Magnetic Trapping of Neutral Atoms

A neutral atom carries no charge, so the Lorentz force has nothing to grip — yet it can be held motionless in free space by a magnetic field alone. The lever is the atom's magnetic moment, and the physics is borrowed wholesale from two results this chapter already owns. From the Zeeman effect comes the energy of a hyperfine state in a field, $\Delta E = m_F g_F \mu_B |\mathbf{B}|$ ; from the adiabatic theorem of quantum mechanics comes the crucial simplification that a slowly moving atom stays locked in one $m_F$ manifold, so that vector energy becomes a scalar potential $U(\mathbf{r}) = m_F g_F \mu_B |\mathbf{B}(\mathbf{r})|$ shaped like the field magnitude itself. A gradient in $|\mathbf{B}|$ is then a conservative force on the atom's centre of mass, and — unlike the optical dipole and radiation-pressure traps of the preceding articles — it scatters no photons, so it is free of recoil heating and light shifts. That cleanliness is exactly what the last, coldest stages of cooling demand. Two hard constraints shape every real trap. Earnshaw's theorem, applied to a source-free static field, permits a local minimum of $|\mathbf{B}|$ but forbids a local maximum, so only weak-field-seeking states ( $m_F g_F > 0$ ) can be confined, and only at field minima. And wherever $|\mathbf{B}|$ actually reaches zero, the Larmor precession that enforces adiabatic following stalls, and atoms spin-flip into untrapped states — the Majorana loss that guts the simplest quadrupole trap precisely as the gas gets cold. The Ioffe-Pritchard geometry answers both: a nonzero bias $B_0$ lifts the field floor above zero and rounds the potential into a harmonic bowl. That plugged, Majorana-safe trap, combined with radio-frequency evaporation, carried the first dilute gases to Bose-Einstein condensation and remains the backbone of ultracold-atom physics.

Key takeaways

An atom of magnetic moment $\boldsymbol{\mu}$ has energy $U = -\boldsymbol{\mu}\cdot\mathbf{B}$ ; if it follows the field adiabatically this collapses to the scalar $U = m_F g_F \mu_B |\mathbf{B}(\mathbf{r})|$ , and a gradient in $|\mathbf{B}|$ is a real force on its centre of mass.
Earnshaw's theorem forbids a static field maximum in free space, so only field minima trap, and only weak-field-seeking states with $m_F g_F > 0$ are held.
A spherical quadrupole gives the tightest, simplest trap but has an exact field zero at its centre where adiabaticity fails and Majorana spin flips eject atoms — and the loss rate grows as the cloud gets colder.
The Ioffe-Pritchard trap adds a nonzero bias $B_0$ , turning the potential into a harmonic bowl with a $|\mathbf{B}|_{\min} = B_0 > 0$ floor; this defeats Majorana loss and is the standard stage for evaporation to quantum degeneracy.
Concrete anchors: it takes only $B' \approx 15\ \mathrm{G/cm}$ to levitate rubidium against gravity, a $100\ \mathrm{G}$ field is a $\sim 6.7\ \mathrm{mK}$ deep well, and the Larmor rate is $2\pi \times 1.40\ \mathrm{MHz}$ per gauss.
Key idea: adiabatic following converts the vector interaction $-\boldsymbol{\mu}\cdot\mathbf{B}$ into a scalar potential $m_F g_F \mu_B |\mathbf{B}(\mathbf{r})|$ that tracks field magnitude — so trap only field minima with weak-field-seekers, and keep $|\mathbf{B}| \neq 0$ everywhere to defeat Majorana spin flips.

The core insight

Adiabatic following turns the vector interaction $-\boldsymbol{\mu}\cdot\mathbf{B}$ into a scalar potential $m_F g_F \mu_B |\mathbf{B}(\mathbf{r})|$ that tracks the field magnitude; trap only field minima with weak-field-seekers, and keep $|\mathbf{B}| \neq 0$ everywhere to defeat Majorana spin flips.

Where this lives today

Magnetic traps remain the conservative-confinement stage that turns a laser-cooled cloud into a quantum gas. In a typical apparatus, atoms collected and cooled in a magneto-optical trap are optically pumped into a single weak-field-seeking state, the light is extinguished, and an Ioffe-Pritchard or plugged quadrupole field is ramped up to catch them without scattering a single further photon — the essential handoff described in the article on the magneto-optical trap. There, forced radio-frequency evaporation selectively spin-flips the hottest atoms out of the trap, driving the phase-space density up by orders of magnitude until, as detailed in the articles on evaporative cooling and Bose-Einstein condensation, the gas crosses into quantum degeneracy. The same principle has been miniaturised into atom chips: micro-fabricated current-carrying wires patterned on a substrate produce Ioffe-type microtraps only microns from the surface, reaching trap frequencies of many kilohertz with currents of an amp or two. These chips make compact, portable BEC sources, magnetic waveguides for guided atom optics, and on-chip matter-wave interferometers — the physics of inertial sensors and transportable atomic clocks that are now leaving the laboratory. Whether the coils span a metre or a millimetre, the operating principle is unchanged: a weak-field-seeking atom, adiabatically following a field with a nonzero minimum, resting in a Majorana-safe valley of $|\mathbf{B}|$ .

The physical picture

Picture the atom as a tiny compass needle that also carries energy: aligned with the field it sits low, anti-aligned it sits high, and the energy is $-\boldsymbol{\mu}\cdot\mathbf{B}$ . Now let the atom drift through a field that changes direction from place to place. If the needle precesses fast enough to re-point toward the local field before the atom has moved appreciably, then the direction of $\mathbf{B}$ is irrelevant — only its strength enters the energy. The atom therefore rolls in a potential landscape whose shape is a faithful copy of $|\mathbf{B}(\mathbf{r})|$ : valleys where the field is weak, hills where it is strong. That single idea already fixes what kinds of trap are possible. Because a source-free magnetic field cannot have a local maximum of its magnitude (Earnshaw's theorem — the same result that forbids purely electrostatic trapping of charges), you cannot build a hilltop for a strong-field seeker to sit on. You can only dig a valley in $|\mathbf{B}|$ , and only atoms whose energy rises with field — the weak-field seekers, $m_F g_F > 0$ — will settle at the bottom of it. Everything about magnetic trapping follows from "you may build valleys, not peaks."

Figure

Two magnetic trap potentials along a line through the centre. The quadrupole gives a linear cusp with an exact field zero at r = 0 (the Majorana hole), while the Ioffe-Pritchard trap raises a smooth harmonic bowl to a nonzero floor μ′B₀ with bias B₀ = 1 G and radial gradient B′ = 1.5 T/m = 150 G/cm

The compass keeps its bearing only while it precesses. Right at a field zero the field has no direction to precess about, the needle "gets lost," and as the atom emerges on the far side the needle may find itself pointing the wrong way — now anti-trapped, and flung out. This is the Majorana spin flip, and it is not a minor leak: a cold, dense cloud piles up exactly at the centre, which is exactly where the zero sits. The quadrupole trap, made from two opposed coils, gives the tightest and simplest valley of all, but with a razor-sharp zero at the very bottom. The Ioffe-Pritchard trap files that spike down: it adds a bias field so the floor of the valley never touches zero, trading a little confinement for a needle that never loses its way.

Figure

Field lines of a spherical quadrupole in the z = 0 plane point radially outward with field magnitude B = B′ρ, vanishing at the centre. A moving atom's spin (green needle) follows the local direction while the Larmor rate ω_L = μ_B·B/ħ = 2π × 1.40 MHz per gauss stays large, but near the zero the field slews faster than the spin can track and the needle flips (red)

The core equations

The starting point is the same Zeeman energy that the prerequisite on magnetic moments derives, but now read as a function of position rather than of a fixed applied field. An atom in hyperfine state $|F, m_F\rangle$ has moment $\boldsymbol{\mu} = -g_F \mu_B \mathbf{F}/\hbar$ ; quantized along the local field it acquires energy proportional to $|\mathbf{B}|$ . Because the atom moves slowly enough to stay in one $m_F$ manifold, this internal energy acts on the centre of mass as an external potential.

U(\mathbf{r}) = -\boldsymbol{\mu}\cdot\mathbf{B}(\mathbf{r}) = m_F\,g_F\,\mu_B\,|\mathbf{B}(\mathbf{r})| \qquad (1)

where: $\boldsymbol{\mu}$ — atomic magnetic moment, $\boldsymbol{\mu} = -g_F \mu_B \mathbf{F}/\hbar$ ; $m_F$ — projection of total angular momentum $F$ along the local field; $g_F$ — hyperfine Landé g-factor; $\mu_B = 9.274\times10^{-24}\ \mathrm{J\,T^{-1}}$ — Bohr magneton; $|\mathbf{B}(\mathbf{r})|$ — local field magnitude. The vector coupling has collapsed to a scalar that depends only on how strong the field is, not on where it points — this is the whole trick, and it is only valid in the adiabatic regime spelled out below. A potential means a force. Differentiating (1) shows the atom is pushed along the gradient of the field magnitude, weighted by the effective moment $\mu' \equiv m_F g_F \mu_B$ . This is the force that must balance gravity and provide confinement, and its sign is set entirely by $m_F g_F$ .

\mathbf{F} = -\nabla U = -m_F\,g_F\,\mu_B\,\nabla|\mathbf{B}(\mathbf{r})|, \qquad m_F g_F > 0\ \ (\text{trappable}) \qquad (2)

where: $\mathbf{F}$ — conservative force on the centre of mass; $\nabla|\mathbf{B}|$ — gradient of the field magnitude (never of a component); $m_F g_F > 0$ — the weak-field-seeking condition, energy rising with field so the atom is driven toward minima. Strong-field seekers ( $m_F g_F < 0$ ) would need a field maximum, which Earnshaw's theorem forbids in free space — hence they cannot be trapped statically. The scalar reduction is not free: it holds only if the atom's spin can re-orient toward the changing field faster than the field direction changes in its own frame. The rate of re-orientation is the Larmor precession frequency, and the adiabatic condition compares it with the angular slew rate $|\dot{\hat{\mathbf{B}}}|$ the moving atom sees.

\omega_L = \frac{\mu_B|\mathbf{B}|}{\hbar} = 2\pi\times 1.40\ \mathrm{MHz}\ \text{per gauss}\ ;\qquad \text{adiabatic if}\ \ \omega_L \gg \left|\frac{\mathrm{d}\hat{\mathbf{B}}}{\mathrm{d}t}\right| \qquad (3)

where: $\omega_L$ — Larmor precession frequency; $\mathrm{d}\hat{\mathbf{B}}/\mathrm{d}t$ — rate of change of field direction seen by the moving atom; $\hbar$ — reduced Planck constant. The numerical coefficient $\mu_B/h = 1.40\ \mathrm{MHz/G}$ is a fixed atomic constant. Adiabaticity fails wherever $|\mathbf{B}|\to 0$ , because there $\omega_L\to 0$ while the slew rate stays finite — the origin of Majorana loss. The simplest field with a minimum is the spherical quadrupole of two opposed (anti-Helmholtz) coils. Near the centre the field is linear in position, and the divergence-free condition $\nabla\cdot\mathbf{B}=0$ forces the axial gradient to be twice the radial one, which puts a factor of four under the axial term of the magnitude.

\mathbf{B} = B'\,(x\,\hat{x} + y\,\hat{y} - 2z\,\hat{z}), \qquad |\mathbf{B}(\mathbf{r})| = B'\sqrt{x^2 + y^2 + 4z^2} \qquad (4)

where: $B'$ — radial field gradient (T/m); the $-2$ on $\hat{z}$ enforces $\partial_x B_x + \partial_y B_y + \partial_z B_z = 0$ ; the factor $4$ under $z^2$ makes the axial confinement twice as steep as the radial. The magnitude vanishes only at the origin, a single point — but that point is a cusp with an exact zero, the Majorana hole. The Ioffe-Pritchard configuration superimposes longitudinal bars (giving a radial gradient $B'$ and a saddle) on axial pinch coils that set a nonzero bias $B_0$ and an axial curvature $B''$ . Expanding $|\mathbf{B}|$ to second order about the centre gives a smooth harmonic bowl whose floor is $B_0 > 0$ , so the field never reaches zero.

|\mathbf{B}| \simeq B_0 + \tfrac12 B'' z^2 + \left(\frac{B'^2}{2B_0} - \frac{B''}{4}\right)\rho^2, \qquad \rho^2 = x^2 + y^2 \qquad (5)

where: $B_0$ — nonzero axial bias at the centre; $B'$ — radial gradient from the Ioffe bars; $B''$ — axial curvature from the pinch coils; $\rho$ — radial coordinate. The transverse field adds in quadrature to the axial bias, which is why the radial term carries $B'^2/2B_0$ : a larger bias $B_0$ softens the radial confinement. Reading the curvatures of (5) as harmonic springs through $U = \mu'|\mathbf{B}|$ gives the two trap frequencies that characterise the cigar-shaped potential, with $\mu' = m_F g_F \mu_B$ the effective moment.

\omega_r = \sqrt{\frac{\mu'}{m}\left(\frac{B'^2}{B_0} - \frac{B''}{2}\right)}, \qquad \omega_z = \sqrt{\frac{\mu' B''}{m}} \qquad (6)

where: $\omega_r, \omega_z$ — radial and axial harmonic oscillation frequencies; $\mu' = m_F g_F \mu_B$ — effective moment; $m$ — atomic mass. Lowering $B_0$ tightens the radial trap (a common compression trick), but if $B_0$ is pushed too small the field floor drops toward zero and Majorana loss returns — the fundamental trade-off of the geometry.

Derivation AFrom the vector interaction to a scalar magnetic potential

The goal is to reduce the full vector coupling $-\boldsymbol{\mu}\cdot\mathbf{B}$ — which in general mixes an atom's spin states as it moves through a field that rotates in direction from point to point — down to a single scalar potential that depends only on the field magnitude $|\mathbf{B}(\mathbf{r})|$ . We begin from the hyperfine Zeeman Hamiltonian, quantize the spin along the local field direction, invoke the adiabatic theorem to lock the atom into one $m_F$ manifold, and read off both the trapping potential and the sign condition that singles out weak-field seekers as the only trappable states.

Step 1. Write the magnetic interaction for a hyperfine level $F$ using $\boldsymbol{\mu} = -g_F\mu_B\mathbf{F}/\hbar$ , so that $H = -\boldsymbol{\mu}\cdot\mathbf{B} = g_F\mu_B\,(\mathbf{F}\cdot\mathbf{B})/\hbar$ . Why this form: it makes explicit that the coupling is between the angular-momentum operator $\mathbf{F}$ and the field vector, so the physics to come is entirely about how $\mathbf{F}$ orients relative to $\mathbf{B}$ .

Step 2. Quantize $\mathbf{F}$ along the local direction $\hat{\mathbf{B}}(\mathbf{r})$ ; the eigenstates $|m_F\rangle$ then satisfy $H|m_F\rangle = m_F g_F\mu_B |\mathbf{B}(\mathbf{r})|\,|m_F\rangle$ . Why the magnitude alone survives: choosing the quantization axis to be the field direction absorbs all information about where $\mathbf{B}$ points, leaving only its length $|\mathbf{B}|$ in the eigenvalue.

Step 3. Invoke the adiabatic theorem: if the field direction changes slowly compared with the Larmor precession, $\omega_L \gg |\dot{\hat{\mathbf{B}}}|$ , an atom prepared in $|m_F\rangle$ stays in $|m_F\rangle$ as it moves, and its internal energy $m_F g_F\mu_B|\mathbf{B}(\mathbf{r})|$ becomes a smooth function of position that the centre of mass feels as an external potential $U(\mathbf{r})$ . Why adiabaticity is essential: without it the atom would sample several $m_F$ levels and the "potential" would be ill-defined; adiabatic following is precisely what makes a single-valued conservative landscape exist.

Step 4. Read off the sign condition. For $m_F g_F > 0$ the energy $U$ rises with $|\mathbf{B}|$ , so the atom is pushed toward field minima and can be trapped (weak-field seeker); for $m_F g_F < 0$ it seeks maxima, which Earnshaw's theorem forbids for a source-free static field, so such states are untrappable by static means. Why this is the whole selection rule: it explains why experiments optically pump into a specific stretched state such as $^{87}\mathrm{Rb}\ |F=2, m_F=+2\rangle$ before loading the magnetic trap. The result reduces to the standard Zeeman shift $\Delta E = m_F g_F\mu_B B$ in a uniform field, and $\mathbf{F} = -\nabla U$ recovers the magnetic-trap force of Metcalf and van der Straten.

Go deeper · Derivation B · 1 minHarmonic frequencies of the Ioffe-Pritchard trap

Derivation BHarmonic frequencies of the Ioffe-Pritchard trap

The aim is to turn the raw coil fields of an Ioffe-Pritchard trap into the two oscillation frequencies $\omega_r$ and $\omega_z$ that an experimentalist actually measures and tunes. We assemble the field from its three ingredients — the transverse Ioffe gradient $B'$ , the axial bias $B_0$ , and the axial pinch curvature $B''$ — expand the magnitude $|\mathbf{B}|$ to second order in the small displacements from the centre, and match the resulting quadratic potential to the harmonic-oscillator form, being careful about how the transverse field adds in quadrature to the bias.

Step 1. Write the field components near the centre as $\mathbf{B}\approx(B'x - \tfrac12 B''xz,\ -B'y - \tfrac12 B''yz,\ B_0 + \tfrac12 B''(z^2 - \tfrac12\rho^2))$ . Why this combination: the Ioffe bars supply the linear transverse gradient $B'$ , while the pinch coils supply the axial bias $B_0$ and its curvature $B''$ ; the cross terms are fixed by $\nabla\cdot\mathbf{B} = \nabla\times\mathbf{B} = 0$ in the current-free trapping region.

Step 2. Form the magnitude. The transverse components are perpendicular to the dominant axial bias, so they add in quadrature: $|\mathbf{B}| = \sqrt{B_z^2 + B'^2\rho^2}\approx B_0 + \tfrac12 B''z^2 + (B'^2/2B_0 - B''/4)\rho^2$ to second order. Why the $B'^2/2B_0$ appears: expanding $\sqrt{B_0^2 + B'^2\rho^2}\approx B_0 + B'^2\rho^2/2B_0$ shows a large bias $B_0$ divides down the radial stiffness — the transverse gradient only weakly lengthens a vector that already points mostly along $z$ .

Step 3. Multiply by the effective moment to get the potential $U = \mu'|\mathbf{B}|$ and match to $U \simeq \mu'B_0 + \tfrac12 m\omega_r^2\rho^2 + \tfrac12 m\omega_z^2 z^2$ , giving $\tfrac12 m\omega_r^2 = \mu'(B'^2/2B_0 - B''/4)$ and $\tfrac12 m\omega_z^2 = \tfrac12\mu'B''$ . Why matching works: near the minimum any smooth potential is quadratic, so its curvatures are literally the spring constants $m\omega^2$ of independent radial and axial oscillators.

Step 4. Solve for the frequencies: $\omega_r = \sqrt{(\mu'/m)(B'^2/B_0 - B''/2)}$ and $\omega_z = \sqrt{\mu'B''/m}$ . Why these are the design knobs: $\omega_r$ can be raised sharply by lowering the bias $B_0$ (radial compression), while $\omega_z$ depends only on the pinch curvature — so the aspect ratio, and hence the cloud shape, is tunable in the lab. As $B_0\to 0$ the radial term diverges and the expansion breaks down, correctly signalling the return of the linear quadrupole limit and its central zero.

Go deeper · Derivation C · 1 minWhy the Majorana loss rate grows as the cloud is cooled

Derivation CWhy the Majorana loss rate grows as the cloud is cooled

The goal is to understand the counter-intuitive fact that a quadrupole trap gets worse at holding atoms as they get colder — the opposite of every classical intuition about deeper confinement. We estimate the size of the small region near the field zero where adiabatic following fails, work out how the thermal velocity sets the atom's dwell time there, and combine them into a scaling law for the loss rate, showing that it climbs as the temperature falls and therefore diverges on the approach to degeneracy.

Step 1. Locate the non-adiabatic region. Adiabaticity fails where the Larmor rate $\omega_L = \mu_B B'r/\hbar$ drops below the direction-slew rate $\sim v/r$ the atom sees; setting them equal gives a hole radius $r_{\rm na}\sim\sqrt{\hbar v/\mu_B B'}$ . Why a finite radius exists: arbitrarily close to the zero the field is arbitrarily weak, so there is always a small sphere in which the spin cannot keep up — its size shrinks with steeper gradient $B'$ but never reaches zero.

Step 2. Set the atom speed by temperature: $v\sim\sqrt{k_B T/m}$ from equipartition. Why temperature enters twice: it fixes both how fast atoms cross the hole and, through the trap's linear potential, how tightly the cloud is squeezed around the very point where the hole sits.

Step 3. Combine the flux of atoms through the small hole with $r_{\rm na}$ to estimate $\Gamma_{\rm Maj}\sim(\hbar/m)(\mu_B B'/k_B T)^2$ . Why the sign of the temperature dependence flips intuition: the rate scales as $T^{-2}$ , so halving the temperature quadruples the loss — colder clouds are denser at the centre and dwell longer in the hole, both of which feed the leak.

Step 4. Draw the consequence: because $\Gamma_{\rm Maj}$ diverges as $T\to 0$ , one cannot evaporate a pure quadrupole trap all the way to quantum degeneracy. Why this drove real hardware: it is exactly why the first condensates used a plugged geometry — an Ioffe bias, a rotating time-orbiting potential, or a blue-detuned optical plug — to lift the field floor off zero before the final cooling stage. The $T^{-2}$ scaling matches the losses measured by Petrich and colleagues and the Ketterle Varenna lectures.

Limiting cases

Case	Result	Meaning
Adiabatic (slow-atom) limit	$\omega_L = \dfrac{\mu_B\|\mathbf{B}\|}{\hbar} \gg \left\vert\dfrac{\mathrm{d}\theta}{\mathrm{d}t}\right\vert$	Spin follows the field; the potential is the scalar $m_F g_F\mu_B\|\mathbf{B}\|$ and trapping is conservative and lossless.
Non-adiabatic (near field zero)	$\vert\mathbf{B}\vert\to 0 \Rightarrow \omega_L\to 0$	Precession cannot keep up; Majorana spin flips send atoms into $m_F g_F < 0$ states that are expelled — the dominant loss in quadrupole traps.
Harmonic (Ioffe-Pritchard near minimum)	$U \simeq \mu'B_0 + \tfrac12 m\omega_r^2\rho^2 + \tfrac12 m\omega_z^2 z^2$	Small oscillations about a nonzero floor $B_0$ ; smooth, Majorana-safe, ideal for evaporation. Here $\mu' = m_F g_F\mu_B$ .
Linear (quadrupole far from zero)	$U \simeq \mu'B'\sqrt{x^2 + y^2 + 4z^2}$	Linear cusp confinement; tighter than harmonic for equal gradient but reintroduces the central zero.
Weak-field-seeker requirement	$m_F g_F > 0$	Set by Earnshaw: only field minima exist in free space, so only states whose energy rises with $\vert\mathbf{B}\vert$ are held.

Worked examples

1 · The gradient that levitates rubidium (confidence builder)

A cloud of $^{87}$ Rb is prepared in the weak-field-seeking stretched state $|F=2, m_F=+2\rangle$ , for which $g_F = +\tfrac12$ , so $m_F g_F = 1$ and the effective moment is $\mu' = \mu_B$ . What radial magnetic-field gradient $B'$ is needed for the trapping force to balance gravity — the minimum gradient to hold the atoms up at all?

Step 1. Set the magnetic force equal to the weight. The upward magnetic force at gradient $B'$ is $F = \mu' B' = \mu_B B'$ , and it must balance $mg$ : $\mu_B B' = mg$ . This is the levitation condition; below it the atoms simply fall out of the bottom of the trap.

Step 2. Insert the numbers for rubidium. The mass is $m = 87\times1.6605\times10^{-27} = 1.445\times10^{-25}\ \mathrm{kg}$ , so the weight is $mg = 1.445\times10^{-25}\times9.81 = 1.417\times10^{-24}\ \mathrm{N}$ .

Step 3. Solve for the gradient: $B' = mg/\mu_B = 1.417\times10^{-24}/9.274\times10^{-24} = 0.153\ \mathrm{T/m}$ . Converting, $1\ \mathrm{T/m} = 100\ \mathrm{G/cm}$ , so $B' = 0.153\ \mathrm{T/m} = 15.3\ \mathrm{G/cm}$ . This modest gradient — a few tens of amps through centimetre-scale coils — is why magnetic trapping of alkali atoms is so practical, and why "about 15 G/cm to levitate rubidium" is a number every cold-atom experimentalist carries in their head.

2 · Frequencies of a real Ioffe-Pritchard trap

An Ioffe-Pritchard trap for the same $^{87}$ Rb state ( $\mu' = \mu_B$ ) runs with radial gradient $B' = 1.5\ \mathrm{T/m}$ , axial bias $B_0 = 1.0\ \mathrm{G} = 1.0\times10^{-4}\ \mathrm{T}$ , and axial curvature $B'' = 50\ \mathrm{T/m^2}$ . Find the radial and axial trap frequencies and the cloud's aspect ratio.

Step 1. Compute the moment-to-mass ratio that appears in every frequency: $\mu'/m = 9.274\times10^{-24}/1.445\times10^{-25} = 64.2\ \mathrm{m^2\,s^{-2}\,T^{-1}}$ . Carrying this once saves repeating it for both axes.

Step 2. Radial frequency from equation (6). The radial stiffness factor is $B'^2/B_0 - B''/2 = (1.5)^2/10^{-4} - 25 = 2.25\times10^4 - 25 \approx 2.25\times10^4\ \mathrm{T/m^2}$ (the curvature term is negligible against the bias term). Then $\omega_r = \sqrt{64.2\times2.25\times10^4} = 1.20\times10^3\ \mathrm{rad/s}$ , so $f_r = \omega_r/2\pi = 191\ \mathrm{Hz}$ .

Step 3. Axial frequency: $\omega_z = \sqrt{(\mu'/m)\,B''} = \sqrt{64.2\times50} = 56.7\ \mathrm{rad/s}$ , so $f_z = 9.0\ \mathrm{Hz}$ . The aspect ratio is $\omega_r/\omega_z = 191/9.0 \approx 21$ : a strongly elongated, cigar-shaped cloud, exactly the shape characteristic of Ioffe-Pritchard condensates. Note how sensitively $f_r$ depends on $B_0$ — halving the bias to $0.5\ \mathrm{G}$ raises $f_r$ by $\sqrt{2}$ to about $270\ \mathrm{Hz}$ , the standard radial-compression manoeuvre before evaporation.

3 · Trap depth versus the Larmor floor

A quadrupole coil produces a field of $100\ \mathrm{G} = 0.01\ \mathrm{T}$ at the edge of the trapping region, again for $\mu' = \mu_B$ . How deep is the trap in temperature units, and what is the Larmor frequency at a $1\ \mathrm{G}$ bias floor?

Step 1. Trap depth is the potential difference between the field maximum and the zero: $U = \mu' B = \mu_B\times0.01 = 9.27\times10^{-26}\ \mathrm{J}$ . In temperature units $U/k_B = 9.27\times10^{-26}/1.381\times10^{-23} = 6.7\times10^{-3}\ \mathrm{K}$ , so the well is about $6.7\ \mathrm{mK}$ deep. That comfortably holds a sub-millikelvin cloud handed over from a magneto-optical trap (typically $\sim 100\ \mu\mathrm{K}$ ), which is why a MOT is the standard loading stage.

Step 2. Larmor frequency at a $1\ \mathrm{G}$ floor: $\omega_L = \mu_B B/\hbar = 9.274\times10^{-24}\times10^{-4}/1.055\times10^{-34} = 8.79\times10^6\ \mathrm{rad/s}$ , i.e. $f_L = 1.40\ \mathrm{MHz}$ — consistent with the $1.40\ \mathrm{MHz/G}$ constant. This is why an Ioffe bias of even a single gauss suffices to defeat Majorana loss: at $1.4\ \mathrm{MHz}$ the spin re-points millions of times per second, vastly faster than the kilohertz-scale rate at which a trapped atom traverses the field, so adiabatic following is never in doubt.

Physics of the portrait

Magnetic trapping is where the theory of atomic moments met the engineering that would eventually make quantum gases. Its defining geometry, and much of the human lineage that led to Bose-Einstein condensation, traces to one MIT laboratory.

Origins · 1 minDavid Pritchard and how this was discovered

David Pritchard

1941– · proposed the Ioffe-Pritchard trap and mentored the makers of the first condensates

David E. Pritchard, born in 1941, is the atomic physicist whose name sits on the workhorse magnetic trap of the field. In a 1983 paper on cooling neutral atoms for precision spectroscopy he proposed superimposing a set of current-carrying bars — a geometry Mikhail Ioffe had introduced in 1962 for confining plasmas — onto axial pinch coils, producing a field with a nonzero minimum and thus a harmonic, zero-free potential. The combination is the Ioffe-Pritchard trap, and its central virtue, a $|\mathbf{B}| > 0$ floor that keeps the Larmor precession from ever stalling, is precisely what makes it safe against the Majorana spin flips that plague the simpler quadrupole. Pritchard's influence runs deeper than one clever geometry, though. His MIT group was the training ground for the experimentalists who realised gaseous condensation: Eric Cornell was his doctoral student, and Wolfgang Ketterle was a postdoc in his laboratory before inheriting it and building the apparatus that produced sodium BEC in an optically plugged trap — both would share, with Carl Wieman, the 2001 Nobel Prize in Physics. Beyond cold atoms, Pritchard pioneered atom interferometry, using the wave nature of whole atoms to make precision measurements, and developed single-ion Penning-trap mass spectrometry accurate to parts in $10^{11}$ . The through-line of his career is the same instinct visible in the trap that bears his name: build the cleanest possible environment, remove every avoidable perturbation, and let the atoms reveal what they will.

History

1962

— Ioffe introduces current-bar magnetic confinement for laboratory plasmas

1983

— Pritchard proposes the Ioffe-Pritchard trap for neutral atoms

1985

— Migdall, Prodan and Phillips magnetically trap neutral sodium in a quadrupole field, the first neutral-atom magnetic trap

1995

— Petrich and colleagues introduce the time-orbiting-potential trap to plug the Majorana hole, and the first condensates are produced in magnetic traps

2001

— Cornell, Ketterle and Wieman share the Nobel Prize for gaseous Bose-Einstein condensation

Going further

Texts. Foot, Atomic Physics (2005), Ch. 10 — magnetic trapping, the Ioffe-Pritchard geometry, and Majorana losses · Metcalf and van der Straten, Laser Cooling and Trapping (1999), Ch. 10 · Pethick and Smith, Bose-Einstein Condensation in Dilute Gases (2008), Ch. 4 · Ketterle, Durfee and Stamper-Kurn, Varenna Lectures (1999), on making, probing, and understanding condensates · Demtröder, Laser Spectroscopy, Vol. 2, on the cooling and trapping of atoms.

Provenance. The equations here follow Foot (Ch. 10), Metcalf and van der Straten (Ch. 10), and Pethick and Smith (Ch. 4). A verifier should confirm: (i) the adiabatic potential $U = m_F g_F\mu_B|\mathbf{B}|$ and the trapping condition $m_F g_F > 0$ ; (ii) the quadrupole magnitude $|\mathbf{B}| = B'\sqrt{x^2 + y^2 + 4z^2}$ from $\nabla\cdot\mathbf{B} = 0$ ; (iii) the Ioffe-Pritchard second-order expansion and frequencies $\omega_r, \omega_z$ ; (iv) $\mu_B = 9.274\times10^{-24}\ \mathrm{J\,T^{-1}}$ and the constant $\mu_B/h = 1.40\ \mathrm{MHz/G}$ ; (v) the $T^{-2}$ Majorana loss scaling. The worked numbers are independently recomputed: the $^{87}$ Rb levitation gradient $B' = mg/\mu_B = 15.3\ \mathrm{G/cm}$ ; the Ioffe-Pritchard frequencies $f_r = 191\ \mathrm{Hz}$ , $f_z = 9.0\ \mathrm{Hz}$ for $B' = 1.5\ \mathrm{T/m}$ , $B_0 = 1\ \mathrm{G}$ , $B'' = 50\ \mathrm{T/m^2}$ ; and the $6.7\ \mathrm{mK}$ depth of a $100\ \mathrm{G}$ well. State-specific values ( $^{87}$ Rb $|F=2, m_F=+2\rangle$ , $g_F = +\tfrac12$ ) are standard tabulated data; no g-factors were fabricated.

Builds on: Zeeman effect and magnetic moments · Magneto-Optical Trap · Sub-Doppler and Sisyphus Cooling. Leads to: Evaporative Cooling · Bose-Einstein Condensation in Dilute Gases · Feshbach Resonances.

Ask the tutor: Want to work through the RF-evaporation ramp that takes one of these Ioffe-Pritchard traps from a millikelvin cloud down to a Bose-Einstein condensate?

Why can a magnetic trap only hold weak-field-seeking states and not strong-field seekers?

Because of Earnshaw's theorem applied to magnetostatics. In a current-free region the field obeys $\nabla\cdot\mathbf{B} = 0$ and $\nabla\times\mathbf{B} = 0$ , and one can prove that $|\mathbf{B}|$ may have a local minimum but never a local maximum in free space. A weak-field seeker ( $m_F g_F > 0$ ) has energy that rises with $|\mathbf{B}|$ , so it settles at a minimum — which is allowed. A strong-field seeker would need a field maximum to sit at, and no such maximum exists. This is a strict theorem, not a technological limitation: you cannot engineer a static field maximum in vacuum no matter how you arrange the coils, which is why every static magnetic trap works with minima and weak-field seekers.

What exactly is a Majorana spin flip, and why is it named after Majorana?

As an atom moves through the trap, its spin precesses about the local field and stays aligned with it — that is adiabatic following. Near a point where $|\mathbf{B}| = 0$ the field direction can swing through a large angle in an instant while the Larmor precession rate $\omega_L = \mu_B|\mathbf{B}|/\hbar$ has fallen to zero, so the spin cannot keep up. The atom emerges in a superposition that includes untrapped $m_F$ states, and those are expelled. Ettore Majorana solved the underlying problem — the probability that a spin follows a rapidly reorienting field — in 1932, and the non-adiabatic transition between magnetic sublevels near a field zero bears his name.

If a quadrupole trap is the tightest, why not just use a colder MOT to keep atoms away from the zero?

Cooling makes the problem worse, not better. In a linear quadrupole the cloud size shrinks as it cools, so the atoms crowd toward the centre — exactly where the field zero is. The loss rate scales as $\Gamma_{\rm Maj}\sim(\hbar/m)(\mu_B B'/k_B T)^2$ , growing as $T^{-2}$ : halve the temperature and you quadruple the leak. This is the specific reason a bare quadrupole cannot be evaporated to Bose-Einstein condensation. The fix is not colder atoms but a plugged geometry — an Ioffe bias, a rotating TOP field, or a repulsive optical plug beam — that removes the accessible zero before the final, coldest cooling stage.

Why does a larger bias field B₀ make the Ioffe-Pritchard trap weaker radially?

Because the transverse field adds in quadrature to the axial bias. Near the centre the total magnitude is $|\mathbf{B}| = \sqrt{B_0^2 + B'^2\rho^2}\approx B_0 + B'^2\rho^2/2B_0$ . The radial stiffness is that $B'^2/2B_0$ coefficient, which is divided by $B_0$ : a strong bias means the field already points almost entirely along the axis, so a small transverse component barely lengthens it, and the radial potential is shallow. Lowering $B_0$ therefore tightens the radial trap, which is the standard compression trick before evaporation. But you cannot lower it indefinitely — as $B_0\to 0$ the floor of the trap drops toward zero and Majorana loss returns, so there is an optimum bias balancing confinement against loss.

How is a magnetic trap loaded in the first place if the atoms start out hot?

It is loaded from a magneto-optical trap (MOT). The MOT and optical molasses first cool atoms to sub-millikelvin temperatures using laser light, then the atoms are optically pumped into a single weak-field-seeking hyperfine state such as $^{87}$ Rb $|F=2, m_F=+2\rangle$ . The cooling lasers are switched off and the magnetic field is ramped up to catch the cloud. The handoff works because the magnetic well is deep enough: a $100\ \mathrm{G}$ field is about $6.7\ \mathrm{mK}$ deep, far deeper than the $\sim 100\ \mu\mathrm{K}$ temperature of the laser-cooled cloud. From that point on no photons are scattered, so the atoms can be cooled much further by evaporation without recoil heating or light shifts.

What is a TOP trap, and how does it differ from the Ioffe-Pritchard solution?

A time-orbiting-potential (TOP) trap solves the quadrupole-zero problem by motion rather than by geometry. A small, uniform bias field is added on top of the quadrupole and rotated at a few kilohertz. This makes the instantaneous field zero orbit in a circle whose radius is chosen larger than the cloud, so no atom ever sits at the zero long enough to flip: the rotation is fast compared with the atomic motion but slow compared with the Larmor frequency. The time-averaged potential is a smooth harmonic bowl. The Ioffe-Pritchard trap instead removes the zero statically, with a permanent bias field. Both give Majorana-safe harmonic traps; TOP traps produced the very first rubidium condensate, while Ioffe-Pritchard geometries dominate today for their static simplicity.

Does the atom really feel a force from the field magnitude and not from the field itself?

Yes, and that is the subtle heart of the mechanism. A charged particle feels the Lorentz force from $\mathbf{B}$ directly, but a neutral atom is unaffected by that — it feels only the gradient of its internal energy. Once adiabatic following locks the atom into a fixed $m_F$ manifold, that internal energy is $U = m_F g_F\mu_B|\mathbf{B}|$ , a function of the scalar magnitude alone. The force $\mathbf{F} = -\nabla U = -m_F g_F\mu_B\nabla|\mathbf{B}|$ therefore points along the gradient of $|\mathbf{B}|$ , toward weaker or stronger field depending on the sign of $m_F g_F$ . The direction the field points does not enter the force at all — only how its strength varies in space, which is why the potential landscape is a literal map of $|\mathbf{B}(\mathbf{r})|$ .