Slow and Cold Atomic Beams

Almost every experiment in modern atomic physics begins by losing a fight with thermodynamics. An oven heated to a few hundred kelvin sprays out atoms at the most-probable speed of a bullet — around $800\ \mathrm{m\,s^{-1}}$ for sodium — with a velocity spread just as wide, while the traps, clocks, and interferometers downstream can only hold atoms moving at walking pace. Bridging that gap is the job of the slow and cold atomic beam, and the tool that does it is borrowed wholesale from the radiation-pressure article that precedes this one: the scattering force, in which each absorbed photon delivers a momentum kick $\hbar k$ against the atom's motion while the following spontaneous emission averages to nothing. Cycle that absorb-and-re-emit process fast enough and the resulting deceleration reaches $\sim 10^{5}$ – $10^{6}\ \mathrm{m\,s^{-2}}$ , enough to stop a thermal sodium atom inside a metre. The force is enormous, but it is also exquisitely fussy. It only acts on atoms whose Doppler shift $kv$ places them within a natural linewidth $\Gamma$ of resonance, and the very act of slowing changes $kv$ by hundreds of linewidths, so an unassisted atom falls out of resonance almost at once and coasts. The whole engineering of a slow beam is therefore the engineering of staying resonant while the velocity drops — a compensating Zeeman shift that tracks position in the Zeeman slower, or a laser frequency swept in time in chirp cooling. A separate transverse molasses collimates the diverging beam, multiplying its brightness. This article treats the effusive source, the deceleration ceiling, the resonance-tracking condition, and the collimation stage as one connected machine, and it sets the loading conditions for the magneto-optical trap, evaporative cooling, and the quantum-degenerate gases that come after.

Key takeaways

A thermal beam effuses from an oven at hundreds of metres per second with a broad flux distribution $f(v)\propto v^{3}e^{-mv^{2}/2k_BT}$ ; using it to feed traps and clocks demands both longitudinal deceleration and transverse collimation.
A near-resonant counter-propagating laser decelerates atoms through the scattering force, with a hard ceiling $a_{\max}=\hbar k\Gamma/(2m)\approx9\times10^5\,{\rm m\,s^{-2}}$ for sodium, enough to stop a $1000\,{\rm m\,s^{-1}}$ atom over a distance of about $0.55\,{\rm m}$ .
The central obstacle is that slowing shifts the atom out of resonance by the changing Doppler shift $kv$ ; the Zeeman slower cancels it with a tapered magnetic field and chirp cooling cancels it with a swept laser frequency.
A two-dimensional optical molasses acting transverse to the beam damps the sideways velocity spread toward the Doppler limit, collimating the flux and multiplying on-axis brightness — often the single largest gain for downstream trap loading.
Figures of merit are the flux (atoms per second), the final velocity (a few to tens of $\mathrm{m\,s^{-1}}$ ), the phase-space brightness, and the capture velocity of the trap being fed.
Key idea: a resonant laser exerts a large but strictly velocity-selective force, so producing a useful slow beam means keeping the decelerating atoms on resonance — via a Zeeman field or a frequency chirp — while a transverse molasses collimates the survivors.

The core insight

A resonant laser exerts a large but strictly velocity-selective force; producing a useful slow beam means keeping the decelerating atoms on resonance — through a position-tailored Zeeman field or a time-swept laser frequency — while a transverse molasses collimates the survivors.

Where this lives today

Walk up to almost any ultracold-atom experiment and the first thing on the vacuum chamber is a slow-beam source. In a common arrangement, a Zeeman slower — a metre-scale tapered magnetic-field structure wrapped around a beam tube, fed by an atomic oven at one end and a single fixed-frequency slowing laser at the other — delivers $10^{10}$ to $10^{11}$ atoms per second at around $20\ \mathrm{m\,s^{-1}}$ into a magneto-optical trap, raising the trapped atom number by many orders of magnitude over what an unslowed thermal beam could load. Rubidium and cesium machines often skip the solenoid and use a two-dimensional MOT as a bright cold source instead, while strontium and ytterbium optical-lattice clocks --- among the leading candidates for a future redefinition of the second at the $10^{-18}$ level — rely on Zeeman slowers plus narrow-line cooling stages to reach the microkelvin temperatures required for precision interrogation. The same front end feeds the whole apparatus of quantum-degenerate matter. A large MOT loaded from a slow beam is transferred to a magnetic or optical dipole trap and evaporatively cooled toward Bose–Einstein condensation or a degenerate Fermi gas, and the slow-beam flux sets a hard limit on the final atom number and condensate size. Light-pulse atom interferometers for inertial sensing and tests of the equivalence principle depend on the narrow momentum spread of a slow, velocity-selected beam, because their phase sensitivity grows as the square of the interrogation time and residual divergence caps that time. From frequency standards to gravimeters to prototype gravitational-wave detectors, the slow and cold atomic beam is the quiet first stage on which the rest is built.

The physical picture

Picture the atom as a ratchet driven by a hail of photons. It moves toward a laser beam; every photon it absorbs from that beam kicks it backward by one recoil velocity $v_r=\hbar k/m\approx 3\ \mathrm{cm\,s^{-1}}$ for sodium, and the fluorescence photon that follows flies off in a random direction, so over many cycles only the absorption imparts net momentum. The kicks are tiny, but they come fast: at saturation the atom can scatter at half the natural decay rate $\Gamma/2$ , and repeating the cycle some $30{,}000$ times strips away $\sim 1000\ \mathrm{m\,s^{-1}}$ . The catch is that the ratchet only remains effective while the atom stays near resonance.

Figure

The photon ratchet: an atom moving right at v absorbs counter-propagating photons, each delivering a recoil kick −ħk with v_r ≈ 3 cm s⁻¹, while spontaneous emission is random and averages to zero over many cycles. At saturation the scattering rate is Γ∕2, and the deceleration ceiling is a_max = ħkΓ∕2m ≈ 9×10⁵ m s⁻² for sodium, requiring about 3×10⁴ photons to stop a 1000 m s⁻¹ atom.

The resonance condition is the entire story. An atom moving toward the laser at speed $v$ sees the light blue-shifted by $kv$ , so to be resonant the laser must sit red-detuned by exactly that amount. As the atom slows, $kv$ shrinks, and the detuning that keeps it resonant must shrink with it. A Zeeman slower supplies a compensating shift that varies with position along the beam; chirp cooling supplies one that varies with time. Both are nothing more than two ways of enforcing the single equation $\delta(v)=0$ along the atom's trajectory — a magnetic field profile in one case, a frequency ramp in the other. The second figure shows the payoff: a velocity that falls as a clean square root of position, shadowed by a magnetic field tailored to keep the atom locked on resonance the whole way down.

Figure

The resonance-tracked velocity profile of an ideal sodium Zeeman slower: constant deceleration a = 0.5 a_max gives v(z) = v₀√(1−z∕L₀) with v₀ = 1000 m s⁻¹ and slowing length L₀ ≈ 1.1 m (blue, left axis), while the tapered field B(z) = B_bias + B₀√(1−z∕L₀) with B₀ ≈ 1213 G tracks the shrinking Doppler shift to hold δ = 0 at every z (amber, right axis)

The effusive source itself is best thought of as a firehose with a huge velocity spread in every direction. Longitudinal slowing tames the forward spread, but the transverse spread makes the beam fan out and spill against the walls of the apparatus long before it reaches the trap. A two-dimensional optical molasses, two orthogonal pairs of counter-propagating beams acting perpendicular to the axis, damps the sideways velocities toward the Doppler limit, squeezing the divergence and multiplying the on-axis brightness — frequently the biggest single win in the whole loading chain. Together the two stages set how cold and how bright the beam can be: the transverse temperature bottoms out near the Doppler limit $k_BT_D=\hbar\Gamma/2$ (a few hundred microkelvin, lower with sub-Doppler mechanisms), while the longitudinal "temperature" is really a narrow velocity slice hand-picked by the resonance-tracking scheme.

The core equations

The source distribution is where the whole problem starts. Atoms leave the oven not with the Maxwell–Boltzmann speed distribution of the gas inside, but with an extra factor of $v$ , because fast atoms cross the aperture more often than slow ones. The result is a flux distribution weighted toward higher speeds, which is exactly why the beam is such a difficult thing to slow.

f(v)\,\mathrm{d}v \propto v^{3}\exp\!\left(-\frac{mv^{2}}{2k_BT}\right)\mathrm{d}v \qquad (1)

where: $f(v)$ — flux velocity distribution in the beam (atoms per unit velocity); $v$ — atomic speed along the beam; $m$ — atomic mass; $T$ — oven temperature; $k_B$ — Boltzmann constant. The most-probable beam speed is $v_p=\sqrt{3k_BT/m}$ , larger than the gas value $\sqrt{2k_BT/m}$ — for sodium at $600\ \mathrm{K}$ this is about $800\ \mathrm{m\,s^{-1}}$ , setting the capture velocity the slower must accept. The force that decelerates that beam is the scattering force. It is the photon momentum $\hbar k$ multiplied by the rate at which photons are scattered, and that rate is a Lorentzian in the atom's detuning: it peaks when the moving atom sees the light exactly on resonance and collapses within a linewidth on either side. This narrow-band character is the source of every complication that follows.

F_{\mathrm{sc}}=\hbar k\,\frac{\Gamma}{2}\,\frac{s_0}{1+s_0+\left(2\delta/\Gamma\right)^{2}} \qquad (2)

where: $\hbar k$ — photon momentum per absorption; $\Gamma$ — natural linewidth (excited-state decay rate); $s_0=I/I_{\mathrm{sat}}$ — on-resonance saturation parameter; $\delta=\delta_L-kv$ — detuning the moving atom actually sees, with $\delta_L$ the laser detuning from the rest resonance and $kv$ the Doppler shift. Read physically, the force is a resonance: switch on the Doppler shift and it vanishes unless the laser detuning is tuned to cancel it. Push the intensity as high as you like and the force does not grow without bound. The excited-state population saturates at one-half, because stimulated emission returns photons as fast as absorption removes them, so the scattering rate cannot exceed $\Gamma/2$ and the deceleration hits a ceiling set only by the atom and the wavelength.

a_{\max}=\lim_{s_0\to\infty}\frac{F_{\mathrm{sc}}}{m}=\frac{\hbar k\,\Gamma}{2m}=\frac{v_r\Gamma}{2} \qquad (3)

where: $v_r=\hbar k/m$ — recoil velocity, the single-photon velocity kick; $\Gamma/2$ — maximum photon-scattering rate at saturation. For sodium ( $\lambda=589\ \mathrm{nm}$ , $\Gamma/2\pi\approx 9.8\ \mathrm{MHz}$ ) this evaluates to $a_{\max}\approx 9\times10^{5}\ \mathrm{m\,s^{-2}}$ , some $10^{5}$ times gravity, and no laser power can exceed this radiative-force limit. To keep the atom scattering at that ferocious rate all the way down, the total detuning must vanish at every instant. In the Zeeman slower this is arranged in space: a magnetic field $B(z)$ tunes the transition through the Zeeman effect so that its shift plus the fixed laser detuning exactly cancels the local Doppler shift $kv(z)$ .

\delta_L-kv+\frac{\mu' B(z)}{\hbar}=0 \qquad (4)

where: $kv$ — Doppler shift of the counter-propagating laser; $\mu' B(z)/\hbar$ — Zeeman shift of the cooling transition at position $z$ ; $\mu'=(g_e m_e-g_g m_g)\mu_B$ — effective magnetic moment of the transition; $\delta_L$ — fixed laser detuning from the zero-field resonance. The equation says the field must be strong where the atom is fast and weak where it is slow, tapering along the beam. Combine the resonance condition with constant-deceleration kinematics and both the velocity and the field acquire the same square-root shape. The velocity profile is fixed by energy conservation; the field profile is forced to mirror it so that the shrinking Doppler shift is cancelled point by point.

v(z)=v_0\sqrt{1-\frac{z}{L_0}},\qquad B(z)=B_{\mathrm{bias}}+B_0\sqrt{1-\frac{z}{L_0}} \qquad (5)

where: $v_0$ — capture velocity at the slower entrance; $L_0=v_0^2/2a$ — design slowing length for deceleration $a=\eta a_{\max}$ ; $B_0=\hbar k v_0/\mu'$ — field amplitude matching the full entrance Doppler shift (about $1213\ \mathrm{G}$ for $v_0=1000\ \mathrm{m\,s^{-1}}$ sodium); $B_{\mathrm{bias}}$ — constant offset that absorbs the fixed laser detuning. The square root is the fingerprint of constant deceleration. Chirp cooling reaches the same end by the other route: hold the field at zero and sweep the laser frequency in time so the detuning follows a chosen velocity class as it decelerates. The required sweep rate is just the rate at which that class's Doppler shift changes.

\frac{\mathrm{d}\delta_L}{\mathrm{d}t}=-k\,\frac{\mathrm{d}v}{\mathrm{d}t}=k\,a=k\,\eta a_{\max} \qquad (6)

where: $\mathrm{d}\delta_L/\mathrm{d}t$ — required laser-frequency chirp rate; $a=\eta a_{\max}$ — chosen deceleration with safety factor $\eta\lesssim0.7$ ; $ka$ — rate at which the Doppler shift falls as the atom slows. For sodium at $\eta=0.5$ this is roughly $0.8\ \mathrm{GHz}$ per millisecond, sweeping about $1.7\ \mathrm{GHz}$ across the whole slowing event. Finally, slowing does nothing for the transverse spread, so a separate two-dimensional molasses collimates the beam. The best transverse temperature is the Doppler limit, and the residual divergence is the ratio of the transverse thermal speed to the forward speed.

T_\perp\ge T_D=\frac{\hbar\Gamma}{2k_B},\qquad \theta_{\mathrm{div}}\simeq\frac{1}{v_\parallel}\sqrt{\frac{k_BT_\perp}{m}} \qquad (7)

where: $T_\perp$ — transverse temperature after collimation; $T_D$ — Doppler limit for the transverse stage ( $\approx 240\ \mu\mathrm{K}$ for Na, $146\ \mu\mathrm{K}$ for Rb); $\theta_{\mathrm{div}}$ — residual half-divergence; $v_\parallel$ — longitudinal velocity of the collimated beam. Because $\theta_{\mathrm{div}}\propto1/v_\parallel$ , a slow beam is intrinsically harder to keep collimated — which is why the transverse stage matters most exactly where the longitudinal stage has done its job.

Derivation AWhy the deceleration has a hard ceiling

The goal is to show that the deceleration a laser can exert on an atom is capped at a value set only by the wavelength and the natural linewidth, and that no amount of laser intensity or clever detuning can exceed it. This is the single most important number in slow-beam design because it fixes the minimum length of the whole apparatus, so we trace it from the two-level scattering rate down to the saturated force, and confirm that the ceiling is a statement about populations, not about power.

Step 1. The scattering force is the photon momentum times the scattering rate, $F_{\mathrm{sc}}=\hbar k\,R_{\mathrm{sc}}$ , and the rate is the decay rate times the excited-state population, $R_{\mathrm{sc}}=\Gamma\rho_{ee}$ , with $\rho_{ee}=\tfrac12 s_0/[1+s_0+(2\delta/\Gamma)^2]$ from the steady-state solution of the optical Bloch equations. Why: every net momentum kick comes from an absorption followed by spontaneous emission, so the force is literally counting fluorescence photons per second.

Step 2. Take the strong-drive limit $s_0\to\infty$ at resonance ( $\delta=0$ ): the fraction becomes $s_0/(1+s_0)\to1$ , so $\rho_{ee}\to\tfrac12$ . The excited state can never hold more than half the population, because once it is half full the stimulated emission rate equals the absorption rate and the two saturate against each other. Why: this is the physical origin of the ceiling — a two-level atom in bright light spends at most half its time excited, no matter how bright the light.

Step 3. Substitute the saturated population into the force: $a_{\max}=\hbar k\,\Gamma\rho_{ee}^{\max}/m=\hbar k\Gamma/2m=v_r\Gamma/2$ . For sodium, $v_r=\hbar k/m\approx 2.95\ \mathrm{cm\,s^{-1}}$ and $\Gamma=2\pi\times9.8\ \mathrm{MHz}$ give $a_{\max}\approx 9.1\times10^{5}\ \mathrm{m\,s^{-2}}$ . Why: the ceiling depends only on $k$ and $\Gamma$ , both fixed by the atom and the cooling transition, so it is a property of the element, not of the apparatus.

Step 4. Because running at $a_{\max}$ leaves no margin for field errors or intensity droop, real devices choose $a=\eta a_{\max}$ with $\eta\approx0.5$ ; the atom then decelerates safely below the ceiling and tolerates the imperfections that would otherwise let it slip out of resonance and coast. Why: the ceiling is a wall, not a target — designing against it, not at it, is what makes the slower robust.

Go deeper · Derivation B · 1 minThe slowing budget — length, time, and photon count

Derivation BThe slowing budget — length, time, and photon count

Given the deceleration ceiling, the goal here is to turn it into the three numbers a designer actually needs: how long the slower must be, how long an atom spends inside it, and how many photons it must scatter on the way through. We assume the resonance condition is held throughout — by a Zeeman field or a chirp — so the atom decelerates at a constant $a=\eta a_{\max}$ from an entrance velocity $v_0$ to a final velocity $v_f$ , and everything follows from constant-acceleration kinematics plus the fact that each photon removes exactly one recoil velocity.

Step 1. Constant deceleration gives $v_f^2=v_0^2-2aL$ , so the slowing length is $L=(v_0^2-v_f^2)/2a$ . Why: this is just work-energy for a constant force, and it shows the length scales as the square of the velocity to be removed — halving the capture velocity quarters the apparatus.

Step 2. From $v_f=v_0-at$ , the slowing time is $t_{\mathrm{slow}}=(v_0-v_f)/a$ . Why: the dwell time sets how far the atom drifts transversely while decelerating, and hence how much the beam spreads before the molasses can act on it.

Step 3. Each spontaneously scattered photon removes one recoil velocity on average, so the photon count is $N_{\mathrm{ph}}=\Delta v/v_r=m(v_0-v_f)/\hbar k$ . This number is set purely by the total velocity change — it is identical whether a Zeeman field or a frequency chirp keeps the atom resonant. Why: the scheme chooses how to stay on resonance, not how many photons the momentum change costs; that is fixed by conservation of momentum alone.

Step 4. Put in sodium numbers with $v_0=1000\ \mathrm{m\,s^{-1}}$ , $v_f\approx0$ , $v_r\approx0.0295\ \mathrm{m\,s^{-1}}$ , $a_{\max}\approx9.1\times10^{5}\ \mathrm{m\,s^{-2}}$ , and $\eta=0.5$ : the length is $L\approx1.1\ \mathrm{m}$ , the time is $t_{\mathrm{slow}}\approx2.2\ \mathrm{ms}$ , and the count is $N_{\mathrm{ph}}\approx3.4\times10^{4}$ . Why: these are the canonical design figures of a sodium Zeeman slower, and they show why the device is roughly a metre long and why the cooling transition must be genuinely closed to survive tens of thousands of cycles.

Go deeper · Derivation C · 1 minThe tapered magnetic field from constant deceleration

Derivation CThe tapered magnetic field from constant deceleration

The goal is to prove that maintaining resonance under constant deceleration forces the magnetic field to fall as a square root of position, and to read off its amplitude and length. The counter-propagating laser runs at a single fixed frequency, so the only free knob is the field: at every point along the beam the sum of the atom's Doppler shift and its field-induced Zeeman shift must equal the fixed laser detuning, and imposing that at all $z$ pins down the whole profile $B(z)$ .

Step 1. Write resonance at every point, $\delta_L-kv(z)+\mu'B(z)/\hbar=0$ , so that the fixed laser detuning, the shrinking Doppler shift, and the Zeeman shift cancel continuously. Why: this is the defining requirement of a Zeeman slower — the atom must never leave resonance, so the equation must hold as an identity in $z$ , not just at one spot.

Step 2. Insert the constant-deceleration velocity profile $v(z)=v_0\sqrt{1-z/L_0}$ with $L_0=v_0^2/2a$ , which is the kinematic result derived above rewritten in terms of position. Why: the velocity is not free — it is dictated by the chosen deceleration — so the field has no choice but to accommodate whatever shape the kinematics hands it.

Step 3. Solve for the field: $\mu'B(z)/\hbar=kv(z)-\delta_L=kv_0\sqrt{1-z/L_0}-\delta_L$ , giving $B(z)=B_{\mathrm{bias}}+B_0\sqrt{1-z/L_0}$ with $B_0=\hbar kv_0/\mu'$ and $B_{\mathrm{bias}}=-\hbar\delta_L/\mu'$ . For $v_0=1000\ \mathrm{m\,s^{-1}}$ sodium the Doppler shift is $kv_0/2\pi\approx1.7\ \mathrm{GHz}$ , requiring $B_0\approx1213\ \mathrm{G}$ . Why: the field inherits the exact square-root shape of the velocity, and the constant laser detuning becomes nothing more than a rigid bias offset added to that curve.

Step 4. A large negative $\delta_L$ turns the profile into a spin-flip slower, where $B$ crosses zero and reverses sign; this halves the peak field the magnet must produce and moves the exit resonance well away from the trap region, so slow atoms are not re-accelerated as they leave. Why: the freedom in $\delta_L$ is a real engineering lever, trading peak field against where the beam decouples from the light.

Limiting cases

The performance envelope of any slow-beam apparatus is fenced in by a handful of clean limits:

Case	Result	Meaning
Maximum deceleration	$a_{\max}=\dfrac{\hbar k\,\Gamma}{2m}$	Set by the saturated scattering force ( $\rho_{ee}\to1/2$ ); $\approx9\times10^{5}\ \mathrm{m\,s^{-2}}$ for Na, independent of laser power once saturated.
Minimum slowing length	$L_{\min}=\dfrac{v_0^2-v_f^2}{2a_{\max}}$	$\approx0.55\ \mathrm{m}$ for Na from $1000\ \mathrm{m\,s^{-1}}$ to rest; real slowers run at $\eta a_{\max}$ and are longer ( $\sim1.1\ \mathrm{m}$ at $\eta=0.5$ ).
Doppler temperature (transverse)	$k_BT_D=\dfrac{\hbar\Gamma}{2}$	Coldest a Doppler molasses reaches; $T_D\approx240\ \mu\mathrm{K}$ (Na), $146\ \mu\mathrm{K}$ (Rb). Sub-Doppler cooling goes lower still.
Recoil velocity	$v_r=\dfrac{\hbar k}{m}$	Single-photon velocity granularity, $\approx3\ \mathrm{cm\,s^{-1}}$ for Na; sets the ultimate longitudinal resolution of any radiative slowing.
Weak drive ( $s_0\ll1$ )	$F_{\mathrm{sc}}\to\hbar k\dfrac{\Gamma}{2}s_0$	Force linear in intensity, far below the ceiling; wasteful for slowing, so slowers run near saturation.

Common trapDesigning a slower to run right at the deceleration ceiling

The maximum deceleration $a_{\max}=\hbar k\Gamma/2m$ is a hard wall, not a set-point. Because the excited-state population saturates at one-half, no detuning and no intensity can push the deceleration past it — throwing more laser power at the problem buys nothing once the transition is saturated. The tempting mistake is to design the slower for $a=a_{\max}$ to make it as short as possible, but an atom decelerating at the very ceiling has zero margin: the slightest dip in field or intensity drops the real deceleration below the design value, the atom fails to slow fast enough, its Doppler shift no longer matches the local field, and it slides out of resonance and coasts — lost. This is why every practical device runs at $a=\eta a_{\max}$ with $\eta\approx0.5$ , deliberately leaving headroom so that field and intensity imperfections cannot push a resonant atom off the trajectory.

Worked examples

1 · The full slowing budget for a sodium atom (confidence builder)

A sodium oven produces a beam with entrance velocity $v_0=1000\ \mathrm{m\,s^{-1}}$ . Using the D2 line ( $\lambda=589\ \mathrm{nm}$ , $\Gamma/2\pi=9.8\ \mathrm{MHz}$ , $m=3.82\times10^{-26}\ \mathrm{kg}$ ), find the deceleration ceiling, the minimum slowing length, the slowing time, and the number of photons needed to stop the atom. Then repeat the length at the practical value $\eta=0.5$ .

Step 1. Photon momentum and recoil velocity. The wavenumber is $k=2\pi/\lambda=1.067\times10^{7}\ \mathrm{m^{-1}}$ , so $\hbar k=1.13\times10^{-27}\ \mathrm{kg\,m\,s^{-1}}$ and $v_r=\hbar k/m=1.13\times10^{-27}/3.82\times10^{-26}=\mathbf{2.95\times10^{-2}\ m\,s^{-1}}$ , about three centimetres per second per photon.

Step 2. Deceleration ceiling. With $\Gamma=2\pi\times9.8\times10^{6}=6.15\times10^{7}\ \mathrm{s^{-1}}$ , the maximum deceleration is $a_{\max}=v_r\Gamma/2=(2.95\times10^{-2})(3.08\times10^{7})=\mathbf{9.1\times10^{5}\ m\,s^{-2}}$ , roughly $9\times10^{4}$ times $g$ .

Step 3. Minimum length and stopping time at the ceiling. $L_{\min}=v_0^2/2a_{\max}=(10^{6})/(1.81\times10^{6})=\mathbf{0.55\ m}$ , and $t=v_0/a_{\max}=1000/9.1\times10^{5}=1.1\ \mathrm{ms}$ . At the practical $a=0.5\,a_{\max}$ the length doubles to $L=\mathbf{1.1\ m}$ and the time to $t_{\mathrm{slow}}=\mathbf{2.2\ ms}$ .

Step 4. Photon count. $N_{\mathrm{ph}}=v_0/v_r=1000/2.95\times10^{-2}=\mathbf{3.4\times10^{4}}$ photons — independent of $\eta$ or of whether a Zeeman field or a chirp holds resonance. So a sodium atom is stopped in about a metre and two milliseconds by scattering some 34,000 photons, and the deceleration cannot exceed $9\times10^{5}\ \mathrm{m\,s^{-2}}$ .

2 · The peak field of the Zeeman slower

For the same $v_0=1000\ \mathrm{m\,s^{-1}}$ sodium beam, find the entrance Doppler shift and the field amplitude $B_0$ needed to cancel it, taking the effective moment $\mu'\approx\mu_B=9.27\times10^{-24}\ \mathrm{J\,T^{-1}}$ .

Step 1. Entrance Doppler shift. $kv_0=(1.067\times10^{7})(1000)=1.07\times10^{10}\ \mathrm{rad\,s^{-1}}$ , i.e. $kv_0/2\pi=\mathbf{1.70\ GHz}$ — roughly $170$ natural linewidths, which is why the atom cannot stay resonant on its own.

Step 2. Field to match. Setting the Zeeman shift equal to the Doppler shift, $\mu'B_0/\hbar=kv_0$ , gives $B_0=\hbar kv_0/\mu'=(1.13\times10^{-27}\times1000)/(9.27\times10^{-24})=0.121\ \mathrm{T}=\mathbf{1213\ G}$ .

Step 3. Interpretation. A field of about $1200\,{\rm G}$ at the entrance, tapering as $\sqrt{1-z/L_0}$ toward the exit, keeps the shrinking Doppler shift cancelled the whole way. A spin-flip design lets $B$ cross zero, so the coil need only produce about half this peak on each side. The slower must supply roughly $1.2\ \mathrm{kG}$ at its entrance to cancel a $1.7\ \mathrm{GHz}$ Doppler shift.

3 · Residual divergence after transverse cooling

A collimated sodium beam leaves the transverse molasses at the Doppler limit $T_\perp=T_D=240\ \mu\mathrm{K}$ with a forward velocity $v_\parallel=30\ \mathrm{m\,s^{-1}}$ . Estimate its half-divergence.

Step 1. Transverse thermal speed. $\sqrt{k_BT_\perp/m}=\sqrt{(1.38\times10^{-23})(2.4\times10^{-4})/(3.82\times10^{-26})}=\sqrt{8.67\times10^{-2}}\approx\mathbf{0.29\ m\,s^{-1}}$ .

Step 2. Divergence. $\theta_{\mathrm{div}}\simeq0.29/30=\mathbf{9.7\ mrad}\approx0.56^{\circ}$ . Without the molasses the same beam, with a transverse spread of order the longitudinal one, would diverge by tens of degrees and never reach the trap.

Step 3. The $1/v_\parallel$ scaling bites here: had the beam been slowed further to $v_\parallel=10\ \mathrm{m\,s^{-1}}$ , the same transverse temperature would give $\theta_{\mathrm{div}}\approx29\ \mathrm{mrad}$ . Transverse cooling turns a beam that would spray across the chamber into one that stays inside a $\sim10\ \mathrm{mrad}$ cone — the collimation gain that dominates trap loading.

Physics of the portrait

The slow beam is the invention that turned laser cooling from a proposal into a laboratory. Its story runs through one physicist more than any other, who built the first device that stopped an atomic beam and shared a Nobel Prize for the field it opened.

Origins · 1 minWilliam D. Phillips and how this was discovered

William D. Phillips

b. 1948 · laser cooling and trapping of neutral atoms

William Daniel Phillips built the machine this article is about. In the early 1980s, working at the US National Bureau of Standards (now NIST), he and Harold Metcalf recognized that the fatal flaw in decelerating an atomic beam with light — the atom's Doppler shift sliding it out of resonance as it slows — could be defeated with a spatially tapered magnetic field, tuning the atomic transition through the Zeeman effect to track the falling Doppler shift point by point. The resulting Zeeman slower, demonstrated in 1982, was the first apparatus to bring a beam of neutral atoms from oven speeds down toward rest, and it remains a standard front end in cold-atom laboratories worldwide. Phillips went further, showing in the late 1980s that optical molasses cooled sodium far below the Doppler limit that theory then predicted — a discrepancy whose resolution, Sisyphus and polarization-gradient cooling, was worked out by Claude Cohen-Tannoudji and Jean Dalibard. For turning the pressure of light into a practical tool for slowing, cooling, and trapping atoms, Phillips shared the 1997 Nobel Prize in Physics with Steven Chu and Cohen-Tannoudji. His slow beams are the unglamorous but indispensable first stage of nearly every experiment that has since produced a Bose–Einstein condensate, an optical-lattice clock, or an atom interferometer.

History

1933

— Frisch deflects a sodium beam with resonant light, demonstrating single-photon recoil

1975

— Hänsch and Schawlow propose Doppler laser cooling of neutral atoms

1982

— Phillips and Metcalf decelerate a sodium beam with a tapered-field Zeeman slower

1985

— chirp-cooling deceleration is demonstrated and optical molasses is realized at Bell Labs

1997

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

Going further

Texts. Foot, Atomic Physics, ch. 9 (Laser Cooling and Trapping) · Metcalf and van der Straten, Laser Cooling and Trapping (Springer, 1999) · Demtröder, Laser Spectroscopy, Vol. 2, ch. on optical cooling · Loudon, The Quantum Theory of Light, 3rd ed., ch. 1 (radiative processes) · Phillips, Rev. Mod. Phys. 70, 721 (1998), Nobel lecture on laser cooling. The key numbers used throughout this article are: the deceleration ceiling $a_{\max}=\hbar k\Gamma/2m$ with its saturation origin ( $\rho_{ee}\to1/2$ ); the sodium values $v_r\approx3\ \mathrm{cm\,s^{-1}}$ , $a_{\max}\approx9\times10^{5}\ \mathrm{m\,s^{-2}}$ , $L\approx1.1\ \mathrm{m}$ at $\eta=0.5$ , $t_{\mathrm{slow}}\approx2.2\ \mathrm{ms}$ , and $N_{\mathrm{ph}}\approx3.4\times10^{4}$ (using $\lambda=589\ \mathrm{nm}$ , $\Gamma/2\pi\approx9.8\ \mathrm{MHz}$ ); the $\sqrt{1-z/L_0}$ velocity and field profiles with the resonance condition; the peak field $B_0\approx1213\ \mathrm{G}$ from $kv_0/2\pi\approx1.7\ \mathrm{GHz}$ with $\mu'\approx\mu_B$ ; the Doppler limits $T_D\approx240\ \mu\mathrm{K}$ (Na) and $146\ \mu\mathrm{K}$ (Rb) from $k_BT_D=\hbar\Gamma/2$ ; and the $v^3$ effusive flux law. No non-standard numbers are introduced.

Prerequisites. Radiation Pressure and the Scattering Force · Doppler Cooling and the Doppler Limit · Optical Molasses. Leads to. The Zeeman Slower · The Magneto-Optical Trap · Evaporative Cooling · Bose–Einstein Condensation in Dilute Gases.

Ask the tutor: Ask me to walk through why a sodium atom needs about 34,000 photons and a metre of slower to stop, or how the Zeeman field is tapered to keep it on resonance the whole way down.

Why does a thermal atomic beam need slowing at all — isn't an oven beam already collimated?

Geometric collimation and velocity are different things. A pinhole or a set of apertures narrows the directions atoms travel, giving a well-aimed beam, but it does nothing to the speed: the atoms still stream out at the oven's most-probable beam speed, around $800\ \mathrm{m\,s^{-1}}$ for sodium at $600\ \mathrm{K}$ . A magneto-optical trap can only capture atoms below its capture velocity, a few tens of $\mathrm{m\,s^{-1}}$ , and the fraction of a thermal beam already that slow is vanishingly small. Slowing is what shifts a usable population of atoms below the capture velocity; collimation and slowing are complementary, not substitutes.

What is the difference between a Zeeman slower and chirp cooling?

Both defeat the same enemy — the Doppler shift that slides a decelerating atom out of resonance — but they do it in different variables. The Zeeman slower cancels the shift in space: a magnetic field tapered along the beam tunes the atomic transition so that a single fixed-frequency laser stays resonant at every position, and the device produces a continuous slow beam. Chirp cooling cancels the shift in time: the field is zero and the laser frequency is swept to follow a chosen velocity class as it slows, which naturally produces pulses of slow atoms rather than a steady stream. The Zeeman slower gives continuous flux at the cost of a bulky tapered magnet; chirp cooling needs only a fast frequency ramp but delivers bunches.

Why can't more laser power make the atoms decelerate faster?

Because the force is limited by how often the atom can scatter, not by how much light is available. The scattering force is $\hbar k$ times the scattering rate $\Gamma\rho_{ee}$ , and $\rho_{ee}$ , the excited-state population, cannot exceed one-half. Once the transition is saturated, adding intensity just drives stimulated emission, which returns a photon in the same direction it came from and produces no net kick. The scattering rate therefore tops out at $\Gamma/2$ and the deceleration at $a_{\max}=\hbar k\Gamma/2m$ . Wanting a shorter slower means wanting a larger $a_{\max}$ , and the only way to get it is a different transition — a larger $\Gamma$ or a shorter wavelength — not more power.

Is the number of photons really independent of the slowing scheme?

Yes, and it follows straight from momentum conservation. To change an atom's velocity by $\Delta v$ you must remove momentum $m\,\Delta v$ , and each spontaneously scattered photon removes, on average, exactly one recoil momentum $\hbar k$ . So the photon count is $N_{\mathrm{ph}}=m\,\Delta v/\hbar k=\Delta v/v_r$ regardless of whether a Zeeman field or a frequency chirp keeps the atom resonant. The scheme only decides how the atom stays on resonance while it scatters those photons — it cannot change the number. For sodium from $1000\ \mathrm{m\,s^{-1}}$ this is about $34{,}000$ photons, which is also why the cooling transition must be closed: an atom that leaks to a dark state after a few thousand cycles never finishes slowing.

What sets the final velocity and temperature of the beam?

Two different things, longitudinal and transverse. The final forward velocity is chosen, not thermal: the resonance-tracking scheme decides where to stop pushing, so a Zeeman slower can be designed to deliver, say, $20\ \mathrm{m\,s^{-1}}$ by ending its taper at the corresponding field. The final transverse temperature is genuinely thermal and is set by the collimation stage, bottoming out at the Doppler limit $T_D=\hbar\Gamma/2k_B$ — about $240\ \mu\mathrm{K}$ for sodium — unless sub-Doppler polarization-gradient cooling is used to push lower. So "how cold is the beam" has two answers: a chosen narrow longitudinal velocity slice, and a Doppler-limited transverse spread.

Why is a slower beam harder to keep collimated?

Because divergence is a ratio of speeds. The half-divergence is roughly the transverse thermal speed divided by the forward speed, $\theta_{\mathrm{div}}\simeq\sqrt{k_BT_\perp/m}\,/\,v_\parallel$ . The transverse thermal speed is fixed by the collimation temperature and does not care how slow the beam is, but the forward speed appears in the denominator — so halving $v_\parallel$ doubles the divergence for the same transverse temperature. This is the awkward trade at the heart of slow-beam design: the slower you make the beam so a trap can catch it, the faster it fans out, which is exactly why the transverse molasses is most critical at the low-velocity end and why very slow beams demand aggressive collimation right before the trap.