. The optimum sits at $\delta=-\Gamma/2$ , giving $T_D=\hbar\Gamma/2k_B$ : about 235 μK for sodium and 146 μK for rubidium-87. Derivation 2 builds it from the diffusion balance. Newton's equation with only a linear friction term has no oscillation and no equilibrium position — just an exponential collapse of velocity. Its time constant is the single most useful number for an experimentalist: how fast molasses acts.

m\dot v=-\alpha v\;\Rightarrow\;v(t)=v_0\,e^{-t/\tau},\qquad \tau=\frac{m}{\alpha} \qquad (5)

where: $\tau$ — 1/e velocity-damping time, typically tens of microseconds; $m$ — atomic mass; $v_0$ — initial speed. There is no restoring term, so the motion is overdamped in the extreme: velocity decays but position does not return. Two velocity scales bound the process, and a third relation says why the cloud leaks away. Fast atoms outside the capture range are Doppler-shifted out of resonance and never cool; the recoil velocity is the granularity of the kicks; and with no spatial force the centre of mass diffuses like a Brownian particle.

v_c\sim\frac{\Gamma}{k},\qquad v_r=\frac{\hbar k}{m},\qquad \langle x^2\rangle=2D_x t,\quad D_x=\frac{k_B T}{\alpha} \qquad (6)

where: $v_c$ — capture velocity, the fastest atom still cooled; $v_r$ — single-photon recoil velocity, the smallest possible velocity change; $D_x$ — spatial diffusion coefficient (Einstein relation); $\langle x^2\rangle$ — mean-square displacement, growing without bound. The escape time from a region of size $L$ is $\sim L^2/2D_x$ — a fraction of a second, all the "confinement" molasses offers.

Derivation AThe molasses friction coefficient from Doppler-shifted forces

The goal is to turn the intuitive "headwind" argument into the explicit friction coefficient $\alpha$ , by writing the two single-beam forces of a counter-propagating pair with their Doppler-shifted detunings, adding them, and expanding to first order in velocity. Three things should fall out cleanly: the zeroth-order forces cancel exactly for a stationary atom, the surviving term is strictly linear in $v$ (a viscous friction), and its coefficient is positive — genuinely cooling — only for red detuning. That sign is the whole reason molasses works.

Step 1. Write the two beams. Take the pair along $x$ ; the beam in $-x$ pushes the atom in $-x$ at detuning $\delta-kv$ , the beam in $+x$ pushes in $+x$ at detuning $\delta+kv$ : $F_\pm=\pm\hbar k\frac{\Gamma}{2}\,s/[1+s+(2(\delta\mp kv)/\Gamma)^2]$ . Why: each carries the stationary-atom force of equation (1), evaluated at the frequency the moving atom actually sees — the only place velocity enters.

Step 2. Add them, $F=F_++F_-$ . At $v=0$ the Doppler shifts vanish, the two Lorentzians are identical, and the equal-and-opposite pushes cancel exactly. Why: there is no zeroth-order force, so any net effect is entirely a velocity effect — the cooling is first order in $v$ .

Step 3. Expand for slow atoms, $kv\ll\Gamma$ : $F\approx\left.\dfrac{\partial(F_++F_-)}{\partial v}\right|_{0}v\equiv-\alpha v$ . The two beams' derivatives add rather than cancel because their velocity dependence has opposite sign. Why: a force proportional to $-v$ is exactly a viscous drag, so a slow atom decelerates exponentially rather than bouncing.

Step 4. Read off $\alpha=-8\hbar k^2 s\,(2\delta/\Gamma)/[1+s+(2\delta/\Gamma)^2]^2$ . The factor $2\delta/\Gamma$ carries the sign of $\delta$ , so $\alpha>0$ (cooling) demands red detuning; a blue-detuned pair gives $\alpha<0$ , anti-friction. Why: the derivation delivers a design rule — tune below resonance, or the molasses becomes an atom cannon. *Verified: reproduces Foot Eq. 9.6 and Metcalf and van der Straten Ch. 7; sign confirmed positive for $\delta<0$ by expanding the two-beam Lorentzian.*

Go deeper · Derivation B · 1 minThe Doppler limit from the cooling–heating balance

Derivation BThe Doppler limit from the cooling–heating balance

The goal is to explain why cooling stops at a finite temperature rather than absolute zero, and to compute where. The friction of Derivation 1 removes energy smoothly, but photons are discrete: every absorption and emission delivers a random $\hbar k$ kick, so the momentum random-walks and pumps energy back in. We treat the momentum as a Brownian variable driven by a deterministic drag plus a random force, set heating equal to cooling in steady state, and minimize over detuning. The answer, $k_BT_D=\hbar\Gamma/2$ , is one of the cleanest results in atomic physics.

Step 1. Heating rate. Random kicks grow the mean-square momentum as $\langle p^2\rangle=2D_p t$ with $D_p\approx(\hbar k)^2 R_{\text{scatt}}$ , so energy is injected at $(\mathrm dE/\mathrm dt)_{\text{heat}}=D_p/m$ . Why: a random walk of step $\hbar k$ taken $R_{\text{scatt}}$ times per second diffuses at exactly this rate — the irreducible price of using real photons.

Step 2. Cooling rate. Friction removes kinetic energy at $(\mathrm dE/\mathrm dt)_{\text{cool}}=-\alpha\langle v^2\rangle=-(2\alpha/m)E_{\text{kin}}$ . Why: power is force times velocity, so a force $-\alpha v$ dissipates at rate $\alpha v^2$ — the faster the atom, the harder molasses brakes it.

Step 3. Balance. In steady state heating equals cooling, giving $k_B T=D_p/\alpha$ — an Einstein relation tying temperature to the ratio of diffusion and drag. Why: the same fluctuation-dissipation logic that fixes the temperature of Brownian pollen fixes the temperature of an atom in light.

Step 4. Substitute and minimize. This gives $k_BT=(\hbar\Gamma/2)\,[1+s+(2\delta/\Gamma)^2]/(4|\delta|/\Gamma)$ ; minimizing over $|\delta|$ at low $s$ sends $2|\delta|/\Gamma\to1$ , i.e. $\delta=-\Gamma/2$ , and $k_BT_D=\hbar\Gamma/2$ . Why: the detuning that minimizes temperature is not the one that maximizes friction — this is the true optimum for a cold cloud. *Verified: $T_D=\hbar\Gamma/2k_B$ gives 235 μK for Na ( $\Gamma/2\pi=9.79$ MHz) and 146 μK for $^{87}$ Rb ( $\Gamma/2\pi=6.07$ MHz), recomputed from tabulated linewidths.*

Go deeper · Derivation C · 1 minTwo different optimal detunings — maximum friction versus minimum temperature

Derivation CTwo different optimal detunings — maximum friction versus minimum temperature

The goal is to resolve a subtlety that trips up almost everyone: the claim that molasses works best at $\delta=-\Gamma/2$ is only half true, because two distinct quantities are optimized. The friction coefficient $\alpha$ , which sets how fast the cloud cools and how wide the capture range is, peaks at one detuning; the steady-state temperature, which sets how cold it ends up, is minimized at another. They do not coincide — the difference is a fast molasses versus a cold one — and we locate both extrema explicitly at low intensity.

Step 1. Maximize the friction. With $u=2\delta/\Gamma$ and $s\to0$ , $\alpha\propto -u/(1+u^2)^2$ ; setting the derivative to zero gives $3u^2=1$ , so $u=-1/\sqrt3$ and $\delta=-\Gamma/(2\sqrt3)\approx-0.29\,\Gamma$ . Why: friction is the slope of the force at $v=0$ , steepest a little closer to resonance than the half-width, where the Lorentzian's flank turns over fastest.

Step 2. Compare. At $-0.29\,\Gamma$ the coefficient is about 30 percent larger than at $-\Gamma/2$ — noticeable but not dramatic. Why: the two detunings are close enough that using $-\Gamma/2$ costs little friction, which is why the folklore "detune by half a linewidth" survives.

Step 3. Minimize the temperature. From equation (4), $k_BT\propto(1+u^2)/(2|u|)$ at low $s$ ; minimizing gives $|u|=1$ , i.e. $\delta=-\Gamma/2$ . Why: though friction is slightly stronger nearer resonance, heating grows faster there, so the coldest cloud sits exactly at half a linewidth.

Step 4. Reconcile. Detune shallow ( $-0.29\,\Gamma$ ) to grab and brake atoms fast; detune to $-\Gamma/2$ for the lowest Doppler temperature; in practice one works near $-\Gamma/2$ and beyond, once sub-Doppler mechanisms take over. Why: a real molasses must both capture a wide velocity class and end cold, so both extrema matter. *Verified: $\alpha$ -maximum at $2\delta/\Gamma=-1/\sqrt3$ and $T$ -minimum at $2\delta/\Gamma=-1$ from $\mathrm d/\mathrm d\delta$ of the low- $s$ expressions; ratio $\alpha(-0.29\Gamma)/\alpha(-0.5\Gamma)\approx1.30$ .*

Limiting cases

Case	Result	Meaning
Slow atom ( $kv\ll\Gamma$ )	$F\approx-\alpha v$	viscous friction; velocity decays exponentially with $\tau=m/\alpha$
Fast atom ( $kv\gtrsim\Gamma$ )	force turns over and falls	Doppler-shifted out of resonance; sets a finite capture velocity $v_c\sim\Gamma/k$
Maximum friction	$\delta=-\Gamma/(2\sqrt3)\approx-0.29\,\Gamma$	steepest force slope — fastest cooling and widest capture, at low $s$
Minimum temperature	$\delta=-\Gamma/2$ , $k_BT_D=\hbar\Gamma/2$	the Doppler limit; coldest two-level equilibrium
Blue detuning ( $\delta>0$ )	$\alpha<0$ , $F=+\|\alpha\|v$	anti-friction: the light heats and expels the cloud
Long times, no field gradient	$\langle x^2\rangle=2D_x t$	position diffuses without bound — molasses cools but never traps

Common trapMolasses traps atoms — it does not

This is the single most common confusion, and it matters. Optical molasses exerts a force that depends only on velocity, $\vec F=-\alpha\vec v$ ; nothing depends on where the atom is. A damper is not a trap. Turn it on and the cloud cools in place, but its centre of mass random-walks out of the beam overlap in a fraction of a second — no restoring force pulls a straying atom back. To trap, add a magnetic-field gradient and $\sigma^+/\sigma^-$ polarizations, converting molasses into a magneto-optical trap where the Zeeman shift makes the force position-dependent. A related trap: the Doppler limit $\hbar\Gamma/2k_B$ is not the coldest molasses reaches. Real multilevel atoms in polarization gradients cool by the Sisyphus mechanism well below it — sodium was measured near 40 μK against a 235 μK prediction — so a temperature below $T_D$ signals sub-Doppler cooling, not an error.

Worked examples

1 · How cold, and how slow, is Doppler-limited rubidium? (the concrete anchor)

The workhorse cold-atom species is $^{87}$ Rb on its D2 line: $\lambda=780$ nm, $\Gamma/2\pi=6.07$ MHz, $m=87\,u$ . Find the Doppler-limit temperature, the corresponding one-dimensional rms speed, and compare that speed to the single-photon recoil velocity.

Step 1. Doppler limit. With $\Gamma=2\pi\times6.07\times10^6=3.81\times10^7\ \mathrm{s^{-1}}$ , $k_BT_D=\hbar\Gamma/2$ gives $T_D=\hbar\Gamma/2k_B=(1.055\times10^{-34})(3.81\times10^7)/(2\times1.381\times10^{-23})=\mathbf{146\ \mu K}$ . That is six orders of magnitude below room temperature.

Step 2. How fast do the atoms move there? The 1D rms speed is $v_{\text{rms}}=\sqrt{k_BT_D/m}$ with $m=87\times1.66\times10^{-27}=1.44\times10^{-25}$ kg: $v_{\text{rms}}=\sqrt{(1.381\times10^{-23})(1.46\times10^{-4})/(1.44\times10^{-25})}=\mathbf{0.118\ m/s}$ , about 12 cm/s — a walking pace slowed ten-thousandfold.

Step 3. Compare to the recoil velocity $v_r=\hbar k/m$ with $k=2\pi/\lambda=8.05\times10^6\ \mathrm{m^{-1}}$ : $v_r=\mathbf{5.9\ mm/s}$ . So at the Doppler limit the cloud is still about $v_{\text{rms}}/v_r\approx\mathbf{20}$ recoils wide — real room below, exactly what sub-Doppler cooling exploits. T_D = 146 μK, v_rms ≈ 12 cm/s ≈ 20 recoil velocities.

2 · How quickly does molasses stop an atom?

For the same $^{87}$ Rb molasses at $\delta=-\Gamma/2$ and saturation $s=0.2$ per beam, estimate the friction coefficient, the velocity-damping time $\tau$ , and the capture velocity.

Step 1. Friction coefficient. At $\delta=-\Gamma/2$ , $2\delta/\Gamma=-1$ , so $\alpha=-8\hbar k^2 s(-1)/[1+s+1]^2=8\hbar k^2 s/(2+s)^2$ . With $\hbar k^2=(1.055\times10^{-34})(8.05\times10^6)^2=6.84\times10^{-21}$ and $s=0.2$ : $\alpha=8(6.84\times10^{-21})(0.2)/(2.2)^2=\mathbf{2.3\times10^{-21}\ kg/s}$ .

Step 2. Damping time. $\tau=m/\alpha=(1.44\times10^{-25})/(2.3\times10^{-21})=\mathbf{64\ \mu s}$ . An atom's velocity falls by $1/e$ every 64 microseconds, so after a few hundred microseconds it is thermalized to the molasses temperature — this is why molasses is a fast stage, not a slow bake.

Step 3. Capture velocity. The linear friction only holds while $kv\lesssim\Gamma$ ; beyond $v_c\sim\Gamma/k=(3.81\times10^7)/(8.05\times10^6)=\mathbf{4.7\ m/s}$ the atom is Doppler-shifted off resonance and slips through. So molasses grabs atoms only up to a few metres per second — which is why it must be fed by a Zeeman slower or a MOT, not a room-temperature vapour directly. Note the photon budget: stopping an atom from $v_c$ takes $v_c/v_r\approx\mathbf{800}$ scattered photons, delivered in about 0.5 ms — which is why a closed cycling transition (plus a repumper) is essential. α ≈ 2.3×10⁻²¹ kg/s, τ ≈ 64 μs, v_c ≈ 4.7 m/s, ≈ 800 photons to stop.

Physics of the portrait

Optical molasses is one of those rare ideas that arrived with its name already perfect. The person who named it, and first watched a cloud of sodium hang glowing in the crossed beams, spent his career turning the trick into a toolkit — and then ran a national energy programme with it on his résumé.

Origins · 1 minSteven Chu and how this was discovered

Steven Chu

1948– · Bell Labs and Stanford · Nobel Prize 1997 · laser cooling and trapping of neutral atoms

Steven Chu was a young group leader at AT&T Bell Laboratories in 1985 when he and his collaborators — Leo Hollberg, John Bjorkholm, Alex Cable and Arthur Ashkin — first cooled sodium atoms in three orthogonal pairs of red-detuned beams and measured a temperature near the Doppler limit, around 240 μK. Chu coined the phrase "optical molasses" for the way the atoms behaved as though suspended in a viscous fluid, and the image stuck because it is exactly right: the force is a velocity-proportional drag. The demonstration built on the 1975 Hänsch–Schawlow proposal of Doppler cooling and on Ashkin's optical-trapping work, and it turned a theoretical scheme into a routine every cold-atom lab now runs daily. When Chu, Claude Cohen-Tannoudji and William Phillips shared the 1997 Nobel Prize in Physics "for the development of methods to cool and trap atoms with laser light," molasses was the shared foundation of all three citations — Phillips having measured the surprising sub-Doppler temperatures, and Cohen-Tannoudji and Jean Dalibard having explained them. Chu later moved to Stanford and served as United States Secretary of Energy from 2009 to 2013 — a reminder that a viscous-fluid analogy for atoms can carry a physicist a very long way.

History

1975

— Hänsch and Schawlow propose Doppler cooling of neutral atoms with counter-propagating laser beams

1985

— Chu, Hollberg, Bjorkholm, Cable and Ashkin demonstrate optical molasses in sodium at Bell Labs

1988

— Lett and colleagues at NIST measure molasses temperatures around 40 μK, far below the Doppler limit

1989

— Dalibard and Cohen-Tannoudji explain sub-Doppler Sisyphus and polarization-gradient cooling

1997

— Chu, Cohen-Tannoudji and Phillips share the Nobel Prize for laser cooling and trapping

Going further

Texts. Foot, Atomic Physics, Ch. 9 — the cleanest textbook derivation of $\alpha$ and $T_D$ · Metcalf and van der Straten, Laser Cooling and Trapping, Ch. 7 — the standard monograph · Demtröder, Laser Spectroscopy, Vol. 2, Ch. 9 · Loudon, The Quantum Theory of Light, 3rd ed. · Saleh and Teich, Fundamentals of Photonics.

Provenance. The friction coefficient, single-beam scattering force and Doppler limit are standard two-level results following Foot Ch. 9 and Metcalf and van der Straten Ch. 7; a verifier should recover $\alpha$ by Taylor-expanding the two-beam Lorentzian and confirm $T_D=\hbar\Gamma/2k_B$ from tabulated linewidths (Na $\Gamma/2\pi=9.79$ MHz gives 235 μK; $^{87}$ Rb $\Gamma/2\pi=6.07$ MHz gives 146 μK, Steck alkali data). The two-optima result — $\alpha$ maximized at $\delta=-\Gamma/(2\sqrt3)$ , $T$ minimized at $\delta=-\Gamma/2$ — follows by differentiating the low-intensity expressions. Historical facts (Chu's 1985 demonstration, the 1988 sub-Doppler measurements, the 1997 Nobel Prize) are per Chu et al., Phys. Rev. Lett. 55, 48 (1985); Lett et al., Phys. Rev. Lett. 61, 169 (1988); and the Nobel citations.

Builds on. Radiation Pressure and the Scattering Force · Doppler Cooling and the Doppler Limit · Dipole Force and the Optical Dipole Potential. Leads to. Sub-Doppler and Sisyphus Cooling · The Recoil Limit and Sub-Recoil Cooling · The Magneto-Optical Trap · Evaporative Cooling · Bose-Einstein Condensation in Dilute Gases.

Ask the tutor: Ask me to show why the coldest molasses detuning ( $-\Gamma/2$ ) differs from the fastest-cooling one ( $-0.29\,\Gamma$ ), or to trace exactly how six beams turn one Doppler shift into three-dimensional friction.

Why exactly does red detuning cool but blue detuning heat?

It comes down to which beam an atom scatters more from. For red detuning ( $\delta<0$ ) the light sits below resonance, so moving toward a beam Doppler-shifts it up, toward the peak, raising its scattering rate; the beam behind is shifted away and scatters less. The stronger push opposes the motion — a headwind, $\alpha>0$ . Blue detuning reverses this: moving toward a beam shifts it past resonance, so now the trailing beam scatters harder and the push is a tailwind that accelerates the atom, $\alpha<0$ , and the cloud heats and blows apart. The sign of $\delta$ is the sign of the friction.

If molasses cannot trap, why is it so useful?

Because trapping and cooling are different jobs, and molasses is the best cooler. A magneto-optical trap holds atoms but its light configuration is a compromise; the coldest temperatures come from switching off the magnetic field and running a brief pure-molasses phase, where polarization-gradient cooling drives the cloud far below the Doppler limit. Molasses buys a dense, microkelvin sample in milliseconds — you then transfer it into a conservative trap that does confine, and let evaporation finish toward degeneracy. Molasses is not a storage device; it is the refrigeration stage that prepares atoms cold enough for the traps that follow.

Where does the heating that limits cooling actually come from?

From the graininess of light. The average force cools smoothly, but each photon is absorbed and then spontaneously re-emitted in a random direction, and each event delivers a discrete $\hbar k$ kick. Because the directions are random, the kicks do not average to zero in their square — the momentum executes a random walk, which heats, at a rate set by the scattering rate. Cooling scales with the friction $\alpha$ ; heating scales with the diffusion coefficient $D_p$ . Their balance, $k_BT=D_p/\alpha$ , is a genuine Einstein fluctuation-dissipation relation, and it never reaches zero because cooling needs scattering, and scattering brings recoil heating.

What is the capture velocity and why does it matter?

The friction $\vec F=-\alpha\vec v$ is only linear while the Doppler shift $kv$ stays small compared with $\Gamma$ . Once an atom moves faster than $v_c\sim\Gamma/k$ — a few metres per second for rubidium — the Doppler shift throws it out of resonance, the scattering collapses, and it sails through barely touched. This is why molasses cannot cool a room-temperature vapour directly: thermal atoms move at hundreds of metres per second, far too fast to catch. A Zeeman slower or the wide capture range of a MOT first brings atoms into the metres-per-second class, and only then does molasses take over.

Is molasses actually cooling, or just slowing the atoms down?

It is genuinely cooling — temperature, not just average speed, goes down. Cooling means shrinking the spread of velocities, and that is exactly what a velocity-proportional friction does: it pulls every atom's velocity toward zero regardless of direction. An atom moving left is braked, an atom moving right is braked, and the width of the distribution — which is what temperature measures — collapses. Slowing a single atom is not cooling; cooling is what happens to an ensemble when the friction narrows its velocity histogram. The Doppler limit is where that narrowing is balanced by recoil heating, so the width settles at the value $\hbar\Gamma/2k_B$ .

Why is a closed cycling transition needed for Doppler cooling?

Because cooling takes hundreds to thousands of absorption–emission cycles, and every one must return the atom to the same ground state ready to absorb again. If a decay path leads into a ground level the cooling laser cannot address — a "dark state" — the atom falls in after a few cycles and stops scattering, stalling the cooling. Real alkali atoms have hyperfine structure that provides exactly such leaks, so practical molasses always includes a "repumper" laser that returns atoms to the cycling transition. The requirement is quantitative: with a per-cycle leak probability of even $10^{-3}$ , an atom needing 800 scatters would almost certainly be lost.