Character Tables and the Orthogonality Theorem

Character tables are square arrays encoding the traces of the irreducible matrix representations of a finite group; the Great Orthogonality Theorem states that the matrix elements (and hence the characters) of distinct irreducible representations are mutually orthogonal as vectors over the group. Together they reduce all questions about how a group acts linearly into simple arithmetic of integers. Used pervasively in chemistry (molecular vibrations, spectroscopy), crystallography, and quantum mechanics. Foundational work by Frobenius (1896) and Schur (1905).

A representation of a finite group $G$ is a homomorphism $ρ : G \to G L (V)$ sending group elements to invertible matrices; its character is the function $χ (g) = tr ρ (g)$ . The Great Orthogonality Theorem asserts that the matrix entries of the irreducible representations form an orthogonal basis of the space of complex functions on $G$ , and the characters of inequivalent irreducibles are orthonormal under a natural inner product. This matters because it makes representations finite, classifiable data: a character table — rows indexed by irreducibles, columns by conjugacy classes — completely captures the representation theory of $G$ . The surprising fact is that mere traces (single numbers per class) suffice to determine a representation up to isomorphism, and that the number of irreducibles equals the number of conjugacy classes — a deep coincidence between two seemingly unrelated counts.

The Core Insight

The space $C [G]$ of all complex functions on a finite group carries a natural Hermitian inner product. The matrix coefficients of irreducible representations are orthogonal vectors in this space, and their lengths are fixed by the dimension of the representation. Taking traces collapses this to class functions (functions constant on conjugacy classes), where the irreducible characters form a complete orthonormal basis. Hence representation theory of finite groups is, at heart, Euclidean geometry in a finite-dimensional space whose coordinates are group elements.

Formal Statement

Representation, Irreducibility, Character

Let $G$ be a finite group of order $∣ G ∣ = n$ .

A (complex linear) representation is a homomorphism $ρ : G \to G L (V)$ where $V$ is a finite-dimensional $C$ -vector space of dimension $d = dim V$ , called the degree or dimension of $ρ$ .
$ρ$ is irreducible if $V$ has no $G$ -invariant subspace other than ${0}$ and $V$ .
The character of $ρ$ is $χ_{ρ} : G \to C$ , $χ_{ρ} (g) = tr ρ (g)$ .
Two representations are equivalent ( $≅$ ) if related by a change of basis: $ρ^{'} (g) = T ρ (g) T^{- 1}$ for a fixed invertible $T$ .

We write $ρ_{ij} (g)$ for the $(i, j)$ matrix entry of $ρ (g)$ in a chosen basis.

The natural inner product on functions $f_{1}, f_{2} : G \to C$ is

⟨ f_{1}, f_{2} ⟩ = \frac{1}{∣ G ∣} g \in G \sum f_{1} (g) \overline{f_{2} (g)} .

⟳average over the group to obtain a

G

-invariant Hermitian form; the factor

1/∣ G ∣

normalises so the trivial character has norm 1.

Great (Schur) Orthogonality of Matrix Coefficients

Let $ρ^{(α)}$ and $ρ^{(β)}$ be inequivalent irreducible unitary representations of $G$ , of dimensions $d_{α}, d_{β}$ . Then for all index choices,

\frac{1}{∣ G ∣} g \in G \sum ρ_{ij}^{(α)} (g) \overline{ρ_{k l}^{(β)} (g)} = \frac{1}{d _{α}} δ_{α β} δ_{ik} δ_{j l} .

⟳apply Schur's Lemma to the averaged operator

M = \frac{1}{∣ G ∣} \sum_{g} ρ^{(α)} (g) A ρ^{(β)} (g)^{- 1}

for arbitrary

A

; it intertwines the two representations.

First Orthogonality Relation for Characters

For irreducible characters $χ_{α}, χ_{β}$ ,

⟨ χ_{α}, χ_{β} ⟩ = \frac{1}{∣ G ∣} g \in G \sum χ_{α} (g) \overline{χ_{β} (g)} = δ_{α β} .

⟳set

i = j

and

k = l

in [2] and sum over

i

and

k

; the right-hand side gives

\frac{1}{d _{α}} δ_{α β} \sum_{i} \sum_{k} δ_{ik} δ_{ik} = δ_{α β}

Because characters are constant on conjugacy classes, [3] is more efficiently written as a sum over classes. Let $C_{1}, \dots, C_{r}$ be the conjugacy classes with representatives $g_{t}$ and sizes $∣ C_{t} ∣$ :

⟨ χ_{α}, χ_{β} ⟩ = \frac{1}{∣ G ∣} t = 1 \sum r ∣ C_{t} ∣ χ_{α} (g_{t}) \overline{χ_{β} (g_{t})} = δ_{α β} .

⟳group identical terms — every element of a class

C_{t}

contributes the same product, so multiply by

∣ C_{t} ∣

Second Orthogonality Relation (Column Orthogonality)

For two conjugacy classes with representatives $g_{s}, g_{t}$ , summing over all irreducible characters,

α \sum χ_{α} (g_{s}) \overline{χ_{α} (g_{t})} = δ_{s t} \frac{∣ G ∣}{∣ C _{t} ∣} = δ_{s t} ∣ Z (g_{t}) ∣,

where

Z (g_{t})

is the centraliser of

g_{t}

⟳the character table matrix (suitably weighted) is unitary; column orthogonality is the dual of row orthogonality. Equivalently, the rows being orthonormal and forming a complete set forces the columns to be orthogonal.

Character Table

The character table of $G$ is the $r \times r$ array whose rows are the irreducible characters $χ_{1}, \dots, χ_{r}$ and whose columns are the conjugacy classes $C_{1}, \dots, C_{r}$ , with entry $χ_{α} (g_{t})$ . Here a fundamental theorem guarantees:

(number of irreducible representations) = (number of conjugacy classes) = r,

so the table is square. Moreover the dimensions satisfy

α = 1 \sum r d_{α}^{2} = ∣ G ∣.

⟳[6] follows because irreducible characters form a basis of the class functions, a space of dimension

r

; [7] is column orthogonality [5] evaluated at the identity class, where

χ_{α} (e) = d_{α}

and

∣ C_{e} ∣ = 1

Why It Holds / Derivation

Great Orthogonality from Schur's Lemma

Step 1 — Unitarisation. Every finite-group representation is equivalent to a unitary one. Define

H = \sum_{g} ρ (g)^{†} ρ (g)

; it is positive-definite Hermitian,

G

-invariant, and conjugating by

H^{1/2}

makes

ρ

unitary. So assume

ρ^{(α)}, ρ^{(β)}

unitary:

ρ (g)^{- 1} = ρ (g)^{†}

, i.e.

ρ_{k l} (g)^{- 1 entry} = \overline{ρ_{l k} (g)}

. Step 2 — Build an intertwiner. For any

d_{α} \times d_{β}

matrix

A

, define

M = \frac{1}{∣ G ∣} g \in G \sum ρ^{(α)} (g) A ρ^{(β)} (g)^{- 1} .

A direct computation using the homomorphism property shows

ρ^{(α)} (h) M = M ρ^{(β)} (h)

for all

h

(reindex the sum by

g \mapsto h^{- 1} g

). Thus

M

intertwines the two representations. Step 3 — Apply Schur's Lemma.

If $α \neq = β$ (inequivalent irreducibles): Schur's Lemma forces $M = 0$ for every $A$ .
If $α = β$ : $M$ commutes with an irreducible representation, so $M = λ I$ . Taking traces, $λ = \frac{tr A}{d _{α}}$ .

Step 4 — Extract matrix elements. Choose

A = E_{j l}

(the matrix with a single 1 in position

(j, l)

). Then the

(i, k)

entry of

M

M_{ik} = \frac{1}{∣ G ∣} g \sum ρ_{ij}^{(α)} (g) (ρ^{(β)} (g)^{- 1})_{l k} = \frac{1}{∣ G ∣} g \sum ρ_{ij}^{(α)} (g) \overline{ρ_{k l}^{(β)} (g)},

using unitarity in the last step. Setting this equal to the Schur conclusion (

0

α \neq = β

;

\frac{1}{d _{α}} δ_{ik} δ_{j l}

α = β

) yields exactly [2].

■

Completeness — Irreducible Characters Span the Class Functions

Step 1. Let

f

be a class function orthogonal to every

χ_{α}

. For each irreducible

ρ^{(α)}

form

T_{α} = \sum_{g} \overline{f (g)} ρ^{(α)} (g)

. Because

f

is a class function,

T_{α}

commutes with all

ρ^{(α)} (h)

, so by Schur

T_{α} = μ I

with

μ = \frac{∣ G ∣}{d _{α}} ⟨ f, χ_{α} ⟩ = 0

. Step 2. Apply this to the regular representation, whose decomposition contains every irreducible. The operator

\sum_{g} \overline{f (g)} ρ_{reg} (g)

is zero, but evaluating on the basis vector

e_{e}

gives

\sum_{g} \overline{f (g)} e_{g} = 0

, forcing

f \equiv 0

. Step 3. Therefore the

χ_{α}

are orthonormal (by [3]) and complete in the

r

-dimensional space of class functions. A complete orthonormal set has exactly

r

elements, proving [6].

■

Ferdinand Georg Frobenius | 1849–1917

Frobenius single-handedly created the character theory of finite groups in a burst of papers from 1896–1900, motivated by a cryptic question from Dedekind about factoring the "group determinant." He defined characters as solutions to certain functional equations before he had matrix representations, then with Issai Schur placed the whole theory on the firm matrix-theoretic footing we use today. The orthogonality relations are, in spirit, his discovery; Schur's Lemma (1905) supplied the clean modern proof. Frobenius held the chair at Berlin and was famously rigorous and combative — but his character tables remain unchanged after 125 years.

Building Intuition

Think of the group algebra $C [G]$ as a vector space with one coordinate axis per group element — an $∣ G ∣$ -dimensional space. The Great Orthogonality Theorem says the matrix-coefficient functions $ρ_{ij}^{(α)}$ are mutually perpendicular vectors in this space, like the standard basis but adapted to the group's symmetry. There are exactly $\sum d_{α}^{2} = ∣ G ∣$ of them, so they form a complete orthogonal basis — equation [7] is just "the dimensions add up."

Characters are the shadows of representations: you throw away all off-diagonal information and keep only the trace, one number per conjugacy class. Remarkably, no information about which representation you have is lost — orthogonality means the characters are still independent. The picture to keep is that of an orthonormal frame (the irreducible characters) inside the function space, and any representation's character is just a vector with non-negative integer coordinates in that frame.

Irreducible characters as orthonormal axes in the space of class functions; an arbitrary character decomposes with integer multiplicities.

<svg viewBox="0 0 520 360" xmlns="http://www.w3.org/2000/svg">
  <rect width="520" height="360" fill="#0a0e17"/>
  <!-- axes -->
  <line x1="60" y1="300" x2="460" y2="300" stroke="#94a3b8" stroke-width="1.5"/>
  <line x1="60" y1="300" x2="60" y2="40" stroke="#94a3b8" stroke-width="1.5"/>
  <line x1="60" y1="300" x2="200" y2="180" stroke="#94a3b8" stroke-width="1.5"/>
  <text x="465" y="304" fill="#e2e8f0" font-size="13">χ₁ (trivial)</text>
  <text x="40" y="36" fill="#e2e8f0" font-size="13">χ₂</text>
  <text x="205" y="178" fill="#e2e8f0" font-size="13">χ₃</text>
  <!-- decomposed character vector -->
  <line x1="60" y1="300" x2="360" y2="120" stroke="#d4a853" stroke-width="2.5"/>
  <polygon points="360,120 350,124 356,132" fill="#d4a853"/>
  <text x="368" y="116" fill="#d4a853" font-size="13">χ = 2χ₁ + χ₂ + χ₃</text>
  <!-- projection guides -->
  <line x1="360" y1="120" x2="360" y2="300" stroke="#94a3b8" stroke-width="0.8" stroke-dasharray="4 3"/>
  <line x1="360" y1="120" x2="60" y2="120" stroke="#94a3b8" stroke-width="0.8" stroke-dasharray="4 3"/>
  <text x="62" y="116" fill="#e2e8f0" font-size="11">⟨χ,χ₂⟩=1</text>
  <text x="320" y="316" fill="#e2e8f0" font-size="11">⟨χ,χ₁⟩=2</text>
  <text x="100" y="340" fill="#e2e8f0" font-size="12">Class-function space: dim = # conjugacy classes</text>
</svg>

Worked Examples

Building the Character Table of $S_3$

Given.

G = S_{3}

∣ G ∣ = 6

. Conjugacy classes:

{e}

(size 1), three transpositions

{(12), (13), (23)}

(size 3), two 3-cycles

{(123), (132)}

(size 2). So

r = 3

classes

\Rightarrow

3 irreducibles. Step 1 — Dimensions. From [7]:

d_{1}^{2} + d_{2}^{2} + d_{3}^{2} = 6

. Every group has the trivial 1-D rep,

d_{1} = 1

. The only positive-integer solution is

1 + 1 + 4 = 6

, so dimensions are

(1, 1, 2)

. Step 2 — One-dimensional reps. Trivial

χ_{1} = (1, 1, 1)

. The sign representation

χ_{2}

sends even permutations to

+ 1

, odd to

- 1

χ_{2} = (1, - 1, 1)

on classes

(e, transposition, 3-cycle)

. Step 3 — The 2-D rep by orthogonality. Write

χ_{3} = (2, a, b)

since

χ_{3} (e) = d_{3} = 2

. Use [4]:

$⟨ χ_{3}, χ_{1} ⟩ = \frac{1}{6} (1 \cdot 2 + 3 \cdot a + 2 \cdot b) = 0 \Rightarrow 2 + 3 a + 2 b = 0$ .
$⟨ χ_{3}, χ_{2} ⟩ = \frac{1}{6} (2 - 3 a + 2 b) = 0 \Rightarrow 2 - 3 a + 2 b = 0$ .

Subtracting gives $6 a = 0 \Rightarrow a = 0$ ; then $2 + 2 b = 0 \Rightarrow b = - 1$ .

Step 4 — Verify normalisation.

⟨ χ_{3}, χ_{3} ⟩ = \frac{1}{6} (4 + 3 \cdot 0 + 2 \cdot 1) = \frac{6}{6} = 1.

✓ Boxed answer.

χ_{1} χ_{2} χ_{3} {e} 112 3 (ab) 1 - 1 0 2 (ab c) 11 - 1

(dimensionless — characters are pure complex numbers).

Decomposing a Reducible Representation of $S_3$

Given. The natural permutation representation

P

S_{3}

C^{3}

(permuting coordinates). Its character counts fixed points:

χ_{P} (g) = # {fixed points of g}

. Step 1 — Compute the character.

χ_{P} (e) = 3

; a transposition fixes 1 point

\Rightarrow χ_{P} = 1

; a 3-cycle fixes 0

\Rightarrow χ_{P} = 0

. So

χ_{P} = (3, 1, 0)

. Step 2 — Project onto each irreducible using

m_{α} = ⟨ χ_{P}, χ_{α} ⟩

via [4]:

$m_{1} = \frac{1}{6} (1 \cdot 3 \cdot 1 + 3 \cdot 1 \cdot 1 + 2 \cdot 0 \cdot 1) = \frac{1}{6} (3 + 3 + 0) = 1.$
$m_{2} = \frac{1}{6} (3 \cdot 1 + 3 \cdot 1 \cdot (- 1) + 0) = \frac{1}{6} (3 - 3) = 0.$
$m_{3} = \frac{1}{6} (3 \cdot 2 + 3 \cdot 1 \cdot 0 + 2 \cdot 0 \cdot (- 1)) = \frac{1}{6} (6) = 1.$

Step 3 — Check. Reconstructed character

1 \cdot χ_{1} + 0 \cdot χ_{2} + 1 \cdot χ_{3} = (1, 1, 1) + (2, 0, - 1) = (3, 1, 0) = χ_{P}

. ✓ Boxed answer.

P ≅ ρ_{triv} \oplus ρ_{std}

, i.e. one copy of the trivial and one copy of the 2-D standard representation; the trivial summand is the line

(1, 1, 1)

fixed by all permutations.

Common Misconceptions

The inner product needs a complex conjugate. Equation [3] uses $\overline{χ_{β} (g)}$ . For real characters this is harmless, but with complex characters (e.g. $Z /3$ , roots of unity) omitting it gives wrong, often non-integer, multiplicities.
Character $\neq =$ representation map. $χ$ is not a homomorphism: $χ (g h) \neq = χ (g) χ (h)$ in general. It is only a class function.
"Square table" is a theorem, not a definition. That #irreducibles = #conjugacy classes [6] requires proof; it is special to finite groups (and works only over fields like $C$ with enough roots of unity — an algebraically closed field of characteristic not dividing $∣ G ∣$ ).
Equal dimensions don't imply equivalence, but equal characters do. Two representations are isomorphic iff their characters coincide on every class.

Characters over $\mathbb{R}$ fail orthogonality counts

For the cyclic group $Z /3 = {e, g, g^{2}}$ , the three irreducible complex characters are $χ_{k} (g^{j}) = ω^{j k}$ with $ω = e^{2 π i /3}$ , $k = 0, 1, 2$ . They are orthonormal under [3]. But if you insist on real representations, two of these fuse into a single 2-D real irreducible, and you get only two real irreducibles — fewer than the three conjugacy classes. Lesson: [6] and the orthogonality relations as stated require an algebraically closed field; over $R$ the count and the table change.

Connections

Character orthogonality is the finite-group avatar of Fourier analysis: for $G = Z / n$ the characters are $χ_{k} (j) = e^{2 π ij k / n}$ and [3] is exactly the orthogonality of the discrete Fourier basis. See Fourier Series and Transforms and Representation Theory of Finite Groups. The whole machinery rests on Schur's Lemma and the Maschke complete-reducibility theorem, which guarantees decomposition into irreducibles. The decomposition multiplicities link to Group Actions and Permutation Representations (Burnside-type counting), and the algebra structure to the Group Algebra and Wedderburn Decomposition, where $C [G] ≅ ⨁_{α} M_{d_{α}} (C)$ — equation [7] is the dimension count of that decomposition.

Real-World Applications

Molecular vibrational spectroscopy (chemistry). The point group of a molecule (e.g. $C_{2 v}$ for water, $T_{d}$ for methane) has a character table; reducing the $3 N$ -dimensional motion representation against it sorts the normal modes into IR-active and Raman-active symmetry species. This is the standard undergraduate symmetry-adapted-linear-combinations method.
Selection rules in quantum mechanics. A transition matrix element $⟨ ψ_{f} ∣ \hat{O} ∣ ψ_{i} ⟩$ vanishes unless the trivial representation appears in the triple product $χ_{f} \otimes χ_{O} \otimes χ_{i}$ — computed directly from the character table.
Crystallography / solid-state physics. The 32 crystallographic point groups and 230 space groups use character tables to classify band degeneracies, phonon modes, and allowed optical transitions.
Ligand-field & crystal-field theory. Splitting of $d$ -orbitals in an octahedral or tetrahedral field is read off by reducing the $ℓ = 2$ representation against the relevant table ( $O_{h}$ , $T_{d}$ ).
Hückel and group-theoretic MO theory. Symmetry-adapted basis functions block-diagonalise the Hamiltonian, dramatically simplifying eigenvalue problems for symmetric molecules like benzene ( $D_{6 h}$ ).
Error-correcting codes & combinatorics. Character sums over finite abelian groups underpin the analysis of difference sets, MacWilliams identities, and the discrete Fourier transform used in coding.
Particle physics. Branching rules and Clebsch–Gordan decompositions for symmetry groups (the finite-group analogue scales up to compact Lie groups) classify multiplets.

Historical Development

1854 — Cayley abstracts the notion of a group; the objects whose representations will later be tabulated come into focus.
1896 — Dedekind communicates the factorisation of the "group determinant" to Frobenius, posing the puzzle that births character theory.
1896–1897 — Frobenius defines group characters via functional equations and proves the first orthogonality relations, initially without explicit matrix representations.
1897 — Frobenius introduces matrix representations proper and links them to his characters.
1899 — Frobenius proves induced characters and reciprocity (Frobenius reciprocity), extending the orthogonality toolkit.
1905 — Issai Schur publishes Schur's Lemma, giving the clean conceptual proof of the Great Orthogonality Theorem used today.
1911 — Burnside's Theory of Groups of Finite Order (2nd ed.) systematises character tables for the English-speaking world and proves the $p^{a} q^{b}$ solvability theorem using characters.
1929 — Wigner imports group representation theory into quantum mechanics; the character table becomes a working physicist's tool.
1930s–40s — Mulliken standardises the notation ( $A_{1}, B_{2}, E, T_{2}, \dots$ ) and layout of chemical character tables, the form printed in every spectroscopy textbook.

Subtleties & Limits

Character tables are powerful but do not determine the group: there exist non-isomorphic groups with identical character tables (the dihedral group $D_{4}$ and the quaternion group $Q_{8}$ are the classic order-8 example — same table, different groups). So the table is a near-invariant, not a complete one. Open and active directions include: the modular representation theory of Brauer (characteristic $p ∣ ∣ G ∣$ ), where Maschke fails, decompositions need not be unique, and "Brauer characters" replace ordinary ones; the McKay conjecture and its 2024 proof by Cabanes–Späth (building on the classification of finite simple groups), relating counts of irreducible characters of a group and of normaliser subgroups; and the broad project of reading deep structural and counting properties of a group off purely numerical features of its character table. Determining exactly what the table does and does not know about $G$ remains a rich research theme.