Fine Structure

Fine structure is the relativistic and spin-dependent splitting of electronic atomic levels beyond the leading nonrelativistic Coulomb or central-field spectrum. In hydrogen, its leading one-electron terms are the relativistic kinetic-energy correction, spin–orbit coupling, and the Darwin contact term. Their sum reorganizes the spectrum by total electronic angular momentum $j$.

In many-electron atoms, fine structure also includes relativistic one- and two-electron interactions that split electronic terms into levels of different total $J$. The result depends on screening, configuration interaction, and the appropriate LS, jj, or intermediate coupling scheme.

Required background. Hydrogenic states supply the unperturbed spectrum. Spin–orbit coupling gives the angular algebra, and degenerate perturbation theory explains how to resolve a degenerate manifold.

Relativistic splitting of electronic levels

Fine structure differs from nuclear-spin hyperfine structure, radiative Lamb shifts, and shifts caused by external fields. Spin–orbit coupling often gives the most visible splitting within a non-$S$ term, but mass–velocity and Darwin terms are also essential. For an $S$ state the spin–orbit term vanishes; the other relativistic shifts remain.

Expansion Parameter and Scale

For a hydrogenic ion of charge $Z$, the characteristic estimates are

$$\frac{v}{c}\sim Z\alpha, \qquad p\sim m_ecZ\alpha,$$

where $\alpha$ is the fine-structure constant. The gross binding scale is

$$E_{\mathrm{gross}} \sim m_ec^2(Z\alpha)^2,$$

while the leading fine-structure scale is

$$E_{\mathrm{fs}} \sim m_ec^2(Z\alpha)^4.$$

Thus

$$\frac{E_{\mathrm{fs}}}{E_{\mathrm{gross}}} \sim (Z\alpha)^2.$$

This power counting explains why fine structure is small in hydrogen and grows rapidly along a hydrogenic sequence. In neutral many-electron atoms, a universal $Z^4$ rule is not reliable because shielding and orbital penetration change the effective field sampled by each electron.

The expansion assumes $Z\alpha\ll1$ and momenta small compared with $m_ec$. Heavy highly charged ions require a nonperturbative relativistic starting point, followed by recoil, finite-nuclear-size, Breit, and radiative corrections appropriate to the desired accuracy.

Effective One-Electron Hamiltonian

Consider a spin-$1/2$ electron in a static central electrostatic potential energy $V(r)$, with no magnetic field. The nonrelativistic reference is

$$H_0=\frac{\mathbf p^2}{2m_e}+V(r).$$

Through order $1/c^2$, the leading one-electron fine-structure operator is written

$$H_{\mathrm{fs}}^{(1)} =H_{\mathrm{mv}}+H_{\mathrm{so}}+H_D.$$

This form is obtained from the low-energy expansion of relativistic spinor dynamics, such as a Foldy–Wouthuysen transformation. It is an effective Hamiltonian with a specified order, not three independent fundamental forces.

Relativistic Kinetic-Energy Correction

The exact free-particle kinetic energy is

$$T=\sqrt{m_e^2c^4+p^2c^2}-m_ec^2.$$

Expanding in $p/(m_ec)$ gives

$$T =\frac{p^2}{2m_e} -\frac{p^4}{8m_e^3c^2} +O(c^{-4}).$$

The leading mass–velocity operator is therefore

$$H_{\mathrm{mv}} =-\frac{p^4}{8m_e^3c^2}.$$

Its expectation value is negative: relativistic kinematics makes a bound level more negative relative to the nonrelativistic reference at this order. Because $p^4$ weights high-momentum components strongly, compact and penetrating orbitals are especially sensitive.

The name “mass–velocity” is historical shorthand. One should not replace $m_e$ by a classical speed-dependent mass inside the Schrödinger equation; the operator follows from expanding the relativistic dispersion relation.

Spin–Orbit Term

For a central potential, including the Thomas factor, the leading one-electron operator is

$$H_{\mathrm{so}} =\frac{1}{2m_e^2c^2} \frac{1}{r}\frac{dV}{dr} \mathbf L\cdot\mathbf S.$$

The angular eigenvalue in a state $|\ell,s;j,m_j\rangle$ is

$$\begin{aligned} \langle\mathbf L\cdot\mathbf S\rangle &=\frac{\hbar^2}{2}\Lambda_{\ell s j}, \\ \Lambda_{\ell s j} &=j(j+1)-\ell(\ell+1) -s(s+1). \end{aligned}$$

For an electron, $s=1/2$ and $j=\ell\pm1/2$ when $\ell>0$. Rotational symmetry preserves degeneracy among the $2j+1$ values of $m_j$ while permitting different $j$ multiplets to split.

For a point Coulomb potential,

$$V(r)=-\frac{Ze^2}{4\pi\epsilon_0r},$$

so

$$H_{\mathrm{so}} =\frac{Ze^2}{8\pi\epsilon_0m_e^2c^2r^3} \mathbf L\cdot\mathbf S.$$

The apparent $r^{-3}$ singularity is harmless for the nonzero-$\ell$ hydrogenic expectation values to which this operator applies. For $\ell=0$, $\mathbf L=0$ and the term vanishes. Finite nuclei and relativistic wavefunctions are required when nuclear-region details matter beyond this expansion.

Darwin Term

The Darwin operator is

$$H_D =\frac{\hbar^2}{8m_e^2c^2}\nabla^2V.$$

For a point Coulomb potential, the distributional identity $\nabla^2(1/r)=-4\pi\delta^3(\mathbf r)$ gives

$$H_D =\frac{Ze^2\hbar^2}{8\epsilon_0m_e^2c^2} \delta^3(\mathbf r).$$

Only states with nonzero probability density at the origin receive this point-nucleus contact shift. In the nonrelativistic hydrogenic basis, that means $S$ states. The Darwin term is positive for an attractive point-Coulomb potential and partly offsets the negative mass–velocity shift.

Descriptions in terms of rapid relativistic position fluctuations can provide intuition, but the controlled statement is operator-level: the contact term emerges in the low-energy expansion of the Dirac theory. For a finite nucleus, the delta function is replaced by sensitivity to the nuclear charge distribution.

Hydrogenic Fine Structure

Take an infinitely heavy point nucleus and treat $H_{\mathrm{mv}}$, $H_{\mathrm{so}}$, and $H_D$ consistently through order $(Z\alpha)^4$. Their sum gives

$$\begin{aligned} E_{nj} &=-\frac{m_ec^2(Z\alpha)^2}{2n^2} +\Delta E_{nj}^{\mathrm{fs}},\\ \Delta E_{nj}^{\mathrm{fs}} &=-\frac{m_ec^2(Z\alpha)^4}{2n^4} \left(\frac{n}{j+1/2}-\frac34\right). \end{aligned}$$

The cancellation can be checked from the radial expectation values. Let $Q=m_ec^2(Z\alpha)^4$ and $\Lambda=j(j+1)-\ell(\ell+1)-3/4$. For $\ell>0$,

$$\begin{aligned} \Delta E_{\mathrm{mv}}/Q &=\frac{3}{8n^4}-\frac{1}{2n^3(\ell+1/2)},\\ \Delta E_{\mathrm{so}}/Q &=\frac{\Lambda}{4n^3\ell(\ell+1/2)(\ell+1)},\\ \Delta E_D&=0. \end{aligned}$$

The first line follows by using $\langle p^4\rangle=4m_e^2\langle(E_n-V)^2\rangle$, the virial theorem, and $\langle r^{-2}\rangle$. For $\ell=0$, use the same mass–velocity expression, zero spin–orbit term, and $\Delta E_D/Q=1/(2n^3)$. Adding the appropriate terms removes the separate $\ell$ dependence and leaves only $n,j$. This is a central structural result, not an accident of notation. It agrees with the expansion of the exact point-Coulomb Dirac energy.

What remains degenerate

Because the energy depends on $j$ rather than separately on $\ell$, states such as

$$nS_{1/2} \quad\text{and}\quad nP_{1/2}$$

remain degenerate in the ideal Dirac–Coulomb problem. Radiative corrections, recoil, and nuclear structure break that degeneracy. The famous $2S_{1/2}$–$2P_{1/2}$ separation is therefore a Lamb-shift question, not a leading fine-structure splitting; Lamb Shift Overview explains those corrections.

Schematic $n=2$ hierarchy; the two panels use independent energy offsets and exaggerated separations. Top: fine structure lowers every level and separates $2P_{3/2}$ from the Dirac-degenerate $2S_{1/2}$ and $2P_{1/2}$ pair. Bottom: the Lamb shift lifts that degeneracy; proton spin $I=1/2$ then gives two hyperfine levels for each electronic level. These later corrections lie outside the strict fine-structure Hamiltonian.

The n = 2 fine-structure interval

For $Z=1$, the leading difference within the $2P$ term is

$$E_{2P_{3/2}}-E_{2P_{1/2}} =\frac{m_ec^2\alpha^4}{32}.$$

This expression is an infinite-nuclear-mass, leading-order result. Comparison with precision spectroscopy requires reduced-mass and recoil effects, radiative corrections, proton structure, and a careful definition of the measured line centroid.

Many-Electron Fine Structure

The nonrelativistic electronic Hamiltonian for a fixed nucleus is

$$H_{\mathrm{NR}} =\sum_i\left[ \frac{\mathbf p_i^2}{2m_e}+V_{\mathrm{nuc}}(r_i) \right] +\sum_{i<j} \frac{e^2}{4\pi\epsilon_0r_{ij}}.$$

Electron–electron repulsion first creates configurations, terms, and correlation structure. Relativistic corrections then act within and between those many-electron states. A schematic Breit–Pauli organization is

$$\begin{aligned} H_{\mathrm{BP}} &=H_{\mathrm{NR}}+H_{1e}^{\mathrm{rel}} +H_{2e}^{\mathrm{rel}},\\ H_{1e}^{\mathrm{rel}} &=\sum_i(H_{\mathrm{mv},i}+H_{D,i}+H_{\mathrm{so},i}),\\ H_{2e}^{\mathrm{rel}} &=H_{\mathrm{oo}}+H_{\mathrm{ss}} +H_{\mathrm{soo}}+\cdots. \end{aligned}$$

Here $H_{\mathrm{oo}}$ denotes orbit–orbit terms, $H_{\mathrm{ss}}$ spin–spin terms, and $H_{\mathrm{soo}}$ spin–other-orbit terms. Authors package contact and mutual spin–orbit contributions differently, so an abbreviation such as “Breit–Pauli” must be accompanied by the actual operator set.

LS coupling

In many light atoms, residual electrostatic interactions organize orbital and spin angular momenta before spin-dependent terms become important:

$$\begin{aligned} \mathbf L&=\sum_i\mathbf l_i,\\ \mathbf S&=\sum_i\mathbf s_i,\\ \mathbf J&=\mathbf L+\mathbf S. \end{aligned}$$

Levels are labeled by term symbols $^{2S+1}L_J$. LS Coupling develops the atomic energy-scale criterion, state-count construction, Landé diagnostics, and same-$J^\pi$ mixing that determine whether those labels are physically useful. If the fine-structure operator within one well-isolated term reduces to

$$H_{\mathrm{fs}}^{\mathrm{term}} =A\,\frac{\mathbf L\cdot\mathbf S}{\hbar^2},$$

with $A$ in energy units and dimensionful angular momenta, then

$$\begin{aligned} K_{LSJ} &=J(J+1)-L(L+1) -S(S+1), \\ E_J &=E_{LS}+\frac{A}{2}K_{LSJ}. \end{aligned}$$

Adjacent intervals obey the Landé interval rule,

$$E_J-E_{J-1}=AJ.$$

This rule is a diagnostic of pure LS coupling, not a universal law. Configuration interaction, spin–spin terms, term mixing, and relativistic radial changes produce deviations.

Term centroid

For the pure $A\mathbf L\cdot\mathbf S/\hbar^2$ model, the degeneracy-weighted fine-structure shift averages to zero. Define

$$\overline E_{LS} =\frac{\sum_J(2J+1)E_J} {\sum_J(2J+1)}.$$

Then $\overline E_{LS}=E_{LS}$. This center-of-gravity relation is useful when comparing a relativistically split multiplet with a nonrelativistic term energy. It can fail as a literal identification when fine structure mixes other terms or when the averaging omits levels.

jj and intermediate coupling

In a jj-coupled description, each electron first forms

$$\mathbf j_i=\mathbf l_i+\mathbf s_i,$$

and the $\mathbf j_i$ are combined into total $\mathbf J$. This basis is useful when one-electron spin–orbit splittings compete strongly with residual electrostatic term separations.

jj Coupling develops the relativistic-subshell occupations, equivalent-electron state counts, $9j$ relation to the LS basis, and energy, magnetic, and transition diagnostics needed to test that description.

Most real atoms are not in an exact LS or jj limit. Their eigenstates are mixtures of configuration-state functions with the same conserved total $J$ and parity. “Intermediate coupling” is not a new interaction; it is the need to diagonalize competing electrostatic and relativistic operators without assuming one ideal coupling order.

Level labels then identify dominant components and can exchange character through avoided crossings. A trustworthy table gives mixing coefficients or at least flags ambiguous classifications.

Alkali doublets

An alkali atom has a single valence electron outside a closed shell, so its low-lying $nP$ levels form $P_{1/2}$ and $P_{3/2}$ partners. The principal $D_1$ and $D_2$ lines terminate on those fine-structure levels. Their separation illustrates the trend toward stronger relativistic splitting in heavier atoms, but quantitative values require the actual screened, correlated, and relativistic valence wavefunction. Alkali Atoms explains how those manifolds enter spectroscopy and optical control.

Connection to the Dirac Equation

For one electron in an electrostatic potential energy $V=q\phi$ (the time component of an electromagnetic four-potential), the Dirac Hamiltonian is

$$H_{\mathrm{Dirac}} =c\boldsymbol{\alpha}\cdot\mathbf p +\beta m_ec^2+V(r),$$

where $\boldsymbol{\alpha}$ and $\beta$ are Dirac matrices, not the fine-structure constant. Its four-component spinor combines spin and relativistic kinematics from the start.

For a point Coulomb potential, the bound-state energy depends on $n$ and $j$ through the Dirac angular quantum number. Expanding the positive-energy result in $Z\alpha$ reproduces the combined mass–velocity, spin–orbit, and Darwin shift above. The individual low-energy terms depend on the chosen transformation and order, while their predicted spectrum is the invariant comparison.

A Dirac calculation changes radial functions and transition matrix elements as well as energies. For many electrons, Dirac–Hartree–Fock and correlated Dirac–Coulomb methods provide relativistic starting points; Breit, recoil, finite-nuclear-size, and QED corrections must then be included consistently. A prescribed-potential Dirac equation alone does not include radiative self-energy or vacuum polarization.

Spectroscopic Interpretation

Level interval versus line separation

A fine-structure interval is an energy difference between two atomic levels. A measured spectrum usually contains transition frequencies,

$$h\nu_{ab}=E_a-E_b.$$

The separation of two lines equals an upper-state fine-structure interval only if they share the same lower level and if unresolved hyperfine, isotope, field, and line-shape shifts are handled consistently. Otherwise a difference of transition frequencies combines more than one level interval.

Unresolved hyperfine components, isotope mixtures, and external fields can alter a fitted line centroid. Compare the same isotope, field limit, and centroid definition. Selection rules determine angular zeros, not energy splittings or observed intensities.

The low-order result also has a precision limit: replacing $m_e$ by a reduced mass does not include all relativistic recoil; finite nuclear size matters for penetrating states; QED changes both the Lamb shift and fine-structure intervals. For large $Z\alpha$, begin with a relativistic Hamiltonian rather than extrapolating the $1/c^2$ expansion. Near-degenerate terms require diagonalization even when the relativistic operators themselves are small.

Exercises

1. Relativistic dispersion and power counting

Expand the free relativistic kinetic energy through order $p^4$. Then use $p\sim m_ecZ\alpha$ to show the scale of the first correction relative to the gross kinetic energy.

Solution

Write

$$T=m_ec^2 \left[\sqrt{1+\frac{p^2}{m_e^2c^2}}-1\right].$$

Using $\sqrt{1+x}=1+x/2-x^2/8+O(x^3)$ gives

$$T=\frac{p^2}{2m_e} -\frac{p^4}{8m_e^3c^2} +O(p^6).$$

The gross term scales as

$$\frac{p^2}{2m_e} \sim m_ec^2(Z\alpha)^2,$$

whereas the correction scales as

$$\frac{p^4}{8m_e^3c^2} \sim m_ec^2(Z\alpha)^4.$$

Their ratio is of order $(Z\alpha)^2$, as expected for the leading relativistic correction.

2. Darwin shift and S states

Show that a point-Coulomb Darwin term has zero first-order expectation value in a nonrelativistic hydrogenic state with $\ell>0$. For a hydrogenic $1S$ state, use

$$|\psi_{1S}(0)|^2=\frac{Z^3}{\pi a_0^3}$$

to show that the infinite-mass Darwin shift is

$$\Delta E_D(1S) =\frac12m_ec^2(Z\alpha)^4.$$

Solution

The expectation value of the contact operator is

$$\langle\delta^3(\mathbf r)\rangle =|\psi(0)|^2.$$

Hydrogenic radial wavefunctions behave as $R_{n\ell}(r)\propto r^\ell$ near the origin. Thus $\psi(0)=0$ for $\ell>0$, and their point-nucleus Darwin shift vanishes.

For $1S$,

$$\begin{aligned} \Delta E_D &=\frac{Ze^2\hbar^2}{8\epsilon_0m_e^2c^2} \frac{Z^3}{\pi a_0^3}\\ &=\frac12m_ec^2(Z\alpha)^4, \end{aligned}$$

where $e^2/(4\pi\epsilon_0)=\alpha\hbar c$ and $a_0=\hbar/(m_ec\alpha)$ were used. The positive contact shift is only one contribution; it must be combined with the mass–velocity term to obtain the full $1S$ fine correction.

3. The ³P Landé intervals

For a pure $^3P$ term, $L=1$ and $S=1$. Using $H=A\mathbf L\cdot\mathbf S/\hbar^2$, find the shifts for $J=0,1,2$, verify the Landé interval rule, and check the degeneracy-weighted centroid.

Solution

Since $L(L+1)=S(S+1)=2$,

$$\Delta E_J =\frac A2[J(J+1)-4].$$

Therefore

$$\begin{array}{c|ccc} J&0&1&2\\ \hline \Delta E_J&-2A&-A&A \end{array}$$

and

$$E_1-E_0=A, \qquad E_2-E_1=2A.$$

These are $AJ$ for $J=1$ and $2$. With degeneracies $2J+1=1,3,5$,

$$\frac{1(-2A)+3(-A)+5(A)}{1+3+5}=0.$$

The fine-structure shifts leave the term center of gravity unchanged in this ideal model. Significant experimental departures from the $1:2$ interval ratio diagnose additional operators or term mixing.

4. Hydrogen n = 2 splitting

Use the leading hydrogenic fine-structure formula to compute $\Delta E_{2,1/2}^{\mathrm{fs}}$ and $\Delta E_{2,3/2}^{\mathrm{fs}}$. Show that the $2P_{3/2}$–$2P_{1/2}$ interval is $m_ec^2\alpha^4/32$. What does the same formula predict for $2S_{1/2}$ versus $2P_{1/2}$?

Solution

For $n=2$, the common coefficient is $-m_ec^2\alpha^4/32$. For $j=1/2$,

$$\frac{n}{j+1/2}-\frac34 =2-\frac34=\frac54,$$

so

$$\Delta E_{2,1/2}^{\mathrm{fs}} =-\frac{5m_ec^2\alpha^4}{128}.$$

For $j=3/2$,

$$\frac{n}{j+1/2}-\frac34 =1-\frac34=\frac14,$$

giving

$$\Delta E_{2,3/2}^{\mathrm{fs}} =-\frac{m_ec^2\alpha^4}{128}.$$

The difference is

$$E_{2P_{3/2}}-E_{2P_{1/2}} =\frac{m_ec^2\alpha^4}{32}.$$

Both $2S_{1/2}$ and $2P_{1/2}$ have the same $n=2,j=1/2$ in the ideal result, so they remain degenerate. Their observed separation is led by radiative and associated Lamb-shift physics.

5. Classify the correction

Classify each contribution as gross electronic, fine, radiative/Lamb, hyperfine, isotope/recoil, or external-field structure: (a) $-p^4/(8m_e^3c^2)$; (b) $A\mathbf I\cdot\mathbf J/\hbar^2$; (c) electron self-energy; (d) $-\mathbf d\cdot\mathbf E$ from an applied field; (e) finite reduced mass; (f) spin–other-orbit interaction.

Solution

Contribution	Classification	Reason
(a) relativistic $p^4$ term	fine structure	leading relativistic electronic kinetic correction
(b) $A\mathbf I\cdot\mathbf J/\hbar^2$	hyperfine structure	couples nuclear and electronic angular momenta
(c) electron self-energy	radiative/Lamb	QED loop correction beyond Dirac–Coulomb theory
(d) applied electric field	external-field Stark structure	depends on an imposed field
(e) finite reduced mass	isotope/recoil	depends on nuclear mass; leading kinematic recoil
(f) spin–other-orbit	fine structure	relativistic electronic two-body interaction

The categories organize a calculation but do not make contributions numerically independent. Precision work must use a consistent effective Hamiltonian so recoil, nuclear, Breit, and radiative pieces are not double counted.

References

• H. A. Bethe, “The Electromagnetic Shift of Energy Levels,” Physical Review 72, 339–341 (1947), doi:10.1103/PhysRev.72.339.
• H. A. Bethe and E. E. Salpeter, Quantum Mechanics of One- and Two-Electron Atoms (Springer, 1957), Chapters 4–6.
• B. H. Bransden and C. J. Joachain, Physics of Atoms and Molecules, 2nd ed. (Pearson, 2003), Chapters 5–8.
• G. Breit, “The Effect of Retardation on the Interaction of Two Electrons,” Physical Review 34, 553–573 (1929), doi:10.1103/PhysRev.34.553.
• E. U. Condon and G. H. Shortley, The Theory of Atomic Spectra (Cambridge University Press, 1935), Chapters 6–12.
• M. I. Eides, H. Grotch, and V. A. Shelyuto, “Theory of Light Hydrogenlike Atoms,” Physics Reports 342, 63–261 (2001), doi:10.1016/S0370-1573(00)00077-6.
• L. L. Foldy and S. A. Wouthuysen, “On the Dirac Theory of Spin 1/2 Particles and Its Non-Relativistic Limit,” Physical Review 78, 29–36 (1950), doi:10.1103/PhysRev.78.29.
• C. J. Foot, Atomic Physics (Oxford University Press, 2005), Chapters 2–5.
• I. P. Grant, Relativistic Quantum Theory of Atoms and Molecules: Theory and Computation (Springer, 2007), doi:10.1007/978-0-387-35069-1.
• W. R. Johnson, Atomic Structure Theory: Lectures on Atomic Physics (Springer, 2007), doi:10.1007/978-3-540-68013-0.
• 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.
• J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed. (Cambridge University Press, 2021), sections on the Dirac equation and nonrelativistic reduction.