Atomic, Molecular & Optical Physics
Feshbach Resonance: Tuning Atomic Interactions with a Magnetic Field
Turn a knob on a magnet and you can make two atoms attract, repel, or ignore each other entirely — and even bind them into a fragile molecule 1,000 times larger than an ordinary one. That is what a Feshbach resonance delivers: near a magnetic field of order 100–1000 gauss, the s-wave scattering length a that governs low-energy collisions sweeps continuously from −∞ through zero to +∞, so the effective interaction between ultracold atoms becomes a tunable experimental parameter.
Physically, a Feshbach resonance occurs when the collision energy of two free atoms in an "open" channel is tuned into degeneracy with a bound molecular state in an energetically "closed" channel. Because the two channels have different magnetic moments, a magnetic field Zeeman-shifts them relative to each other, and near the crossing the coupling between them dramatically rewrites the scattering. It is the single most important control knob in cold-atom physics.
- RegimeUltracold collisions, T ~ nK–µK, s-wave threshold
- Key relationa(B) = a_bg [1 − Δ/(B − B₀)]
- Proposed / observedTiesinga–Verhaar–Stoof 1993; observed Inouye/Ketterle 1998
- Characteristic scaleB₀ ~ 100–1000 G; widths Δ ~ mG to ~300 G
- Realized in²³Na, ⁶Li (834 G), ⁴⁰K, ⁸⁷Rb, ¹³³Cs, and mixtures
- Matters forBEC-BCS crossover, unitary Fermi gas, ultracold molecules, Efimov states
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What a Feshbach Resonance Is and Why It Matters
At temperatures near absolute zero, two colliding atoms interact almost entirely through the lowest partial wave (s-wave, ℓ = 0), and their entire low-energy scattering is captured by one number: the scattering length a. The cross section is σ = 4πa² (or 8πa² for identical bosons), and a sets the sign and strength of the mean-field interaction, g = 4πℏ²a/m. A Feshbach resonance is a scattering resonance in which a — and hence the effective interaction — can be tuned to any value, including ±∞, by an external magnetic field.
This turns interaction strength into a dial. Attractive (a < 0) or repulsive (a > 0) gases, non-interacting ideal gases (a = 0), and the strongly-correlated "unitary" regime (|a| → ∞) all become accessible in one apparatus by sweeping B. That control underlies the experimental study of the BEC-BCS crossover, the unitary Fermi gas, bright matter-wave solitons, Efimov trimers, and the bottom-up assembly of ultracold molecules. No other system lets you change a fundamental coupling constant continuously in real time.
The Mechanism: Open and Closed Channels
Two atoms approach along a molecular potential set by their spin state — this is the open channel, energetically accessible at the collision energy E ≈ 0. A different spin configuration defines a closed channel whose potential asymptote sits above E, so free atoms cannot escape into it, but it can support a molecular bound state. The essential physics: if that closed-channel bound state lies near the open-channel threshold, it hybridizes with the scattering continuum and resonantly distorts it.
The two channels have different total magnetic moments, so their relative energy shifts with field as ΔE(B) ≈ δµ·(B − B₀), where δµ is the differential magnetic moment (typically a fraction of a Bohr magneton µ_B). Tuning B brings the bound state through the threshold. The coupling that mixes open and closed channels is the atomic hyperfine interaction plus exchange — the same off-diagonal terms Herman Feshbach treated with projection operators (P for open, Q for closed) in his 1958 unified theory of nuclear reactions. As the bound state crosses threshold, the scattering phase shift jumps by π, and a diverges.
The Key Equation, Widths and Scales
Near an isolated resonance the scattering length follows the dispersive form
a(B) = a_bg [ 1 − Δ / (B − B₀) ],
where a_bg is the background (off-resonant) scattering length, B₀ the resonance center, and Δ the magnetic width — the field separation between the pole (a → ±∞ at B₀) and the zero crossing (a = 0 at B₀ + Δ). On the a > 0 side a weakly-bound "Feshbach molecule" exists with universal binding energy E_b = ℏ²/(m a²) for a much larger than the van der Waals length; this halo dimer's size ∼ a/2 can reach hundreds of nm.
Scales span orders of magnitude: ⁶Li has a very broad resonance at B₀ ≈ 834 G with a_bg ≈ −1405 a₀ and Δ ≈ 300 G, whereas many resonances are narrow (Δ ∼ mG–G). Resonances are classed by the dimensionless strength s_res = a_bg·δµ·Δ / (ā·Ē) with ā the mean scattering length; broad (entrance-channel-dominated, s_res ≫ 1) resonances give universal, single-channel physics, narrow ones do not.
How It Is Realized and Measured
The first magnetically-tuned Feshbach resonance was observed by Inouye, Ketterle and co-workers in 1998 in an optically-trapped ²³Na Bose-Einstein condensate (Nature 392, 151), which revealed resonances near 853 G and 907 G through a factor-of-ten dispersive change in a and a sharp enhancement of inelastic trap loss. The idea of using such resonances to tune interactions had been proposed by Tiesinga, Verhaar and Stoof in 1993 (Phys. Rev. A 47, 4114).
Experimental signatures are direct. (1) Loss spectroscopy: three-body recombination rate scales roughly as a⁴, so atom number plummets near B₀, mapping resonance positions. (2) Cross-dimensional or thermalization measurements extract |a|. (3) Molecule association: a slow adiabatic ramp of B across B₀ from the a < 0 to the a > 0 side converts pairs of atoms into weakly-bound dimers, detected by RF or optical spectroscopy of the binding energy E_b(B). (4) Interferometric and RF-association measurements pin down E_b and thus a to high precision. The comprehensive reference is Chin, Grimm, Julienne and Tiesinga, Rev. Mod. Phys. 82, 1225 (2010).
Where It Operates and Distinctions From Related Effects
Magnetic Feshbach resonances have been mapped in essentially every laser-coolable species with nonzero nuclear/electronic spin — ²³Na, ⁶Li, ⁷Li, ⁴⁰K, ⁸⁵Rb, ⁸⁷Rb, ¹³³Cs, ¹⁶⁸Er, ¹⁶²Dy — and in heteronuclear mixtures used to build polar molecules (e.g. ⁴⁰K–⁸⁷Rb, ²³Na–⁶Li). Fermionic ⁶Li and ⁴⁰K, with their broad resonances, are the workhorses of strongly-interacting Fermi-gas physics.
It should be distinguished from a shape resonance, a single-channel quasi-bound state trapped behind a centrifugal barrier (not magnetically tunable). Its atomic-physics ancestor is the Fano resonance — Ugo Fano's continuum–discrete interference (autoionization, 1935/1961) — the same mathematics viewed from the interference side; the combined name Fano-Feshbach is common. An optical Feshbach resonance uses a laser-coupled excited molecular level instead of a magnetically-shifted ground state, at the cost of spontaneous-emission loss, and is valuable for species lacking usable magnetic resonances. Feshbach resonances also exist in higher partial waves (p-wave), which are dipolar-anisotropic and lossier.
Applications, Open Questions and Significance
The tunability enabled landmark results: the BEC-BCS crossover, where a two-spin Fermi gas is smoothly tuned from a molecular BEC (a > 0) through the unitary regime (|a| → ∞) to BCS-paired superfluidity (a < 0), realized in ⁶Li and ⁴⁰K around 2004. At unitarity the only length is the interparticle spacing, so thermodynamics becomes universal — governed by the Bertsch parameter ξ ≈ 0.37 — making cold atoms a clean analog of neutron-star matter and quark-gluon plasma. Feshbach resonances also produce the giant three-body Efimov states, whose log-periodic scaling (factor ≈ 22.7 in a) was first seen via loss resonances in ¹³³Cs (Grimm group, 2006).
They are the gateway to ultracold ground-state polar molecules (magnetoassociation then STIRAP) for quantum chemistry and quantum simulation. Open frontiers include controlling losses near broad resonances, resonances in dipolar lanthanides where hundreds of overlapping resonances appear "chaotic," atom-ion Feshbach resonances (observed 2021), and RF- or laser-dressed "synthetic" resonances. Few tools have reshaped a field as thoroughly.
| Feature | Feshbach (magnetic) resonance | Shape resonance | Optical Feshbach resonance |
|---|---|---|---|
| Bound state involved | Closed-channel molecular state below threshold | Quasi-bound state behind a centrifugal/potential barrier in the same channel | Excited-state molecular level reached by a laser photon |
| Tuning knob | Static magnetic field B (via differential Zeeman shift δµ) | Fixed by the potential; not tunable in situ | Laser frequency and intensity |
| Coupling origin | Hyperfine + exchange coupling between channels | Single-channel barrier tunneling | Photoassociation dipole coupling |
| Loss / heating | Low (dominant near broad resonances); strong 3-body loss at unitarity | Set by barrier lifetime | High — spontaneous emission from excited state |
| Where used | BEC-BCS crossover, tunable a, molecule assembly | Cold-collision spectroscopy, cross-section peaks | Species with no usable magnetic resonance (e.g. some alkaline earths) |
Frequently asked questions
Why does the scattering length diverge exactly at the resonance?
As the magnetic field brings the closed-channel molecular bound state into degeneracy with the open-channel threshold (E → 0), the s-wave phase shift jumps by π. In effective-range terms, the divergence corresponds to the bound state sitting precisely at zero energy: a → ±∞. Just below resonance (a > 0) that state is a real weakly-bound molecule; just above (a < 0) it becomes a virtual state. The sign flip across B₀ is why interactions switch from effectively repulsive to attractive.
What is the difference between a broad and a narrow Feshbach resonance?
It is quantified by the resonance strength parameter s_res ∝ a_bg·δµ·Δ. Broad (entrance-channel-dominated, s_res ≫ 1) resonances have a wide field width and their near-threshold physics is universal — the closed-channel fraction stays tiny and single-channel effective-range theory applies. Narrow (closed-channel-dominated) resonances have large closed-channel admixture, an energy-dependent scattering length, and multichannel physics that cannot be reduced to a alone. ⁶Li's 834 G resonance is a classic broad one.
How is a Feshbach molecule made and how big is it?
Starting on the a < 0 side with free atom pairs, you ramp the magnetic field adiabatically across B₀ onto the a > 0 side; the pair is converted into the weakly-bound dimer that emerges below threshold. Its binding energy is universal, E_b = ℏ²/(m a²), and its size scales as a/2. Since a can reach thousands of Bohr radii, these "halo" dimers can be hundreds of nanometers across — enormous compared with a chemical bond (~0.1 nm).
Who discovered the Feshbach resonance, and why is it named after Feshbach?
Herman Feshbach developed the projection-operator formalism (open/closed channels via P and Q operators) in his 1958 "Unified theory of nuclear reactions," originally for compound-nucleus scattering. Ugo Fano gave the equivalent continuum–discrete interference picture in atomic physics (1935/1961), so the name Fano-Feshbach is also used. The application to magnetically tuning ultracold atomic interactions was proposed by Tiesinga, Verhaar and Stoof in 1993 and first observed by the Ketterle group in 1998.
Why do atoms get lost near a Feshbach resonance?
The dominant loss is three-body recombination: two atoms form a deeply-bound molecule while a third carries off the released binding energy, and all three usually leave the trap. The rate scales roughly as a⁴, so it grows explosively as |a| increases toward resonance. This a⁴ scaling both limits how close to unitarity one can hold a dense gas and provides the loss-spectroscopy signature used to locate resonances. Fermi gases are more stable because Pauli suppression reduces three-body events.
How does the Feshbach resonance enable the BEC-BCS crossover?
In a two-spin-state Fermi gas you tune B across a broad resonance. On the a > 0 side, atoms pair into bosonic Feshbach molecules that Bose-condense (BEC). On the a < 0 side they form long-range BCS Cooper pairs, a fermionic superfluid described by BCS theory. At unitarity (|a| → ∞) the system is strongly-interacting and scale-invariant, with universal thermodynamics set by the Bertsch parameter ξ ≈ 0.37. The resonance smoothly connects these limits — the paradigm of tunable strong correlations.