Zeeman Slower

A Zeeman slower takes one idea from the article before it — the radiation-pressure scattering force, the net push a two-level atom feels when it repeatedly absorbs photons from one direction and re-emits them into random ones — and solves the single problem that force cannot solve on its own: it is fiercely velocity-selective. An atom streaming out of a hot oven at several hundred metres per second Doppler-shifts a counter-propagating laser far off resonance; it scatters efficiently only inside a narrow velocity window a few m/s wide, of order $\Gamma/k$ . Begin to slow such an atom and it immediately red-shifts itself out of resonance, the force switches off, and deceleration stalls after a handful of photons. The device invented by Phillips and Metcalf in 1982 defeats this with a trick of exquisite economy: a magnetic field that is tapered along the beam axis shifts the atomic transition, through the Zeeman effect, by exactly the amount needed to cancel the shrinking Doppler shift. A chosen velocity class therefore stays glued to resonance as it decelerates, and a single fixed-frequency laser slows atoms all the way down the tube. The whole design collapses to one bookkeeping equation — laser detuning plus Doppler shift plus Zeeman shift equals zero at every point — and demanding constant deceleration turns it into the slower's signature square-root field profile. The deceleration itself is enormous, of order $10^{5}$ to $10^{6}\,\mathrm{m/s^2}$ , enough to stop a sodium atom in about a metre after roughly thirty thousand photon kicks delivered in a couple of milliseconds. Zeeman slowers are the unglamorous first stage of most cold-atom machines: they hand a magneto-optical trap a bright, continuous flux of atoms slow enough to catch, and from there optical molasses, magnetic trapping, and evaporative cooling reach Bose-Einstein condensation and degenerate Fermi gases.

Key takeaways

A Zeeman slower decelerates a thermal atomic beam with radiation pressure from one counter-propagating laser, the scattering force of resonant absorption followed by isotropic spontaneous emission.
As the atom slows its Doppler shift $kv$ shrinks; a spatially tapered magnetic field tunes the atomic transition by the Zeeman effect so Doppler and Zeeman shifts cancel and the atom stays on resonance the whole way.
Setting the total detuning to zero, $\delta_L-kv(z)+\mu'B(z)/\hbar=0$ , and demanding constant deceleration fixes the field taper $B(z)=B_{\mathrm{bias}}+B_0\sqrt{1-z/z_0}$ .
The deceleration ceiling is $a_{\max}=\hbar k\Gamma/2m\approx9\times10^{5}\,\mathrm{m/s^2}$ for sodium, slowing atoms from about $1000\,\mathrm{m/s}$ to tens of m/s over roughly one metre.
Real slowers run below the ceiling at a safety factor $\eta\sim0.5$ so a bunch of atoms stays trapped in the moving resonance; the slowed beam feeds a magneto-optical trap, the gateway to BEC and quantum-gas experiments.
Key idea: deceleration continuously red-shifts an atom's Doppler-shifted resonance, and a spatially tapered Zeeman field shifts the atomic transition by exactly the compensating amount, so a whole velocity class stays resonant with a single fixed-frequency laser and decelerates smoothly over the length of the slower.

The core insight

Deceleration continuously red-shifts an atom's Doppler-shifted resonance; a spatially tapered Zeeman field shifts the atomic transition by exactly the compensating amount, so a whole velocity class stays resonant with a single fixed-frequency laser and decelerates smoothly over the length of the slower.

Where this lives today

Walk up to almost any alkali or alkaline-earth quantum-gas apparatus and the first thing the atoms meet after leaving the oven is a Zeeman slower: a metre-long tapered solenoid, often wound in a spin-flip (zero-crossing) geometry that splits the field into positive and negative sections so the peak field magnitude is halved and the exit field is nearly zero, keeping stray field out of the downstream trap and cutting coil power. It delivers a continuous flux of $10^{9}$ to $10^{11}$ atoms per second slow enough for a magneto-optical trap to catch, which is why it remains the standard front end for machines that go on to Bose-Einstein condensation, degenerate Fermi gases, and optical-lattice quantum simulators. The design also stretches to species the alkalis never demanded. Strontium and ytterbium optical-lattice atomic clocks — today's most accurate timekeepers, good to parts in $10^{18}$ — are loaded through Zeeman slowers on the broad singlet line before a narrow intercombination-line stage takes over, and highly magnetic lanthanides such as erbium and dysprosium, used to make dipolar quantum gases, are slowed the same way. Where length is at a premium some groups swap the long solenoid for permanent-magnet arrays or a compact two-dimensional magneto-optical trap, but the underlying idea — shift the atom's resonance in space to chase its falling Doppler shift — is Phillips and Metcalf's, unchanged since 1982.

The physical picture

Imagine trying to stop a rolling marble with gusts from a fan that only pushes hard when the marble moves at one particular speed. Blow at it head-on and the first gust lands squarely — but the very act of pushing changes the marble's speed, the next gust arrives mistuned, and the pushing dies. The cure is to retune the fan continuously so its resonant speed tracks the marble down to rest. In a Zeeman slower the atom is the marble, the counter-propagating laser is the fan, and the tapered magnetic field is the retuner: it slides the atom's internal resonant speed downward in exact lockstep with the real deceleration, so the push never lets go. The push is granular. Each absorbed photon delivers one recoil kick of $\hbar k/m$ , about $3\,\mathrm{cm/s}$ for sodium, and the atom must absorb tens of thousands of them. Because spontaneous emission is isotropic, those recoils average to a clean force along the beam while their transverse fluctuations slowly heat and spread the beam sideways — the reason real machines add transverse cooling. To bleed $1000\,\mathrm{m/s}$ off a sodium atom takes roughly $v_0m/\hbar k\approx3\times10^{4}$ absorption-emission cycles, and at the maximum scattering rate of half a linewidth that whole ballet finishes in about a millisecond.

Figure

Anatomy of a Zeeman slower: a sodium beam leaves the oven near 1000 m/s and is decelerated by a counter-propagating 589 nm laser as it crosses a magnetic field whose tapered part shrinks as √(1 − z/z₀), exiting near 30 m/s to load a magneto-optical trap

The square-root shape is not a designer's whim; it is forced by kinematics. Constant deceleration makes the kinetic energy fall linearly with position, so $v^2$ is linear in $z$ and $v\propto\sqrt{z_0-z}$ . Since the Zeeman shift must track the Doppler shift $kv$ to hold resonance, the field inherits the identical square-root taper. The subtle payoff is phase stability: because the design runs below the maximum force, an atom that is momentarily too fast for its position sits on the steep flank of its resonance, scatters harder than the ideal atom, and is pulled back onto the trajectory — the moving resonance drags a whole bunch of atoms along with it, exactly as a synchrotron bucket bunches particles.

The core equations

The engine is the radiation-pressure scattering force, the momentum $\hbar k$ delivered per photon times the rate at which photons are scattered. For a two-level atom driven by a single beam, that rate is a Lorentzian in the effective detuning, saturating as the intensity climbs. This is the quantity every other equation either maximizes or holds constant.

F_{\mathrm{sc}} = \hbar k\, R_{\mathrm{sc}} = \hbar k\,\frac{\Gamma}{2}\,\frac{s_0}{1 + s_0 + \left(\dfrac{2\delta_{\mathrm{eff}}}{\Gamma}\right)^2} \qquad (1)

where: $\hbar k$ — photon momentum, with $k=2\pi/\lambda$ ; $\Gamma$ — natural linewidth, the excited-state decay rate; $s_0=I/I_{\mathrm{sat}}$ — on-resonance saturation parameter; $\delta_{\mathrm{eff}}$ — effective detuning, gathering the laser, Doppler, and Zeeman contributions. The force peaks when $\delta_{\mathrm{eff}}=0$ and collapses once the atom is more than a linewidth off resonance. Slowing works only if $\delta_{\mathrm{eff}}$ can be held near zero as $v$ changes. In the atom's frame a counter-propagating beam is Doppler-shifted up by $kv$ , while the field lowers the transition frequency through the Zeeman effect. Setting the sum to zero is the master design condition of the whole device.

\delta_{\mathrm{eff}} = \delta_L - k v(z) + \frac{\mu' B(z)}{\hbar} = 0 \qquad (2)

where: $\delta_L=\omega_L-\omega_0$ — laser detuning from the zero-field transition; $-kv$ — Doppler shift for the counter-propagating beam; $\mu'=(g_e m_e - g_g m_g)\mu_B$ — differential magnetic moment of the cooling transition (near $\mu_B$ for the sodium $\sigma^+$ line); $B(z)$ — axial field. Read it as a running balance: whatever Doppler shift the shrinking velocity gives up, the field must give back through $\mu'B/\hbar$ . The force cannot grow without bound, because a two-level atom can hold at most half its population in the excited state. As $s_0\to\infty$ the scattering rate saturates at $\Gamma/2$ , and the force divided by mass fixes the largest deceleration the method can ever reach.

a_{\max} = \frac{F_{\max}}{m} = \frac{\hbar k \Gamma}{2m} \qquad (3)

where: $m$ — atomic mass; $F_{\max}=\hbar k\Gamma/2$ — saturated force. For sodium ( $\lambda=589\,\mathrm{nm}$ , $\Gamma/2\pi=9.79\,\mathrm{MHz}$ ) this is $a_{\max}\approx9.1\times10^{5}\,\mathrm{m/s^2}$ , nearly $10^{5}$ times gravity. The ceiling scales with $\Gamma$ , so broad transitions slow atoms hardest. Choosing a constant design deceleration $a=\eta a_{\max}$ makes the kinematics trivial: $v^2$ falls linearly with $z$ , giving the square-root velocity profile that defines the ideal trajectory.

v(z) = v_0\sqrt{1 - \frac{z}{z_0}},\qquad z_0=\frac{v_0^2}{2a}=\frac{m v_0^2}{\eta\,\hbar k\Gamma} \qquad (4)

where: $v_0$ — capture velocity at the entrance; $z_0$ — stopping length at the design deceleration; $\eta$ — safety factor. Every atom entering at or below $v_0$ is decelerated; faster atoms never lock onto the resonance and coast through unslowed. Substituting that profile back into the resonance condition (2) gives the field the slower must produce — a fixed bias plus a square-root taper whose amplitude compensates the full initial Doppler shift.

B(z) = B_{\mathrm{bias}} + B_0\sqrt{1 - \frac{z}{z_0}},\qquad B_0 = \frac{\hbar k v_0}{\mu'},\qquad B_{\mathrm{bias}}=-\frac{\hbar\delta_L}{\mu'} \qquad (5)

where: $B_0$ — tapered amplitude cancelling the entrance Doppler shift; $B_{\mathrm{bias}}$ — constant offset set by the laser detuning. For sodium at $v_0=1000\,\mathrm{m/s}$ , $B_0\approx0.12\,\mathrm{T}=1213\,\mathrm{G}$ : the field must sweep more than a kilogauss along the tube. Finally, the atomic cost. Each cycle removes one photon momentum, so the number of scattered photons is the initial momentum in units of $\hbar k$ , and the run must proceed on a closed cycling transition that survives that many round trips.

N_{\mathrm{ph}} \approx \frac{m v_0}{\hbar k},\qquad t_{\mathrm{slow}}\approx\frac{v_0}{a}=\frac{N_{\mathrm{ph}}}{\eta\,\Gamma/2} \qquad (6)

where: $N_{\mathrm{ph}}$ — absorption-emission cycles; $t_{\mathrm{slow}}$ — slowing time. For sodium at $1000\,\mathrm{m/s}$ this is about $3.4\times10^{4}$ photons in roughly $2\,\mathrm{ms}$ — a demand that forces a repump laser to plug any leak out of the cycling transition.

Derivation AResonance tracking and the square-root field profile

The goal is to show, from the scattering force alone, why a magnetic field shaped like a square root keeps a decelerating atom pinned to resonance with one fixed-frequency laser, and to pin down the amplitude of that field in real units. We work in the atom's rest frame, write the three frequency shifts that determine whether the laser is resonant, demand that their sum stay zero along the whole trajectory, and then feed in the velocity profile that constant deceleration produces. The taper falls out algebraically, and evaluating its amplitude for sodium reveals just how strong a field the tube must carry.

Step 1. In the atom's frame the counter-propagating laser (wavevector $-k\hat z$ , atom moving $+v\hat z$ ) appears blue-shifted by $+kv$ , so the effective detuning is $\delta_{\mathrm{eff}}=\delta_L-kv(z)+\mu'B(z)/\hbar$ , and resonant slowing requires $\delta_{\mathrm{eff}}=0$ at every $z$ . Why: the force in equation (1) is sharply peaked at zero detuning, so keeping the atom useful over the whole tube means holding this sum at zero, not merely at the entrance.

Step 2. Impose constant deceleration $a$ . Then $v^2(z)=v_0^2-2az$ , so $v(z)=v_0\sqrt{1-z/z_0}$ with $z_0=v_0^2/2a$ , and the atom reaches $v=0$ exactly at $z=z_0$ . Why: a uniform force integrates to a velocity-squared that is linear in position, which is the mathematical origin of the square root — nothing about magnetism yet, only kinematics.

Step 3. Solve the resonance condition for the field, $\mu'B(z)/\hbar=kv(z)-\delta_L$ , and substitute the profile to get $B(z)=\dfrac{\hbar k v_0}{\mu'}\sqrt{1-z/z_0}-\dfrac{\hbar\delta_L}{\mu'}$ . Why: the Zeeman shift must supply, point by point, exactly the Doppler shift the slowing atom is losing, up to the fixed laser offset — so the field copies the velocity's square-root shape.

Step 4. Read off the amplitude $B_0=\hbar k v_0/\mu'$ and the bias $B_{\mathrm{bias}}=-\hbar\delta_L/\mu'$ . For sodium with $\mu'\approx\mu_B$ , $k=2\pi/589\,\mathrm{nm}$ , and $v_0=1000\,\mathrm{m/s}$ , this gives $B_0\approx0.121\,\mathrm{T}=1213\,\mathrm{G}$ . Why: putting numbers in shows the design is demanding but achievable — a kilogauss-scale field swept smoothly over a metre, the reason slowers are wound as long tapered solenoids. *As a check: $B_0=\hbar k v_0/\mu_B=1213\,\mathrm{G}$ for sodium.*

Go deeper · Derivation B · 1 minThe maximum deceleration and the minimum slower length

Derivation BThe maximum deceleration and the minimum slower length

The aim here is to turn the saturation of the scattering force into a hard bound on how short a slower can be, so a designer knows the trade before winding a single coil. A two-level atom cannot be more than half in the excited state under incoherent driving, which caps the scattering rate; that cap caps the force, the force caps the deceleration, and the deceleration sets the stopping distance. We push each cap through to the next and evaluate the final length for sodium, then explain why practical machines deliberately run longer than the bound.

Step 1. Take the scattering rate to its ceiling: as $s_0\to\infty$ the Lorentzian denominator is dominated by $s_0$ and $R_{\mathrm{sc}}\to\Gamma/2$ . Why: emission recycles the atom to the ground state, and stimulated processes cannot invert a two-level population, so half a linewidth is the fastest an atom can scatter no matter how bright the beam.

Step 2. Convert the saturated force to deceleration, $a_{\max}=F_{\max}/m=\hbar k\Gamma/2m$ . For sodium this is $9.06\times10^{5}\,\mathrm{m/s^2}$ . Why: this single number, fixed by $\lambda$ , $\Gamma$ , and $m$ , is the atom's answer to how hard light can push it, and it governs every length estimate that follows.

Step 3. Feed $a_{\max}$ into the kinematic stopping length $z_0=v_0^2/2a$ to get the floor $z_{0,\min}=v_0^2/2a_{\max}=m v_0^2/(\hbar k\Gamma)$ . For sodium at $v_0=1000\,\mathrm{m/s}$ this is about $0.55\,\mathrm{m}$ . Why: no Zeeman slower for this species and capture velocity can be shorter than this, since it already assumes the atom scatters at the theoretical maximum the whole way.

Step 4. Restore the safety factor: real designs use $a=\eta a_{\max}$ with $\eta\sim0.5$ , so $z_0=z_{0,\min}/\eta\approx1.1\,\mathrm{m}$ . Why: running at the ceiling leaves no restoring margin, so the length is knowingly traded for phase stability — the device is made longer precisely so it can actually catch atoms. *As a check: $a_{\max}=\hbar k\Gamma/2m=9.06\times10^{5}\,\mathrm{m/s^2}$ and $z_{0,\min}=0.55\,\mathrm{m}$ for sodium.*

Go deeper · Derivation C · 1 minPhoton budget, slowing time, and the cycling transition

Derivation CPhoton budget, slowing time, and the cycling transition

This derivation counts the raw currency of laser slowing — photons — and turns the count into a time and a hard engineering constraint. Momentum conservation says each absorbed photon removes exactly $\hbar k$ from the atom along the beam, so stopping the atom is simply a matter of how many such kicks its initial momentum is worth. Dividing that count by the fastest sustainable scattering rate gives the slowing time, and comparing the count to the number of photons a transition can scatter before an atom leaks into a dark state explains why every slower carries a repump laser.

Step 1. Equate momentum removed to momentum carried: $N_{\mathrm{ph}}\hbar k = m v_0$ , so $N_{\mathrm{ph}}=m v_0/\hbar k$ . For sodium at $1000\,\mathrm{m/s}$ , with recoil velocity $\hbar k/m\approx2.95\,\mathrm{cm/s}$ , this is $N_{\mathrm{ph}}\approx3.4\times10^{4}$ . Why: the transverse recoils of the isotropic emission average to zero along the axis, so only the directed absorption kicks accumulate — the emission is what makes it a net force at all.

Step 2. Divide by the scattering rate to get the time. At the design rate $\eta\Gamma/2$ with $\eta=0.5$ (so $1.5\times10^{7}\,\mathrm{s^{-1}}$ ), $t_{\mathrm{slow}}=N_{\mathrm{ph}}/(\eta\Gamma/2)\approx2.2\,\mathrm{ms}$ , matching $v_0/a$ directly. Why: the two routes to the time — counting photons and integrating the deceleration — must agree, and their agreement is a clean check on the whole picture.

Step 3. Demand a closed cycle. Scattering $3.4\times10^{4}$ photons is impossible if the atom can leak into an uncoupled hyperfine state after a few thousand cycles, so the cooling transition must be nearly closed and a repump laser returns any escapees. Why: the photon budget is not just a number, it is the reason a simple one-laser slower almost always needs a second beam — a leak of one in $10^{4}$ per cycle would empty the beam before it stopped. *As a check: $N_{\mathrm{ph}}=m v_0/\hbar k=3.4\times10^{4}$ and $t_{\mathrm{slow}}=2.2\,\mathrm{ms}$ for sodium at $v_0=1000\,\mathrm{m/s}$ , $\eta=0.5$ ; recoil velocity $\hbar k/m=2.95\,\mathrm{cm/s}$ .*

Limiting cases

Case	Result	Meaning
Saturated drive	$a_{\max}=\dfrac{\hbar k\Gamma}{2m}$	High intensity $s_0\gg1$ : the scattering rate saturates at $\Gamma/2$ , giving the hardest possible push and the shortest possible slower.
Weak drive	$R_{\mathrm{sc}}\approx\dfrac{\Gamma}{2}\,\dfrac{s_0}{1+(2\delta_{\mathrm{eff}}/\Gamma)^2}$	Low intensity $s_0\ll1$ : force is linear in intensity with a Lorentzian velocity dependence, so the slower must be made longer.
Off resonance	$\delta_{\mathrm{eff}}\gg\Gamma$	If the field mistunes the atom by more than a linewidth the force becomes negligible and the atom coasts at fixed velocity, dropping out of the deceleration.
Narrow-line species	$a_{\max}\propto\Gamma$	Strontium, ytterbium, and lanthanides have small $\Gamma$ on their intercombination lines, so slowing is done on their broad line; length grows as the linewidth shrinks.
Design length	$z_0=\dfrac{v_0^2}{2\eta a_{\max}}=\dfrac{m v_0^2}{\eta\,\hbar k\Gamma}$	Total length for capture velocity $v_0$ and safety factor $\eta<1$ ; higher $v_0$ or lower $\eta$ both stretch the device.

Common trapThe slower cools the beam the way a scoop cools soup

It is tempting to call a Zeeman slower a cooler, but at the level of the whole beam it mostly decelerates and compresses a single velocity class in the lab frame — it sweeps atoms down toward low velocity rather than shrinking the spread of an equilibrium distribution. True cooling, a narrowing of the momentum spread, is a secondary effect that sharpens near the exit; the primary job is bulk deceleration. Two practical corollaries follow. First, an atom pushed all the way to $v=0$ inside the tube will be blown backward or sideways by the still-resonant beam, so the design must leave a deliberate nonzero exit velocity. Second, the near-resonant slowing laser points straight at the downstream trap, so it must be detuned, geometrically dodged, or extinguished by a zero-crossing field so it does not shove the freshly caught atoms out of the magneto-optical trap.

Worked examples

1 · Designing a sodium slower (confidence builder)

Take sodium ( $\lambda=589\,\mathrm{nm}$ , $\Gamma/2\pi=9.79\,\mathrm{MHz}$ , $m=23\,\mathrm{u}$ ), a capture velocity $v_0=1000\,\mathrm{m/s}$ , and a safety factor $\eta=0.5$ . Find the maximum deceleration, the design deceleration, the required length, and the number of photons and time to stop an atom.

Step 1. Maximum deceleration. With $k=2\pi/\lambda=1.067\times10^{7}\,\mathrm{m^{-1}}$ , $\hbar k=1.125\times10^{-27}\,\mathrm{kg\,m/s}$ , $\Gamma=2\pi(9.79\,\mathrm{MHz})=6.15\times10^{7}\,\mathrm{s^{-1}}$ , and $m=3.82\times10^{-26}\,\mathrm{kg}$ , equation (3) gives $a_{\max}=\hbar k\Gamma/2m=\mathbf{9.06\times10^{5}\,\mathrm{m/s^2}}$ .

Step 2. Design deceleration and length. At $\eta=0.5$ , $a=\eta a_{\max}=4.53\times10^{5}\,\mathrm{m/s^2}$ . The stopping length from equation (4) is $z_0=v_0^2/2a=(10^{6})/(9.06\times10^{5})=\mathbf{1.10\,\mathrm{m}}$ . Running at the ceiling instead ( $\eta=1$ ) would shrink it to $0.55\,\mathrm{m}$ , the hard floor from Derivation 2.

Step 3. Photon budget. From equation (6), $N_{\mathrm{ph}}=m v_0/\hbar k=(3.82\times10^{-26})(1000)/(1.125\times10^{-27})=\mathbf{3.4\times10^{4}}$ photons, each a $2.95\,\mathrm{cm/s}$ recoil.

Step 4. Slowing time. Either $t=v_0/a=1000/(4.53\times10^{5})=2.2\,\mathrm{ms}$ , or $t=N_{\mathrm{ph}}/(\eta\Gamma/2)=(3.4\times10^{4})/(1.5\times10^{7})=2.2\,\mathrm{ms}$ — the two agree. A sodium atom is stopped in about 1.1 m, 3.4×10⁴ photons, and 2.2 ms.

2 · The field taper and the Doppler-Zeeman balance

For the same sodium beam, find the Doppler shift that must be compensated at the entrance, the field amplitude $B_0$ (take $\mu'=\mu_B$ ), and verify that the Zeeman shift at $B_0$ matches the Doppler shift.

Step 1. Entrance Doppler shift. The counter-propagating beam is shifted by $kv_0$ ; as a frequency, $kv_0/2\pi=(1.067\times10^{7})(1000)/2\pi=\mathbf{1.70\,\mathrm{GHz}}$ . That is the tuning the field must supply at $z=0$ .

Step 2. Field amplitude. From equation (5), $B_0=\hbar k v_0/\mu_B=(1.125\times10^{-27})(1000)/(9.274\times10^{-24})=0.121\,\mathrm{T}=\mathbf{1213\,\mathrm{G}}$ . The tapered part of the field must sweep from about $1213\,\mathrm{G}$ down to near zero across $z_0$ .

Step 3. Consistency. The Zeeman tuning coefficient is $\mu_B/h=1.40\,\mathrm{MHz/G}$ , so at $B_0=1213\,\mathrm{G}$ the shift is $(1.40\,\mathrm{MHz/G})(1213\,\mathrm{G})=1.70\,\mathrm{GHz}$ — exactly the entrance Doppler shift, as the resonance condition demands. B₀ = 1213 G tunes the transition by 1.70 GHz, cancelling the 1.70 GHz Doppler shift at v₀.

3 · Transverse heating from random recoils

Each of the $N_{\mathrm{ph}}=3.4\times10^{4}$ spontaneous emissions kicks the atom in a random direction. Estimate the resulting root-mean-square transverse velocity and comment on beam divergence.

Step 1. The emission kicks execute a random walk in momentum. After $N$ isotropic recoils the mean-square momentum per Cartesian axis is $N(\hbar k)^2/3$ , so the rms velocity per transverse axis is $v_\perp=v_{\mathrm{rec}}\sqrt{N/3}$ with $v_{\mathrm{rec}}=\hbar k/m=2.95\,\mathrm{cm/s}$ .

Step 2. Insert the numbers: $v_\perp=(0.0295\,\mathrm{m/s})\sqrt{3.4\times10^{4}/3}=(0.0295)(106)=\mathbf{3.1\,\mathrm{m/s}}$ per transverse axis.

Step 3. Over the roughly $2\,\mathrm{ms}$ of flight this spread barely blurs the trajectory, but relative to a $30\,\mathrm{m/s}$ exit velocity a few m/s of transverse motion is a large divergence angle, so the slow beam fans out before reaching the trap. The beam picks up ~3 m/s of transverse spread, which is why slowers are paired with transverse cooling to keep the flux collimated.

Physics of the portrait

The Zeeman slower is a rare example of a device named for the physics it exploits rather than its inventor — but the inventor is unmistakable, and he built it to solve a problem everyone else had run into and given up on.

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 first Zeeman slower at the U.S. National Bureau of Standards (now NIST) and reported it with Harold Metcalf in 1982, decelerating a sodium beam by winding a solenoid whose tapered field held the atoms on resonance as they slowed. The obstacle they overcame was exactly the velocity selectivity of the scattering force: earlier attempts to slow beams stalled after the atoms Doppler-shifted themselves out of the laser, and a rival approach chirped the laser frequency instead of shaping a field. Phillips's insight was that a static, spatially varying magnetic field could do the retuning in space rather than in time, producing a continuous slow beam rather than a pulsed one. He went on to find optical molasses far colder than the Doppler limit anyone had predicted, forcing the theory of sub-Doppler cooling, and for this body of work on cooling and trapping atoms with light he shared the 1997 Nobel Prize with Steven Chu and Claude Cohen-Tannoudji. The slower remains his most-copied piece of hardware: some version of it sits at the head of a large fraction of the world's cold-atom laboratories.

History

1975

— Hänsch and Schawlow propose Doppler cooling of neutral atoms with laser light

1982

— Phillips and Metcalf slow a sodium beam with a tapered magnetic field, the first Zeeman slower

1985

— Chu and co-workers demonstrate optical molasses at Bell Labs

1995

— Ketterle's group reaches Bose-Einstein condensation in sodium, its beam fed through a Zeeman slower

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), ch. 6 · Phillips and Metcalf, Phys. Rev. Lett. 48, 596 (1982) · Demtröder, Laser Spectroscopy, Vol. 2 · Saleh and Teich, Fundamentals of Photonics (light forces on atoms). The key anchor numbers for sodium ( $\lambda=589\,\mathrm{nm}$ , $\Gamma/2\pi=9.79\,\mathrm{MHz}$ , $m=23\,\mathrm{u}$ ) are: $a_{\max}=9.1\times10^{5}\,\mathrm{m/s^2}$ , $B_0=\hbar k v_0/\mu_B=1213\,\mathrm{G}$ , $z_0=v_0^2/2\eta a_{\max}=1.1\,\mathrm{m}$ , $N_{\mathrm{ph}}=m v_0/\hbar k=3.4\times10^{4}$ , and the Doppler-Zeeman balance $\mu_B B_0/h=1.70\,\mathrm{GHz}=kv_0/2\pi$ .

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

Ask the tutor: Ask me to walk through the phase-stability argument step by step, or to redesign the field taper for a different species such as rubidium or strontium.

Why does the magnetic field need to change along the slower at all?

Because the thing knocking the atom off resonance changes along the slower. As the atom decelerates, its Doppler shift $kv$ shrinks, so a fixed-frequency laser that was resonant at the entrance would be badly mistuned a few centimetres later. The field's job is to supply, through the Zeeman effect, exactly the tuning the shrinking Doppler shift gives up: near the entrance the atom is fast and needs a large compensating shift, near the exit it is slow and needs almost none. A uniform field could hold resonance at one velocity only; a tapered field holds it continuously as the velocity sweeps from $v_0$ to nearly zero, which is the whole point of the device.

What is the safety factor η and why not just run at η = 1?

The safety factor is the fraction of the maximum force the design actually uses, $a=\eta a_{\max}$ , typically around 0.5. Running at $\eta=1$ puts the atom at the very peak of its force-versus-detuning curve, where any small deviation — a slightly fast atom, a field imperfection — can only reduce the force, and the atom falls behind the ideal trajectory and never recovers. At $\eta$ below 1 the atom sits on the flank of the resonance with headroom above: an atom that runs a little fast scatters a little harder and is pulled back onto the profile. That restoring behaviour, called phase stability, is what lets the slower capture a bunch of atoms rather than a single perfectly-tuned one, at the modest cost of a longer tube.

Is a Zeeman slower actually cooling the atoms?

Not in the strict sense, at least not primarily. Cooling means shrinking the spread of velocities; the slower mainly shifts one velocity class bodily downward in the lab frame, decelerating rather than narrowing. There is some genuine compression of the velocity spread near the exit as the resonance bunches atoms, but the dominant action is bulk deceleration of a chosen slice of the beam. That distinction matters downstream: the slowed atoms are still warm by cold-atom standards (tens of m/s), and the real cooling to microkelvin temperatures happens later, in optical molasses and evaporative stages. The slower's contribution is to hand those stages a beam slow enough to be captured at all.

Why does the slowing laser cause trouble for the trap it is feeding?

The slowing beam is nearly resonant and points straight down the axis at the magneto-optical trap it is loading. Atoms that have just been slowed to a few m/s are exactly the atoms that beam can push hardest, so left unchecked it blows them back out of the trap or heats them transversely. Designers defeat this in several ways: they leave a deliberate nonzero exit velocity so the atoms are still moving forward, they detune the laser so it is off-resonance in the low-field trap region, they use a spin-flip (zero-crossing) field so the exit field is small and the slowing transition is dark there, or they bring the beam in slightly off-axis. Getting this hand-off right is one of the fussier parts of building a slower.

Why do slowers need a repump laser?

Because an atom has to scatter tens of thousands of photons to stop, and few transitions are perfectly closed. In sodium and the other alkalis the excited state can occasionally decay to the wrong ground-state hyperfine level, which is dark to the cooling laser; once there, the atom stops scattering and coasts through unslowed. Even a leak of one part in $10^{4}$ per cycle would empty the useful beam before it reached the exit, since $N_{\mathrm{ph}}\approx3\times10^{4}$ . A repump laser tuned to the leaked-into state pumps those atoms back into the cycling transition, keeping the cycle effectively closed for the whole flight. The photon budget, not any subtlety of the field, is what forces this second beam.

Increasing-field, decreasing-field, spin-flip — what is the difference?

They are three ways to arrange the same square-root taper relative to the field's zero. An increasing-field slower runs the field growing along $z$ with a positive bias; a decreasing-field slower runs it the other way. The spin-flip (zero-crossing) design splits the field into a positive section and a negative section that cross through zero in the middle. The spin-flip has two practical wins: the peak magnitude of the field is roughly halved, which cuts coil power and heat, and the exit field is near zero, which keeps stray field away from the downstream trap. Its cost is a small region near the zero crossing where the atom's magnetic sublevel can flip and briefly go dark, needing careful repumping. Most modern machines choose the spin-flip geometry for the low exit field alone.