Character Tables and Decomposition of Representations

A finite group hides all of its linear algebra inside one small grid of integers and roots of unity: the character table. Each irreducible representation (irrep) is a way the group can act by matrices that cannot be block-diagonalized further; each is summarized by its character $\chi^{(\alpha)}(g)=\mathrm{Tr}\,D^{(\alpha)}(g)$ , the trace of its matrices. Because the trace is invariant under conjugation, $\chi(hgh^{-1})=\chi(g)$ , the character is a class function — constant on each conjugacy class. The table stacks these class functions as rows (irreps) against columns (conjugacy classes), and a counting theorem forces it to be square: the number of inequivalent irreps equals the number of classes. Nothing about the actual matrices is kept, yet everything needed to decompose, count invariants, and impose selection rules survives. The engine that makes the table useful is orthogonality. The rows are orthonormal under a group-averaged inner product — a consequence of Schur's lemmas and the great orthogonality theorem — and the columns obey a dual relation. From these follow the dimension sum rule $\sum_\alpha d_\alpha^2=|G|$ and the fact that every irrep appears in the regular representation exactly $d_\alpha$ times. The operational payoff is the reduction formula: any representation, however reducible, breaks into irreps with multiplicities $a_\alpha=\langle\chi^{(\alpha)},\chi\rangle$ read off from traces alone. This turns the structural question "how does this thing decompose?" into arithmetic — and because the answers are integers, the arithmetic checks itself.

Key takeaways

A character table arranges the irreducible characters $\chi^{(\alpha)}$ of a finite group $G$ in a square array, rows indexed by irreps and columns by conjugacy classes, weighted by class sizes $n_C$ .
Row orthogonality $\tfrac{1}{|G|}\sum_C n_C\,\chi^{(\alpha)}(C)^*\chi^{(\beta)}(C)=\delta_{\alpha\beta}$ makes the rows an orthonormal basis of class functions, forcing #irreps = #classes.
The reduction formula $a_\alpha=\tfrac{1}{|G|}\sum_C n_C\,\chi^{(\alpha)}(C)^*\chi(C)$ extracts how many times each irrep appears in any representation from its character alone — never the matrices.
The dimension sum rule $\sum_\alpha d_\alpha^2=|G|$ and the regular-representation multiplicity $a_\alpha=d_\alpha$ fix the size of the table and many of its entries.
A representation is irreducible iff $\tfrac{1}{|G|}\sum_C n_C|\chi(C)|^2=1$ ; worked through, the machine sends the $C_{3v}$ Cartesian rep to $A_1\oplus E$ and ammonia's $3N$ motion to vibrations $2A_1\oplus 2E$ .
Key idea: characters are class functions whose table rows are orthonormal, so projecting any reducible character onto each irreducible character (the reduction formula) yields integer multiplicities — decomposing a representation becomes finite linear algebra done with traces.

The core insight

Characters are class functions whose rows in the character table are orthonormal; projecting any reducible character onto each irreducible character (the reduction formula) returns integer multiplicities, so decomposing any representation becomes a finite linear-algebra computation carried out with traces alone.

Where this lives today

The character table stopped being a hand tool decades ago and became an engine inside software. Every quantum-chemistry package — Gaussian, ORCA, Q-Chem — assigns Mulliken symmetry labels to molecular orbitals and normal modes by running exactly the reduction formula of Example 2 on a molecule's point group, then uses the trivial-irrep selection rule to predict which vibrational bands are infrared or Raman active before a single integral is evaluated. The same machine, scaled to the 230 crystallographic space groups, drives modern condensed-matter physics: electronic bands at high-symmetry points carry irreps of the little group, and the 2017 "topological quantum chemistry" program of Bradlyn, Bernevig and collaborators recast the search for topological insulators as a question about which band representations are induced from atomic orbitals and which are not — a character-theoretic diagnosis now automated on the Bilbao Crystallographic Server to screen tens of thousands of materials. The pattern is always Frobenius's: turn a hard structural question into an integer read off a table.

The physical picture

Think of the irreducible characters as an orthonormal basis of pure tones for the space of class functions. Any representation you meet in a physics problem — the $3N$ displacements of a molecule, the $2\ell+1$ orbitals of an atomic level in a crystal, a tensor product of two symmetry species — has a character, and that character is a chord: a superposition of pure tones with integer loudness. The reduction formula is the Fourier-like projection that reads how loud each tone is, and because dimensions are integers, every answer must be a whole number — a built-in consistency check that catches an arithmetic slip on sight. The first figure shows the smallest complete instance of the basis: the character table of $C_{3v}$ , the symmetry group of ammonia, laid out as the square grid the counting theorem guarantees.

Figure

The character table of C₃ᵥ is a square 3×3 grid because the group has 3 conjugacy classes and hence 3 irreducible representations; the identity column lists the dimensions 1, 1, 2 and Σ dα² = 1 + 1 + 4 = 6 = the group order

The reason a class function is the right object is geometric: conjugacy sorts the group elements into shape bins — all rotations by the same angle, all reflections of one kind — and a trace cannot tell two elements in the same bin apart, so the honest data are the per-class values, one per column. Feeding a reducible character into the reduction formula spreads it back over the pure tones as a list of symmetry labels with integer weights. The second figure runs this projection on the case that opens the worked examples: the three Cartesian displacements of a $C_{3v}$ molecule, whose character $(3,0,1)$ resolves into one copy of $A_1$ and one of $E$ .

Figure

Decomposition as projection: the reducible Cartesian character χ = 3, 0, 1 of C₃ᵥ projected onto each irreducible character by the reduction formula gives multiplicities a = 1, 0, 1, so the 3D representation resolves into A₁ ⊕ E — a spectrum of pure symmetry tones

The core equations

Everything begins with the character and the single fact that makes it powerful: the trace does not see conjugation, so the character is a class function needing only one value per class.

\chi^{(\alpha)}(g)=\mathrm{Tr}\,D^{(\alpha)}(g),\qquad \chi^{(\alpha)}(hgh^{-1})=\chi^{(\alpha)}(g) \qquad (1)

where: $D^{(\alpha)}(g)$ — the matrix of irrep $\alpha$ representing $g$ ; $\chi^{(\alpha)}(g)$ — its character (trace); $h$ — any group element, showing invariance under conjugation. Reading: because $\mathrm{Tr}(ABA^{-1})=\mathrm{Tr}\,B$ , the character is constant on each class, fully specified by one value per class. The rows are orthonormal under the group-averaged inner product — the central theorem, and what makes the table square and the reduction formula work.

\langle\chi^{(\alpha)},\chi^{(\beta)}\rangle \equiv \frac{1}{|G|}\sum_{C} n_C\,\chi^{(\alpha)}(C)^*\,\chi^{(\beta)}(C)=\delta_{\alpha\beta} \qquad (2)

where: $|G|$ — the order of the group; $n_C$ — the number of elements in conjugacy class $C$ ; $\delta_{\alpha\beta}$ — the Kronecker delta over irreps. Reading: distinct irreps are orthogonal and each has unit norm, so the irreducible characters form an orthonormal set of class functions. Dual to the rows, the columns are orthogonal too — the practical tool for completing a table when a row is missing.

\sum_{\alpha} \chi^{(\alpha)}(C)^*\,\chi^{(\alpha)}(C')=\frac{|G|}{n_C}\,\delta_{CC'} \qquad (3)

where: $C,C'$ — two conjugacy classes; $n_C$ — the size of class $C$ ; $\delta_{CC'}$ — unity when the classes coincide, else zero. Reading: the columns, weighted by $|G|/n_C$ , are mutually orthogonal, making the number of independent columns equal the number of rows. Any representation is a direct sum of irreps with non-negative integer multiplicities, and its character is the matching sum of irreducible characters.

D=\bigoplus_\alpha a_\alpha\,D^{(\alpha)},\qquad \chi(C)=\sum_\alpha a_\alpha\,\chi^{(\alpha)}(C) \qquad (4)

where: $a_\alpha$ — the multiplicity (a non-negative integer) of irrep $\alpha$ ; $\chi(C)$ — the character of $D$ on class $C$ . Reading: characters add under direct sums, so the character of a big representation is a linear combination of table rows with the multiplicities as coefficients. The reduction (decomposition) formula inverts equation (4) by projecting onto row $\alpha$ — the workhorse of the entire subject.

a_\alpha=\frac{1}{|G|}\sum_{C} n_C\,\chi^{(\alpha)}(C)^*\,\chi(C) \qquad (5)

where: $a_\alpha$ — the multiplicity of irrep $\alpha$ in $D$ , which must come out a non-negative integer; $\chi^{(\alpha)}(C)^*$ — the complex conjugate of the irreducible character (it matters when characters are complex). Reading: to find how many copies of $\alpha$ live inside $D$ , take the class-weighted overlap of $D$ 's character with row $\alpha$ ; a fractional answer signals an error. The norm of a character diagnoses irreducibility on the spot, without ever splitting the representation.

\frac{1}{|G|}\sum_{C} n_C\,|\chi(C)|^2 = \sum_\alpha a_\alpha^2 \ \ge 1,\qquad =1\iff D\ \text{irreducible} \qquad (6)

where: $|\chi(C)|^2$ — the squared modulus of the character; $\sum_\alpha a_\alpha^2$ — the sum of squared multiplicities. Reading: the self-overlap equals $\sum a_\alpha^2$ , which is $1$ exactly when a single irrep appears once, and a larger integer flags a reducible representation. Two further identities close the toolkit: the dimension sum rule, which fixes the table's size, and the tensor-product rule behind every selection rule.

\sum_{\alpha} d_\alpha^2 = |G|,\quad d_\alpha=\chi^{(\alpha)}(e);\qquad \chi^{D_1\otimes D_2}(g)=\chi^{D_1}(g)\,\chi^{D_2}(g) \qquad (7)

where: $d_\alpha$ — the dimension of irrep $\alpha$ , equal to its character on the identity $e$ ; $D_1\otimes D_2$ — the tensor (Kronecker) product representation. Reading: the identity column lists the dimensions and their squares total $|G|$ ; product characters multiply class by class, so a product reduces by the same formula (5).

Derivation ARow orthogonality of characters from the great orthogonality theorem

The goal is to derive the row-orthogonality relation (2) — that the irreducible characters are orthonormal — starting from the great orthogonality theorem for matrix elements, itself the direct output of Schur's lemmas. The strategy is to collapse the matrix-element relation to a statement about traces by setting the row and column indices equal and summing, then repackage the sum over group elements as a class-weighted sum. Nothing new is assumed; every step is index bookkeeping, and the payoff is the single most important equation in the theory.

Step 1. Begin with the great orthogonality theorem, $\sum_g D^{(\alpha)}(g)_{ij}^*\,D^{(\beta)}(g)_{kl}=\tfrac{|G|}{d_\alpha}\,\delta_{\alpha\beta}\,\delta_{ik}\,\delta_{jl}$ , where $i,j,k,l$ label matrix entries. Why: this is the raw content of Schur's lemmas — matrix elements of inequivalent irreps are orthogonal as functions on the group, with a normalization $|G|/d_\alpha$ set by the second lemma.

Step 2. Trace both factors by setting $i=j$ and $k=l$ and summing over them, so the left side becomes $\sum_g \chi^{(\alpha)}(g)^*\chi^{(\beta)}(g)$ and the right becomes $\tfrac{|G|}{d_\alpha}\delta_{\alpha\beta}\sum_{i,k}\delta_{ik}\delta_{ik}=\tfrac{|G|}{d_\alpha}\delta_{\alpha\beta}\,d_\alpha=|G|\,\delta_{\alpha\beta}$ . Why: summing $\delta_{ik}\delta_{ik}=\sum_i 1=d_\alpha$ exactly cancels the $1/d_\alpha$ , so the dimension-dependent factor disappears and the characters carry a clean normalization.

Step 3. Convert the sum over elements to a sum over classes: since $\chi$ is a class function, $\sum_g f(g)=\sum_C n_C f(C)$ , and dividing by $|G|$ gives $\tfrac{1}{|G|}\sum_C n_C\,\chi^{(\alpha)}(C)^*\chi^{(\beta)}(C)=\delta_{\alpha\beta}$ . Why: the class-size weight $n_C$ is not optional — it accounts for the many group elements collapsed into each column, and dropping it is the single commonest error in the whole subject.

Go deeper · Derivation B · 1 minThe reduction formula by projecting onto an irreducible character

Derivation BThe reduction formula by projecting onto an irreducible character

The goal is to establish the decomposition formula (5), which returns the multiplicity of each irrep in an arbitrary representation from its character alone. The idea is exactly that of Fourier analysis: the irreducible characters are an orthonormal basis, and any class function is expanded in it with coefficients found by inner products. Applied to a representation's character, those coefficients are the multiplicities, and orthonormality guarantees that only one term survives the projection — converting a structural question into a finite sum of table entries.

Step 1. Expand the given character in irreducible characters using equation (4): $\chi=\sum_\beta a_\beta\,\chi^{(\beta)}$ , so that $\langle\chi^{(\alpha)},\chi\rangle=\sum_\beta a_\beta\,\langle\chi^{(\alpha)},\chi^{(\beta)}\rangle$ . Why: the character of a direct sum is the sum of characters, and the inner product is linear, so the overlap distributes across the irreducible pieces.

Step 2. Apply row orthonormality, $\langle\chi^{(\alpha)},\chi^{(\beta)}\rangle=\delta_{\alpha\beta}$ , so every term with $\beta\neq\alpha$ vanishes and only $\beta=\alpha$ survives, leaving $a_\alpha=\langle\chi^{(\alpha)},\chi\rangle$ . Why: orthonormality is what isolates a single coefficient — this is the discrete analogue of extracting a Fourier coefficient by integrating against one basis function.

Step 3. Write the surviving inner product explicitly as $a_\alpha=\tfrac{1}{|G|}\sum_C n_C\,\chi^{(\alpha)}(C)^*\chi(C)$ , a finite sum over the columns of the table. Why: because each $a_\alpha$ must be a non-negative integer, the formula is self-checking — a fractional or negative result means a wrong table entry, a dropped $n_C$ , or a missing conjugate.

Go deeper · Derivation C · 1 minThe regular representation and the dimension sum rule

Derivation CThe regular representation and the dimension sum rule

The goal is to prove the dimension sum rule $\sum_\alpha d_\alpha^2=|G|$ by decomposing one special representation — the regular representation, in which the group acts on itself by left multiplication. The plan is to compute its character directly (it is almost entirely zero), feed it into the reduction formula, discover that each irrep appears exactly $d_\alpha$ times, and then match total dimensions on both sides. The sum rule that falls out both fixes how many rows the table can have and provides a stringent check on any proposed set of dimensions.

Step 1. Compute the regular character. Left multiplication by $g$ permutes the $|G|$ basis vectors labelled by group elements, with a fixed point only when $g=e$ , so $\chi^{\mathrm{reg}}(g)=|G|$ for $g=e$ and $0$ otherwise. Why: a permutation matrix's trace counts its fixed points, and $gh=h$ forces $g=e$ , so only the identity contributes any diagonal.

Step 2. Reduce it. In $a_\alpha^{\mathrm{reg}}=\tfrac{1}{|G|}\sum_C n_C\,\chi^{(\alpha)}(C)^*\chi^{\mathrm{reg}}(C)$ only the identity class ( $n_e=1$ ) survives, because the regular character is zero on every other class. Why: the reduction formula weights each class by the reducible character there, and that character is supported entirely on the identity.

Step 3. Evaluate the one surviving term: $a_\alpha^{\mathrm{reg}}=\tfrac{1}{|G|}\cdot 1\cdot \chi^{(\alpha)}(e)^*\cdot|G|=\chi^{(\alpha)}(e)=d_\alpha$ . Why: the identity character of an irrep is its dimension, since $D^{(\alpha)}(e)$ is the $d_\alpha\times d_\alpha$ unit matrix — so every irrep occurs in the regular representation dimension-many times.

Step 4. Match dimensions: $|G|=\dim\,\mathrm{reg}=\sum_\alpha a_\alpha^{\mathrm{reg}}\,d_\alpha=\sum_\alpha d_\alpha^2$ . Why: the regular representation has dimension $|G|$ by construction, and summing the dimensions of its irreducible pieces gives the sum rule — for $C_{3v}$ , $1^2+1^2+2^2=6=|G|$ , exactly.

Limiting cases

The orthogonality relations collapse to a set of standing constraints that recur in every application; each row below is a special case worth memorizing.

Case	Result	Meaning
Trivial irrep	$\chi^{(1)}(C)=1\ \ \forall C$	Every group has the one-dimensional trivial rep; its multiplicity $a_1=\tfrac{1}{\|G\|}\sum_C n_C\chi(C)$ counts the $G$ -invariant vectors — the seat of every selection rule.
Identity column	$\chi^{(\alpha)}(e)=d_\alpha$	The first column lists the irrep dimensions; squaring and summing gives $\sum_\alpha d_\alpha^2=\|G\|$ .
Regular representation	$\chi^{\mathrm{reg}}(g)=\|G\|\,\delta_{g,e}$	Each irrep appears $a_\alpha=d_\alpha$ times, so the group's own action already contains every irrep.
Abelian group	$d_\alpha=1\ \forall\alpha,\ \ \#\text{irreps}=\|G\|$	Every element is its own class, all irreps are one-dimensional, and the table is that of a product of cyclic groups.
Irreducible input	$\tfrac{1}{\|G\|}\sum_C n_C\|\chi(C)\|^2=1$	The character norm equals one precisely when the representation cannot be split further.

Common trapthe class-weight and complex-conjugate trap

The most persistent mistake in decomposition is treating the character inner product as a plain sum. The correct projection is $a_\alpha=\tfrac{1}{|G|}\sum_C n_C\,\chi^{(\alpha)}(C)^*\chi(C)$ — a sum over classes, each weighted by its size $n_C$ , with a complex conjugate on the irreducible character. Forgetting $n_C$ silently under-counts every multi-element class (in $C_{3v}$ the reflection class alone carries weight three), so multiplicities come out fractional. Dropping the conjugate does no harm for real characters but corrupts every group with complex entries — the cyclic groups $C_n$ for $n\ge3$ , where characters are genuine roots of unity. The tell is always the same: if any $a_\alpha$ is not a non-negative integer, one of these two weights is missing, or a table entry or class ordering is wrong. Re-derive, do not round.

Worked examples

1 · Decomposing the Cartesian representation of C₃ᵥ (confidence builder)

The three Cartesian displacements $(x,y,z)$ of an ammonia-like molecule furnish a 3-dimensional representation of $C_{3v}$ . Decompose it into irreps. The classes are $E\,(n=1)$ , $2C_3\,(n=2)$ , $3\sigma_v\,(n=3)$ , so $|G|=6$ ; the table rows are $\chi^{A_1}=(1,1,1)$ , $\chi^{A_2}=(1,1,-1)$ , $\chi^{E}=(2,-1,0)$ .

Step 1. Find the character of the 3D representation. A proper rotation by angle $\theta$ acts on $(x,y,z)$ with trace $1+2\cos\theta$ ; a reflection has trace $+1$ . Thus $\chi(E)=1+2\cos 0=3$ , $\chi(C_3)=1+2\cos 120°=1+2(-\tfrac12)=\mathbf{0}$ , and $\chi(\sigma_v)=+1$ , giving the reducible character $\chi=(3,0,1)$ .

Step 2. Project onto $A_1$ : $a_{A_1}=\tfrac{1}{6}\big[1\cdot1\cdot3+2\cdot1\cdot0+3\cdot1\cdot1\big]=\tfrac{3+0+3}{6}=\mathbf{1}$ . One copy of the totally symmetric irrep.

Step 3. Project onto $A_2$ : $a_{A_2}=\tfrac{1}{6}\big[1\cdot1\cdot3+2\cdot1\cdot0+3\cdot(-1)\cdot1\big]=\tfrac{3+0-3}{6}=\mathbf{0}$ . No $A_2$ content.

Step 4. Project onto $E$ : $a_{E}=\tfrac{1}{6}\big[1\cdot2\cdot3+2\cdot(-1)\cdot0+3\cdot0\cdot1\big]=\tfrac{6+0+0}{6}=\mathbf{1}$ . One copy of the two-dimensional irrep. Dimension check: $1\cdot1+1\cdot2=3$ . Result: Γxyz = A₁ ⊕ E. Physically, $z$ (along the symmetry axis) transforms as $A_1$ and the in-plane pair $(x,y)$ together as $E$ .

2 · Full normal-mode analysis of ammonia, NH₃ (fully worked)

Ammonia has four atoms in $C_{3v}$ symmetry, so its $3N=12$ Cartesian displacements form a large reducible representation. Find the symmetry species of its normal vibrations.

Step 1. Build the $3N$ character from $\chi(R)=N_{\mathrm{unmoved}}(R)\times\xi(R)$ , where the per-atom contribution $\xi$ is $1+2\cos\theta$ for a proper operation and $-1+2\cos\theta$ for an improper one. Under $E$ all four atoms stay and $\xi=3$ , so $\chi(E)=4\times3=\mathbf{12}$ . Under $C_3$ only the axial nitrogen is unmoved and $\xi=1+2\cos120°=0$ , so $\chi(C_3)=\mathbf{0}$ . Under each $\sigma_v$ the nitrogen and the one in-plane hydrogen stay ( $N_{\mathrm{unmoved}}=2$ ) with $\xi=-1+2\cos0=1$ , so $\chi(\sigma_v)=\mathbf{2}$ . Thus $\Gamma_{\mathrm{tot}}=(12,0,2)$ .

Step 2. Reduce $\Gamma_{\mathrm{tot}}$ with formula (5), weights $n_C=(1,2,3)$ : $a_{A_1}=\tfrac{1}{6}[1\cdot1\cdot12+2\cdot1\cdot0+3\cdot1\cdot2]=\tfrac{18}{6}=\mathbf{3}$ ; $a_{A_2}=\tfrac{1}{6}[1\cdot1\cdot12+2\cdot1\cdot0+3\cdot(-1)\cdot2]=\tfrac{6}{6}=\mathbf{1}$ ; $a_{E}=\tfrac{1}{6}[1\cdot2\cdot12+2\cdot(-1)\cdot0+3\cdot0\cdot2]=\tfrac{24}{6}=\mathbf{4}$ . So $\Gamma_{\mathrm{tot}}=3A_1\oplus A_2\oplus 4E$ , dimension $3+1+8=12$ . ✓

Step 3. Remove the rigid-body motions. Translations span $\Gamma_{\mathrm{trans}}=\Gamma_{xyz}=A_1\oplus E$ (from Example 1), and rotations span $\Gamma_{\mathrm{rot}}=A_2\oplus E$ (with $R_z\sim A_2$ and $(R_x,R_y)\sim E$ ).

Step 4. Subtract: $\Gamma_{\mathrm{vib}}=\Gamma_{\mathrm{tot}}-\Gamma_{\mathrm{trans}}-\Gamma_{\mathrm{rot}}=(3A_1+A_2+4E)-(A_1+E)-(A_2+E)=\mathbf{2A_1\oplus 2E}$ . Dimension $2\cdot1+2\cdot2=6=3N-6$ , correct for a non-linear four-atom molecule. Result: NH₃ has vibrations 2A₁ ⊕ 2E — four distinct frequencies (symmetric stretch and umbrella bend, both $A_1$ ; asymmetric stretch and bend, both doubly-degenerate $E$ ). In $C_{3v}$ both $A_1$ and $E$ are infrared- and Raman-active, so all four bands appear in both spectra — as observed.

Physics of the portrait

Character theory was not born in physics at all; it was invented by a pure algebraist chasing a question about group determinants, and only later became the daily language of spectroscopy and solid-state physics.

Origins · 1 minGeorg Frobenius and how this was discovered

Georg Frobenius

1849–1917 · the invention of group characters and their orthogonality

Ferdinand Georg Frobenius created the character theory of finite groups almost single-handedly in a burst of papers beginning in 1896, prompted by a correspondence with Dedekind about factoring the "group determinant". Where earlier workers had matrices, Frobenius realized that a single number attached to each conjugacy class — the character — already carried the essential information, and he proved the orthogonality relations that make the character table a rigid, self-checking object. Within two years he had the full apparatus: irreducible characters, the reduction of the regular representation, the dimension sum rule $\sum_\alpha d_\alpha^2=|G|$ , and the induction of characters from subgroups that still bears his name. Issai Schur, his student, distilled the proofs to the two lemmas now taught first. For three decades this was regarded as beautiful but useless mathematics. Then quantum mechanics arrived, and the class functions Frobenius had built to factor a determinant turned out to be exactly the labels nature uses for atomic terms, molecular vibrations, and crystal-field levels — Wigner and Bethe simply read the physics off the tables. Frobenius died in Berlin in 1917, a decade before the characters he defined became the working vocabulary of physics.

History

1896

— Frobenius defines the characters of a finite group and proves their orthogonality relations

1897

— Frobenius founds the representation theory of finite groups; Burnside's treatise on group theory appears

1905

— Schur reduces the theory to two lemmas, sharpening the great orthogonality theorem

1929

— Bethe splits atomic terms in a crystal field by reducing rotation-group levels under a point group

1931

— Wigner applies characters to atomic spectra in his Gruppentheorie und ihre Anwendung

Going further

Texts. Tinkham, Group Theory and Quantum Mechanics, Chs. 3–4 (characters and applications) · Stone & Goldbart, Mathematics for Physics, representation-theory chapter · Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., group-theory chapter · Hassani, Mathematical Physics, 2nd ed., Part on group representation theory · Hamermesh, Group Theory and Its Application to Physical Problems, Ch. 3 · Nakahara, Geometry, Topology and Physics, for the Lie-group extension. The key numerical benchmarks are: row orthogonality follows from the great orthogonality theorem via traces; the regular-representation result gives $a_\alpha=d_\alpha$ and hence $\sum_\alpha d_\alpha^2=|G|$ ; the $C_{3v}$ table entries are $A_1=(1,1,1)$ , $A_2=(1,1,-1)$ , $E=(2,-1,0)$ with class sizes $(1,2,3)$ , order $6$ , and Cartesian character $(3,0,1)$ giving $A_1\oplus E$ ; and the ammonia analysis yields $\Gamma_{\mathrm{tot}}=(12,0,2)\to 3A_1+A_2+4E$ and $\Gamma_{\mathrm{vib}}=2A_1+2E$ .

Prerequisites. Conjugacy Classes and Normal Subgroups · Group Representations · Reducible and Irreducible Representations · Schur's Lemmas · Characters and Orthogonality Theorems. Leads to. Tensor Products and Clebsch–Gordan Decomposition · The Rotation Group SO(3) · SU(2) and the Spin Double Cover.

Ask the tutor: Hand me any molecule's point group or any reducible character and I will walk you through the reduction formula class by class until the integer multiplicities fall out.

Why must the number of irreducible representations equal the number of conjugacy classes?

Both count the dimension of the same vector space: the class functions on the group. A class function assigns one number per conjugacy class, so its dimension is the number of classes. Separately, the irreducible characters are linearly independent (row orthogonality) and they span the class functions. A basis has a fixed size, so the number of irreducible characters equals the number of classes — which is exactly why the character table is square.

What is the difference between the reduction formula and just staring at the character table?

The table lists only the irreducible characters — the "pure tones". A representation you meet in practice, like the $3N$ displacements of a molecule, is a reducible mixture whose character is some other vector, not a table row. The reduction formula is the projection that tells you which rows, and how many of each, add up to your vector. Without it you would be guessing combinations; with it the multiplicities drop out as a mechanical class-weighted sum. It is the difference between owning a dictionary and being able to translate a sentence.

Why does the complex conjugate appear, and when does it actually matter?

The character inner product is a genuine Hermitian inner product on class functions, and Hermitian inner products conjugate one argument. For groups whose characters are all real — every dihedral group, the symmetric groups, most point groups used in molecular physics — the conjugate changes nothing and is easy to forget. It becomes essential for groups with genuinely complex characters, above all the cyclic groups $C_n$ with $n\ge 3$ , whose irreducible characters are powers of the root of unity $e^{2\pi i/n}$ . There, omitting the conjugate produces complex or fractional "multiplicities" that expose the mistake at once.

How do selection rules come out of this machinery?

A matrix element $\langle f|\hat O|i\rangle$ is a number, and a number cannot change under symmetry — so it must vanish unless the integrand contains the totally symmetric (trivial) irrep. Using $\chi^{A\otimes B}=\chi^A\chi^B$ , form the product character of $\Gamma_f^*\otimes\Gamma_O\otimes\Gamma_i$ and apply the reduction formula to the trivial irrep. If that multiplicity is zero, the transition is forbidden by symmetry alone, no integral evaluated. This single argument underlies infrared and Raman activity, atomic dipole selection rules, and much of spectroscopy.

Can the reduction formula ever give a wrong answer if I do the arithmetic correctly?

No — that is its virtue. Given the correct character table, reducible character, and class sizes, the multiplicities are uniquely determined and guaranteed to be non-negative integers. This is why the formula is self-checking: a non-integer or negative result is proof of an input error, never a feature of the physics. The only genuine subtlety is upstream, in building the reducible character correctly — for instance getting the per-atom contributions and the count of unmoved atoms right in a vibrational analysis. Once that vector is correct, the decomposition cannot mislead you.

Does any of this survive for continuous groups like SO(3)?

Yes, with sums becoming integrals. For a compact continuous group the characters remain class functions — for $SO(3)$ a class is a rotation angle $\theta$ , and the spin- $\ell$ character is $\chi_\ell(\theta)=\sin\big((2\ell+1)\theta/2\big)/\sin(\theta/2)$ . Orthogonality holds against the invariant (Haar) measure, and the reduction formula becomes an integral over $\theta$ with weight $\tfrac{1}{\pi}\sin^2(\theta/2)\,d\theta$ . Decomposing a product of two such representations reproduces the Clebsch–Gordan series — the continuous cousin of everything here, and the subject of the nodes this one leads to.