Selection Rules for Molecular Transitions

A molecule holds energy in a ladder of rotational, vibrational, and electronic levels, and in principle any pair could be joined by a photon. In practice almost none are: point a spectrometer at a gas and you see a sparse forest of sharp lines in a sea of silence. Selection rules are the bookkeeping of which transitions appear. They take a single result from quantum mechanics: by time-dependent perturbation theory and Fermi's golden rule, the rate of an electric-dipole transition is proportional to the modulus squared of one integral, the transition-dipole moment ∣⟨ψf∣μ^∣ψi⟩∣2|\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle|^2. Whenever symmetry forces that integral to vanish, the transition is forbidden to first order and the line is dark. Two further inputs make the integral tractable. The Born–Oppenheimer approximation factorizes the state into electronic, vibrational, and rotational parts, so the three kinds of rule follow almost independently — a permanent dipole for pure rotation, a changing dipole for vibration, orbital, spin, and parity relationships for electronic jumps. Group theory supplies the rest: an integral over all space survives only if its integrand contains the totally symmetric representation, turning the vanishing of the transition moment into a statement about parity, angular momentum, and spin. The rules are never absolute — they hold to the extent their approximations do, and their controlled failures (overtones, phosphorescence, the forbidden atmospheric oxygen bands) are as informative as the rules themselves.

Key takeaways
  • A transition is electric-dipole allowed only if the transition moment ⟨ψf∣μ^∣ψi⟩\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle is nonzero; symmetry decides when it vanishes, and the intensity goes as its modulus squared.
  • Pure-rotation (microwave) absorption needs a permanent dipole μ0≠0\mu_0\ne0 and obeys ΔJ=±1\Delta J=\pm1; homonuclear diatomics like N2\mathrm{N_2} and O2\mathrm{O_2} have no rotational or infrared fundamental spectrum.
  • Vibrational (infrared) transitions need a changing dipole (dμ/dq)0≠0(\mathrm{d}\mu/\mathrm{d}q)_0\ne0; the harmonic rule Δv=±1\Delta v=\pm1 relaxes into weak overtones Δv=±2,±3,…\Delta v=\pm2,\pm3,\dots through anharmonicity.
  • Electronic transitions in diatomics obey ΔΛ=0,±1\Delta\Lambda=0,\pm1, ΔS=0\Delta S=0, the Laporte rule g↔ug\leftrightarrow u, and Σ+ ⁣↔ ⁣Σ+\Sigma^+\!\leftrightarrow\!\Sigma^+, Σ− ⁣↔ ⁣Σ−\Sigma^-\!\leftrightarrow\!\Sigma^-; the Franck–Condon factor sets the vibrational envelope.
  • Raman scattering complements the infrared: it needs a changing polarizability (dα/dq)0≠0(\mathrm{d}\alpha/\mathrm{d}q)_0\ne0, giving ΔJ=0,±2\Delta J=0,\pm2, so infrared-silent modes can still be Raman-active.
  • Key idea: a transition is electric-dipole allowed if and only if ⟨ψf∣μ^∣ψi⟩≠0\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle\ne0; parity, angular momentum, and spin symmetry tell you exactly when that integral is forced to zero.
The core insight

A transition is electric-dipole allowed if and only if the transition moment ⟨ψf∣μ^∣ψi⟩\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle is nonzero; symmetry — parity, angular momentum, spin — tells you when this single integral vanishes, and its modulus squared sets the line intensity.

Where this lives today

Selection rules are the silent grammar of modern molecular measurement. In radio astronomy the ΔJ=±1\Delta J=\pm1 rule makes the CO line at 115 GHz115\ \mathrm{GHz} the workhorse tracer of cold gas: because H2\mathrm{H_2} has no permanent dipole and no rotational spectrum, astronomers cannot see the dominant molecule in a cloud directly and instead map its faint CO companion in telescopes such as ALMA. The same rules explain the greenhouse effect — N2\mathrm{N_2} and O2\mathrm{O_2} are infrared-transparent because their fundamentals are dipole-forbidden, while CO2\mathrm{CO_2}, H2O\mathrm{H_2O}, and CH4\mathrm{CH_4} absorb heat through dipole-active modes that satellite spectrometers invert to retrieve concentrations. The dipole–polarizability complementarity underpins a whole toolkit: infrared and Raman spectrometers sit side by side because a mode dark to one is often bright to the other, and the mutual exclusion rule reads out whether a molecule has a centre of inversion. The controlled failures matter as much — organic light-emitting diodes engineer spin-orbit coupling to harvest the spin-forbidden triplet emission, and the forbidden oxygen bands sense the oxygen column of Earth and of exoplanets. Reading a spectrum is, in the end, reading which rules a molecule obeys and which it is gently breaking.

The physical picture

Light couples to the motion of charge. An oscillating field exchanges energy with a molecule most effectively when the molecule's dipole is changing at the driving frequency. If a motion leaves the dipole unchanged — the rotation of homonuclear N2\mathrm{N_2}, or the symmetric stretch of CO2\mathrm{CO_2} where both bonds lengthen in lockstep — the field finds no "handle" and the transition is dark. So the first question for any level pair is whether the molecule presents a dipole handle to the light; the first figure sorts two familiar molecules on that test.

Figure
The dipole handle sorts which motions couple to light. HCl carries a permanent dipole μ0 ≈ 1.08 D, so its rotation modulates the dipole and it absorbs in the microwave with ΔJ = ±1. Homonuclear N₂ has μ = 0 and dμ/dq = 0, so it has no microwave and no infrared fundamental
Does the motion modulate the dipole? That is the first selection rule HCl — heteronuclear N₂ — homonuclear H Cl δ+ δ− μ₀ ≈ 1.08 D rotation turns the dipole → ΔJ = ±1 microwave active N N μ = 0 · dμ/dq = 0 no microwave, no IR fundamental electric-dipole dark

Once a molecule passes the dipole test, symmetry does the fine bookkeeping. Integrating an odd function over all space gives zero, and the dipole operator is odd under inversion, so it connects only states of opposite parity — the Laporte rule g↔ug\leftrightarrow u. The angular part of the dipole is a rank-1 spherical tensor, so it changes the rotational quantum number by at most one unit, giving ΔJ=±1\Delta J=\pm1 and a ladder of equally spaced lines; spin lives in a sector the spatial dipole cannot touch, hence ΔS=0\Delta S=0. The rules form a hierarchy: parity and spin rest on good symmetries and are nearly rigorous, whereas Δv=±1\Delta v=\pm1 leans on the harmonic idealization and leaks when it fails. The second figure shows the reward for a molecule that passes every test — the clean rotational spectrum of carbon monoxide.

Figure
The pure-rotation spectrum of CO is a comb of equally spaced lines at 2B(J+1). With B ≈ 1.93 cm⁻¹ the spacing is 2B ≈ 3.86 cm⁻¹, and the first line J = 0 → 1 sits at 115.8 GHz (2.6 mm), the famous CO line mapped by radio telescopes
CO rotational spectrum: equal spacing 2B ≈ 3.86 cm⁻¹, first line at 115.8 GHz 3.86 7.72 11.59 15.45 19.31 23.17 transition wavenumber (cm⁻¹) absorption intensity 2B = 3.86 cm⁻¹ J = 0 → 1 115.8 GHz · 2.6 mm J = 2 → 3

Picture the three domains as filters in series: a rovibronic line survives only if it passes the electronic, vibrational (Franck–Condon), and rotational rules at once. That product of independent gates is what you resolve in a high-resolution band, and reading it backwards reconstructs the geometry of a molecule far too small to see.

The core equations

Everything begins with one quantity. Fermi's golden rule makes the transition rate the modulus squared of the perturbation matrix element, and in the electric-dipole approximation the perturbation is −μ^⋅E-\hat{\boldsymbol\mu}\cdot\mathbf E. The whole subject is when that one integral is nonzero.

Wi→f∝∣⟨ψf∣μ^∣ψi⟩∣2,μ^=−e∑krk+∑nZne Rn(1)W_{i\to f}\propto |\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle|^2,\qquad \hat{\boldsymbol\mu}=-e\sum_k \mathbf r_k+\sum_n Z_n e\,\mathbf R_n \qquad (1)

where: Wi→fW_{i\to f} — electric-dipole transition rate; μ^\hat{\boldsymbol\mu} — dipole operator over electrons and nuclei; rk,Rn\mathbf r_k,\mathbf R_n — electron and nuclear positions; e,Zne,Z_n — elementary and nuclear charge. A symmetry that forces this single number to zero forbids the line outright, whatever the photon energy. The integral runs over all space, so its value is invariant under every symmetry operation of the molecule. Only the totally symmetric part of the integrand survives, and group theory makes that explicit.

⟨ψf∣μ^∣ψi⟩≠0  ⟺  Γf⊗Γμ⊗Γi⊃Γsym(2)\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle\ne0 \iff \Gamma_f\otimes\Gamma_\mu\otimes\Gamma_i \supset \Gamma_{\text{sym}} \qquad (2)

where: Γi,Γf\Gamma_i,\Gamma_f — representations of the states; Γμ\Gamma_\mu — representation spanned by (x,y,z)(x,y,z); Γsym\Gamma_{\text{sym}} — totally symmetric representation. The line is allowed only if Γf⊗Γi\Gamma_f\otimes\Gamma_i contains a Cartesian component of the dipole — the one criterion behind the parity, angular-momentum, and spin rules alike. Apply the criterion to a rotating molecule with a body-fixed dipole. The dipole must be permanent, and the rank-1 character of its angular part limits JJ to a single-unit change.

μ0≠0,ΔJ=±1,ΔMJ=0,±1(3)\mu_0\ne0,\qquad \Delta J=\pm1,\qquad \Delta M_J=0,\pm1 \qquad (3)

where: μ0\mu_0 — permanent dipole magnitude; JJ — rotational quantum number; MJM_J — projection of J\mathbf J on the field axis. No permanent dipole means no microwave spectrum; where one exists, the lines land at 2B(J+1)2B(J+1) with uniform spacing 2B2B. For vibration the question is whether the dipole changes as the bond stretches. On expanding μ(q)\mu(q), the constant term drives no transition and the linear term carries the fundamental.

(dμdq)0≠0,Δv=±1 (harmonic)(4)\left(\frac{\mathrm{d}\mu}{\mathrm{d}q}\right)_0\ne0,\qquad \Delta v=\pm1\ (\text{harmonic}) \qquad (4)

where: qq — normal-mode coordinate; vv — vibrational quantum number; (dμ/dq)0(\mathrm{d}\mu/\mathrm{d}q)_0 — dipole slope at equilibrium. Infrared activity needs a changing dipole, not a permanent one — which is why CO2\mathrm{CO_2}, with μ0=0\mu_0=0, still absorbs through its bending and antisymmetric-stretch modes. Electronic transitions in a diatomic must respect its good quantum numbers. These are the orbital-angular-momentum projection, the total spin, the inversion parity, and the reflection label of the Σ\Sigma states.

ΔΛ=0,±1;ΔS=0;g↔u;Σ+ ⁣↔ ⁣Σ+, Σ− ⁣↔ ⁣Σ−(5)\Delta\Lambda=0,\pm1;\quad \Delta S=0;\quad g\leftrightarrow u;\quad \Sigma^+\!\leftrightarrow\!\Sigma^+,\ \Sigma^-\!\leftrightarrow\!\Sigma^- \qquad (5)

where: Λ\Lambda — orbital-angular-momentum projection on the axis; SS — total electronic spin; g/ug/u — inversion parity (homonuclear molecules); Σ±\Sigma^\pm — reflection label of a Σ\Sigma state. These pick out the strong UV–visible bands and mark the spin-forbidden ones that survive only as weak phosphorescence. A band's vibrational structure is set separately. The nuclear part of the dipole matrix element factors out under the Condon approximation, leaving the intensity set by the overlap of the two vibrational wavefunctions.

Iv′ ⁣←v′′∝∣⟨χv′∣χv′′⟩∣2(6)I_{v'\!\leftarrow v''}\propto |\langle\chi_{v'}|\chi_{v''}\rangle|^2 \qquad (6)

where: χv\chi_v — vibrational wavefunction of an electronic state; II — relative vibronic-band intensity; ∣⟨χv′∣χv′′⟩∣2|\langle\chi_{v'}|\chi_{v''}\rangle|^2 — the Franck–Condon factor. The vibrational progression pictures how far the equilibrium bond length shifts between the two states. Raman scattering swaps the operator. It is governed by the polarizability, a second-rank tensor, so its rules mirror the infrared ones and reach what the dipole cannot.

(dαdq)0≠0,ΔJ=0,±2,Δv=±1(7)\left(\frac{\mathrm{d}\alpha}{\mathrm{d}q}\right)_0\ne0,\qquad \Delta J=0,\pm2,\qquad \Delta v=\pm1 \qquad (7)

where: α\alpha — polarizability tensor; (dα/dq)0(\mathrm{d}\alpha/\mathrm{d}q)_0 — its derivative at equilibrium; ΔJ=±2\Delta J=\pm2 — the O and S branches. Because the operator is rank 2, rotational Raman lines are spaced by 4B4B, and infrared-silent modes like the CO2\mathrm{CO_2} symmetric stretch still scatter.

Derivation AWhy a symmetry can switch a transition off

The goal is to turn "forbidden by symmetry" into a precise, checkable criterion. Starting from the golden rule, we show the intensity is fixed by one integral, forced to zero whenever the product of state and operator symmetries lacks the totally symmetric representation — an argument identical for parity, angular momentum, and spin, so the named rules become special cases of one idea.

Step 1. Start from Fermi's golden rule, Wi→f=2πℏ∣⟨f∣H^′∣i⟩∣2ρ(Ef)W_{i\to f}=\tfrac{2\pi}{\hbar}|\langle f|\hat H'|i\rangle|^2\rho(E_f), and use the electric-dipole coupling H^′=−μ^⋅E\hat H'=-\hat{\boldsymbol\mu}\cdot\mathbf E, giving W∝∣⟨ψf∣μ^∣ψi⟩∣2W\propto|\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle|^2. Why: this isolates a single, symmetry-analyzable integral as the arbiter of allowedness.

Step 2. Any integral over all space is unchanged by every symmetry operation; a non-invariant quantity must equal something other than itself, hence zero. Why: this converts "is the integral nonzero" into "does the integrand have a totally symmetric part".

Step 3. The integrand ψf∗ μ^ ψi\psi_f^*\,\hat{\boldsymbol\mu}\,\psi_i transforms as Γf⊗Γμ⊗Γi\Gamma_f\otimes\Gamma_\mu\otimes\Gamma_i, which contains Γsym\Gamma_{\text{sym}} only if Γf⊗Γi\Gamma_f\otimes\Gamma_i contains a component xx, yy, or zz of Γμ\Gamma_\mu. Why: reduce the triple product and look for the symmetric representation — that is the whole recipe.

Step 4. For parity, μ^\hat{\boldsymbol\mu} is odd under inversion, so a nonzero integral demands opposite-parity states, g↔ug\leftrightarrow u; for spin, μ^\hat{\boldsymbol\mu} touches only spatial coordinates, so ΔS=0\Delta S=0. Why: the two most robust rules drop out at once — "rigorous" rules are those tied to exact symmetries of the operator.

Go deeper · Derivation B · 1 minThe rotational rule ΔJ = ±1 from three spherical harmonics
Derivation BThe rotational rule ΔJ = ±1 from three spherical harmonics

The goal is to derive the pure-rotation rule for a rigid 1Σ^1\Sigma diatomic by evaluating the angular integral in the transition moment. The rigid-rotor states are spherical harmonics ∣J,M⟩=YJM(θ,ϕ)|J,M\rangle=Y_J^M(\theta,\phi), and the body-fixed dipole projected onto laboratory axes is a rank-1 spherical tensor, so the matrix element reduces to an integral of three spherical harmonics whose vanishing conditions are known exactly.

Step 1. The body-fixed dipole μ0z^body\mu_0\hat{\mathbf z}_{\text{body}} has laboratory components going as Y1qY_1^{q} with q=0,±1q=0,\pm1, so the matrix element becomes ⟨J′M′∣μ^q∣JM⟩∝∫(YJ′M′)∗ Y1q YJM dΩ\langle J'M'|\hat\mu_q|JM\rangle\propto\int (Y_{J'}^{M'})^{*}\,Y_1^{q}\,Y_J^{M}\,\mathrm{d}\Omega, a three-harmonic (Gaunt) integral. Why: casting the operator as a rank-1 tensor turns the problem into the clean algebra of spherical-harmonic integrals.

Step 2. It is nonzero only when the triangle condition ∣J−1∣≤J′≤J+1|J-1|\le J'\le J+1 holds and the parity factor J′+J+1J'+J+1 is even. Why: these two conditions are the exact statement of coupling a rank-1 tensor to JJ, and they alone fix the allowed jumps.

Step 3. The triangle allows J′=J−1,J,J+1J'=J-1,J,J+1, but parity kills J′=JJ'=J for a Σ\Sigma state, leaving J′=J±1J'=J\pm1; the MM integral permits ΔM=0,±1\Delta M=0,\pm1. Why: eliminating ΔJ=0\Delta J=0 is why a diatomic shows a clean ladder with no central Q branch — one reappears only when Λ≠0\Lambda\ne0.

Step 4. Absorption from level JJ lands at 2B(J+1)2B(J+1), so the lines are equally spaced by 2B2B, the signature comb of a microwave spectrum. Why: uniform spacing 2B2B reads off the rotational constant and hence the bond length.

Go deeper · Derivation C · 1 minThe vibrational rule from expanding the dipole in the bond coordinate
Derivation CThe vibrational rule from expanding the dipole in the bond coordinate

The goal is to obtain both halves of the infrared rule — the requirement (dμ/dq)0≠0(\mathrm{d}\mu/\mathrm{d}q)_0\ne0 and the harmonic selection Δv=±1\Delta v=\pm1 — by Taylor-expanding the dipole in the vibrational coordinate and feeding it to the harmonic-oscillator matrix elements, seeing along the way where the overtones come from.

Step 1. Expand the dipole, μ(q)=μ0+(dμ/dq)0 q+12(d2μ/dq2)0 q2+⋯\mu(q)=\mu_0+(\mathrm{d}\mu/\mathrm{d}q)_0\,q+\tfrac12(\mathrm{d}^2\mu/\mathrm{d}q^2)_0\,q^2+\cdots, and insert it into ⟨v′∣μ(q)∣v⟩\langle v'|\mu(q)|v\rangle. Why: splitting the dipole into constant, slope, and curvature lets each term be matched to a distinct spectroscopic consequence.

Step 2. The constant term gives μ0⟨v′∣v⟩=μ0 δv′v\mu_0\langle v'|v\rangle=\mu_0\,\delta_{v'v} — it acts only at v′=vv'=v and drives no transition. Why: a permanent dipole alone does nothing for vibration; the dipole must vary with the coordinate.

Step 3. The linear term uses ⟨v′∣q∣v⟩=ℏ/2μω (v+1 δv′,v+1+v δv′,v−1)\langle v'|q|v\rangle=\sqrt{\hbar/2\mu\omega}\,(\sqrt{v+1}\,\delta_{v',v+1}+\sqrt{v}\,\delta_{v',v-1}), which connects only adjacent levels, so it drives Δv=±1\Delta v=\pm1 and needs (dμ/dq)0≠0(\mathrm{d}\mu/\mathrm{d}q)_0\ne0. Why: the ladder-operator structure of qq is the origin of the one-quantum rule — a property of the oscillator, not the chemistry.

Step 4. The quadratic term brings in ⟨v′∣q2∣v⟩\langle v'|q^2|v\rangle, linking v′=v, v±2v'=v,\ v\pm2, so with anharmonic wavefunctions overtones Δv=±2,±3,…\Delta v=\pm2,\pm3,\dots gain weak intensity; and a symmetric motion like the CO2\mathrm{CO_2} symmetric stretch, with (dμ/dq)0=0(\mathrm{d}\mu/\mathrm{d}q)_0=0, is infrared-inactive though still Raman-active. Why: one expansion predicts both the overtones — its controlled breakdown — and which normal modes are dark to the infrared.

Limiting cases

Every rule is an approximation attached to an idealization; the table shows where each holds and the mechanism that relaxes it.

CaseResultMeaning
Harmonic limit (vibration)Δv=±1\Delta v=\pm1Exact for a harmonic oscillator with a dipole linear in qq; the fundamental carries nearly all intensity.
Anharmonic or nonlinear dipoleΔv=±2,±3,…\Delta v=\pm2,\pm3,\dotsAnharmonicity and quadratic terms in μ(q)\mu(q) make overtones weakly allowed, 10−210^{-2}–10−310^{-3} of the fundamental.
Rigid-rotor limit (rotation)ΔJ=±1\Delta J=\pm1Holds for a 1Σ^1\Sigma molecule with μ0≠0\mu_0\ne0; a Q branch (ΔJ=0\Delta J=0) opens when Λ≠0\Lambda\ne0.
Heavy-atom / spin-orbit limitΔS=0\Delta S=0 relaxedSpin-orbit coupling mixes different SS; phosphorescence becomes weakly allowed, more so for heavy atoms.
Centrosymmetric moleculesg↔ug\leftrightarrow u, g↮gg\nleftrightarrow gLaporte is rigorous for fixed nuclei but broken by vibronic coupling to odd-parity modes.
Rank-2 operator (Raman)ΔJ=0,±2\Delta J=0,\pm2The polarizability is second-rank, so rotational lines are spaced by 4B4B and reach dipole-dark modes.

Worked examples

1 · The rotational spectrum of carbon monoxide (confidence builder)

Carbon monoxide has a bond length re=112.83 pmr_e=112.83\ \mathrm{pm} and atomic masses 12.00012.000 and 15.995 u15.995\ \mathrm{u}. Treating it as a rigid rotor with a permanent dipole, find its rotational constant BB, the spacing of its microwave lines, and the frequency of the very first absorption line.

Step 1. Reduced mass and moment of inertia. μ=(12.000×15.995)/(12.000+15.995) u=6.856 u=1.139×10−26 kg\mu=(12.000\times15.995)/(12.000+15.995)\ \mathrm{u}=6.856\ \mathrm{u}=1.139\times10^{-26}\ \mathrm{kg}, and I=μre2=(1.139×10−26)(1.1283×10−10)2=1.449×10−46 kg m2I=\mu r_e^2=(1.139\times10^{-26})(1.1283\times10^{-10})^2=1.449\times10^{-46}\ \mathrm{kg\,m^2}.

Step 2. Rotational constant. In wavenumbers, B=h/(8π2cI)=(6.626×10−34)/[8π2(2.998×1010 cm/s)(1.449×10−46)]=1.93 cm−1B=h/(8\pi^2 c I)=(6.626\times10^{-34})/[8\pi^2(2.998\times10^{10}\ \mathrm{cm/s})(1.449\times10^{-46})]=\mathbf{1.93\ cm^{-1}}. Since ΔJ=±1\Delta J=\pm1, absorption from level JJ falls at 2B(J+1)2B(J+1), so the lines are equally spaced by 2B=3.86 cm−12B=\mathbf{3.86\ cm^{-1}}.

Step 3. The first line. The lowest transition J=0→1J=0\to1 sits at 2B=3.86 cm−12B=3.86\ \mathrm{cm^{-1}}; converting with 1 cm−1=29.979 GHz1\ \mathrm{cm^{-1}}=29.979\ \mathrm{GHz} gives ν=115.8 GHz\nu=\mathbf{115.8\ GHz}, wavelength λ=c/ν=2.6 mm\lambda=c/\nu=\mathbf{2.6\ mm}. This is the celebrated CO line (observed 115.27 GHz115.27\ \mathrm{GHz}; the offset comes from using rer_e rather than B0B_0) that radio telescopes use to map molecular clouds across the galaxy — a direct payoff of one selection rule. B ≈ 1.93 cm⁻¹, spacing 2B ≈ 3.86 cm⁻¹, and the J = 0 → 1 line sits at 115.8 GHz (2.6 mm).

2 · Why nitrogen is microwave-dark but Raman-bright

Nitrogen is homonuclear with rotational constant B=1.99 cm−1B=1.99\ \mathrm{cm^{-1}}. Explain why it produces no pure-rotation microwave spectrum, then find the spacing and the first Stokes shift of its rotational Raman spectrum.

Step 1. The microwave test. Pure-rotation absorption requires μ0≠0\mu_0\ne0, but N2\mathrm{N_2} has two identical nuclei and no charge separation, so μ0=0\mu_0=0 exactly and rule (3) forbids every microwave line — the spectrum is not weak but absent.

Step 2. Switch operators. Raman scattering is driven by the polarizability, and even a homonuclear bond is more polarizable along the axis than across it, so (dα/dq)0≠0(\mathrm{d}\alpha/\mathrm{d}q)_0\ne0; being rank 2, the operator gives ΔJ=0,±2\Delta J=0,\pm2.

Step 3. Line positions. For the S branch (ΔJ=+2\Delta J=+2) the Stokes shift is B(4J+6)B(4J+6): the lowest, J=0→2J=0\to2, sits at 6B=6(1.99)=11.9 cm−16B=6(1.99)=\mathbf{11.9\ cm^{-1}}, and successive S lines are spaced by 4B=7.96 cm−14B=\mathbf{7.96\ cm^{-1}} — twice the 2B2B of a dipole spectrum, the fingerprint of the rank-2 operator. N₂ has μ₀ = 0 so no microwave spectrum, but its rotational Raman lines are spaced by 4B ≈ 7.96 cm⁻¹, the first Stokes line at 6B ≈ 11.9 cm⁻¹.

Physics of the portrait

Molecular selection rules are the work of a generation that learned to read band spectra — and one figure catalogued them so completely his volumes are still the first place a spectroscopist looks.

Origins · 1 minGerhard Herzberg and how this was discovered
GH
Gerhard Herzberg
1904–1999 · molecular spectroscopy and selection rules

Gerhard Herzberg was the physicist who made molecular spectra legible. Trained in Germany and forced to emigrate in 1935, he built at the National Research Council in Ottawa the world's leading laboratory for molecular spectroscopy. His achievement was less a single discovery than a synthesis: across three volumes of Molecular Spectra and Molecular Structure — diatomics, then polyatomic infrared and Raman spectra, then polyatomic electronic spectra — he assembled the selection rules for rotation, vibration, and electronic transitions into the working reference the field still uses. He showed how to reason from a molecule's symmetry to the lines it must and must not show, and back from a spectrum to its geometry, identifying transient free radicals such as CH2\mathrm{CH_2} and CH3\mathrm{CH_3} from the faint bands that survive symmetry breaking. The Nobel Committee awarded him the 1971 Prize in Chemistry "for his contributions to the knowledge of electronic structure and geometry of molecules, particularly free radicals" — the rules governing which transitions appear being the very rules by which molecules are identified, in the laboratory and across interstellar space.

History
1924
— Laporte finds the parity (g ↔ u) selection rule while analyzing the iron-arc spectrum
1927
— Hund's coupling cases classify diatomic angular momenta, fixing the ΔΛ electronic rules
1928
— Condon recasts Franck's intensity idea in quantum form as the Franck-Condon principle
1950
— Herzberg publishes Spectra of Diatomic Molecules, codifying the molecular selection rules
1971
— Herzberg receives the Nobel Prize in Chemistry for molecular electronic structure

Going further

Texts. Foot, Atomic Physics (2005), Ch. 2 and 7 · Herzberg, Molecular Spectra and Molecular Structure, Vols. I–III — the definitive reference for molecular selection rules · Demtröder, Molecular Physics / Laser Spectroscopy · Loudon, The Quantum Theory of Light, 3rd ed., Ch. 1–2 · Saleh & Teich, Fundamentals of Photonics, atom–field interaction · Hecht, Optics — dipole radiation and absorption.

Provenance. The master criterion follows from Fermi's golden rule in the dipole approximation; the rotational rule from the three-spherical-harmonic (Gaunt) integral (ΔJ=±1\Delta J=\pm1, spacing 2B2B); the vibrational rule from Taylor-expanding μ(q)\mu(q) with harmonic matrix elements ⟨v±1∣q∣v⟩=ℏ/2μω (v+1,v)\langle v\pm1|q|v\rangle=\sqrt{\hbar/2\mu\omega}\,(\sqrt{v+1},\sqrt{v}); the electronic rules in standard diatomic form; and Franck–Condon factorization under the Condon approximation. Numbers recomputed here: CO with re=112.83 pmr_e=112.83\ \mathrm{pm} gives B=1.93 cm−1B=1.93\ \mathrm{cm^{-1}} and J=0→1J=0\to1 at 115.8 GHz115.8\ \mathrm{GHz} (observed 115.27 GHz115.27\ \mathrm{GHz}); N2\mathrm{N_2} with B=1.99 cm−1B=1.99\ \mathrm{cm^{-1}} gives Raman 4B=7.96 cm−14B=7.96\ \mathrm{cm^{-1}}, first Stokes 6B=11.9 cm−16B=11.9\ \mathrm{cm^{-1}}. Cross-check against Foot, Demtröder, and Herzberg; no non-standard values are asserted.

Prerequisites. The Born–Oppenheimer Approximation · The Rigid Rotor Model · The Harmonic Oscillator Model of Vibration · Term Symbols and Molecular Electronic States. Leads to. Vibration-Rotation Spectra and P, Q, R Branches · The Franck–Condon Principle · Raman Scattering and the Rotational and Vibrational Raman Effect · Electronic Transitions in Molecules.

Ask the tutor: Give me any molecule and a proposed transition, and I will walk you through the dipole-handle test, the symmetry rule that decides it, and the number you would actually measure.

What is actually meant by a "forbidden" transition — does it never happen?

Forbidden means forbidden to first order in the electric dipole, not impossible — the leading integral ⟨ψf∣μ^∣ψi⟩\langle\psi_f|\hat{\boldsymbol\mu}|\psi_i\rangle vanishes by symmetry, but weaker channels remain. Magnetic-dipole and electric-quadrupole interactions, vibronic coupling that momentarily lowers the symmetry, and spin-orbit mixing can each lend small but real intensity. The atmospheric oxygen bands and molecular phosphorescence are exactly such transitions — electric-dipole forbidden yet observable over long paths or long times. So "forbidden" means a rate small by orders of magnitude, not zero.

Why does a homonuclear molecule like N₂ have no infrared or microwave spectrum at all?

Both spectra rely on a changing electric dipole, which a homonuclear molecule cannot supply. With two identical nuclei there is no charge separation, so the permanent dipole is zero and the microwave rule μ0≠0\mu_0\ne0 fails. When the bond stretches the molecule stays symmetric, so (dμ/dq)0=0(\mathrm{d}\mu/\mathrm{d}q)_0=0 too, killing the infrared fundamental. This is not a weak spectrum but an absent one. The molecule is not invisible, however: rotational and vibrational Raman scattering, driven by the polarizability, are fully allowed — which is how N2\mathrm{N_2} and O2\mathrm{O_2} are studied.

How can CO₂ be a greenhouse gas if it has no dipole moment?

Because greenhouse absorption is a vibrational process, and infrared activity needs a changing dipole, not a permanent one. Carbon dioxide is centrosymmetric, so its permanent dipole is zero and it shows no microwave spectrum. But two vibrations break that symmetry: the bending mode and the antisymmetric stretch each give the molecule a nonzero, oscillating dipole, so (dμ/dq)0≠0(\mathrm{d}\mu/\mathrm{d}q)_0\ne0 and they absorb strongly near 667667 and 2349 cm−12349\ \mathrm{cm^{-1}}. Only the symmetric stretch, which keeps the dipole at zero, is infrared-silent.

What is the difference between the infrared and Raman selection rules?

Different operators, so they probe complementary motions. Infrared absorption is an electric-dipole process needing (dμ/dq)0≠0(\mathrm{d}\mu/\mathrm{d}q)_0\ne0, with ΔJ=±1\Delta J=\pm1. Raman scattering is governed by the polarizability, needing (dα/dq)0≠0(\mathrm{d}\alpha/\mathrm{d}q)_0\ne0, with ΔJ=0,±2\Delta J=0,\pm2 because the polarizability is rank 2. In a centrosymmetric molecule the two become mutually exclusive: no mode is both infrared- and Raman-active — a direct experimental test for a centre of inversion.

Where does the rule ΔJ = ±1 actually come from?

From the tensor rank of the dipole operator. The angular part of a permanent dipole, projected onto the lab frame, transforms as a rank-1 spherical tensor Y1qY_1^q, so the rotational matrix element is an integral of three spherical harmonics. Such an integral is nonzero only when the angular momenta obey a triangle condition and a parity condition, which permit J′=J±1J'=J\pm1 and forbid J′=JJ'=J for a Σ\Sigma state. The "one" in ΔJ=±1\Delta J=\pm1 is the "1" of the rank-1 tensor; the rank-2 polarizability instead gives ΔJ=0,±2\Delta J=0,\pm2.

Why is phosphorescence so much weaker and slower than fluorescence?

Because it breaks the spin rule ΔS=0\Delta S=0. Fluorescence is a spin-allowed singlet-to-singlet decay, fast on nanosecond timescales. Phosphorescence is the spin-forbidden triplet-to-singlet decay T1→S0T_1\to S_0, which the electric dipole cannot drive because it does not act on spin. It becomes weakly allowed only through spin-orbit coupling, which mixes a little singlet character into the triplet; the mixing is small, so the rate is small and the state lives for milliseconds to seconds. That mixing grows steeply with nuclear charge — the heavy-atom effect exploited in light-emitting devices.

Before this

Where this leads