Hyperfine Structure

Hyperfine structure is the splitting and mixing of atomic levels caused by electromagnetic interactions between the nucleus and the electrons. The leading terms couple the nuclear magnetic dipole moment and electric quadrupole moment to electronic fields at the nucleus. They resolve a fine-structure level of electronic angular momentum $J$ into levels labeled by the total atomic angular momentum

$$\mathbf F=\mathbf I+\mathbf J,$$

where $I$ is the nuclear spin.

The word hyperfine denotes a layer of structure finer than electronic fine structure. It is not another name for spin–orbit coupling: fine structure couples electronic degrees of freedom and is present even for a spinless nucleus, while hyperfine structure depends on the isotope’s nuclear spin and electromagnetic moments.

Required background. Hydrogenic states and alkali atoms supply the electronic levels. Total angular momentum and magnetic moments supply the coupling and sign conventions.

The formulas below first fix an electronic level $|\gamma J\rangle$ and resolve its nuclear-spin coupling. This is justified when hyperfine matrix elements are small compared with neighboring electronic separations. Otherwise off-diagonal hyperfine interactions mix different $J$ levels and must also be included.

Nuclear Spin and Electromagnetic Moments

The nucleus has angular momentum operators

$$\begin{aligned} I^2|I m_I\rangle &= \hbar^2I(I+1)|I m_I\rangle, \\ I_z|I m_I\rangle &= \hbar m_I|I m_I\rangle. \end{aligned}$$

Its magnetic dipole moment is conventionally written

$$\boldsymbol\mu_I = g_I\mu_N\frac{\mathbf I}{\hbar}, \qquad \mu_N = \frac{e\hbar}{2m_p},$$

where $g_I$ is the nuclear $g$ factor and $\mu_N$ is the nuclear magneton. Nuclear magnetic moments are typically smaller than electronic moments because $\mu_N/\mu_B=m_e/m_p$.

The magnetic moment need not be parallel to $\mathbf I$ with a positive coefficient; the sign is carried by $g_I$. Isotopes of the same element can have different $I$, $g_I$, and quadrupole moment, so their hyperfine spectra can differ even when their electronic configurations are nearly identical.

Electric quadrupole moment

A nucleus with $I\ge1$ may possess a nonspherical electric charge distribution characterized by a spectroscopic quadrupole moment $Q$. A spin-$0$ or spin-$1/2$ nucleus cannot carry a permanent rank-2 moment in a stationary angular-momentum eigenstate.

The measured electric-quadrupole hyperfine constant is not $Q$ alone. It is proportional to the product of the nuclear quadrupole moment and the electronic electric-field gradient at the nucleus. Extracting one from the other requires either independent nuclear information or an atomic-structure calculation.

Higher nuclear multipoles, such as a magnetic octupole moment, can produce still smaller terms. They matter in selected precision studies but are outside the standard $A$–$B$ model developed here.

Coupling I and J

Because $\mathbf I$ and $\mathbf J$ act on nuclear and electronic factors, the uncoupled basis is

$$|I m_I\rangle\otimes|\gamma J m_J\rangle.$$

For a rotationally invariant zero-field hyperfine Hamiltonian, the coupled basis is more useful:

$$|\gamma;I J;F m_F\rangle,$$

with

$$F=|I-J|,|I-J|+1,\ldots,I+J$$

and

$$m_F=-F,-F+1,\ldots,F.$$

The state transformation is a Clebsch–Gordan expansion,

$$|I J;F m_F\rangle = \sum_{m_I,m_J} C^{F m_F}_{I m_I,J m_J} \times |I m_I\rangle |J m_J\rangle.$$

The dimensions agree:

$$(2I+1)(2J+1) = \sum_{F=|I-J|}^{I+J}(2F+1).$$

At zero external field, rotational symmetry makes all $2F+1$ values of $m_F$ degenerate. Hyperfine coupling correlates nuclear and electronic projections; it does not in general preserve $m_I$ and $m_J$ separately.

For $I=3/2$, $J=1/2$, and $A>0$, the fivefold $F=2$ level lies at $3A/4$ and the threefold $F=1$ level at $-5A/4$. Their weighted centroid is zero, although their midpoint is not. Generally $E_+-E_-=A(I+1/2)$; negative $A$ reverses the order, and the transition energy is the absolute difference.

Magnetic-Dipole Hyperfine Interaction

Inside an isolated fine-structure level, rotational symmetry makes the leading magnetic-dipole interaction a scalar. With $A$ defined as an energy,

$$H_{\mathrm{M1}} = \frac{A}{\hbar^2} \mathbf I\cdot\mathbf J.$$

The angular identity

$$2\mathbf I\cdot\mathbf J = F^2-I^2-J^2$$

immediately gives

$$E_F^{\mathrm{M1}} = \frac{A}{2}K,$$

where

$$K \equiv F(F+1)-I(I+1)-J(J+1).$$

The constant $A$ includes the nuclear magnetic moment and the electronic magnetic field at the nucleus for the chosen state. It may be positive or negative. The formula predicts angular dependence after $A$ is known; it does not calculate $A$ from first principles.

Hyperfine interval rule

For adjacent $F$ manifolds in the pure magnetic-dipole model,

$$\begin{aligned} E_F-E_{F-1} &= \frac{A}{2} \left[K_F-K_{F-1}\right], \\ &=AF. \end{aligned}$$

Thus adjacent intervals are proportional to the larger $F$. Electric-quadrupole terms, higher multipoles, second-order mixing, and external fields violate this simple interval rule.

The degeneracy-weighted magnetic-dipole centroid is unchanged:

$$\sum_F(2F+1)E_F^{\mathrm{M1}}=0.$$

This trace statement is useful when reconstructing a fine-structure centroid from resolved hyperfine levels. An unweighted average is generally wrong.

Microscopic magnetic contributions

The effective constant $A$ can contain several electronic interactions:

Contribution	Electronic structure	Typical importance
orbital magnetic field	electron orbital current coupled to $\boldsymbol\mu_I$	states with nonzero orbital angular momentum
spin-dipolar field	anisotropic dipole–dipole coupling of electron and nucleus	non-$S$ density and open shells
Fermi contact term	electron spin density at the nucleus	dominant for many $S$ states
relativistic and correlation terms	small components, core polarization, configuration mixing	essential in heavy and many-electron atoms

For a one-electron $S$ state, the leading contact interaction can be written

$$H_{\mathrm F} = \frac{2\mu_0}{3} g_e\mu_Bg_I\mu_N \delta^3(\mathbf r) \frac{\mathbf I\cdot\mathbf S}{\hbar^2},$$

where $g_e$ denotes the positive magnitude convention for the electron spin $g$ factor. Its first-order constant is proportional to

$$A_{nS}^{(0)} \propto g_I\mu_N g_e\mu_B |\psi_{nS}(0)|^2.$$

This explains the strong contact sensitivity of hydrogenic and alkali $S$ states. It does not justify inserting a hydrogenic wavefunction into a heavy neutral atom. Core polarization, exchange, relativistic penetration, finite magnetization, and electronic correlation alter the spin density at the nucleus.

Electric-Quadrupole Hyperfine Interaction

The electric-quadrupole interaction couples rank-2 nuclear and electronic tensors. Schematically,

$$H_{\mathrm{E2}} = \sum_{q=-2}^{2} (-1)^q T_q^{(2)}(\mathrm{nuc}) T_{-q}^{(2)}(\mathrm{el}).$$

Within an isolated $I,J$ manifold, the conventional first-order energy is

$$\begin{aligned} E_F^{\mathrm{E2}} &= B\frac{N_{IJF}}{D_{IJ}}, \\ N_{IJF} &= \frac34K(K+1) -I(I+1)J(J+1), \\ D_{IJ} &= 2I(2I-1)J(2J-1). \end{aligned}$$

The $B$ term exists only when both angular momenta can support the required rank-2 coupling:

$$I\ge1, \qquad J\ge1.$$

For $J=1/2$, the electronic quadrupole expectation vanishes, even if the nucleus has $I\ge1$. This is why alkali $nS_{1/2}$ and $nP_{1/2}$ levels have no first-order electric-quadrupole $B$ term, whereas $nP_{3/2}$ levels can.

The standard dipole-plus-quadrupole model is therefore

$$E_F^{\mathrm{hfs}} = \frac{A}{2}K +E_F^{\mathrm{E2}}.$$

Different databases may tabulate $A$ and $B$ in frequency units, writing $A/h$ and $B/h$ implicitly. Always inspect the convention before substituting a number into an energy formula.

Hydrogen Ground-State Hyperfine Splitting

The hydrogen $1S_{1/2}$ ground state has

$$I=\frac12, \qquad J=\frac12,$$

so the allowed total angular momenta are $F=0$ and $F=1$. For the pure dipole Hamiltonian,

$$\begin{aligned} E_{F=1} &=\frac{A}{4}, \\ E_{F=0} &=-\frac{3A}{4}. \end{aligned}$$

Their difference is

$$\Delta E_{1S}^{\mathrm{hfs}}=A.$$

For ordinary hydrogen, $A>0$, so the $F=1$ triplet lies above the $F=0$ singlet. The evaluated transition frequency quoted by NIST is

$$\nu_{21} = 1420.405\,751\,766\,7(10) \ \mathrm{MHz}.$$

Using $\lambda=c/\nu$ gives approximately

$$\lambda_{21}\simeq21.106\ \mathrm{cm},$$

the celebrated 21 cm line. It is a weak magnetic-dipole transition, but the vast amount of neutral hydrogen in astrophysical environments makes it an exceptionally important probe.

Beyond the Fermi result

The contact model gives the leading scale, not the final prediction. Precision hydrogen theory includes:

• the electron anomalous magnetic moment and other QED corrections;
• relativistic and recoil terms;
• finite proton magnetization and charge distributions;
• the Zemach correction, which couples charge and magnetic form factors;
• proton polarizability and other two-photon-exchange effects;
• weak and hadronic vacuum-polarization contributions at much smaller scales.

The measured interval is extraordinarily precise, while theoretical interpretation is limited more strongly by proton structure. Hyperfine spectroscopy therefore probes a different nuclear combination from the Lamb shift: the Lamb shift is primarily sensitive to charge-radius effects, whereas hyperfine structure is especially sensitive to magnetic structure and spin-dependent two-photon exchange.

The words singlet and triplet describe the coupling of electron and proton spins here. The particles are distinguishable, so this terminology does not impose the exchange-symmetry constraints that apply to two identical electrons.

Alkali Ground-State Doublets

An alkali ground state is approximately $nS_{1/2}$, so $J=1/2$. If $I\ne0$, it splits into

$$F_-=I-\frac12, \qquad F_+=I+\frac12.$$

The dipole energies are

$$\begin{aligned} E_{F_+} &=\frac{AI}{2}, \\ E_{F_-} &=-\frac{A(I+1)}{2}, \end{aligned}$$

so the signed difference $E_{F_+}-E_{F_-}$ is

$$\Delta E_{\mathrm{hfs}} = A\left(I+\frac12\right).$$

Representative microwave scales are:

Isotope and state	Nuclear spin	Hyperfine levels	Ground interval
$^1$H $1S_{1/2}$	$1/2$	$F=0,1$	$1.4204057517667\ \mathrm{GHz}$
$^{85}$Rb $5S_{1/2}$	$5/2$	$F=2,3$	about $3.036\ \mathrm{GHz}$
$^{87}$Rb $5S_{1/2}$	$3/2$	$F=1,2$	about $6.835\ \mathrm{GHz}$
$^{133}$Cs $6S_{1/2}$	$7/2$	$F=3,4$	exactly $9.192631770\ \mathrm{GHz}$ as an SI defining frequency

The rubidium entries are rounded orientation values. For numerical hyperfine constants, use isotope- and state-specific measurements compiled by Allegrini, Arimondo, and Orozco or the convention-consistent data sheets by Steck; a general electronic-level database need not resolve hyperfine components.

Excited alkali levels

For $nP_{1/2}$, $J=1/2$ again eliminates the first-order $B$ term. For $nP_{3/2}$, $J=3/2$ permits electric-quadrupole structure when $I\ge1$. Optical $D$ lines therefore split into multiple $F\to F'$ components with isotope-dependent spacings and line strengths.

The contact interaction is no longer the entire story in excited or non-$S$ states. Orbital, spin-dipolar, relativistic, and many-body contributions can dominate. Hyperfine constants are stringent tests of electronic wavefunctions near the nucleus and of core-polarization calculations.

Hyperfine anomalies

For identical electronic structure and pointlike nuclei, $A$ is proportional to $g_I=\mu_I/(I\mu_N)$, not simply to the tabulated magnetic moment $\mu_I$ when isotopes have different spins. Finite nuclear magnetization and charge distributions produce departures from the $g_I$ ratio known as hyperfine anomalies; ordinary mass-dependent electronic changes must also be controlled. They are physical isotope-dependent corrections, not necessarily inconsistent measurements.

Hyperfine Transitions

Within one electronic state, transitions between hyperfine manifolds are typically driven by microwave magnetic fields. In the weak-field coupled basis, the leading magnetic-dipole rules are

$$\Delta F=0,\pm1, \qquad F=0\not\leftrightarrow F'=0,$$

with

$$\Delta m_F=0,\pm1$$

selected by the microwave polarization relative to the quantization axis. These are angular selection rules, not guarantees of equal line strength.

For optical electric-dipole transitions between hyperfine-resolved electronic levels, the same $F$ and $m_F$ rules appear after the electronic reduced matrix element is recoupled to nuclear spin. Atomic Selection Rules gives the operational $6j$ recoupling and polarization rules; their general derivation belongs to the Wigner–Eckart Theorem and Selection Rules pages.

Hyperfine transitions can be exceptionally narrow because the two states share the same electronic configuration and the transition is not an allowed optical electric-dipole decay. Narrowness alone does not make a transition an ideal clock: it must also be preparable, interrogable, reproducible, and insensitive to environmental perturbations.

Hyperfine Transitions in Clocks

For $^{133}$Cs, $I=7/2$ and the $6S_{1/2}$ ground state has $F=3$ and $F=4$. The microwave clock transition uses the states

$$|F=3,m_F=0\rangle \longleftrightarrow |F=4,m_F=0\rangle.$$

At weak magnetic field, a hyperfine Zeeman sublevel has the leading shift

$$\Delta E_Z^{(1)} \simeq g_F\mu_Bm_FB.$$

Each clock state has $m_F=0$, so its diagonal first-order shift vanishes. This makes the transition first-order insensitive near zero field. It is not magnetically insensitive to all orders: the field mixes the two $m_F=0$ hyperfine states and produces a quadratic Zeeman shift.

The SI second

The SI second is defined by fixing the numerical value of the unperturbed $^{133}$Cs ground-state hyperfine transition frequency:

$$\Delta\nu_{\mathrm{Cs}} = 9\,192\,631\,770\ \mathrm{Hz}$$

exactly. The word unperturbed is crucial. A laboratory clock measures atoms in finite fields, at finite temperature, with collisions, motion, microwave interrogation, and gravitational potential. A primary frequency standard evaluates and corrects its realization toward the defining ideal.

For a cesium standard, the main corrections include quadratic Zeeman, blackbody Stark, collisional, and microwave interrogation shifts. Comparing geographically separated clocks also requires gravitational-redshift corrections. Atomic Clocks explains how interrogation and the apparatus determine the measured frequency.

Magnetic-Field Regimes

In a static axially symmetric magnetic field, $m_F$ remains exact while $F$ is an approximate label when the Zeeman interaction is weak compared with the hyperfine interval. A practical comparison is

$$g_J\mu_BB \ll \Delta E_{\mathrm{hfs}}.$$

In an intermediate field, states with the same conserved $m_F=m_I+m_J$ mix. One should diagonalize

$$H = H_{\mathrm{hfs}} -\boldsymbol\mu_J\cdot\mathbf B -\boldsymbol\mu_I\cdot\mathbf B$$

in fixed-$m_F$ blocks. In the strong-field hyperfine Paschen–Back regime, the uncoupled labels $m_I,m_J$ become more useful. Zeeman Effect in Atoms derives the Breit–Rabi crossover and polarization-resolved spectra; Zeeman Effect as a Perturbation Example develops the general perturbative field logic.

Avoid assigning a single field-independent $g_F$ through this crossover. The eigenvectors change with $B$, and avoided crossings occur between states with the same exact symmetries.

Extracting hyperfine constants

A $J=1/2$ doublet determines the magnitude of $A$ from one interval; assigning the $F$ levels also determines its sign. When both $A$ and $B$ contribute, fit several levels with the complete formula. Reconstruct the electronic centroid with degeneracy weights, rather than averaging observed optical peak positions: peaks can share different upper and lower endpoints and have population-dependent intensities.

The first-order $A$–$B$ model presumes an isolated $J$ level. Nearby electronic levels produce second-order shifts through off-diagonal hyperfine matrix elements, so precision fits must distinguish those shifts from effective fitted constants. Fields, optical pumping, collisions, and unresolved Zeeman components can also alter the measured resonance.

Exercises

1. Separate dipole and quadrupole constants

For $I=J=1$, measured signed intervals are $(E_1-E_0)/h=5\;\mathrm{MHz}$ and $(E_2-E_1)/h=46\;\mathrm{MHz}$. Neglect higher-order terms. Find $A/h$ and $B/h$, and verify the level centroid.

Solution

For $F=0,1,2$, the values of $K$ are $-4,-2,2$. The level formula gives

$$E_0=-2A+\frac52B,\quad E_1=-A-\frac54B,\quad E_2=A+\frac14B.$$

Writing $d_1=E_1-E_0$ and $d_2=E_2-E_1$ gives

$$B=\frac{d_2-2d_1}{9},\qquad A=d_1+\frac{15}{4}B.$$

Thus $B/h=4\;\mathrm{MHz}$, $A/h=20\;\mathrm{MHz}$, and the three levels are $E_F/h=(-30,-25,21)\;\mathrm{MHz}$. Their weighted sum is $1(-30)+3(-25)+5(21)=0$. Neither measured gap alone equals $A/h$.

2. Derive the J = 1/2 doublet

For general nuclear spin $I\ge1/2$ and $J=1/2$, use $E_F=AK/2$ to derive the two energies, their separation, and their degeneracy-weighted centroid.

Solution

For $F_+=I+1/2$,

$$K_+=I, \qquad E_+=\frac{AI}{2}.$$

For $F_-=I-1/2$,

$$K_-=-(I+1), \qquad E_-=-\frac{A(I+1)}{2}.$$

Therefore

$$E_+-E_- = A\left(I+\frac12\right).$$

The degeneracies are $2I+2$ and $2I$. Their weighted sum is

$$\begin{aligned} (2I+2)E_+ +(2I)E_- &=AI(I+1)-AI(I+1), \\ &=0. \end{aligned}$$

Thus the pure dipole interaction splits the level about its degeneracy-weighted electronic centroid.

3. Diagnose a quadrupole term

Decide whether a first-order electric-quadrupole hyperfine constant $B$ can occur for each pair: (a) $I=1/2,J=3/2$; (b) $I=3/2,J=1/2$; (c) $I=3/2,J=3/2$; (d) $I=0,J=2$. Explain each answer without evaluating the energy formula.

Solution

A diagonal rank-2 coupling requires both $I\ge1$ and $J\ge1$.

Case	$B$ term?	Reason
(a)	no	the spin-$1/2$ nucleus has no permanent quadrupole moment
(b)	no	a $J=1/2$ electronic manifold cannot support the required rank-2 expectation
(c)	yes	both angular momenta are at least one
(d)	no	an $I=0$ nucleus has neither magnetic-dipole nor electric-quadrupole orientation

Hyperfine mixing with other $J$ levels can produce higher-order effects even when a first-order diagonal $B$ constant vanishes, but that does not change this selection rule.

4. Recover the 21 cm wavelength

Using $c=2.99792458\times10^8\ \mathrm{m\,s^{-1}}$ and $\nu=1.4204057518\times10^9\ \mathrm{s^{-1}}$, compute the vacuum wavelength of the hydrogen ground-state hyperfine transition.

Solution

The wavelength is

$$\begin{aligned} \lambda &=\frac{c}{\nu}, \\ &\simeq \frac{2.99792458\times10^8} {1.4204057518\times10^9} \ \mathrm m, \\ &\simeq0.21106\ \mathrm m. \end{aligned}$$

Thus

$$\lambda\simeq21.106\ \mathrm{cm}.$$

The nickname “21 cm line” is a rounded wavelength label; precision work uses the transition frequency and its stated convention.

5. Why clock states still need a bias field

The caesium clock transition uses $m_F=0$ states, for which the first-order Zeeman formula vanishes. Why is a small bias magnetic field still useful, and why must its magnitude be measured?

Solution

A bias field defines a quantization axis and separates the $m_F\ne0$ Zeeman components. That makes state preparation, polarization control, and identification of the clock resonance more robust.

The field cannot be ignored because it mixes the two $m_F=0$ hyperfine states and produces a quadratic Zeeman shift. Spatial inhomogeneity also makes different atomic trajectories sample different shifts. A primary standard therefore measures the field, often using field-sensitive Zeeman transitions, and corrects the clock frequency to the unperturbed limit with an uncertainty.

References

• M. Allegrini, E. Arimondo, and L. A. Orozco, “Survey of Hyperfine Structure Measurements in Alkali Atoms,” Journal of Physical and Chemical Reference Data 51, 043102 (2022), doi:10.1063/5.0098061.
• E. Arimondo, M. Inguscio, and P. Violino, “Experimental Determinations of the Hyperfine Structure in the Alkali Atoms,” Reviews of Modern Physics 49, 31–75 (1977), doi:10.1103/RevModPhys.49.31.
• G. Breit and I. I. Rabi, “Measurement of Nuclear Spin,” Physical Review 38, 2082–2083 (1931), doi:10.1103/PhysRev.38.2082.2.
• Bureau International des Poids et Mesures, SI Base Unit: Second, defining $\Delta\nu_{\mathrm{Cs}}=9\,192\,631\,770\ \mathrm{Hz}$ exactly.
• C. E. Carlson, V. Nazaryan, and K. Griffioen, “Proton Structure Corrections to Electronic and Muonic Hydrogen Hyperfine Splitting,” Physical Review A 78, 022517 (2008), doi:10.1103/PhysRevA.78.022517.
• E. Fermi, “Über die magnetischen Momente der Atomkerne,” Zeitschrift für Physik 60, 320–333 (1930), doi:10.1007/BF01339933.
• C. J. Foot, Atomic Physics (Oxford University Press, 2005), Chapters 2, 6, and 7.
• A. Kramida, Y. Ralchenko, J. Reader, and the NIST ASD Team, NIST Atomic Spectra Database, National Institute of Standards and Technology, doi:10.18434/T4W30F.
• National Institute of Standards and Technology, Atomic Spectroscopy: Atomic States, Shells, and Configurations, including the evaluated $^1$H ground-state hyperfine interval.
• I. I. Sobelman, Atomic Spectra and Radiative Transitions, 2nd ed. (Springer, 1992), doi:10.1007/978-3-642-76907-8.
• D. A. Steck, Alkali D Line Data, isotope-resolved reference sheets with constants, conventions, and source histories.
• G. K. Woodgate, Elementary Atomic Structure, 2nd ed. (Oxford University Press, 1980), Chapters 9–12.