How tight is the bound n_i ≤ [G : Z(G)]?

Tight exactly for groups where a single degree dominates. For an extraspecial group of order p^3 the degree p meets the bound p^2 exactly, and for Q_8 the degree 2 meets 4. For groups with trivial centre and many small degrees, such as A_5, it is loose: it permits degrees up to 7 while the largest is 5.

Integrality and the Degrees of Irreducible Characters

Q: Why does the argument need algebraic integers rather than ordinary integers?

Because character values are generally irrational: for A_5 they involve 5. The quantity m_g_i(g)/n_i lands in the ring A of algebraic integers, not in Z. The rationality only reappears at the last step, when the specific combination |G|/n_i or [G:Z(G)]/n_i turns out to be rational and A Q = Z can be used.

Executive Summary

The degrees $n_{1}, \dots, n_{r}$ of the irreducible characters of a finite group are constrained far more tightly than the identity $\sum_{i} n_{i}^{2} = | G |$ suggests. In characteristic zero each $n_{i}$ divides $| G |$ , and by Schur's theorem each divides the index $[G : Z (G)]$ of the centre, with $n_{i}^{2} \leq [G : Z (G)]$ as a bonus.

The mechanism is arithmetic rather than group-theoretic. The integral centre $Z (ℤ G) = ⨁_{g} ℤ C_{g}$ is a commutative ring that is finitely generated as an abelian group, so all of its elements are integral over $ℤ$ ; pushing that integrality through the central characters is the whole argument.

$n_{i} ∣ | G |$ Frobenius

$n_{i} ∣ [G : Z (G)]$ Schur

$n_{i}^{2} \leq [G : Z (G)]$ Bound

$ℤ$

A \cap ℚ

Overview

Everything on this page rests on one hypothesis that the previous two pages did not need: $char k = 0$ . Over such a $k$ the ring $A$ of algebraic integers of $k$ is available, and the crucial arithmetic fact $A \cap ℚ = ℤ$ turns statements about algebraic integers into statements about divisibility.

ω_{i} (C_{g}) = \frac{m_{g} χ_{i} (g)}{n_{i}} \in A, χ_{i} (g) \in A,

(8.18)

The two integrality statements. The first is about the ring structure of the centre; the second is about eigenvalues being roots of unity.

Combining them with the two formulas of the Central Idempotents and Characters page gives $| G | / n_{i} \in A \cap ℚ = ℤ$ . Schur's refinement runs the same argument on a coarser partition of $G$ , one built from conjugacy and multiplication by the centre.

The lemma $(8.20)$ closing the page — a sum of $n$ complex numbers of modulus $1$ has modulus at most $n$ , with equality only when they coincide — is the analytic counterpart, and is the tool used on the Integral Group Rings page.

Learning Objectives

Explain why every element of $Z (ℤ G)$ is integral over $ℤ$ .
Prove that $m_{g} χ_{i} (g) / n_{i}$ and $χ_{i} (g)$ both lie in $A$ .
Deduce Frobenius's theorem that $n_{i}$ divides $| G |$ .
State Schur's theorem and follow the equivalence-class argument that proves it.
Prove the dimension bound $n_{i}^{2} \leq [G : Z (G)]$ from surjectivity of the representation map.
Test all three statements on the alternating group of degree $5$ and on the quaternion group of order $8$ .

Definitions

Standing hypotheses for this page: G is a finite group, k is a splitting field for G of characteristic zero, and A is the ring of algebraic integers of k. Notation for simple modules, degrees, characters, idempotents and class sums is as on the Central Idempotents and Characters page.

$A$: The elements of $k$ satisfying a monic polynomial over $ℤ$ . $A$ is a ring, and $A \cap ℚ = ℤ$ because $ℤ$ is integrally closed.
$Z (ℤ G)$: The centre of the integral group ring: the free abelian group on the class sums, closed under multiplication because $C_{g} C_{h}$ is a non-negative integral combination of class sums.
$ω_{i}$: The central character $Z (k G) \to k$ with $ω_{i} (e_{j}) = δ_{i j}$ and $ω_{i} (C_{g}) = m_{g} χ_{i} (g) / n_{i}$ .
$Z (G)$: The centre of the group. Note the clash of notation with the centre of a ring; $Z (G)$ always means the group centre on this page.
$[G : H]$: The index of a subgroup, $| G | / | H |$ .

Lemma—Integrality criterion

Let $S$ be a commutative ring containing $ℤ$ and let $x \in S$ . If $x$ lies in a subring $T \subseteq S$ that is finitely generated as an abelian group, then $x$ is integral over $ℤ$ . If $ϕ : S \to S'$ is a ring homomorphism, then $ϕ (x)$ is integral over $ℤ$ as well.

The first assertion is the standard determinant-trick characterisation of integrality; the second is immediate from applying $ϕ$ to a monic equation.

Core Concepts

Two independent sources of algebraic integers

The theorem has two halves, and they come from genuinely different places.

Source 1

The integral centre is a finite ring extension

$Z (ℤ G) = ⨁_{g} ℤ C_{g}$ is a commutative ring, finitely generated as an abelian group by the $r$ class sums. So every $C_{g}$ is integral over $ℤ$ , and so is its image under any central character.

Source 2

Character values are sums of roots of unity

If $g$ has order $m$ then its matrix $T$ on any module satisfies $T^{m} = I$ , so $T$ is diagonalisable with eigenvalues $m$ -th roots of unity. A character value is a sum of such, hence an algebraic integer.

Why integrality gives divisibility

An algebraic integer that happens to be rational is an ordinary integer. So any time an argument produces a rational number of the form $| G | / n_{i}$ inside $A$ , divisibility follows for free. The engineering of the proof consists entirely in arranging for that rational number to appear as a coordinate of something manifestly integral.

C_{g}

integral over

ℤ

ω_{i} (C_{g}) \in A

| G | e_{i} / n_{i} \in \sum A C_{g}

| G | / n_{i} \in A \cap ℚ

n_{i} ∣ | G |

What Schur's argument adds

Frobenius's bound uses only the conjugacy partition of $G$ . Schur's uses a coarser partition: declare $a \sim g$ when $a = g_{1} h_{1}$ with $g_{1}$ conjugate to $g$ and $h_{1} \in Z (G)$ . Because a central element acts on an absolutely irreducible module as a scalar, the function $g \mapsto χ (g^{- 1}) χ (g)$ is constant on these larger classes, and the first orthogonality relation can be summed over them instead. The factor $| Z (G) |$ that comes out is exactly the improvement.

Key Results

Theorem(8.18)Integrality in the centre

Let $G$ be a finite group, $k$ a splitting field for $G$ with $char k = 0$ , and $A$ the ring of algebraic integers of $k$ . Then:

$C_{g} \in \sum_{i = 1}^{r} A e_{i}$ for every $g \in G$ ; equivalently $ω_{i} (C_{g}) = m_{g} χ_{i} (g) / n_{i} \in A$ for all $i$ and $g$ .
$| G | e_{i} / n_{i} \in \sum_{g} A C_{g}$ for every $i$ , the sum running over class representatives.

Proof

(1). Work inside $Z (k G)$ and consider the subring $Z (ℤ G) = ⨁_{g} ℤ C_{g}$ . It is a ring because a product of two class sums is again central with integer coefficients, and it is finitely generated as an abelian group by the $r$ class sums. Hence every $C_{g}$ is integral over $ℤ$ . Now $Z (k G) ≅ \prod_{i = 1}^{r} k e_{i}$ as $k$ -algebras, and the $i$ -th coordinate map is the ring homomorphism $ω_{i}$ . Ring homomorphisms carry integral elements to integral elements, so $ω_{i} (C_{g})$ is integral over $ℤ$ and lies in $k$ , hence in $A$ . Its value is $m_{g} χ_{i} (g) / n_{i}$ by $(8.15) (2)$ .

(2). First, every character value is in $A$ : if $g$ has order $m$ and $T$ is the matrix of $g$ acting on a finite-dimensional $k G$ -module $M$ in some $k$ -basis, then $T^{m} = I$ , so the eigenvalues of $T$ in an algebraic closure are $m$ -th roots of unity, and $χ_{M} (g)$ , being their sum, is an algebraic integer lying in $k$ .

Now multiply $(8.15) (1)$ by $| G | / n_{i}$ :

\frac{| G |}{n_{i}} e_{i} = \sum_{g \in G} χ_{i} (g^{- 1}) g = \sum_{t = 1}^{r} χ_{i} (g_{t}^{- 1}) C_{g_{t}},

where the second equality groups the sum by conjugacy classes, legitimate because $χ_{i}$ is a class function and inversion preserves classes. All coefficients $χ_{i} (g_{t}^{- 1})$ lie in $A$ , which is (2).

Corollary(8.18′)Frobenius divisibility

Under the hypotheses of $(8.18)$ , each degree $n_{i}$ divides $| G |$ in $ℤ$ .

Proof

By (2), $| G | e_{i} / n_{i} \in \sum_{g} A C_{g}$ ; by (1), each $C_{g} \in \sum_{j} A e_{j}$ . Composing, $| G | e_{i} / n_{i} \in \sum_{j} A e_{j}$ . The $e_{j}$ are a $k$ -basis of $Z (k G)$ , so comparing $e_{i}$ -coordinates gives $| G | / n_{i} \in A$ . But $| G | / n_{i}$ is rational, and $A \cap ℚ = ℤ$ . Hence $| G | / n_{i} \in ℤ$ .

Theorem(8.19)Schur: degrees divide the index of the centre

Let $G$ be a finite group, $k$ a splitting field for $G$ with $char k = 0$ , $A$ its ring of algebraic integers, and $H = Z (G)$ the centre of $G$ . Then for every irreducible degree $n_{i}$ :

$n_{i}$ divides $[G : H]$ ;
$n_{i}^{2} \leq [G : H]$ , that is $n_{i} \leq \sqrt{[G : H]}$ ;
if $G$ is a $p$ -group, then $n_{i}^{2}$ divides $[G : H]$ .

Proof

Write $χ = χ_{i}$ , $n = n_{i}$ , $M = M_{i}$ . We may assume $G$ acts faithfully on $M$ : replacing $G$ by $\bar{G} = G / K$ with $K$ the kernel of the action changes neither $χ$ nor $n$ , and $[\bar{G} : Z (\bar{G})]$ divides $[G : Z (G)]$ because $G / Z (G)$ surjects onto $\bar{G} / Z (\bar{G})$ .

(1). Since $k$ splits $G$ , ${End}_{k G} (M) = k$ , so each $h \in H$ acts on $M$ as a scalar $μ (h) \in k ∖ {0}$ ; faithfulness makes $μ : H \to k^{\times}$ injective. For $g \in G$ and $h \in H$ the matrix of $g h$ is $μ (h)$ times that of $g$ , so $χ (g h) = μ (h) χ (g)$ .

Define $a \sim g$ to mean $a = g_{1} h_{1}$ with $g_{1}$ conjugate to $g$ and $h_{1} \in H$ . This is an equivalence relation, and $χ (a^{- 1}) χ (a)$ is constant on its classes: with $a = g_{1} h_{1}$ ,

χ (a^{- 1}) χ (a) = μ (h_{1}^{- 1}) χ (g_{1}^{- 1}) \cdot μ (h_{1}) χ (g_{1}) = χ (g_{1}^{- 1}) χ (g_{1}) = χ (g^{- 1}) χ (g) .

If $χ (g) \neq 0$ the class $C' (g)$ of $g$ has exactly $m_{g} | H |$ elements. Indeed it has at most that many, and the factorisation $a = g_{1} h_{1}$ is unique: if also $a = g_{2} h_{2}$ then $χ (g_{1}) μ (h_{1}) = χ (g_{2}) μ (h_{2})$ with $χ (g_{1}) = χ (g_{2}) = χ (g) \neq 0$ , so $μ (h_{1}) = μ (h_{2})$ , so $h_{1} = h_{2}$ by injectivity and $g_{1} = g_{2}$ .

Let $g_{1}, \dots, g_{s}$ represent the $\sim$ -classes on which $χ$ does not vanish. The first orthogonality relation, with the terms where $χ$ vanishes discarded, gives

| G | = \sum_{g \in G} χ (g^{- 1}) χ (g) = \sum_{j = 1}^{s} | C' (g_{j}) | χ (g_{j}^{- 1}) χ (g_{j}) = | H | \sum_{j = 1}^{s} m_{g_{j}} χ (g_{j}^{- 1}) χ (g_{j}) .

Divide by $| H | n$ :

\frac{[G : H]}{n} = \sum_{j = 1}^{s} (\frac{m_{g_{j}} χ (g_{j})}{n}) χ (g_{j}^{- 1}) \in A,

since each bracket lies in $A$ by $(8.18) (1)$ and each $χ (g_{j}^{- 1})$ lies in $A$ by $(8.18) (2)$ . The left-hand side is rational, so it lies in $A \cap ℚ = ℤ$ , proving (1).

(2). Let $D : k G \to {End}_{k} (M)$ be the representation. Because $k$ splits $G$ and $M$ is simple, $D$ is surjective, so ${D (g) : g \in G}$ spans the $n^{2}$ -dimensional space ${End}_{k} (M)$ . For $h \in H$ we have $D (g h) = μ (h) D (g)$ , so the span is unchanged if $g$ ranges only over a set of coset representatives $t_{1}, \dots, t_{[G : H]}$ of $H$ in $G$ . A spanning set has at least as many elements as the dimension, so $n^{2} \leq [G : H]$ .

(3). If $G$ is a $p$ -group then $n$ is a power of $p$ by (1), so $n^{2}$ and $[G : H]$ are both powers of $p$ ; for powers of a prime the inequality $n^{2} \leq [G : H]$ is the same as $n^{2} ∣ [G : H]$ .

Remark—Itô's refinement

Itô improved $(8.19) (1)$ : for any abelian normal subgroup $H ⊴ G$ , every irreducible degree divides $[G : H]$ . Taking $H = Z (G)$ recovers Schur's statement, and taking $H = 1$ recovers Frobenius's. The proof uses Clifford theory rather than the counting argument above.

Lemma(8.20)Unit-modulus sums

Let $w_{1}, \dots, w_{n} \in ℂ$ satisfy $| w_{1} | = \dots = | w_{n} | = 1$ . Then $| w_{1} + \dots + w_{n} | \leq n$ , with equality if and only if $w_{1} = \dots = w_{n}$ .

Proof

Put $s = w_{1} + \dots + w_{n}$ . If $s = 0$ the inequality is clear and equality would force $n = 0$ . Otherwise choose $λ = | s | / s$ , so $| λ | = 1$ and $λ s = | s |$ is real and positive. Then

| s | = Re (λ s) = \sum_{j = 1}^{n} Re (λ w_{j}) \leq \sum_{j = 1}^{n} | λ w_{j} | = n .

Equality forces $Re (λ w_{j}) = | λ w_{j} | = 1$ for every $j$ , hence $λ w_{j} = 1$ and $w_{j} = λ^{- 1}$ for every $j$ . The converse is immediate.

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Find a finitely generated subring

To prove an element integral, exhibit a subring containing it that is finitely generated as an abelian group. For group rings, $Z (ℤ G)$ is handed to you by the class sums.

Move 2

Push integrality along a homomorphism

A ring map sends integral elements to integral elements. Central characters are ring maps; that is why $m_{g} χ_{i} (g) / n_{i}$ is an algebraic integer even though it is a quotient.

Move 3

Land a rational number in $A$

$A \cap ℚ = ℤ$ converts an integrality statement into a divisibility statement. Every degree theorem here ends with this step.

Schur's proof adds a fourth move worth isolating: coarsen the partition. The first orthogonality relation is a sum over $G$ ; if the summand happens to be constant on larger sets than conjugacy classes, summing over those sets extracts a bigger common factor. Recognising that $χ (g^{- 1}) χ (g)$ is $Z (G)$ -invariant is the entire idea.

Worked Example

The alternating group of degree five

$G = A_{5}$ has order $60$ , trivial centre, and five conjugacy classes represented by $1$ , $(1 2) (3 4)$ , $(1 2 3)$ , $(1 2 3 4 5)$ and $(1 3 5 2 4)$ , of sizes $1, 15, 20, 12, 12$ . Over $k = ℚ (\sqrt{5})$ — a splitting field — the character table is:

Character table of $A_{5}$ ; $ϕ = (1 + \sqrt{5}) / 2$ and $\bar{ϕ} = (1 - \sqrt{5}) / 2$ .
	$1$	$(1 2) (3 4)$	$(1 2 3)$	$(1 2 3 4 5)$	$(1 3 5 2 4)$
class size	$1$	$15$	$20$	$12$	$12$
$χ_{1}$	$1$	$1$	$1$	$1$	$1$
$χ_{2}$	$3$	$- 1$	$0$	$ϕ$	$\bar{ϕ}$
$χ_{3}$	$3$	$- 1$	$0$	$\bar{ϕ}$	$ϕ$
$χ_{4}$	$4$	$0$	$1$	$- 1$	$- 1$
$χ_{5}$	$5$	$1$	$- 1$	$0$	$0$

Degrees $1, 3, 3, 4, 5$ : check $1 + 9 + 9 + 16 + 25 = 60 = | G |$ , and every degree divides $60$ , as $(8.18')$ requires. Since $Z (A_{5}) = 1$ , Schur's theorem here says nothing beyond Frobenius's, and the bound $n_{i}^{2} \leq 60$ is satisfied with room to spare — note $5^{2} = 25$ does not divide $60$ , so $(8.19) (3)$ genuinely needs the $p$ -group hypothesis.

Now verify $(8.18) (1)$ , that $m_{g} χ_{i} (g) / n_{i} \in A$ . For $χ_{5}$ the values are $1 \cdot 5 / 5 = 1$ , $15 \cdot 1 / 5 = 3$ , $20 \cdot (- 1) / 5 = - 4$ , $0$ , $0$ — all rational integers. For $χ_{4}$ : $1$ , $0$ , $20 / 4 = 5$ , $- 12 / 4 = - 3$ , $- 3$ . For $χ_{2}$ the interesting entries appear:

\frac{m_{g} χ_{2} (g)}{n_{2}} |_{g = (1 2 3 4 5)} = \frac{12}{3} \cdot \frac{1 + \sqrt{5}}{2} = 2 + 2 \sqrt{5},

(E.1)

A root of $x^{2} - 4 x - 16 = 0$ , hence in $A$ but not in $ℤ$ — the theorem really is about algebraic integers.

Also $ϕ$ and $\bar{ϕ}$ are themselves algebraic integers, both roots of $x^{2} - x - 1 = 0$ , as $(8.18) (2)$ predicts for values of a character.

The quaternion group of order eight

$G = Q_{8}$ has centre $Z (G) = {\pm 1}$ of order $2$ , so $[G : Z (G)] = 4$ . Its five irreducible complex characters have degrees $1, 1, 1, 1, 2$ , and $1 + 1 + 1 + 1 + 4 = 8 = | G |$ .

Testing the three parts of $(8.19)$ on $Q_{8}$
Degree $n$	$n ∣ [G : Z (G)] = 4$	$n^{2} \leq 4$	$n^{2} ∣ 4$
$1$	yes	$1 \leq 4$	yes
$2$	yes	$4 \leq 4$ , tight	yes

Comparison and Classification

The three degree constraints compared
Statement	Hypotheses	Strength	Fails without
$\sum_{i} n_{i}^{2} = \| G \|$	$k$ splits $G$ , $char k ∤ \| G \|$	an identity, not a bound on individual $n_{i}$	splitting field
$n_{i} ∣ \| G \|$	additionally $char k = 0$	excludes most integers below $\| G \|$	characteristic zero
$n_{i} ∣ [G : Z (G)]$	same	strictly stronger whenever $Z (G) \neq 1$	the centre acting by scalars
$n_{i}^{2} \leq [G : Z (G)]$	same	a size bound, no divisibility	surjectivity of $D$ , i.e. splitting
$n_{i}^{2} ∣ [G : Z (G)]$	additionally $G$ a $p$ -group	sharpest of all	the $p$ -group hypothesis

Which constraint bites for which group
	$\sum n_{i}^{2} = \| G \|$	$n_{i} ∣ \| G \|$	$n_{i} ∣ [G : Z (G)]$ improves on it	$n_{i}^{2} ∣ [G : Z (G)]$
$G$ abelian	yes	yes	no	yes
$G = S_{3}$	yes	yes	no	no
$G = Q_{8}$ or $D_{8}$	yes	yes	yes	yes
$G = A_{5}$	yes	yes	no	no
$G$ extraspecial of order $p^{3}$	yes	yes	yes	yes

Which constraint bites for which group

The middle column is empty exactly when $Z (G) = 1$ , since Schur's theorem then reads $n_{i} ∣ | G |$ . The theorem earns its keep on groups with a large centre — the nilpotent groups where all the interesting degree combinatorics lives.

Relationship Map

The dependency structure is linear: nothing here is proved without the two idempotent formulas, and the first orthogonality relation enters only in Schur's proof.

$(8.15)$ idempotent formulas — the source of everything
- $(8.18) (1)$ : $ω_{i} (C_{g}) \in A$
  - $(8.18')$ : $n_{i} ∣ | G |$
  - $(8.19) (1)$ : $n_{i} ∣ [G : Z (G)]$
  - Burnside's $p^{a} q^{b}$ theorem (outside this text)
- $(8.18) (2)$ : $χ_{i} (g) \in A$
  - $(8.18')$ again
  - $(8.21)$ : central torsion units of $A G$
  - $(8.26)$ : $A G$ has only trivial idempotents
$(8.16)$ (A) first orthogonality — used once
- the sum $| G | = \sum_{g} χ (g^{- 1}) χ (g)$ in Schur's proof
$(7.3)$ surjectivity of $D$ — the density theorem in the split case
- $(8.19) (2)$ : $n_{i}^{2} \leq [G : Z (G)]$

n_{i}^{2} ∣ [G : Z (G)]

(

p

-groups)

n_{i} ∣ [G : Z (G)]

n_{i} ∣ | G |

n_{i} \leq \sqrt{| G |}

Applications and Industry Use

Applications here means where this structure is used — inside mathematics and in the engineering and computing disciplines that consume it.

Finite group theory

Burnside's solvability theorem

That a group of order $p^{a} q^{b}$ is solvable is proved from $(8.18) (1)$ : if $\gcd (m_{g}, n_{i}) = 1$ and $χ_{i} (g) \neq 0$ then $χ_{i} (g) / n_{i}$ is an algebraic integer of modulus at most $1$ , forcing $g$ to act by a scalar. The unit-modulus lemma $(8.20)$ is the analytic half of that argument.

Computational group theory

Pruning degree searches

When a character table is being constructed, $n_{i} ∣ | G |$ and $\sum_{i} n_{i}^{2} = | G |$ reduce the search for the degree multiset to a small integer-partition problem — usually to a unique answer for groups of modest order.

Quantum chemistry and physics

Bounded degeneracies

Energy-level degeneracies of a system with symmetry group $G$ are the degrees $n_{i}$ . Schur's bound $n_{i} \leq \sqrt{[G : Z (G)]}$ caps the possible degeneracy from the symmetry group alone, before any Hamiltonian is written down.

Number theory

Artin L-functions

Degrees of Galois representations enter conductor and functional-equation bookkeeping; the divisibility constraints restrict which degrees can occur for a given Galois group.

The honest summary is that these results are structural constraints, not construction recipes. They tell you which degree patterns are impossible, which is exactly what a classification argument needs.

Failure Modes and Common Mistakes

Do not confuse the two centres. $Z (G)$ is a subgroup of $G$ ; $Z (k G)$ is a subring of the group algebra. They are related — $Z (G)$ embeds in the unit group of $Z (k G)$ — but $\dim_{k} Z (k G) = r$ , not $| Z (G) |$ .
$m_{g} χ_{i} (g) / n_{i}$ is an algebraic integer, not usually a rational one. Concluding $n_{i} ∣ m_{g} χ_{i} (g)$ in $ℤ$ is wrong as soon as the character values are irrational.
Schur's theorem needs $k$ to split $G$ : the argument uses that central elements act by scalars, which is exactly ${End}_{k G} (M) = k$ .
The bound $n_{i} \leq \sqrt{[G : Z (G)]}$ is often far from tight. For $A_{5}$ it gives $n_{i} \leq 7$ while the true maximum is $5$ ; it is a structural constraint, not an estimate.
Do not read $(8.18) (2)$ as saying $e_{i}$ has algebraic-integer coefficients. It says $| G | e_{i} / n_{i}$ does; $e_{i}$ itself has denominators.

Historical Notes and Lessons Learned

1896Frobenius introduces charactersGroup characters are defined via the group determinant, and the divisibility of the degrees by the group order is established almost immediately.
1897–1905Burnside and Schur reformulateRepresentations by matrices replace the group determinant. Schur proves that each degree divides the index of the centre; Burnside proves the solvability of groups of order p^a q^b using the same integrality technique.
1929Noether's module-theoretic recastingRepresentations become modules over the group algebra, and the integrality arguments are recognised as statements about the centre of an order in a semisimple algebra.
1951Itô's refinementThe index of the centre is replaced by the index of any abelian normal subgroup, using Clifford theory rather than the counting argument.
1963–1970sDegrees as a classification toolCharacter degree patterns become a standard invariant in the classification of finite simple groups, and results such as the Itô–Michler theorem tie degree divisibility to the structure of Sylow subgroups.

The methodological lesson is that arithmetic constraints came before structural ones. Long before anyone could classify groups, the ring of algebraic integers was already telling group theorists which representation degrees were possible — and that information was strong enough to prove theorems, such as Burnside's, whose statements mention no representations at all.

Quick Reference

Hypotheses

G

finite,

char k = 0

k

splits

G

A

its algebraic integers

Key ring

Z (ℤ G) = ⨁_{g} ℤ C_{g}

, finitely generated over

ℤ

$(8.18) (1)$

m_{g} χ_{i} (g) / n_{i} \in A

$(8.18) (2)$

χ_{i} (g) \in A

; equivalently

| G | e_{i} / n_{i} \in \sum_{g} A C_{g}

Frobenius

n_{i} ∣ | G |

Schur

n_{i} ∣ [G : Z (G)]

and

n_{i}^{2} \leq [G : Z (G)]

$p$ -groups

n_{i}^{2} ∣ [G : Z (G)]

Itô

n_{i} ∣ [G : H]

for any abelian

H ⊴ G

$(8.20)$

| w_{1} + \dots + w_{n} | \leq n

for unit-modulus

w_{j}

, equality iff all equal

Degree data for small groups
$G$	$\| G \|$	$\| Z (G) \|$	Degrees
$C_{n}$	$n$	$n$	$1$ repeated $n$ times
$S_{3}$	$6$	$1$	$1, 1, 2$
$Q_{8}$ , $D_{8}$	$8$	$2$	$1, 1, 1, 1, 2$
$A_{4}$	$12$	$1$	$1, 1, 1, 3$
$S_{4}$	$24$	$1$	$1, 1, 2, 3, 3$
$A_{5}$	$60$	$1$	$1, 3, 3, 4, 5$

Frequently Asked Questions

Why does the argument need algebraic integers rather than ordinary integers?

Because character values are generally irrational: for $A_{5}$ they involve $\sqrt{5}$ . The quantity $m_{g} χ_{i} (g) / n_{i}$ lands in the ring $A$ of algebraic integers, not in $ℤ$ . The rationality only reappears at the last step, when the specific combination $| G | / n_{i}$ or $[G : Z (G)] / n_{i}$ turns out to be rational and $A \cap ℚ = ℤ$ can be used.

Is the divisibility

n_{i} ∣ | G |

true in characteristic p as well?

The theorem as proved here is a characteristic-zero statement, and the proof does not transfer: it uses the ring of algebraic integers of $k$ . When $char k = p ∤ | G |$ the simple $k G$ -modules have the same degrees as in characteristic zero, so the numbers still divide $| G |$ ; when $p$ divides $| G |$ the simple modules are different objects and the question becomes part of modular representation theory.

Why is the index of the centre the right refinement?

Because the centre is exactly the set of elements that act by scalars on every absolutely irreducible module. Scalars carry no information about the module beyond a phase, so the character data really lives on $G / Z (G)$ , and one should expect the constraints to be indexed by $[G : Z (G)]$ rather than $| G |$ . Itô's theorem shows the same is true for any abelian normal subgroup.

How tight is the bound

n_{i} \leq \sqrt{[G : Z (G)]}

Tight exactly for groups where a single degree dominates. For an extraspecial group of order $p^{3}$ the degree $p$ meets the bound $\sqrt{p^{2}}$ exactly, and for $Q_{8}$ the degree $2$ meets $\sqrt{4}$ . For groups with trivial centre and many small degrees, such as $A_{5}$ , it is loose: it permits degrees up to $7$ while the largest is $5$ .

Where does the unit-modulus lemma get used?

It is not used to prove the degree theorems; Lam records it at this point because it is needed immediately afterwards, in the study of central torsion units and idempotents of integral group rings. The same lemma is the analytic heart of Burnside's $p^{a} q^{b}$ theorem, where it shows that an algebraic integer of the form $χ (g) / n$ with modulus $1$ forces $g$ to act by a scalar.

Does knowing the degrees determine much about the group?

Something, but not everything. The multiset of degrees determines $| G |$ via $\sum n_{i}^{2}$ and detects abelianness (all degrees $1$ ) and nilpotency-related properties, and results such as the Itô–Michler theorem read Sylow structure off degree divisibility. But $D_{8}$ and $Q_{8}$ share both their degrees and their whole character table without being isomorphic.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §8, (8.18)–(8.20).
I. M. Isaacs, Character Theory of Finite Groups, Academic Press, 1976, Chapters 3 and 6.
C. W. Curtis and I. Reiner, Representation Theory of Finite Groups and Associative Algebras, Interscience, 1962, §§32–33.
I. Schur, “Über die Darstellung der endlichen Gruppen durch gebrochene lineare Substitutionen”, Journal für die reine und angewandte Mathematik 127 (1904), 20–50.
N. Itô, “On the degrees of irreducible representations of a finite group”, Nagoya Mathematical Journal 3 (1951).
B. Huppert, Endliche Gruppen I, Grundlehren der mathematischen Wissenschaften 134, Springer-Verlag, 1967, Kapitel V.

AI Suggested Questions

Work through Burnside's p^a q^b theorem and identify exactly where the integrality of the central character values is used.
State and prove Ito's theorem that degrees divide the index of an abelian normal subgroup.
What are the possible character degree multisets for a group of order 32, and which are realised?
How do the degrees of the simple modules in characteristic p relate to the ordinary degrees via the decomposition matrix?
Show that a group all of whose irreducible character degrees are 1 or p has a normal abelian subgroup of small index.
Prove that the algebraic integers in the rationals are exactly the ordinary integers, and where that fails for other number fields.
How does the bound on degrees constrain the maximal degeneracy of energy levels in a molecule with a given point group?

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Two independent sources of algebraic integers

The integral centre is a finite ring extension

Character values are sums of roots of unity

Why integrality gives divisibility

What Schur's argument adds

06Key Results

07Proof Techniques and Method

Find a finitely generated subring

Push integrality along a homomorphism

Land a rational number in A

08Worked Example

The alternating group of degree five

The quaternion group of order eight

09Comparison and Classification

10Relationship Map

11Applications and Industry Use

Burnside's solvability theorem

Pruning degree searches

Bounded degeneracies

Artin L-functions

12Failure Modes and Common Mistakes

13Historical Notes and Lessons Learned

14Quick Reference

15Frequently Asked Questions

16Related KEVOS Topics

17References

18AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

Land a rational number in $A$

Worked Example

Comparison and Classification

Relationship Map

Applications and Industry Use

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions