Atomic Spectra and Selection Rules

Atomic emission and absorption spectroscopy, filtered through exactly these selection rules, remains one of the most direct diagnostic tools across chemistry, astrophysics, and technology. Flame photometry and atomic absorption spectroscopy (AAS), workhorses of trace-metal quantitative analysis in environmental and clinical labs, rely on the fact that each element's allowed resonance lines (like the sodium D-lines) are intense, sharp, and essentially unique fingerprints, while forbidden transitions from the same atom stay dark and do not clutter the spectrum. Inductively coupled plasma optical emission spectroscopy (ICP-OES) instruments used throughout industrial quality control are essentially automated Grotrian-diagram readers, matching observed allowed lines against tabulated atomic energy levels. Astrophysics leans on selection rules even more heavily, because it cannot control the sample. Interstellar and nebular emission is dominated by forbidden lines — [O III], [N II], [O II] — precisely because these environments are so rarefied that collisional de-excitation cannot outrun the (very slow) magnetic-dipole and electric-quadrupole decay channels; measuring their relative intensities gives astronomers electron densities and temperatures of ionized gas clouds. The 21 cm hydrogen hyperfine line, used to map the neutral hydrogen distribution of entire galaxies, is itself a magnetic-dipole "forbidden" transition with a spontaneous emission lifetime of about 11 million years per atom — observable at all only because there is so much hydrogen in the universe. Modern atomic clocks and quantum-information platforms deliberately exploit intercombination and doubly-forbidden lines. The strontium optical lattice clock's famous $^1S_0 \to {}^3P_0$ "clock transition" is forbidden to E1 radiation in bosonic isotopes by every rule derived here ($\Delta J = 0$, $J=0\to0$; $\Delta S \neq 0$); it is made weakly allowed by hyperfine mixing in fermionic isotopes or by an applied magnetic field, giving a transition so narrow (linewidth set by a lifetime of tens of seconds) that it underlies the most precise clocks ever built. Laser cooling and trapping schemes (magneto-optical traps) are engineered entirely around cycling on strongly allowed E1 transitions while relying on nearby forbidden transitions to provide long-lived dark/storage states for qubits. Even fluorescent lighting and neon signage are direct industrial exploitations of the same allowed-transition logic that governs the sodium lamp.

:::quiz[Why can the forbidden [O III] nebular lines be observed in interstellar gas but not in a dense laboratory discharge?]

• Because forbidden transitions do not conserve energy in a lab
• Because the low density of nebular gas prevents collisional de-excitation from outrunning the slow radiative decay ✓
• Because oxygen does not exist as O²⁺ in the laboratory
• Because the selection rules only apply above 10,000 K

Ask the tutor: How would these selection rules change if I needed to include the electric-quadrupole (E2) transition moment explicitly for a heavy-atom forbidden line?

An atom emits or absorbs light by making a transition between two stationary eigenstates of its non-relativistic (plus spin-orbit) Hamiltonian. Time-dependent perturbation theory shows that the rate of a radiative transition is proportional to $|\langle f | \hat{\boldsymbol{\mu}} | i \rangle|^2$, where $\hat{\boldsymbol{\mu}} = -e\sum_i \mathbf{r}_i$ is the electric-dipole operator (sum over all electrons, position measured from the nucleus). This single matrix element is the entire content of atomic spectroscopy's "selection rules" — everything else is a systematic way of figuring out when it must vanish by symmetry. The position operator $\mathbf{r}$ has odd parity: under $\mathbf{r} \to -\mathbf{r}$, it flips sign. Atomic stationary states in a central potential have definite parity, $(-1)^{\sum \ell_i}$, from the product of one-electron orbital parities. Because $\hat{\boldsymbol{\mu}}$ is odd, the integral $\langle f|\hat{\boldsymbol{\mu}}|i\rangle$ is identically zero unless the initial and final total parities differ — this is the Laporte rule, exact for any isolated atom with only electrostatic and spin-orbit interactions (which conserve parity). For a one-electron jump this collapses to $\Delta \ell = \pm 1$: an electron in a $p$ orbital can only radiate to $s$ or $d$, never to another $p$ or to $f$ directly. Beyond parity, the dipole operator carries angular momentum quantum number $\ell=1$ (it transforms as a spherical vector, like the $p$-orbitals themselves, $Y_1^{0,\pm1}$). The Wigner-Eckart theorem then forces triangle-inequality-type selection rules on the total angular momenta: $\Delta L = 0, \pm 1$ and $\Delta J = 0, \pm 1$, with the $0 \to 0$ case forbidden outright since a photon cannot carry zero angular momentum away from a spherically symmetric ($J=0$) state to another spherically symmetric state. Because $\hat{\boldsymbol{\mu}}$ acts only on spatial coordinates and is spin-independent, it cannot change total spin: $\Delta S = 0$ in pure LS coupling. This rule is only as good as LS coupling itself — once spin-orbit coupling mixes states of different $S$, weak "intercombination" lines appear, exactly as seen in mercury (the $^3P_1 \to {}^1S_0$ intersystem line at 253.7 nm) and in astrophysical forbidden lines.

The third piece of the physical picture is fine structure. Spin-orbit coupling adds a term $\hat{H}_{SO} = \xi(r)\,\hat{\mathbf{L}}\cdot\hat{\mathbf{S}}$ to the Hamiltonian, which commutes with $\hat{J}^2$ but not with $\hat{L}^2$ or $\hat{S}^2$ separately (in the strict sense — in LS coupling it is treated as a small perturbation that only weakly mixes different $L,S$). Its effect on a term with definite $L, S$ is to split it into levels of definite $J = |L-S|, \dots, L+S$, each shifted in energy by

This produces the Landé interval rule: the spacing between adjacent $J$ and $J-1$ levels of the same term is proportional to $J$,

which is exactly how spectroscopists extract the spin-orbit coupling constant $\zeta_{nl}$ from measured level spacings without any independent calculation of $\langle 1/r^3\rangle$.

The transition rate between initial state $|i\rangle$ and final state $|f\rangle$ under spontaneous emission is given by the Einstein $A$ coefficient,

where:

• $\omega_{fi} = (E_i - E_f)/\hbar$ is the transition angular frequency,
• $\hat{\boldsymbol{\mu}} = -e\sum_k \mathbf{r}_k$ is the total electric dipole operator,
• $\varepsilon_0$ is the vacuum permittivity.

Reading: the rate is exactly zero whenever the matrix element vanishes by symmetry — that is the entire origin of "forbidden lines." When it is nonzero, the strong $\omega^3$ dependence means high-energy (UV) E1 transitions are enormously faster than low-energy ones of comparable matrix element, which is why some metastable states with only far-infrared or forbidden decay channels survive for seconds. The single-electron orbital selection rule, obtained by expanding $\mathbf{r}$ in spherical harmonics and using the Gaunt/triangle rules for integrals of three spherical harmonics, is

where:

• $\ell$ is the orbital angular momentum quantum number of the single electron undergoing the transition (all other electrons "spectators"),
• $m_\ell$ is its projection, and the three allowed $\Delta m_\ell$ values correspond to linearly polarized ($\Delta m_\ell = 0$) or circularly polarized ($\Delta m_\ell = \pm 1$) light, i.e. to the three spherical components of $\mathbf{r}$.

Reading: the dipole operator literally is a $p$-type ($\ell=1$) tensor operator, so by angular momentum addition an $s$-orbital ($\ell=0$) can only couple to $p$ ($\ell=1$), and a $p$-orbital only to $s$ or $d$ ($\ell = 0, 2$) — never back to another $p$. For the full many-electron atomic term, the LS-coupling selection rules are

where:

• $S, L, J$ are the total spin, orbital, and total angular momentum quantum numbers of the term/level,
• the parenthetical exclusions rule out $0\to0$ because a single photon cannot carry away zero angular momentum between two states that are both spherically symmetric in the relevant sense.

Reading: these rules are hierarchical in reliability — the Laporte rule and $\Delta\ell = \pm1$ are essentially exact (violated only by much weaker magnetic-dipole or electric-quadrupole radiation, or by external fields), $\Delta L$ and $\Delta J$ rules are exact within LS coupling and become approximate once coupling schemes break down (heavy atoms, jj coupling), and $\Delta S = 0$ is the least robust, broken by spin-orbit coupling itself.

Example 1 — Foundational: which hydrogen transitions between $n=3$ and $n=1$ are E1-allowed? The $n=3$ shell of hydrogen contains states with $\ell = 0$ ($3s$), $\ell=1$ ($3p$), $\ell=2$ ($3d$). The $n=1$ shell has only $\ell=0$ ($1s$). Apply $\Delta\ell = \pm1$:

• $3s \to 1s$: $\ell_i = 0 \to \ell_f = 0$, $\Delta\ell = 0$ — forbidden.
• $3p \to 1s$: $\ell_i = 1 \to \ell_f = 0$, $\Delta\ell = -1$ — allowed.
• $3d \to 1s$: $\ell_i = 2 \to \ell_f = 0$, $\Delta\ell = -2$ — forbidden.

So of the three degenerate-in-energy $n=3\to n=1$ possibilities, only $3p\to1s$ radiates by E1, and it does so as the third member of the Lyman series (Lyman-$\gamma$). Its energy, from the Rydberg formula

matching the observed Lyman-$\gamma$ line at 102.57 nm. The $3s$ and $3d$ populations, if somehow prepared (e.g. by electron-impact excitation), must instead decay by much slower routes: $3s\to2p$ ($\Delta\ell=+1$, allowed, in the Balmer series region) or $3d\to2p$ ($\Delta\ell=-1$, allowed), never directly to $1s$.

Limiting-case check. As $n\to\infty$ within a Rydberg series converging on a given lower state, the Rydberg formula gives $\tilde\nu \to R_H(1/n_f^2)$, a finite series limit (the ionization threshold from level $n_f$) — consistent with the well-known convergence of the Lyman, Balmer, etc. series to their respective continuum edges, and no selection-rule content is lost in this limit since $\Delta \ell=\pm1$ applies at every $n$.

Example 2 — Applied: the Mg intercombination line and why it is so weak. Neutral magnesium's ground configuration is $[\text{Ne}]3s^2$, a closed-subshell configuration giving the single term $^1S_0$. Excitation of one $3s$ electron to $3p$ gives the configuration $3s3p$, which by the standard vector-coupling rules yields terms $^1P_1$ and $^3P_{0,1,2}$ (singlet and triplet, since the two electrons are now in different subshells and both spin arrangements are allowed). The strong resonance line is $3s3p\ {}^1P_1 \to 3s^2\ {}^1S_0$ at 285.2 nm: check the rules — $\Delta S = 0$ ($1\to1$, both singlets), $\Delta L = 1\to0$ (allowed, not $0\to0$), $\Delta J = 1\to0$ (allowed), parity changes because one electron moved from $s$ ($\ell=0$) to $p$ ($\ell=1$), $\Delta\ell = \pm1$ for that electron. All rules satisfied — this is a fully allowed, fast E1 line with $A \approx 1.8\times10^8\ \text{s}^{-1}$, lifetime $\tau \approx 5.6$ ns. The intercombination line $3s3p\ {}^3P_1 \to 3s^2\ {}^1S_0$ at 457.1 nm has identical $\Delta L$, $\Delta J$, and parity behavior, but $\Delta S = 1\to0$ — spin changes. In pure LS coupling this line is exactly forbidden. It is observed only because spin-orbit coupling admixes a small amount of $^1P_1$ character into the nominal $^3P_1$ level,

so the effective transition moment is $\alpha\langle {}^1P_1|\hat{\boldsymbol\mu}|{}^1S_0\rangle$, suppressed by the small mixing coefficient. Measured lifetime of the Mg $^3P_1$ level is about 5 ms — roughly $10^6$ times longer than the allowed $^1P_1$ line, exactly the sort of suppression factor $|\alpha|^2$ one expects from a fine-structure-driven admixture in a moderately light atom. This transition is in fact the basis of magnesium (and analogous alkaline-earth) optical clocks, prized precisely because the state is metastable.

Limiting-case check. As the atomic number of the alkaline-earth-like species increases (Mg → Ca → Sr → Yb), spin-orbit coupling grows roughly as $Z^4$ for a given orbital, so the mixing coefficient $\alpha$ grows and the intercombination line becomes progressively less forbidden — consistent with the empirically much shorter intercombination-line lifetimes in heavier alkaline earths and the heavy noble-gas-like ions where "forbidden" lines become almost as bright as allowed ones.

Example 3 — Competition level: relative D-line intensities and the term-level degeneracies. A classic exam question: predict the intensity ratio of the sodium D2 line ($3p\,^2P_{3/2}\to 3s\,^2S_{1/2}$) to the D1 line ($3p\,^2P_{1/2}\to 3s\,^2S_{1/2}$) in an optically thin sodium vapor lamp where the two upper levels are populated statistically (proportional to their degeneracies, as in a high-temperature source with fast collisional mixing between $3p_{3/2}$ and $3p_{1/2}$).

Step 1 — degeneracies. Each level's degeneracy is $g_J = 2J+1$. So $g(^2P_{3/2}) = 2(3/2)+1 = 4$ and $g(^2P_{1/2}) = 2(1/2)+1 = 2$.

Step 2 — relate population to intensity. If populations are statistical (proportional to $g_J$) and each level decays with a comparable Einstein $A$ coefficient (both are pure $\Delta\ell=-1$, $\Delta L=-1$, $\Delta S=0$, $\Delta J = -1$ or $0$ allowed E1 transitions with similar radial matrix elements since they share the same $3p$ radial wavefunction), the emitted intensity ratio is approximately the degeneracy ratio times a small correction from the actual $A$ coefficients (which differ slightly because $\Delta J=1$ vs $\Delta J=0$ carry different angular factors, from the $6j$/Wigner-Eckart geometry). To leading (statistical) order:

Using the standard LS-coupling angular factors for a $^2P_J \to {}^2S_{1/2}$ transition, the ratio of squared reduced matrix elements (line strengths) works out to $S(J=3/2)/S(J=1/2) = 2$ exactly (from the $6j$-symbol algebra for a single-electron $p\to s$ transition), so

(the factor of 2 in line strength per level is already what generates the observed intensity ratio close to 2:1, consistent with the well-known experimental result that the D2 line is roughly twice as bright as D1 in an ordinary sodium lamp under thermal, collision-broadened conditions.)

Step 3 — sanity check against wavelengths. Since $A \propto \omega^3 |\mu|^2$ and $\omega(D_2)$ is only about $0.1\%$ larger than $\omega(D_1)$ (589.0 nm vs. 589.6 nm), the frequency-cubed factor is essentially 1:1 between the two lines, confirming that the observed $\approx 2{:}1$ intensity ratio is overwhelmingly a degeneracy/line-strength effect, not a frequency effect.

Limiting-case check. If instead the sodium vapor is prepared with population only in $3p_{1/2}$ (e.g. selectively pumped by a narrow-band laser resonant with D1 alone, with no collisional relaxation to mix the levels), then $I(D_2) \to 0$ identically, since there is no population left in $^2P_{3/2}$ to emit from — the 2:1 ratio derived above is strictly a consequence of assuming statistical, collision-equilibrated populations, and vanishes as an artifact once that assumption is relaxed.

Spectral selection rules as a systematic, quantum-mechanical subject were forged jointly by atomic spectroscopists and the founders of quantum mechanics in the 1920s, but if one figure must be singled out for connecting term-symbol structure to observed line patterns with a rule that chemists still use verbatim, it is Henry Norris Russell, whose collaboration with Frederick Saunders gave Russell-Saunders (LS) Coupling its name and its predictive power for exactly the multiplet structures discussed here. Russell was an astronomer by primary training (co-inventor of the Hertzsprung-Russell diagram), and his deep familiarity with solar and stellar spectra — vast catalogs of lines that needed classifying into series and multiplets — drove him toward the systematic vector-coupling scheme for atomic terms.