Electric Dipole Selection Rules

Sodium's yellow D lines blaze; the hydrogen 2s state sits in the dark for an eighth of a second — an eternity, $10^{8}$ times longer than its 2p neighbor — unable to radiate a single dipole photon. Nothing about energy forbids it: the gap to 1s is the same 10.2 eV. What decides is a single number, the dipole matrix element $\langle f|\mathbf{r}|i\rangle$ , and symmetry kills it long before dynamics gets a vote. Every rate formula of this chapter — absorption, stimulated emission, the Einstein $A$ coefficient — carries $|\langle f|\mathbf{r}|i\rangle|^2$ as a factor, so when the matrix element vanishes, the entire E1 channel closes at once. This article assembles the complete accounting of when that happens. Stark Effect already proved the $\hat{z}$ -only half — $\Delta\ell=\pm1$ , $\Delta m=0$ — by sweeping every angular pair with $\ell,\ell'\leq3$ ; here we extend the sweep to the full vector operator, resolving $x\pm iy$ into spherical components, and derive the complete E1 rule set: $\Delta\ell=\pm1$ with parity flip, $\Delta m=0,\pm1$ sorted by photon polarization, $\Delta s=0$ because $\mathbf{r}$ never touches spin, and — once fine structure couples $\mathbf{L}$ and $\mathbf{S}$ — $\Delta j=0,\pm1$ with $j=0\to j=0$ banned outright. We prove the parity and azimuthal rungs directly, read the triangle rung off the Wigner–Eckart theorem, and machine-verify the whole set on 768 angular matrix elements. The payoff is an atlas: which hydrogen levels can decay where, why 2s is a dead end, how many fine-structure components hide inside Balmer-α (seven — we count them), and which polarization each $\Delta m$ channel emits. The rules are the grammar of every spectrum you will ever read.

Key takeaways

The E1 transition rate is proportional to $|\langle f|\mathbf{r}|i\rangle|^2$ ; selection rules state exactly when this matrix element vanishes by symmetry — those transitions are "forbidden" at leading order.
For a one-electron atom: $\Delta\ell=\pm1$ (parity must flip, and the rank-1 triangle forbids everything else), $\Delta m=0,\pm1$ (matched to photon polarization: $\pi$ for $\Delta m=0$ , $\sigma^\pm$ for $\Delta m=\pm1$ ), $\Delta s=0$ , any $\Delta n$ .
With spin–orbit coupling the good labels are $j,m_j$ : the rules become $\Delta j=0,\pm1$ with $j=0\to j=0$ strictly forbidden, $\Delta m_j=0,\pm1$ , parity still flips.
The machinery is the Wigner–Eckart theorem: $\mathbf{r}$ is a rank-1 spherical tensor, so its matrix elements inherit Clebsch–Gordan zeros wholesale; parity supplies the independent second rung.
We sweep all 768 angular elements $\langle\ell'm'|r_q|\ell m\rangle$ with $\ell,\ell'\leq3$ symbolically: nonzero exactly on $\Delta\ell=\pm1$ , $m'=m+q$ — the full-vector extension of the $\hat z$ -only sweep in Stark Effect.
Key idea: $\langle f|\mathbf{r}|i\rangle\neq0$ demands (1) a parity flip and (2) the triangle and projection conditions of a rank-1 tensor — together $\Delta\ell=\pm1$ , $\Delta m=0,\pm1$ , $\Delta s=0$ , and for coupled momenta $\Delta j=0,\pm1$ without $0\to0$ .

The core insight

The dipole matrix element $\langle f|\mathbf{r}|i\rangle$ is nonzero only when (1) parity changes and (2) the angular momenta satisfy the triangle and projection conditions of the Wigner–Eckart theorem for a rank-1 tensor. Together these give $\Delta\ell=\pm1$ , $\Delta m=0,\pm1$ , $\Delta s=0$ , parity flip — and $\Delta j=0,\pm1$ with $j=0\to j=0$ forbidden.

Where this lives today

Selection rules are working infrastructure, not history. Every optical atomic clock is built on a deliberate violation hierarchy: a strong E1-allowed line for cooling and readout, and a clock transition chosen precisely because E1 is forbidden — strontium's $^1S_0\to{}^3P_0$ intercombination line owes its millihertz-class natural width to a double ban (spin flip and $J=0\to J=0$ ) that only hyperfine mixing slightly relaxes. Trapped-ion qubits store quantum information in states connected only by forbidden transitions, so spontaneous emission cannot scramble them between gates. And solar astronomers still run Laporte's grammar in reverse: the polarization dictionary of $\pi$ and $\sigma$ components (built in Zeeman Effect, counted in Weak-Field Zeeman Effect) turns spectropolarimetry of magnetically split lines into vector maps of the Sun's field. When a spectral catalogue marks a line with square brackets — [O III] — the brackets are the astronomer's shorthand for "E1-forbidden," a flag planted by the rules assembled here and explored through the slower doors next.

The physical picture

Think of the atom's level diagram as a city map and E1 photons as a bus line with strict boarding rules. A bus departs only between stops whose parity differs — $s\leftrightarrow p$ , $p\leftrightarrow d$ — and only if the angular momentum changes by exactly one notch. Most of the map is therefore unreachable in a single ride: $3s$ cannot drop straight to $1s$ or $2s$ (same parity, $\Delta\ell=0$ ), and $3d$ cannot reach $2s$ ( $\Delta\ell=2$ ). The state $2s$ is the map's famous stranded passenger: the only level below it is $1s$ , and $s\to s$ has no E1 bus at all. It waits $0.12\ \mathrm{s}$ — about $10^{8}$ 2p-lifetimes — until a two-photon backdoor lets it out (Forbidden Transitions: Magnetic Dipole and Electric Quadrupole owns those slower doors).

Figure

The hydrogen E1 route atlas: every allowed decay changes ℓ by one — and 2s is a dead end

The second half of the story is the magnetic quantum number. The three spherical components of $\mathbf{r}$ are three different machines: $z$ leaves $m$ alone, while $x\pm iy$ raise or lower it by one. A photon carries one unit of angular momentum, and which component of that unit it takes from the atom is stamped on its polarization — the $\Delta m=0$ channel radiates $\pi$ -polarized light, the $\Delta m=\pm1$ channels radiate $\sigma^\pm$ . The radiation geometry — which channel you see from which direction, dark axes and circular rings — is the dictionary Zeeman Effect built; what we derive here is the operator side: exactly which $m$ -channels exist and why no others do.

Figure

Three channels, one photon: the spherical components of r sort Δm by polarization

From rate formula to gatekeeper

Everything upstream of this article funnels into one object. In the dipole approximation the atom–field coupling is $-\mathbf{d}\cdot\mathbf{E}$ with $\mathbf{d}=-e\mathbf{r}$ (The Dipole Approximation derives when that replacement is legitimate), and first-order time-dependent perturbation theory makes every transition amplitude proportional to the matrix element of the perturbation between the unperturbed states (First-Order Transition Amplitudes and Probabilities carries the general amplitude). For light driving an atom, that means one factor decides all:

W_{i\to f}\;\propto\;\bigl|\langle f|\mathbf{r}|i\rangle\bigr|^2

where: $|i\rangle,|f\rangle$ are unperturbed atomic eigenstates and the proportionality constant carries the field strength or — for spontaneous decay — the $\omega^3$ phase-space factor assembled in Spontaneous Emission and the Einstein A Coefficient. The physics of this section is binary: either symmetry permits $\langle f|\mathbf{r}|i\rangle\neq0$ , or the E1 rate is exactly zero no matter how bright the lamp. To interrogate the matrix element, resolve the vector into its spherical components, which are rescaled $\ell=1$ spherical harmonics:

r_0=z=r\sqrt{\tfrac{4\pi}{3}}\,Y_1^0,\qquad r_{\pm1}=\mp\tfrac{x\pm iy}{\sqrt2}=r\sqrt{\tfrac{4\pi}{3}}\,Y_1^{\pm1}

where: $q=0,\pm1$ labels the projection of the one unit of angular momentum the operator carries. This is the entire secret: $\mathbf{r}$ is a rank-1 spherical tensor, so its matrix elements between $|\ell m\rangle$ states are Gaunt integrals of three spherical harmonics — objects whose zeros are pure geometry. A hydrogenic matrix element factorizes as

\langle n'\ell'm'|r_q|n\ell m\rangle=\underbrace{\int_0^\infty R_{n'\ell'}\,r\,R_{n\ell}\,r^2\,dr}_{\text{radial: sets the size}}\times\underbrace{\sqrt{\tfrac{4\pi}{3}}\int Y_{\ell'}^{m'*}\,Y_1^q\,Y_\ell^m\,d\Omega}_{\text{angular: sets the rules}}

where: the radial factor is generically nonzero for any $n,n'$ — which is why there is no $\Delta n$ selection rule — while the angular factor enforces every rule below. Selection rules live entirely in the angular integral.

The rule set, assembled

Stark Effect proved the $q=0$ half of the story as its parity selection theorem — $\langle n\ell m|\hat z|n'\ell'm'\rangle\neq0$ demands $\Delta\ell=\pm1$ and $\Delta m=0$ , verified there by sweeping "all 100 pairs with $\ell,\ell'\leq3$ " and anchored by the allowed witness $\langle100|\hat z|210\rangle=128\sqrt2/243\,a_0$ . The derivation below is the full-vector completion: all three components at once, which adds the $\sigma^\pm$ sector that a static field along $z$ never probes.

Derivation AThe Complete E1 Rule Set by Full-Vector Sweep

We derive the electric-dipole selection rules for all three spherical components of $\mathbf{r}$ and certify them by brute-force symbolic integration. The strategy has three independent rungs — an azimuthal integral for $\Delta m$ , a parity argument for " $\Delta\ell$ odd", and the rank-1 triangle for " $|\Delta\ell|\leq1$ " — each killing part of the possibility space; their intersection is the rule set. The trap to watch: parity alone would happily allow $\Delta\ell=\pm3$ , and the triangle alone would allow $\Delta\ell=0$ ; only together do they force exactly $\Delta\ell=\pm1$ .

Step 1. The azimuthal rung. Every factor in the angular integral carries its $\phi$ -dependence as a pure exponential: $Y_\ell^m\propto e^{im\phi}$ , $Y_1^q\propto e^{iq\phi}$ , $Y_{\ell'}^{m'*}\propto e^{-im'\phi}$ . The $\phi$ integral is $\int_0^{2\pi}e^{i(m+q-m')\phi}d\phi=2\pi\,\delta_{m',m+q}$ . Since $q$ only takes the values $0,\pm1$ , this is already the projection rule: $\Delta m=0,\pm1$ , with the channel labeled by which component acted — $z$ gives $\pi$ ( $\Delta m=0$ ), $x\pm iy$ give $\sigma^\pm$ ( $\Delta m=\pm1$ ). Why the photon gets the credit: the operator $r_q$ transfers exactly $q\hbar$ of $z$ -angular momentum to the field. There is no $q=\pm2$ component of a vector, so a single E1 photon can never carry $\Delta m=\pm2$ .

Step 2. The parity rung. Under inversion $\hat P\mathbf{r}\hat P^{-1}=-\mathbf{r}$ , and central-potential states have definite parity $(-1)^\ell$ — the antipodal-map theorem catalogued in Spherical Harmonics. Sandwiching, $\langle f|\mathbf{r}|i\rangle=-(-1)^{\ell+\ell'}\langle f|\mathbf{r}|i\rangle$ , which forces $\ell+\ell'$ odd: parity must flip, and $\Delta\ell$ must be odd. This kills $\Delta\ell=0$ unconditionally — the deeper of the two reasons the 2s→1s door is shut.

Step 3. The triangle rung. The product $Y_1^qY_\ell^m$ decomposes into harmonics of degree $\ell-1$ , $\ell$ , $\ell+1$ only — multiplying by one unit of angular momentum cannot reach further, the same Clebsch–Gordan geometry the Wigner–Eckart theorem packages (next derivation). Orthogonality of the $Y_{\ell'}^{m'}$ then demands $\ell'\in\{\ell-1,\ell,\ell+1\}$ . Intersect with Step 2: $\ell'=\ell$ is parity-killed, $\ell'=\ell\pm3$ is triangle-killed, leaving exactly $\Delta\ell=\pm1$ .

Step 4. The sweep. We certify the intersection by computing every angular integral $\sqrt{4\pi/3}\int Y_{\ell'}^{m'*}Y_1^qY_\ell^m\,d\Omega$ for $\ell,\ell'\leq3$ , all $m,m'$ , all $q$ — 768 elements counting projections. The nonzero set is exactly $\{\Delta\ell=\pm1,\ m'=m+q\}$ : 54 surviving elements, every one with odd $\Delta\ell$ , and the parity-allowed but triangle-forbidden probe $\langle30|r_0|00\rangle$ vanishing on cue. This extends the $\hat z$ -only sweep of Stark Effect to the $\sigma^\pm$ sector it never needed. Verified: the 768-element symbolic sweep lands nonzero exactly on the allowed set (54 elements); the azimuthal integral returns $2\pi\delta_{m',m+q}$ ; $\Delta\ell=3$ probe $=0$ ; every surviving element has odd $\Delta\ell$ — verify script, checks 1–4.

The triangle rung deserves its proper name. The Wigner–Eckart Theorem proved that matrix elements of any spherical tensor factorize as $\langle\alpha'j'm'|T^k_q|\alpha jm\rangle=\langle jm;kq|j'm'\rangle\langle\alpha'j'\|T^k\|\alpha j\rangle/\sqrt{2j'+1}$ — "a Clebsch–Gordan factor carrying all orientation dependence times one reduced matrix element carrying all dynamics — selection rules are the CG factor's zeros," as that article puts it. For $\mathbf{r}$ , take $k=1$ : the CG factor $\langle jm;1q|j'm'\rangle$ vanishes unless the triangle $|j-1|\leq j'\leq j+1$ and projection $m'=m+q$ hold. Every rule in this article is one of those zeros plus the parity rung, which Wigner–Eckart does not know about — the theorem is blind to inversion, which is why both rungs are needed.

\langle n'\ell'm'|r_q|n\ell m\rangle=\langle \ell m;1q|\ell'm'\rangle\,\frac{\langle n'\ell'\|r\|n\ell\rangle}{\sqrt{2\ell'+1}}

where: the reduced element $\langle n'\ell'\|r\|n\ell\rangle$ is one number per $(n\ell,n'\ell')$ pair — compute the $\pi$ route and the $\sigma^\pm$ strengths come free, the ratio method The Wigner–Eckart Theorem runs "on a dipole block, end to end." That is why Example 1 below finds all three 2p→1s channels equal in strength: they share one reduced element and differ only in CG factors of equal magnitude.

Go deeper · Derivation B · 2 minSpin, Total J, and the Zero-to-Zero Ban

Derivation BSpin, Total J, and the Zero-to-Zero Ban

We extend the rules from orbital labels to the labels nature actually uses once fine structure couples $\mathbf L$ and $\mathbf S$ . The goal is three results: the spin spectator rule $\Delta s=0$ , the coupled rule $\Delta j=0,\pm1$ with $\Delta m_j=0,\pm1$ , and the strict ban on $j=0\to j'=0$ . The subtlety worth watching is that $\Delta j=0$ becomes legal even though $\Delta\ell=0$ never is — the parity rung acts on $\ell$ , not on $j$ .

Step 1. Spin is a spectator. The operator $\mathbf{r}$ acts on coordinates only, so on a product state the matrix element factorizes: $\langle f|\mathbf r|i\rangle=\langle\text{space}_f|\mathbf r|\text{space}_i\rangle\,\langle\chi_f|\chi_i\rangle$ . Orthogonality of spin states, $\langle\uparrow|\downarrow\rangle=0$ , kills any spin-flip: $\Delta s=0$ and $\Delta m_s=0$ at the E1 level. Two lines, as promised.

Step 2. Recouple to j. In the $|n\,\ell\,s\,j\,m_j\rangle$ basis, $\mathbf r$ is still a rank-1 tensor under total rotations (it commutes with nothing that would change its rank), so Wigner–Eckart applies verbatim with $j$ in place of $\ell$ : the CG factor $\langle jm_j;1q|j'm_j'\rangle$ demands $\Delta j=0,\pm1$ and $\Delta m_j=0,\pm1$ . Parity, which lives on $\ell$ , independently still demands $\Delta\ell=\pm1$ — so a $\Delta j=0$ line like $3p_{3/2}\to2s_{1/2}$ ? No: check, $\Delta j=1$ there; a true $\Delta j=0$ example is $3d_{3/2}\to2p_{3/2}$ , allowed because $\Delta\ell=1$ carries the parity flip while $j$ stays put.

Step 3. The zero-to-zero ban. For $j=0\to j'=0$ the CG factor is $\langle 00;1q|00\rangle$ , which vanishes for every $q$ : you cannot add one unit of angular momentum to zero and end at zero — the triangle $|0-1|\leq0\leq0+1$ closes to the single point $j'=1$ . The ban is absolute at E1 and survives to all single-photon multipoles (a photon always carries at least one unit); $0\to0$ transitions proceed only by two photons or electron conversion — traffic handled in Forbidden Transitions: Magnetic Dipole and Electric Quadrupole.

Step 4. One more floor: hyperfine F. Coupling the nuclear spin, $\mathbf F=\mathbf I+\mathbf J$ is just another angular-momentum addition (Hyperfine Structure builds it), and the same theorem stamps $\Delta F=0,\pm1$ with $F=0\to F'=0$ forbidden. Every floor of the coupling hierarchy inherits the same rank-1 grammar. Verified: CG sweep over $j,j'\in\{0,\tfrac12,1,\tfrac32,2\}$ — nonzero exactly on the triangle with integer $\Delta j$ , and $\langle00;1q|00\rangle=0$ for all $q$ ; spin-flip elements vanish on $\langle\uparrow|\downarrow\rangle=0$ — verify script, checks 5–6.

Limiting cases

case	rule status	meaning
any $\Delta n$ , including $\Delta n=0$	allowed	radial integral generically nonzero; rules are purely angular (2p→2s transitions exist — at radio frequencies)
weak magnetic field	rules hold in $(j,m_j)$	Zeeman splitting relabels lines; counting per Weak-Field Zeeman Effect: sodium D₁→4, D₂→6 components
strong field (Paschen–Back)	rules hold in $(m_\ell,m_s)$	$\Delta m_\ell=0,\pm1$ , $\Delta m_s=0$ — same grammar, decoupled labels
multi-electron LS coupling	add $\Delta S=0$ , $\Delta L=0,\pm1$ , $L=0\!\nrightarrow\!0$	approximate: spin–orbit mixing lets "intercombination" lines through weakly (Hg 253.7 nm)
$j=0\to j'=0$	forbidden at every single-photon multipole	a photon carries $\geq1\hbar$ ; only two-photon or conversion routes remain
E1 dead ( $\Delta\ell=0,\pm2$ …)	M1/E2 doors open	rates down by $\sim(\alpha Z)^2$ and worse — the subject of the next article

Worked examples

Example 1 — the 2p→1s block, all three channels. Compute every E1 route out of hydrogen 2p and check the isotropy that Spontaneous Emission and the Einstein A Coefficient silently uses when it averages $|\langle f|\mathbf r|i\rangle|^2$ over $m$ .

Step 1. Enumerate. From $2p_m$ ( $m=-1,0,+1$ ) the only lower level is 1s ( $\ell=0$ ), reached by the component $q=-m$ : three routes, $2p_{-1}\xrightarrow{\sigma^-}1s$ , $2p_0\xrightarrow{\pi}1s$ , $2p_{+1}\xrightarrow{\sigma^+}1s$ .

Step 2. Apply. The shared radial integral times the three angular factors gives $|\langle100|r_q|21m\rangle|=128\sqrt2/243\,a_0=0.7449\,a_0$ for each route — the same number Stark Effect pinned for the $\pi$ element, now certified for the $\sigma^\pm$ elements too (one reduced element, equal-magnitude CG factors).

Step 3. Boundary number. The rotational invariant is $\sum_q|\langle100|r_q|21m\rangle|^2=(128\sqrt2/243)^2a_0^2=0.5549\,a_0^2$ , independent of $m$ — so the $m$ -averaged $\frac13\sum$ that enters the $A$ coefficient equals any single route's strength. Every 2p sublevel decays at the same 1.6 ns rate, and the three polarization channels split the photons evenly.

Example 2 — how many lines hide in Balmer-α? The $n=3\to n=2$ line at 656.3 nm looks like one line; fine structure resolves it into components. Count them.

Step 1. Enumerate the fine-structure levels: $n=3$ carries $3s_{1/2},3p_{1/2},3p_{3/2},3d_{3/2},3d_{5/2}$ ; $n=2$ carries $2s_{1/2},2p_{1/2},2p_{3/2}$ .

Step 2. Apply both rungs to all $5\times3=15$ pairs: parity demands $\Delta\ell=\pm1$ (s↔p, p↔s, d↔p), the triangle demands $\Delta j=0,\pm1$ . Survivors: $3s_{1/2}\to2p_{1/2}$ , $3s_{1/2}\to2p_{3/2}$ , $3p_{1/2}\to2s_{1/2}$ , $3p_{3/2}\to2s_{1/2}$ , $3d_{3/2}\to2p_{1/2}$ , $3d_{3/2}\to2p_{3/2}$ , $3d_{5/2}\to2p_{3/2}$ .

Step 3. Boundary check: $3d_{5/2}\to2p_{1/2}$ dies ( $\Delta j=2$ ), $3p\to2p$ dies (parity), $3d\to2s$ dies ( $\Delta\ell=2$ ). Exactly 7 components — the classic Balmer-α fine-structure multiplet, machine-enumerated in the verify script. (The famous Lamb shift lives inside this multiplet, splitting $2s_{1/2}$ from $2p_{1/2}$ .)

The rule that fell — and the one that stands

Origins · 1 minParity and how this was discovered

Parity

1924–1957 · Ann Arbor, Princeton, Washington

Parity entered spectroscopy as bookkeeping. In 1924 Otto Laporte, cataloguing the bewildering iron spectrum, noticed its terms split into two families — gestrichene and ungestrichene, primed and unprimed — with every strong line jumping between families, never within one. Nobody knew why. In 1927 the young Eugene Wigner supplied the reason: the two families are eigenstates of spatial inversion with eigenvalues $\pm1$ , the dipole operator is inversion-odd, and Laporte's empirical rule is just parity conservation in electromagnetism — one of the first theorems ever extracted from a symmetry group in quantum mechanics. The concept seemed untouchable until 1956, when Lee and Yang noticed it had never been tested in weak interactions and Chien-Shiung Wu's cobalt-60 experiment found beta electrons preferring one hemisphere — parity falls in weak decays, published January 1957. The electromagnetic verdict, though, still stands unbroken: every E1 photon ever measured has changed its atom's parity, exactly as the iron lines whispered in 1924.

History

1918

— Rubinowicz derives Δm = 0, ±1 and the π/σ polarization assignments from angular-momentum conservation, pre-quantum-mechanically

1924

— Laporte sorts the iron spectrum into two term families; strong lines only cross between families

1925

— Goudsmit and Uhlenbeck's electron spin makes the fine-structure doublets and their j-rules intelligible

1927

— Wigner derives the Laporte rule from inversion symmetry: parity becomes a quantum number

1935

— Condon and Shortley's monograph codifies the full E1 rule set for complex spectra

1957

— Wu's cobalt-60 result (run December 1956) breaks parity in weak decays; electromagnetic selection rules survive untouched

Going further

References: Griffiths, Introduction to Quantum Mechanics, 3rd ed., Ch. 11 (spontaneous emission and selection rules) · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., Ch. 5 (Wigner–Eckart theorem; radiative transitions) · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 18 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. II, complement on the dipole approximation · Townsend, A Modern Approach to Quantum Mechanics, 2nd ed., Ch. 14 · Bransden & Joachain, Physics of Atoms and Molecules, 2nd ed., Ch. 4.

Provenance note (verbatim from the source corpus): Selection rules are standard textbook results. A verifier should: (1) reproduce the parity argument $\hat P\mathbf r\hat P^{-1}=-\mathbf r$ giving $\ell_f+\ell_i$ odd; (2) confirm the azimuthal integral $\int_0^{2\pi}e^{i(m_i+q-m_f)\phi}\mathrm d\phi=2\pi\delta_{m_f,m_i+q}$ yielding $\Delta m=0,\pm1$ ; (3) verify the Wigner–Eckart triangle rule for a rank-1 tensor gives $\Delta J=0,\pm1$ and the vanishing of $\langle 0\,0;1\,q|0\,0\rangle$ . Numbers cited: hydrogen $2p$ lifetime $\approx1.6$ ns and $2s$ two-photon lifetime $\approx0.12$ s (standard atomic-physics values); [O III] forbidden lines at 495.9 and 500.7 nm. Cross-check against Griffiths Ch. 11, Sakurai Ch. 5, Bransden & Joachain Ch. 4, Cohen-Tannoudji Vol. II. All of it is done in this article's verify script: the parity sandwich, the azimuthal delta, the CG triangle with the $0\to0$ zero, and the lifetime pair.

Prerequisites: The Dipole Approximation · First-Order Transition Amplitudes and Probabilities · Semiclassical Theory of Atom–Radiation Interaction · Harmonic Perturbations and Resonance. Leads to: Absorption and Stimulated Emission · Fermi's Golden Rule · Density of Final States — and directly onward to Forbidden Transitions: Magnetic Dipole and Electric Quadrupole, where the shut doors reopen slowly.

Why exactly Δℓ = ±1 and not just any odd change

Two separate executioners. Parity only demands odd $\Delta\ell$ — it would allow $\pm3$ . But the operator carries exactly one unit of angular momentum ( $r_q\propto rY_1^q$ ), and coupling $\ell\otimes1$ reaches only $\ell-1,\ell,\ell+1$ : the triangle kills $\pm3$ and beyond, parity kills $0$ and the survivors are $\pm1$ . The verify script's probe $\langle\ell'=3,m'=0|r_0|\ell=0,m=0\rangle$ — parity-legal, triangle-dead — computes to exactly zero.

Is there really no restriction on n at all

None. The radial integral $\int R_{n'\ell'}\,r\,R_{n\ell}\,r^2dr$ has no symmetry to kill it and is generically nonzero for every pair — Lyman-α ( $\Delta n=1$ ), Lyman-β ( $\Delta n=2$ ), and even $\Delta n=0$ transitions like 2s→2p (driven at radio frequencies) all proceed. What $\Delta n$ controls is the size of the matrix element and the $\omega^3$ factor in rates, not the permission.

If E1 is forbidden, how does the state ever decay

Through the terms the dipole approximation dropped. The next order of $e^{i\mathbf k\cdot\mathbf r}$ generates magnetic-dipole (M1) and electric-quadrupole (E2) operators with their own rule sets — Δℓ = 0 becomes legal for M1, Δℓ = ±2 for E2 — at rates suppressed by roughly $(\alpha Z)^2$ . When even those die (hydrogen 2s: E1 dead on parity, M1 dead on radial orthogonality), two-photon emission takes over. Forbidden Transitions: Magnetic Dipole and Electric Quadrupole opens each door in turn.

Do the rules survive when I turn on a magnetic field

Yes — the grammar is rotational and inversion symmetry of the atom's internal dynamics, and a uniform field only relabels the good basis. Weak field: rules read $\Delta j=0,\pm1$ , $\Delta m_j=0,\pm1$ on Zeeman-split sublevels — that is exactly how Weak-Field Zeeman Effect counts sodium D₁ into 4 and D₂ into 6 components. Strong field (Paschen–Back): the labels decouple to $(m_\ell,m_s)$ and the rules become $\Delta m_\ell=0,\pm1$ with $\Delta m_s=0$ . Same theorem, different wardrobe.

What does Δs = 0 mean for real atoms — I have seen spin-forbidden lines in catalogues

$\Delta s=0$ is exact only while spin genuinely decouples from space. Spin–orbit coupling mixes $S$ values in heavy atoms, so "spin-forbidden" intercombination lines exist but are weak — mercury's famous 253.7 nm $^3P_1\to{}^1S_0$ line runs on a small singlet admixture in the triplet state. The heavier the atom, the stronger the mixing, the brighter the intercombination line: a $Z$ -dependent leak, not a broken theorem.

Why is j = 0 to j = 0 banned even for the multipoles that allow Δj = 0

Because the ban is not about $\Delta j$ ; it is about the photon's minimum cargo. A single photon carries at least one unit of angular momentum, and there is no way to add $\geq1$ to $j=0$ and land on $j'=0$ — the CG coefficient $\langle00;kq|00\rangle$ vanishes for every $k\geq1$ . Two photons can conspire (their momenta can cancel), which is why $0\to0$ transitions are the two-photon channel's signature turf.

Why do all three 2p sublevels decay at exactly the same rate

Rotational invariance. The Wigner–Eckart factorization gives every route one shared reduced matrix element; summing $|CG|^2$ over the photon channels out of any $m$ gives the same total. Example 1 exhibits it numerically: $\sum_q|\langle1s|r_q|2p_m\rangle|^2=0.5549\,a_0^2$ for $m=-1,0,+1$ alike. An excited atom has no memory of which way is up unless something external breaks the tie.

Do diagonal matrix elements of r ever survive — a permanent dipole moment

Never, for parity eigenstates: $\langle n\ell m|\mathbf r|n\ell m\rangle$ has $\Delta\ell=0$ , so parity kills it — the corollary Stark Effect states as "every permanent dipole vanishes." Polar molecules dodge this by having nearly degenerate opposite-parity states whose superpositions are long-lived; an isolated atom in an energy eigenstate cannot.