Hyperfine Structure

Hyperfine structure is the physics of the atomic clock. Caesium-133's ground-state hyperfine transition was defined to be $9\,192\,631\,770$ Hz in 1967, making the second — and through it the metre and the volt — a hyperfine measurement; the interval rule and the Breit–Rabi field-independent transition are what make it reproducible. Rubidium and hydrogen masers run on the same principle, and optical-lattice clocks now push hyperfine and related transitions to the $10^{-18}$ level. Beyond timekeeping: NMR and MRI read the hyperfine (contact and dipolar) coupling of nuclear spins to their electronic environment, turning chemical structure into a spectrum. Electron-spin resonance maps unpaired-electron systems the same way. In quantum information, hyperfine states of trapped ions and neutral atoms are long-lived qubits precisely because the clock transition is field-insensitive. And in astrophysics the hydrogen $21$ cm line — the singlet–triplet transition computed above — is how we weigh neutral hydrogen across the universe, the subject of the next article.

Fine structure is $O(\alpha^2)$ relativistic/spin-orbit; hyperfine adds the nuclear moment, weaker by $\mu_N/\mu_B=m_e/m_p$. The suppression is a mass ratio, not another power of α.

Because $\hat H_{hf}\propto\hat{\mathbf I}\cdot\hat{\mathbf J}$, only the total $\mathbf F=\mathbf I+\mathbf J$ commutes with the Hamiltonian; $m_I$ and $m_J$ separately do not.

The contact term is $\propto\delta^3(\mathbf r)$, so its expectation is $\propto|\psi(0)|^2$. Only $\ell=0$ states reach the origin; for $\ell>0$ the centrifugal barrier forces $\psi(0)=0$.

From $E_F=\tfrac A2[F(F+1)-I(I+1)-J(J+1)]$, subtracting adjacent levels leaves $A\,F$ — the spacing grows linearly with the upper $F$, so spacing ratios count the $F$ values.

The $m_F=0$ Breit–Rabi curves are flat at $B=0$ (their separation has zero first derivative in $B$), so the clock transition is field-independent to first order — the reason it defines the second.

An electron orbiting a proton is, from the proton's point of view, a ring of current and a nearby bar magnet (the electron's own spin). It bathes the nucleus in a magnetic field. The proton is itself a magnet — a spin-$\tfrac12$ particle with a magnetic moment — and a magnet in a field has an orientation energy. Flip the proton's moment relative to the electron's field and the atom's energy changes. That energy shift, resolved by orientation, is hyperfine structure. Two things make it tiny and make it special. Tiny: the nuclear magneton $\mu_N=e\hbar/2m_p$ is smaller than the electron's Bohr magneton $\mu_B=e\hbar/2m_e$ by the mass ratio $m_e/m_p\approx1/1836$. A nuclear magnet is nearly two thousand times weaker than an electronic one, which is why hyperfine sits a factor $\sim g_p m_e/m_p\approx1/300$ below the fine structure of Fine Structure of Hydrogen — that article names hyperfine explicitly as "suppressed by $m_e/m_p$" one rung down its $\alpha$-pyramid. Special: for an s-state, the electron's orbital angular momentum is zero, so there is no orbital current and no dipole field at a distance. What is left is the electron's spin sitting literally on top of the point nucleus — the electron cloud has nonzero density at the origin. This is the Fermi contact interaction, a delta-function coupling felt only by states that reach the nucleus. It shares its mathematical skeleton — a $\delta^3(\mathbf{r})$ acting only on s-states — with the Darwin term of Darwin Term, which likewise "touches only s-states, acts at exactly one point." Same structure; different magnet. The Darwin term's magnet is the vacuum; here it is the proton.

The coupling knits nuclear spin into the atom's angular-momentum bookkeeping. Before the nucleus mattered, the good number was $\mathbf{J}=\mathbf{L}+\mathbf{S}$. Now $\hat{\mathbf{I}}\cdot\hat{\mathbf{J}}$ is the interaction, so neither $m_I$ nor $m_J$ is conserved separately — only their sum. The total $\mathbf{F}=\mathbf{I}+\mathbf{J}$ is the new good quantum number, exactly the angular-momentum-addition problem of Coupled and Uncoupled Bases (which forward-names "the hyperfine contact term behind the 21 cm line" as one place its coupling machinery is spent), with $\mathbf{I}$ playing the role of a second spin.

The nuclear magnetic moment is written in nuclear magnetons, the natural unit set one mass-ratio below the electron's, from the moment dictionary of Gyromagnetic Ratio and the Magnetomechanical Effect:

where: $g_I$ is the nuclear $g$-factor (for the bare proton $g_p\approx5.586$), $\mu_N$ the nuclear magneton, and the ratio $\mu_N/\mu_B=m_e/m_p\approx1/1836$ is the entire reason hyperfine structure is small. A proton's magnet is feeble compared to an electron's; the interaction it produces is correspondingly feeble. The moment sits in the magnetic field $\mathbf{B}_e$ that the electron creates at the nucleus, giving orientation energy $\hat{H}_{hf}=-\hat{\boldsymbol{\mu}}_I\cdot\mathbf{B}_e$. Because $\mathbf{B}_e$ is built from the electron's own angular momentum $\hat{\mathbf{J}}$, the interaction reduces — after the angular averaging that any central-field state permits — to a single scalar coupling:

where: $A$ is the hyperfine constant, an energy that packages the strengths $g_I$, the electron's field, and the radial integrals into one number per level. Everything orientation-dependent lives in the operator $\hat{\mathbf{I}}\cdot\hat{\mathbf{J}}$. The magnitude of $A$ waits for the contact calculation below; the pattern of splitting follows from $\hat{\mathbf{I}}\cdot\hat{\mathbf{J}}$ alone. Because $\hat{H}_{hf}\propto\hat{\mathbf{I}}\cdot\hat{\mathbf{J}}$, it commutes with $\hat{F}^2$ and $\hat{F}_z$ but not with $\hat{I}_z$ or $\hat{J}_z$ separately. The label that survives is $F$ — a good quantum number in the exact sense of Good Quantum Numbers and Symmetry-Adapted Bases: the coupled basis $|F,m_F\rangle$ diagonalizes the perturbation with no matrix to invert, because the symmetry does the diagonalization for us. This is $\mathbf{I}\otimes\mathbf{J}$ angular-momentum addition, one more instance of the machinery of Coupled and Uncoupled Bases.

For a state with orbital angular momentum $\ell>0$ the electron avoids the nucleus (the centrifugal barrier of Hydrogenic Wavefunctions and Quantum Numbers keeps $\psi\propto r^\ell$ near the origin), and the hyperfine coupling comes from the ordinary dipole field of the orbiting current. But s-states are different: they alone have nonzero probability density at the nucleus, and there the dipole-field expression is singular. The physically correct term for $\ell=0$ is the Fermi contact interaction — a delta-function coupling proportional to the electron density at the origin:

where: $\hat{\boldsymbol{\mu}}_e=-g_e\mu_B\hat{\mathbf{S}}_e/\hbar$ and $\hat{\boldsymbol{\mu}}_I=g_p\mu_N\hat{\mathbf{S}}_p/\hbar$ are the two moments, and the $\delta^3(\mathbf{r})$ enforces contact. This is the same one-point structure the vacuum produces in Darwin Term — that article's contact interaction is "felt only by states that reach the nucleus," and hyperfine borrows the skeleton with the proton's moment in place of the vacuum's. Taking the expectation value replaces $\delta^3(\mathbf{r})$ by the density at the origin. For the hydrogen $1s$ state that density is $|\psi_{100}(0)|^2=1/(\pi a^3)$ — a value this article computes directly from the normalized ground state (the wavefunction itself is owned by Hydrogenic Wavefunctions and Quantum Numbers, which records that "only s-states reach the nucleus"; the number at the origin is derived here). The spin operators reduce to $\hat{\mathbf{S}}_e\cdot\hat{\mathbf{S}}_p$, whose eigenvalues split the ground state.

Add an external field $\mathbf{B}$ and the story becomes the hyperfine analogue of the Zeeman/Paschen–Back competition. The field couples mainly to the electron moment, energy $\sim\mu_B B$; the hyperfine coupling holds $\mathbf{I}$ and $\mathbf{J}$ together with strength $\sim A$. Which wins sets the good basis. In the weak field ($\mu_B B\ll A$) the hyperfine coupling dominates, $F$ and $m_F$ stay good, and the level fans into $2F+1$ equally spaced sublevels — the hyperfine version of the weak-field Zeeman fan, with a hyperfine $g_F$ obtained by projection (the projection-theorem grammar of Landé g-Factor, which forward-names "Hyperfine Structure's $g_F$" as the clock-level analogue of its electronic $g_J$). In the strong field ($\mu_B B\gg A$) the field tears $\mathbf{J}$ away from $\mathbf{I}$: they precess independently, $m_J$ and $m_I$ become good, $F$ dies. This is exactly the Back–Goudsmit effect — the hyperfine analogue that Strong-Field Paschen-Back Effect sets up but explicitly defers here: it notes "the identical drama replays at the hyperfine scale, where the nuclear spin" decouples, just as $\mathbf{L}$ and $\mathbf{S}$ decoupled at the fine-structure scale in that article. The bridge across both regimes, for the important $I=J=\tfrac12$ ground state, is the Breit–Rabi formula. Diagonalizing the two-level $m_F=0$ block gives a closed form:

where: $\Delta E$ is the zero-field hyperfine gap, and $x$ is the field measured in units of that gap. At $B=0$ the two $m_F=0$ states sit at $+\Delta E/4$ (the $F=1$ member) and $-3\Delta E/4$ (the $F=0$ singlet), reproducing the singlet–triplet gap $\Delta E$. The remarkable feature is the transition between them: because both levels curve the same way to lowest order in $B$, their separation is flat at $B=0$ — a field-independent clock transition $|F=0\rangle\leftrightarrow|F=1,m_F=0\rangle$, insensitive to stray magnetic fields to first order. That immunity is why hyperfine transitions define atomic time.

Example 1 — hydrogen's ground-state splitting, end to end. Compute the $1s$ hyperfine gap, its frequency, and its wavelength from first principles.

Step 1. Enumerate the states: $I=\tfrac12$ (proton), $J=\tfrac12$ (electron spin, $\ell=0$). Coupling gives $F=1$ (triplet) and $F=0$ (singlet). Two levels, spin-dot values $+\tfrac14\hbar^2$ and $-\tfrac34\hbar^2$ — the whole gap is one contact constant.

Step 2. Apply the contact formula with $g_p=5.586$, $m_e/m_p=1/1836$, $\alpha=1/137.036$, $m_ec^2=0.511$ MeV:

Step 3. Convert to a photon. $\nu=\Delta E/h=(5.88\times10^{-6}\ \text{eV})/(4.136\times10^{-15}\ \text{eV}\,\cdot\,\text{s})=\mathbf{1.42\times10^{9}\ Hz}=1420$ MHz. Wavelength $\lambda=c/\nu=(3.00\times10^{8})/(1.42\times10^{9})=\mathbf{0.211\ m}=21.1$ cm.

Step 4. Compare: the measured line is $1420.405751768$ MHz. The simple contact number is right to a few parts in $10^3$ — the residual is reduced-mass, bound-state QED, and finite-proton-size corrections. The exact value and its astrophysical use belong to 21-cm Hydrogen Line.

Example 2 — the interval rule pins nuclear spin. Sodium's $3^2P_{3/2}$ level has electronic $J=\tfrac32$; its hyperfine multiplet shows four sublevels with spacings in the ratio $1:2:3$. What is the nuclear spin $I$?

Step 1. The number of hyperfine levels is $\min(2I+1,2J+1)$. Four levels with $J=\tfrac32$ ($2J+1=4$) means $2I+1\geq4$, so $I\geq\tfrac32$.

Step 2. The interval rule gives spacings $E_F-E_{F-1}=AF$. For $F$ running over $\{0,1,2,3\}$ the spacings are $A\cdot1,\,A\cdot2,\,A\cdot3$ — ratio $\mathbf{1:2:3}$ (verified). This matches the observed pattern only if the top level is $F=3$.

Step 3. With $J=\tfrac32$, reaching $F=3$ requires $I=\tfrac32$ (since $F_{\max}=I+J=3$). So $\mathbf{I=3/2}$ — which is indeed $^{23}$Na's nuclear spin, read from the ratios of the spacings alone, with $A$ unknown.

• Griffiths and Schroeter, Introduction to Quantum Mechanics, 3rd ed., §7.5 — hyperfine splitting in this register.
• Sakurai and Napolitano, Modern Quantum Mechanics, 2nd ed., ch. 5 — the coupled-basis treatment.
• Cohen-Tannoudji, Diu, Laloë, Quantum Mechanics, complement A$_{XII}$ — the contact interaction in full.
• Bethe and Salpeter, Quantum Mechanics of One- and Two-Electron Atoms, §22 — the definitive hydrogenic hyperfine reference.

Provenance note: Equations from Griffiths §7.5, Sakurai & Napolitano ch. 5, Cohen-Tannoudji complement A_XII, and Bethe-Salpeter §22. A verifier should recompute: (1) the hydrogen ground-state splitting $\Delta E=\frac{4}{3}g_p\frac{m_e}{m_p}\alpha^4 m_e c^2\approx5.88\times10^{-6}\,\text{eV}$ giving $\nu=1420\,\text{MHz}$, $\lambda=21\,\text{cm}$ (standard value $1420.405751\,\text{MHz}$); (2) the interval-rule algebra $E_F-E_{F-1}=AF$ from $\mathbf{I}\cdot\mathbf{J}=\tfrac12(F^2-I^2-J^2)$; (3) $|\psi_{100}(0)|^2=1/(\pi a^3)$; (4) constants $g_p\approx5.586$, $\mu_N/\mu_B=m_e/m_p\approx1/1836$, caesium clock frequency 9 192 631 770 Hz (SI definition). — The items this article owns (the $\hat{\mathbf{I}}\cdot\hat{\mathbf{J}}$ eigenvalue, the interval rule, the singlet–triplet eigenvalues, the contact number to $1420$ MHz, and the Breit–Rabi clock transition) are executed in its verify script (22 checks). Edition note: the true hyperfine Breit–Rabi formula (with $m_F$, equation 4) is cashed here — an edition note relocated from the base of Zeeman Effect, whose base had carried an $m_j$-labelled version of the ground-state field dependence; the hyperfine formula belongs with the hyperfine article.

Builds on: Degenerate Perturbation Theory · Good Quantum Numbers and Symmetry-Adapted Bases · Spin-Orbit Coupling · Fine Structure of Hydrogen Leads to: 21-cm Hydrogen Line · Zeeman Effect · Lamb Shift