Selection Rules from Symmetry

Selection rules from symmetry remain the first filter applied to every modern spectroscopic technique. In infrared and Raman instrumentation, the mutual-exclusion rule is routinely used to distinguish centrosymmetric conformers from non-centrosymmetric ones in solution — a molecule that shows a given vibrational frequency in both IR and Raman spectra cannot possess a center of inversion at that geometry, a fast structural diagnostic used in everything from small-molecule conformational analysis to polymer crystallinity studies. In materials science, symmetry-forbidden $d$–$d$ transitions in transition-metal complexes (Laporte-forbidden in centrosymmetric octahedral fields) explain the pale colors of many aqua and ammine complexes, while distortions that break inversion symmetry (Jahn–Teller effects, or coordination to asymmetric ligands) are engineered deliberately to intensify color in dyes, pigments, and photocatalysts. In pharmaceutical and biophysical spectroscopy, circular dichroism and vibrational circular dichroism instruments rely on symmetry-breaking (chirality) precisely because achiral, centrosymmetric-adjacent selection rules would otherwise forbid the relevant rotationally-differential signals; understanding which transitions are symmetry-allowed guides the assignment of chromophore bands in protein and nucleic acid spectra. In surface science and heterogeneous catalysis, surface selection rules (a close cousin of the bulk rules derived here, modified by the symmetry-breaking presence of a metal surface) determine which adsorbate vibrational modes are IR-visible in reflection-absorption spectroscopy, directly informing catalyst characterization. Finally, in quantum information and laser spectroscopy, selection rules dictate which transitions can be driven by a given laser polarization, a design constraint central to state-selective excitation schemes in cold-molecule and precision-spectroscopy experiments.

Ask the tutor: How would the selection-rule analysis change for a Jahn–Teller-active vibronic transition in an octahedral complex?

Time-dependent perturbation theory tells us that the rate of a photon-induced transition between states $i$ and $f$ is proportional to $|\langle \psi_f|\hat\mu|\psi_i\rangle|^2$, the squared transition dipole moment. This quantity is an integral over all space,

where $\hat{\vec\mu} = \sum_k q_k \vec r_k$ sums the charge-weighted position vectors of every particle in the molecule. This is a genuine vector integral with three Cartesian components ($\mu_x, \mu_y, \mu_z$), and the transition is allowed if any one of them is nonzero. The trick that makes group theory useful here is a general theorem about integrals of products of functions over all space: an integral $\int f\,d\tau$ can be nonzero only if the integrand $f$ transforms as (contains) the totally symmetric representation of the molecular point group. This isn't a hand-wave — it follows because any function that transforms as a non-totally-symmetric irreducible representation changes sign or rotates into a different function under some symmetry operation of the group, and the definite integral over all space (which is itself invariant under every symmetry operation, since the operations just relabel the same physical space) would then have to equal minus itself, forcing it to be zero. Now, $\psi_i$, $\psi_f$, and each Cartesian component of $\hat{\vec\mu}$ all transform as some irreducible representation (irrep) of the point group — this is guaranteed because the molecular Hamiltonian commutes with every symmetry operation, so eigenstates can always be chosen to transform irreducibly. The dipole operator components transform exactly like $x$, $y$, $z$ themselves (as basis functions for translation), and every character table lists which irrep each of $x$, $y$, $z$ belongs to. The question of whether $\vec\mu_{fi}$ vanishes is then purely combinatorial: does

Two extremely useful specializations follow immediately. First, for the rotational gross selection rule, we ask whether the molecule possesses a permanent dipole at all — equivalent to asking whether $\hat\mu$ itself (not a transition integral) is nonzero in the ground vibrational/electronic state, which happens only if the molecule lacks a center of symmetry along at least one axis for linear/symmetric cases, or more generally lacks the symmetry that forces $\langle\mu\rangle=0$. Second, for vibrational transitions, the relevant question becomes whether a normal mode $Q_k$ changes the dipole moment, i.e., whether $\Gamma_{Q_k}$ itself matches one of $\Gamma_x,\Gamma_y,\Gamma_z$ — this is the vibrational gross selection rule and is checked with exactly the same direct-product logic, since within the harmonic approximation the relevant integral reduces to $\langle 0|Q_k|1\rangle$ times a symmetry factor.

Once the test framework is set, applying it to rotational, vibrational, and electronic problems is a matter of finding the right irreps — from the rigid rotor's rotational wavefunctions transforming under the permutation-inversion or point group, from normal-mode symmetry analysis via reducible representations of atomic displacements, or from molecular-orbital symmetry labels for electronic states.

The master criterion, from which everything else in this article is derived:

where: $\Gamma_f, \Gamma_i$ are the irreducible representations of the final and initial molecular states; $\Gamma_\mu$ stands for the irrep(s) of whichever Cartesian component ($x$, $y$, or $z$) of the dipole operator is being tested; $\Gamma_{\text{tot.sym.}}$ is the totally symmetric irrep of the point group ($A_1$, $A_{1g}$, $A_{1'}$, $\Sigma_g^+$, etc., depending on the group). Reading: this says a transition can only "turn on" light absorption/emission along a given polarization if combining the symmetry labels of both states with that dipole component gives back the group's "do-nothing" representation — i.e., the product is symmetric under every operation of the group, exactly what's required for a nonvanishing integral over all space. For most molecular ground states, $\Gamma_i = A_1$ (totally symmetric), which simplifies the criterion enormously:

where: this uses the fact that for any group, $\Gamma_a \otimes \Gamma_b$ contains $A_1$ if and only if $\Gamma_a = \Gamma_b$ (irreps are orthonormal under the group's inner product, a direct consequence of the Great Orthogonality Theorem). Reading: if the molecule starts in its totally symmetric ground state, a transition to $\Gamma_f$ is dipole-allowed exactly when $\Gamma_f$ matches the symmetry of $x$, $y$, or $z$ — literally, "look up $\Gamma_f$ in the character table and see whether $x,y,$ or $z$ sits in that row." For vibrational overtones and combination bands, and more generally whenever $\Gamma_i \neq A_1$, one needs the full triple product, and it is often organized via characters:

where: $\chi_\Gamma(R)$ is the character (trace) of irrep $\Gamma$ under symmetry operation $R$; the character of a direct product representation is just the product of the individual characters, operation by operation. Reading: this lets you build the character row for the whole triple product directly from the character table without ever constructing matrices, then reduce it using the standard reduction formula to see whether $A_1$ appears with nonzero coefficient. The reduction formula itself, needed whenever the direct product isn't obviously irreducible:

where: $h$ is the order of the group, $g_R$ is the number of operations in the class of $R$, and $\chi_{A_1}(R)=1$ for all $R$ since the totally symmetric representation has character 1 everywhere. Reading: $n_{A_1}$ counts how many times the totally symmetric representation appears in the (possibly reducible) product representation; if $n_{A_1} \geq 1$, the transition is allowed, if $n_{A_1}=0$, it is strictly forbidden by symmetry.

Example 1 — Foundational: the rotational selection rule for HCl. HCl belongs to point group $C_{\infty v}$. Its permanent dipole moment lies along the molecular ($z$) axis, and it possesses no center of symmetry, so nothing forbids $\mu_z \neq 0$ in the ground state — this settles the gross rotational selection rule immediately: HCl has a pure rotational (microwave) spectrum, unlike homonuclear diatomics such as N₂ (point group $D_{\infty h}$), for which every dipole component is forced to zero by the inversion center regardless of vibrational or rotational state. For the specific selection rule, use rigid-rotor wavefunctions, which for a linear molecule are the spherical harmonics $Y_{J}^{M}(\theta,\phi)$. The transition moment integral to test is

where $\mu_0$ is the (constant) permanent dipole magnitude and $\cos\theta$ is the direction cosine that projects the body-fixed dipole onto the space-fixed $z$-axis (here $z$ transforms as $\Gamma_\mu$ in $C_{\infty v}$, matching $\Sigma^+$). Rather than integrate directly, use the standard recursion/orthogonality property of spherical harmonics (equivalently, a special case of the direct-product test in the continuous rotation group $O(3)$ restricted to $C_{\infty v}$): the product $\cos\theta \cdot Y_J^M$ decomposes, via the Legendre recursion, purely into $Y_{J+1}^M$ and $Y_{J-1}^M$,

with nonzero coefficients $a_{J,M}, b_{J,M}$ given by standard Clebsch–Gordan-type expressions. Because $Y_{J'}^{M'}$ is orthogonal to every $Y_{J}^{M}$ except when $J'$ matches exactly, the overlap integral survives only for $J' = J+1$ or $J' = J-1$; all other $J'$ give zero by orthogonality. This yields the specific rotational selection rule

(the $\Delta M = 0$ piece comes from the $\phi$-integral, since $\cos\theta$ carries no $\phi$-dependence and $\int_0^{2\pi} e^{i(M-M')\phi}\,d\phi = 0$ unless $M=M'$).

Numerical check with asymmetric input. Take the $J=3 \to J=4$ transition of HCl (an off-round, non-adjacent-to-zero case). Using the rigid-rotor energy $E_J = BJ(J+1)$ with $B_{\text{HCl}} = 10.59\ \text{cm}^{-1}$,

This transition is allowed ($\Delta J=+1$), and its frequency should appear in the HCl microwave/far-IR spectrum; by contrast, a hypothetical $J=3\to J=5$ line ($\Delta J=+2$) is strictly forbidden — no such line appears no matter how carefully one searches, because the direct-product/orthogonality argument makes that overlap integral exactly zero, not merely small.

Limiting case. As the dipole moment $\mu_0 \to 0$ (approaching a homonuclear diatomic limit), every transition moment scales linearly with $\mu_0$ and vanishes — consistent with the fact that N₂, O₂, and H₂ show no permitted pure rotational spectrum at all, regardless of $\Delta J$.

Example 2 — Applied: vibrational activity of the CO₂ normal modes. CO₂ is linear, point group $D_{\infty h}$, with four vibrational degrees of freedom ($3N-5 = 3(3)-5 = 4$): the symmetric stretch, the doubly degenerate bend, and the antisymmetric stretch. Reduction of the $3N$-dimensional Cartesian displacement representation (a standard exercise in constructing $\Gamma_{3N}$ from unmoved-atom counts and then subtracting translations and rotations) gives the vibrational representation

corresponding respectively to the symmetric stretch ($\nu_1$), the doubly-degenerate bend ($\nu_2$), and the antisymmetric stretch ($\nu_3$). In $D_{\infty h}$, the character table lists $x,y$ as a degenerate pair transforming as $\Pi_u$, and $z$ transforming as $\Sigma_u^+$. Apply the gross selection rule (mode symmetry must match $x$, $y$, or $z$):

• $\nu_1$ ($\Sigma_g^+$): does not match $\Sigma_u^+$ or $\Pi_u$ — IR-inactive. Physically, stretching both C=O bonds symmetrically keeps the dipole at zero throughout the motion (the two bond dipoles always cancel), consistent with $(\partial\mu/\partial Q_1)_0=0$.
• $\nu_2$ ($\Pi_u$): matches $(x,y)$ exactly — IR-active, doubly degenerate, appears near 667 cm⁻¹.
• $\nu_3$ ($\Sigma_u^+$): matches $z$ exactly — IR-active, appears near 2349 cm⁻¹.

Now check Raman activity, which requires the mode to transform as one of the quadratic functions ($x^2+y^2, z^2$, etc., all listed as $\Sigma_g^+$ or other $g$ representations in $D_{\infty h}$): only $\nu_1$ ($\Sigma_g^+$) matches a Raman tensor component, so only the symmetric stretch is Raman-active; the two IR-active modes are Raman-silent. This is the rule of mutual exclusion: in any molecule with a center of inversion, no normal mode can be both IR- and Raman-active, because $x,y,z$ (IR) always transform as $u$ representations while the Raman tensor components ($x^2,y^2,z^2,xy,\ldots$) always transform as $g$ representations, and a mode has a single, definite parity under $i$.

Numerical check with asymmetric input. The observed frequencies, $\nu_1 = 1333\ \text{cm}^{-1}$ (Raman-only, strong), $\nu_2 = 667\ \text{cm}^{-1}$ (IR, doubly degenerate), $\nu_3 = 2349\ \text{cm}^{-1}$ (IR), are all distinct, non-round numbers — as they must be, since these are physically separate vibrational motions with different force constants and reduced masses, not symmetry-related copies of each other.

Limiting case. If one imagined artificially bending CO₂ permanently into a non-linear geometry (as happens in the excited electronically bent state), the point group drops to $C_{2v}$, the center of inversion is lost, and the mutual-exclusion rule no longer applies — all three vibrational modes of bent CO₂ (now $A_1, A_1, B_1$ or similar, depending on convention) become simultaneously IR- and Raman-active, illustrating that mutual exclusion is strictly a consequence of the inversion center, not a universal law.

Example 3 — Exam/competition level: the symmetry-forbidden benzene ¹A₁g → ¹B₂u transition. Benzene has point group $D_{6h}$. Its lowest-energy singlet π→π* transition (the weak ~260 nm band) is from the ground state $^1A_{1g}$ to the excited state $^1B_{2u}$. Apply the master criterion directly: since $\Gamma_i = A_{1g}$ is totally symmetric, the transition is allowed only if $\Gamma_f = B_{2u}$ matches the symmetry of $x$, $y$, or $z$ in $D_{6h}$. Checking the $D_{6h}$ character table: the in-plane translations $(x,y)$ transform together as the degenerate irrep $E_{1u}$, and $z$ transforms as $A_{2u}$. Neither matches $B_{2u}$:

Therefore the electric-dipole transition moment integral $\langle {}^1B_{2u}|\hat\mu|{}^1A_{1g}\rangle$ is rigorously zero within the rigid, equilibrium-geometry (Born–Oppenheimer, harmonic) approximation — this transition is symmetry-forbidden, exactly the origin of benzene's anomalously weak ~260 nm absorption band (molar absorptivity of order $\varepsilon \sim 200\ \text{L mol}^{-1}\text{cm}^{-1}$, roughly two orders of magnitude weaker than a fully allowed π→π* transition, which typically has $\varepsilon \sim 10^4$). Yet the band is observed, faintly, and this is resolved by vibronic coupling: a non-totally-symmetric vibration can distort the molecule during the transition, changing the effective point group symmetry instantaneously and "borrowing" intensity. The full vibronic transition moment factorizes (to leading order) as a product over electronic and vibrational symmetry labels, and the true selection rule becomes

For benzene, the vibrationally-excited final state can combine $B_{2u}$ (electronic) with a non-totally-symmetric vibration transforming as $e_{2g}$; because

(a standard $D_{6h}$ direct-product decomposition), and $E_{1u}$ does match $(x,y)$, the vibronically-allowed combination $B_{2u}\otimes e_{2g}\otimes E_{1u}(\text{from }\mu)\otimes A_{1g}$ does contain $A_{1g}$. The transition steals its (weak) intensity from simultaneous excitation of an $e_{2g}$ vibrational mode — precisely why the benzene band shows resolvable vibronic fine structure rather than a clean single line, and why its intensity, though nonzero, remains far below a fully-allowed transition.

Numerical sanity check. The oscillator strength $f$ for an allowed π→π* transition in a similarly-sized conjugated system is typically $f \sim 0.5$–1.0; benzene's $^1L_b$ band ($^1A_{1g}\to{}^1B_{2u}$) has $f \approx 0.001$–0.01, a suppression of roughly 50–500-fold — the correct order of magnitude expected for a vibronically-borrowed, symmetry-forbidden transition, not a coincidence but a direct, checkable consequence of the group-theoretical argument above.

Limiting case. If benzene's symmetry were artificially lowered (e.g., by a single substituent breaking $D_{6h}$ down to $C_{2v}$, as in monosubstituted benzenes), the strict forbiddenness is lifted at the purely electronic level — many substituted benzenes show noticeably stronger, less vibronically-structured absorption near this same energy region, consistent with the symmetry argument rather than contradicting it.

The systematic use of group theory to derive spectroscopic selection rules is inseparable from the work of Eugene Wigner, who established the general theorem connecting vanishing integrals to representation theory (what is now often called the Wigner projection/vanishing-integral theorem), and from Hans Bethe, whose 1929 treatment of crystal-field splitting first showed chemists and physicists how to systematically apply point-group character tables to real molecular and ionic problems. Their combined framework — representation theory applied concretely to molecular symmetry — became the standard toolkit later codified for chemists by F. Albert Cotton's classic text on group theory in chemistry, which turned an abstract mathematical result into the practical character-table method every spectroscopy course now teaches.