Symmetry and Spectroscopic Selection Rules

Selection-rule analysis is the reason a synthetic chemist can glance at the carbonyl-stretch region of an infrared spectrum of a metal carbonyl complex and immediately distinguish geometric isomers: $\text{Mo(CO)}_6$ (octahedral, $O_h$) shows a single IR-active $T_{1u}$ carbonyl stretch, while a cis-disubstituted $\text{M(CO)}_4L_2$ complex, lower in symmetry, shows up to four IR-active carbonyl bands — a textbook diagnostic used continuously in inorganic and organometallic laboratories to assign coordination geometry without crystallography. The mutual exclusion rule is the working principle behind confirming whether a small symmetric molecule (SF₆, $\text{XeF}_4$, benzene) truly possesses the inversion center its idealized structure suggests, since even a subtle distortion shows up as a coincidence of IR and Raman bands that should otherwise be strictly forbidden. In materials science, polarized Raman spectroscopy on single crystals exploits not just which modes are Raman-active but which tensor components ($\alpha_{xx}$ vs $\alpha_{xy}$, etc.) are activated, letting researchers assign crystallographic axes and detect phase transitions in ferroelectrics and perovskites by watching selection-rule-forbidden modes suddenly appear as symmetry breaks. Semiconductor process engineers use Raman selection rules to map mechanical strain in silicon wafers, where strain-induced symmetry lowering activates modes forbidden in the unstrained lattice. Pharmaceutical and forensic laboratories rely on IR/Raman selection-rule complementarity for polymorph identification, since different crystal packings of the same drug molecule can have different local site symmetries and therefore different observed vibrational activity patterns even with chemically identical bonds. Astrochemists use the absence of a dipole-allowed rotational or vibrational spectrum (as for homonuclear $\text{H}_2$ or $\text{N}_2$, whose $D_{\infty h}$ symmetry with inversion center forbids the fundamental stretch from IR activity) to explain why these highly abundant interstellar molecules are notoriously difficult to detect directly and must instead be inferred from Raman, collision-induced absorption, or indirect chemical tracers.

Ask the tutor: How would the vibrational selection rules change if XeF₄ were slightly puckered into a non-planar geometry?

A molecule's normal modes are not arbitrary wiggles: each one belongs to an irreducible representation (irrep) of the molecular point group, exactly as an atomic orbital belongs to an irrep of the group that describes the local site symmetry. This classification is what makes group theory useful for spectroscopy, because the interaction between light and matter is itself governed by an operator with well-defined symmetry — the electric dipole moment operator for infrared absorption, and the induced-dipole (polarizability) operator for Raman scattering. Infrared absorption happens because an oscillating electric field $\vec E(t)$ couples to the molecular dipole moment $\hat{\vec\mu}$. Quantum mechanically, the rate of absorption between vibrational states $v=0$ and $v=1$ of a normal mode $Q_k$ is proportional to the square of the transition dipole moment,

If this integral is zero, the transition is forbidden and no absorption line appears (to that order). Because $\hat{\vec\mu} = (\mu_x,\mu_y,\mu_z)$ transforms exactly like the coordinates $x,y,z$ under the operations of the point group, whether $\vec\mu_{fi}$ can be nonzero is decided by whether the normal mode's symmetry matches $x$, $y$, or $z$. Raman scattering is a two-photon process: an incident photon is absorbed and a shifted-frequency photon is re-emitted, mediated by the induced dipole moment $\vec\mu_{ind} = \boldsymbol\alpha \cdot \vec E$, where $\boldsymbol\alpha$ is the molecular polarizability tensor. The relevant transition integral involves matrix elements of $\alpha_{xy}, \alpha_{xx}$, etc. — quantities that transform as the quadratic functions of $x,y,z$ listed in the rightmost column of every character table. A vibration is Raman-active only if its symmetry species matches one of these quadratic forms.

The reason CO₂'s symmetric stretch is invisible in the infrared but shows up brightly in Raman, while the antisymmetric stretch and bend are the reverse, is the g/u parity imposed by the inversion center. Any operator built from $x,y,z$ linearly (the dipole) is odd (ungerade) under inversion; any operator built from products of two coordinates (the polarizability, quadratic in $x,y,z$) is even (gerade). A gerade vibration can only produce a nonzero integral against a gerade operator, so gerade modes are Raman-active and IR-silent; ungerade modes are the opposite. This is the mutual exclusion rule, and it is not a coincidence of CO₂ — it holds for every molecule possessing a center of inversion.

The master statement governing every selection rule, electronic or vibrational, is that an integral over all space of a product of functions vanishes unless the integrand itself is invariant under every symmetry operation of the point group — equivalently, unless its representation contains the totally symmetric irrep.

where:

• $\psi_i, \psi_f$ are the initial and final vibrational (or vibronic) wavefunctions, each transforming as some irrep $\Gamma_i, \Gamma_f$ of the point group,
• $\hat O$ is the perturbing operator (dipole moment $\hat{\vec\mu}$ for IR, polarizability components $\alpha_{ij}$ for Raman), transforming as irrep $\Gamma_{\hat O}$,
• $\Gamma_1$ denotes the totally symmetric representation, and $\supset$ means "contains, upon reduction, at least one copy of."

Reading: this is a necessary condition, not a proof of a large transition moment — symmetry can only tell you the integral is forced to be zero, or that it is allowed to be nonzero. Whether an allowed integral is actually large is a separate, dynamical question. For the vibrational ground state $v=0$ of every normal mode, the vibrational wavefunction is totally symmetric ($\Gamma_i = \Gamma_1$), since the harmonic oscillator ground-state Gaussian $e^{-\alpha Q^2}$ is invariant under every operation that leaves $Q^2$ unchanged. For the fundamental transition $v=0\to v=1$, the excited-state wavefunction transforms as the same irrep as the normal coordinate $Q_k$ itself (a standard result of harmonic-oscillator symmetry theory: the $v=1$ wavefunction $\propto Q_k\, e^{-\alpha Q_k^2}$ inherits the symmetry of $Q_k$). Substituting $\Gamma_i=\Gamma_1$ collapses the master rule to:

where:

• the final equivalence uses the orthogonality property of irreducible representations: the direct product of two irreps contains the totally symmetric representation if and only if the two irreps are identical (for point groups without complex characters this is exact; for those with complex-conjugate pairs it holds pairwise).

Reading: a fundamental vibration is IR-active precisely when its symmetry species is the same irrep as one of $x, y, z$; it is Raman-active precisely when its symmetry species matches one of the quadratic functions. This is why every character table lists $x,y,z$ and the quadratics in a dedicated column — that column is the selection-rule lookup table. Counting how many normal modes belong to each irrep uses the reducible Cartesian displacement representation $\Gamma_{3N}$, built from placing an $(x,y,z)$ displacement vector on every atom, then subtracting translations and rotations:

where:

• $\Gamma_{3N}$ is reduced with the standard reduction formula using characters computed from the "unmoved atoms × contribution per operation" recipe,
• $\Gamma_{trans}$ is the representation spanned by $x,y,z$ (read directly off the character table),
• $\Gamma_{rot}$ is the representation spanned by $R_x, R_y, R_z$ (also read off the table).

Reading: $\Gamma_{vib}$ is the complete symmetry inventory of every vibrational normal mode; once reduced into irreps, each irrep's IR/Raman activity follows immediately from the character table columns, without ever writing down an explicit normal coordinate.

Example 1 — Foundational: water, $C_{2v}$, all three modes classified by hand. Water has $N=3$ atoms, so $3N = 9$ Cartesian degrees of freedom. The $C_{2v}$ character table has classes $E, C_2, \sigma_v(xz), \sigma_v'(yz)$, group order $h=4$. Unmoved-atom counts: under $E$, all 3 atoms are unmoved; under $C_2$ (through O, bisecting H–O–H), only the O atom is unmoved (the two H's swap); under $\sigma_v(xz)$ (the molecular plane, containing both H's and O), all 3 atoms lie in this plane and are unmoved; under $\sigma_v'(yz)$ (perpendicular to the molecular plane, containing only O), only O is unmoved. Contribution per unmoved atom: $E$ gives $+3$ ($1+2\cos0^\circ = 3$); $C_2$ gives $-1$ ($1+2\cos180^\circ=1-2=-1$); $\sigma$ gives $+1$ ($-1+2\cos0^\circ=1$).

Reduce using $n_i = \frac{1}{4}\sum_R \chi_{3N}(R)\chi_i(R)$ with the $C_{2v}$ characters ($A_1: 1,1,1,1$; $A_2: 1,1,-1,-1$; $B_1: 1,-1,1,-1$; $B_2: 1,-1,-1,1$):

So $\Gamma_{3N} = 3A_1 + A_2 + 3B_1 + 2B_2$ (check: $3+1+3+2=9$ ✓). Translations transform as $x,y,z$: in $C_{2v}$, $z\to A_1$, $x\to B_1$, $y\to B_2$ (matching the standard convention with the molecular plane as $xz$), so $\Gamma_{trans} = A_1+B_1+B_2$. Rotations transform as $R_x\to B_2$, $R_y\to B_1$, $R_z\to A_2$, so $\Gamma_{rot} = A_2+B_1+B_2$.

This gives $3N-6 = 3$ modes, matching expectation: $2A_1 + B_1$, i.e., the symmetric stretch ($A_1$), the bend ($A_1$), and the antisymmetric stretch ($B_1$). Selection rules: the $C_{2v}$ character table lists $z\to A_1$, $(x,y)$ individually to $B_1, B_2$, and quadratics $x^2,y^2,z^2\to A_1$, $xy\to A_2$, $xz\to B_1$, $yz\to B_2$. Since $A_1$ matches both $z$ and the quadratics, and $B_1$ matches both $x$ and $xz$, all three vibrations of water are both IR- and Raman-active — consistent with water having no center of inversion, so mutual exclusion does not apply. Observed values: symmetric stretch $\approx 3657\ \text{cm}^{-1}$, bend $\approx 1595\ \text{cm}^{-1}$, antisymmetric stretch $\approx 3756\ \text{cm}^{-1}$, all seen in both IR and Raman spectra of water vapor.

Example 2 — Applied: carbon dioxide, $D_{\infty h}$, deriving mutual exclusion numerically. CO₂ is linear, $N=3$, so $3N=9$, and the vibrational count is $3N-5=4$ (one fewer subtraction than nonlinear molecules, because linear molecules have only two independent rotational degrees of freedom — rotation about the molecular axis is not a physical motion of a collection of point masses on that axis). Using the standard tabulated result for $D_{\infty h}$ AB₂ linear molecules (derivable by the same unmoved-atom method, treated with a limiting sequence over $C_n$ for large $n$), the reducible Cartesian representation reduces to

with translations $\Gamma_{trans} = \Sigma_u^+ + \Pi_u$ (matching $z\to\Sigma_u^+$, $(x,y)\to\Pi_u$) and rotations $\Gamma_{rot} = \Pi_g$ (matching $(R_x,R_y)\to\Pi_g$; there is no $R_z$ for a linear molecule). Subtracting,

Counting degrees of freedom: $\Sigma_g^+$ and $\Sigma_u^+$ are each one-dimensional (1 mode each), $\Pi_u$ is two-dimensional (doubly degenerate bend, 2 modes), total $1+1+2 = 4 = 3N-5$ ✓, matching $3(3)-5=4$. Selection rules from the $D_{\infty h}$ table: $z\to \Sigma_u^+$; $(x,y)\to\Pi_u$; quadratics $(x^2+y^2,z^2)\to\Sigma_g^+$, $(xz,yz)\to\Pi_g$, $(x^2-y^2,xy)\to\Delta_g$.

• $\Sigma_g^+$ (symmetric stretch, $\nu_1$): matches the quadratic $\Sigma_g^+$ → Raman-active; does not match $z$ or $(x,y)$ → IR-inactive.
• $\Sigma_u^+$ (antisymmetric stretch, $\nu_3$): matches $z$ → IR-active; does not match any quadratic (all quadratics are gerade) → Raman-inactive.
• $\Pi_u$ (bend, $\nu_2$, doubly degenerate): matches $(x,y)$ → IR-active; gerade quadratics cannot match a $u$ mode → Raman-inactive.

This exactly reproduces the mutual-exclusion picture in the figure: $\nu_1$ (1388 cm⁻¹) Raman-only, $\nu_2$ (667 cm⁻¹) and $\nu_3$ (2349 cm⁻¹) IR-only, with zero overlap — direct experimental confirmation of the inversion-center argument.

Example 3 — Exam/competition level: $XeF_4$, $D_{4h}$, full reduction and IR/Raman table. $XeF_4$ is square planar, $N=5$ ($1$ Xe + $4$ F), so $3N=15$. The $D_{4h}$ point group has 8 classes: $E, 2C_4, C_2, 2C_2', 2C_2'', i, 2S_4, \sigma_h, 2\sigma_v, 2\sigma_d$ (order $h=16$; here written with the 10-class breakdown standard for $D_{4h}$: $E, 2C_4, C_2, 2C_2', 2C_2'', i, 2S_4, \sigma_h, 2\sigma_v, 2\sigma_d$). Place the molecule in the $xy$-plane with Xe at the origin and the four F atoms along $\pm x, \pm y$. Unmoved atoms per operation: $E$: all 5 unmoved. $C_4$ (about $z$): only Xe unmoved (F atoms cycle). $C_2$ (=$C_4^2$, about $z$): only Xe unmoved (F's swap in pairs, $+x\leftrightarrow-x$, $+y\leftrightarrow-y$; none stay fixed since the F atom positions map to the opposite F position). $C_2'$ (through two opposite F atoms, e.g. the $x$-axis): Xe and the two F's on that axis are unmoved (3 atoms), the other two F's swap. $C_2''$ (through F–Xe–F bisectors, e.g. the line at 45°): only Xe is unmoved (all four F's move). $i$: only Xe unmoved (F's map $+x\to-x$ etc., none fixed). $S_4$: only Xe unmoved. $\sigma_h$ (the molecular $xy$-plane): all 5 atoms lie in this plane, all unmoved. $\sigma_v$ (containing the Xe–F axes, e.g. $xz$ and $yz$ planes): Xe and the 2 F's on that axis lie in the plane (3 atoms unmoved) — there are two such planes ($xz$, $yz$), each fixing a different opposite pair of F's. $\sigma_d$ (diagonal planes, bisecting F–Xe–F angles, containing the $z$-axis and the 45° diagonals): only Xe lies in each diagonal plane (1 atom unmoved), since no F atom sits on the diagonal. Per-atom contribution factors: proper rotation $C_n$: $1+2\cos\theta$; improper $S_n, \sigma, i$: $-1+2\cos\theta$.

So $\Gamma_{3N}$ has characters $(15, 1, -1, -3, -1, -3, -1, 5, 3, 1)$ over classes $(E, 2C_4, C_2, 2C_2', 2C_2'', i, 2S_4, \sigma_h, 2\sigma_v, 2\sigma_d)$. Reducing against the ten $D_{4h}$ irreps ($A_{1g}, A_{2g}, B_{1g}, B_{2g}, E_g, A_{1u}, A_{2u}, B_{1u}, B_{2u}, E_u$) with $h=16$ using $n_i = \frac{1}{16}\sum g_R \chi_{3N}(R)\chi_i(R)$ (carrying out each dot product with the class sizes $1,2,1,2,2,1,2,1,2,2$) yields the standard, textbook-verified result:

Check total dimension: $1+1+1+1+2+2(1)+1+3(2) = 1+1+1+1+2+2+1+6=15$ ✓. Translations: $z\to A_{2u}$, $(x,y)\to E_u$, so $\Gamma_{trans}=A_{2u}+E_u$. Rotations: $R_z\to A_{2g}$, $(R_x,R_y)\to E_g$, so $\Gamma_{rot}=A_{2g}+E_g$.

Dimension count: $1+1+1+2(1)+1+2(2) = 1+1+1+2+1+4=10$, and indeed $3N-6=3(5)-6=9$ — but wait, this needs the correct count: $1(A_{1g})+1(B_{1g})+1(B_{2g})+2\times1(A_{2u})+1(B_{2u})+2\times2(E_u) = 1+1+1+2+1+4 = 10$. Since $3N-6=9$, one $A_{2u}$ must be subtracted again — this is because $2A_{2u}$ appeared in $\Gamma_{3N}$ but only one $A_{2u}$ is translation ($z$), leaving exactly one $A_{2u}$ as a genuine vibration, which the arithmetic above already reflects correctly as "$2A_{2u} - A_{2u} = A_{2u}$" inside $\Gamma_{vib}$. Written out fully and correctly:

Dimension check: $1+1+1+1+1+2(2) = 1+1+1+1+1+4 = 9 = 3(5)-6$ ✓. This is the correct, standard $XeF_4$ vibrational representation (9 vibrational degrees of freedom: 6 distinct frequencies since $E_u$ modes are doubly degenerate, giving $5+2(2)=9$ individual normal coordinates across $A_{1g}, B_{1g}, B_{2g}, A_{2u}, B_{2u}$, and $2\times E_u$). Selection rules from the $D_{4h}$ table:

• $A_{1g}$: matches quadratics $x^2+y^2, z^2$ → Raman-active only (symmetric Xe–F stretch).
• $B_{1g}$: matches quadratic $x^2-y^2$ → Raman-active only (a symmetric stretch combination).
• $B_{2g}$: matches quadratic $xy$ → Raman-active only (in-plane deformation).
• $A_{2u}$: matches $z$ → IR-active only (out-of-plane bend).
• $B_{2u}$: matches nothing in the linear or quadratic columns of $D_{4h}$ → silent in both IR and Raman (this is the "silent mode," experimentally confirmed by the absence of any band assignable to it, found instead only through combination bands or inelastic neutron scattering).
• $E_u$ (appears twice, i.e. two distinct $E_u$ vibrations): matches $(x,y)$ → IR-active only, doubly degenerate each.

Every gerade mode ($A_{1g}, B_{1g}, B_{2g}$) is Raman-only; every ungerade mode ($A_{2u}, B_{2u}, E_u$) is either IR-only or silent — never Raman-active. This is mutual exclusion again, now demonstrated on a five-atom molecule with a genuinely silent mode as a bonus subtlety.

The systematic marriage of group theory to molecular vibration is largely the achievement of Eugene Wigner and, in the specific chemical-spectroscopy form used daily by chemists, of Gerhard Herzberg, whose monumental Infrared and Raman Spectra of Polyatomic Molecules (1945) turned character-table bookkeeping into the standard working tool of experimental spectroscopists. Wigner's earlier and more abstract 1927–1931 work established that quantum states of a symmetric system must transform as irreducible representations of the system's symmetry group, and that transition probabilities between such states are controlled by the vanishing or non-vanishing of group-theoretical matrix elements — the mathematical bedrock beneath every character-table selection rule used in this article.