Splitting Fields for Finite Groups

Q: If k splits G, does k split every subgroup of G?

No. The property passes to quotient groups, because k[G/N] is a quotient algebra of kG and its simple modules are among those of kG. Subgroups give subalgebras, not quotients, and simple kH-modules need not appear among the simple kG-modules.

Q: How does one recognise a non-splitting field in practice?

Compute _i (_k M_i)^2 for the irreducibles you have. If it exceeds |G| - _k rad kG, some D_i is larger than k and k does not split G. If it falls short, either the list is incomplete or, again, a D_i is larger than k. Only exact equality certifies splitting.

Q: What is the connection with the Galois-theoretic notion of a splitting field?

They agree for cyclic groups, where kG ≅ k[x]/(x^n-1) and splitting the algebra is factoring the polynomial into linear factors. In general the algebraic notion is about a ring becoming a product of matrix algebras, which for noncommutative D_i has no polynomial analogue.

Q: Why does the proof of (8.3) go through the prime field?

Because the prime field is perfect, and (7.10) — the existence of a finite splitting field — requires perfectness to guarantee that the algebra remains semisimple after extending to the algebraic closure. An arbitrary k of characteristic p need not be perfect, but its prime field always is, and the compositum step transfers the result back.

Executive Summary

The structure theorem for $k G / rad k G$ involves division algebras $D_{i}$ that are usually not $k$ . A splitting field is a ground field for which they all are. Over such a field every numerical formula simplifies, matrix models of the irreducible representations exist over $k$ , and scalar extension no longer breaks anything.

Two facts make the notion practical. First, a splitting field always exists finitely far away: for any field $k$ and any finite group $G$ , some finite extension of $k$ splits $G$ . Second, Brauer identified one explicitly — adjoin a primitive $m$ th root of unity, where $m$ is the exponent of $G$ . Neither statement is easy, and the second is quoted here rather than proved.

$D_{i} = k$ Defining condition

$\prod_{i} M_{n_{i}} (k)$ Semisimple quotient

$[K : k] < \infty$ Always achievable

$ℚ (ζ_{m})$ Brauer,

m = \exp G

Overview

Fix a finite group $G$ and a field $k$ . By The Structure of kG modulo Its Radical, $k G / rad k G ≅ \prod_{i = 1}^{r} M_{n_{i}} (D_{i})$ with $D_{i} = End (_{k G} M_{i})$ a division algebra over $k$ . Nothing forces $D_{i} = k$ : the cyclic group of order $3$ over $ℚ$ already produces $D_{2} = ℚ (ω)$ , and the quaternion group of order $8$ produces a noncommutative $D_{i}$ .

When some $D_{i}$ is strictly larger than $k$ , the module $M_{i}$ is irreducible but not absolutely irreducible: it decomposes after a suitable extension of scalars. Splitting fields are exactly the ground fields where this does not happen, so that irreducibility is stable under every further extension.

k splits G iff k G / rad (k G) ≅ M_{n_{1}} (k) \times \dots \times M_{n_{r}} (k) .

(8.2a)

The matrix criterion, specialising $(7.7)$ to $R = k G$ . All division algebras have collapsed to $k$ .

Because splitting is detected by $k G / rad k G$ being a product of matrix algebras — a condition invariant under passing to the opposite ring — the notion is left-right symmetric, and there is no need to distinguish left splitting fields from right ones.

Learning Objectives

State Definition $(8.2)$ and unwind it to a statement about $End (_{k G} M_{i})$ .
Give the matrix criterion and the dimension criterion for $k$ to split $G$ .
Prove Theorem $(8.3)$ : a finite extension of any field splits any finite group.
Quote Brauer's theorem correctly, including the characteristic $p$ form.
Show that $ℚ$ and $ℝ$ fail to split the quaternion group of order $8$ and that $ℚ (\sqrt{- 1})$ succeeds.
Explain why splitting fields are stable under further extension but not under restriction.

Definitions

Definition(8.2)Splitting field for a group

Let $G$ be a finite group. A field $k$ is a **splitting field for $G$ ** if the group algebra $k G$ splits over $k$ in the sense of $(7.6)$ — that is, if every simple left $k G$ -module is absolutely irreducible. For an extension $K \supseteq k$ we say $K$ is a splitting field for $G$ when $K G$ splits over $K$ .

Equivalently: $End (_{k G} M) = k$ for every simple left $k G$ -module $M$ .

$R^{K}$: The scalar extension $R \otimes_{k} K$ of a $k$ -algebra $R$ ; for $R = k G$ this is $K G$ .
$M^{K}$: The extended module $M \otimes_{k} K$ , a left $R^{K}$ -module. $M$ is absolutely irreducible iff $M^{K}$ is simple for every extension $K$ .
Exponent $\exp G$: The least $m \geq 1$ with $g^{m} = 1$ for all $g \in G$ ; a divisor of $| G |$ , and the modulus in Brauer's theorem.
Schur index $m_{i}$: For a simple module $M_{i}$ over a semisimple $k$ -algebra with $k$ the centre of $D_{i}$ , the integer with $\dim_{k} D_{i} = m_{i}^{2}$ . It is $1$ exactly when $M_{i}$ is absolutely irreducible.
Prime field: $ℚ$ or $𝔽_{p}$ ; the smallest subfield of $k$ . Prime fields are perfect, which is the hypothesis that starts the proof of $(8.3)$ .

Throughout, G is finite and all representations are finite-dimensional. char k is arbitrary unless stated.

Core Concepts

Three ways to test for splitting

The definition is about endomorphism rings, which are awkward to compute. Two reformulations are what get used.

\dim_{k} k G = \dim_{k} (rad k G) + \sum_{i = 1}^{r} (\dim_{k} M_{i})^{2}

(7.8)

The dimension criterion: $k$ splits $G$ if and only if this holds, with ${M_{i}}$ a full set of simple left $k G$ -modules. Note $| G |$ on the left.

The dimension criterion is the practical one, because the left-hand side is $| G |$ and the right-hand side is computable from any list of irreducibles you can produce. It is also self-certifying: if equality holds, the list is complete and the field splits.

all

D_{i} = k

k G / rad k G ≅ \prod_{i} M_{n_{i}} (k)

| G | = \dim_{k} rad k G + \sum_{i} (\dim_{k} M_{i})^{2}

Splitting is stable upwards, not downwards

If $K$ splits $G$ and $L \supseteq K$ , then $L$ splits $G$ , and the simple $L G$ -modules are exactly the scalar extensions $V^{L}$ of the simple $K G$ -modules $V$ ; this is $(7.14)$ . The converse fails: $ℂ$ splits every finite group, but $ℚ$ splits very few. Descending from a splitting field to a subfield is the difficult direction, and is governed by the Schur index.

Why the algebraic closure is not the end of the story

Every finite group is split by $\bar{k}$ , since a finite-dimensional division algebra over an algebraically closed field is the field itself. That observation is cheap and useless for computation: $\bar{ℚ}$ is infinite-dimensional over $ℚ$ . The content of $(8.3)$ and of Brauer's theorem is that one can stay finite, and in the classical case explicitly cyclotomic.

Key Results

Proposition(7.7)–(7.9)Criteria, specialised to group algebras

Let $G$ be a finite group and $K \supseteq k$ fields. The following are equivalent:

$K$ is a splitting field for $G$ ;
$K G / rad (K G)$ is a finite direct product of matrix algebras over $K$ ;
$\dim_{K} K G = \dim_{K} rad (K G) + \sum_{i} (\dim_{K} V_{i})^{2}$ , where ${V_{i}}$ is a full set of simple left $K G$ -modules.

Moreover $K$ is a splitting field for $G$ if and only if it is a splitting field for the semisimple algebra $K G / rad (K G)$ . In particular the notion is left-right symmetric.

Theorem(8.3)Existence of a finite splitting field

Let $k$ be any field and $G$ any finite group. Then there is a finite extension $K \supseteq k$ which is a splitting field for $G$ .

Proof

Let $k_{0} \subseteq k$ be the prime field of $k$ — either $ℚ$ or $𝔽_{p}$ — and fix an algebraic closure $\bar{k}$ of $k$ . Prime fields are perfect.

Apply $(7.10)$ to the finite-dimensional $k_{0}$ -algebra $k_{0} G$ : an algebra over a perfect field splits over some finite extension of that field. This yields a finite extension $k_{1} \supseteq k_{0}$ , which we may realise inside $\bar{k}$ , that is a splitting field for $k_{0} G$ , hence for $G$ .

Now set $K = k \cdot k_{1}$ , the compositum formed inside $\bar{k}$ . Since $k_{1}$ is a finite extension of $k_{0}$ , it is generated over $k_{0}$ by finitely many algebraic elements, so $K = k (k_{1})$ is a finite extension of $k$ .

Finally $K \supseteq k_{1}$ and $k_{1}$ splits $G$ , so by $(7.14)$ — a splitting field remains a splitting field after any further extension — $K$ is a splitting field for $G$ as well.

Theorem(8.3B)Brauer's splitting field (quoted without proof)

Let $G$ be a finite group of exponent $m$ , and let $ζ_{m}$ be a primitive $m$ th root of unity.

If the prime field is $ℚ$ , then $ℚ (ζ_{m})$ is a splitting field for $G$ .
If the prime field is $𝔽_{p}$ , one may take $ℤ [ζ_{m}] / 𝔭$ for any prime ideal $𝔭$ of $ℤ [ζ_{m}]$ containing $p$ .

This is a deep result — it rests on Brauer's induction theorem — and it is stated here for reference only. Proofs are in Curtis and Reiner.

Corollary(8.3C)Abelian groups over a splitting field

Let $G$ be a finite abelian group and $k$ a splitting field for $G$ . Then every irreducible $k$ -representation of $G$ is $1$ -dimensional, and the irreducible representations correspond bijectively to the group homomorphisms $G \to k^{\times}$ .

Proof

By the abelian case of $(8.1)$ all $n_{i} = 1$ and every $D_{i}$ is a field extension of $k$ ; since $k$ splits $G$ , each $D_{i} = k$ , so $\dim_{k} M_{i} = n_{i} \dim_{k} D_{i} = 1$ . A $1$ -dimensional representation is a homomorphism $G \to G L_{1} (k) = k^{\times}$ , and distinct such homomorphisms give non-isomorphic modules.

Remark—Splitting the group versus splitting a quotient

Since $k [G / N]$ is a quotient algebra of $k G$ for $N ⊴ G$ , the simple $k [G / N]$ -modules form a subset of the simple $k G$ -modules. Hence a splitting field for $G$ splits every quotient group of $G$ . The converse is false, and splitting a subgroup is a genuinely different question.

Proof Techniques and Method

How these arguments work, and which move is worth reusing.

Move 1

Descend to the prime field

$k_{0} G$ is defined over $ℚ$ or $𝔽_{p}$ whatever $k$ is, and prime fields are perfect. Proving something over $k_{0}$ and pushing it up by compositum is the whole of $(8.3)$ .

Move 2

Matrix units are finitely many

If $R^{\bar{k}}$ is a product of matrix algebras, the finitely many matrix units defining the decomposition have entries in a finite extension. Finiteness of a basis is what converts an algebraic-closure statement into a finite-extension statement.

Move 3

Close the dimension count

Produce irreducibles by hand until $\sum (\dim_{k} M_{i})^{2}$ plus the radical dimension reaches $| G |$ . Equality certifies both completeness of the list and splitting; a shortfall means either a missing module or a division algebra.

Move 2 deserves emphasis. It is the standard trick for turning an existence statement over $\bar{k}$ into one over a finite extension, and it recurs whenever a structure is defined by finitely many equations: idempotents, matrix units, or a basis of a subalgebra.

Worked Example

The quaternion group of order 8

Let $G = ⟨ a, b ∣ a^{4} = 1, b^{2} = a^{2}, b a b^{- 1} = a^{- 1} ⟩$ , the quaternion group of order $8$ . Its commutator subgroup is $⟨ a^{2} ⟩$ of order $2$ , so $G / G'$ is the Klein four-group.

Over $k = ℚ$

The four homomorphisms $G / G' \to {\pm 1}$ give four $1$ -dimensional $ℚ G$ -modules $M_{1}, \dots, M_{4}$ , pairwise non-isomorphic. For a fifth, let $D$ be the division algebra of rational quaternions. Identifying $a \mapsto i$ and $b \mapsto j$ embeds $G$ into $D^{\times}$ as ${\pm 1, \pm i, \pm j, \pm k}$ , so $D$ becomes a left $ℚ G$ -module $M_{5}$ of $ℚ$ -dimension $4$ . It is simple: a $ℚ G$ -submodule would be a left ideal of $D$ , because $G$ spans $D$ over $ℚ$ , and $D$ is a division ring.

Now $rad ℚ G = 0$ by Maschke, and $End (_{ℚ G} M_{5}) ≅ D$ , so $n_{5} = \dim_{D} M_{5} = 1$ and $\dim_{ℚ} D = 4$ . The count of $(8.1) (4)$ reads

8 = 0 + 1^{2} \cdot 1 + 1^{2} \cdot 1 + 1^{2} \cdot 1 + 1^{2} \cdot 1 + 1^{2} \cdot 4, ℚ G ≅ ℚ^{4} \times D .

(E.1)

The count closes, so $r = 5$ and the list is complete.

Test the dimension criterion $(7.8)$ : $\sum_{i} (\dim_{ℚ} M_{i})^{2} = 1 + 1 + 1 + 1 + 16 = 20 \neq 8$ . So $ℚ$ is not a splitting field for $G$ — as it must not be, since $D_{5} = D \neq ℚ$ . Replacing $ℚ$ by $ℝ$ changes nothing except that $D$ becomes Hamilton's real quaternions, still a division ring; $ℝ$ does not split $G$ either.

Over $K = ℚ (\sqrt{- 1})$

The four $1$ -dimensional modules stay simple. The fifth does not: the rational quaternion algebra splits over any field containing $\sqrt{- 1}$ , so $D \otimes_{ℚ} K ≅ M_{2} (K)$ , and

K G ≅ K \times K \times K \times K \times M_{2} (K), 8 = 1 + 1 + 1 + 1 + 2^{2} .

(E.2)

Now every factor is a matrix algebra over $K$ itself, so $K$ is a splitting field.

As a left module, $K G ≅ M_{1}' \oplus M_{2}' \oplus M_{3}' \oplus M_{4}' \oplus 2 M_{5}'$ where $M_{i}' = M_{i} \otimes_{ℚ} K$ for $i \leq 4$ and $M_{5}'$ is the unique simple $M_{2} (K)$ -module, of dimension $2$ , occurring with multiplicity $n_{5} = 2$ . Explicitly, up to equivalence,

a ⟼ (\begin{matrix} \sqrt{- 1} & 0 \\ 0 & - \sqrt{- 1} \end{matrix}), b ⟼ (\begin{matrix} 0 & 1 \\ - 1 & 0 \end{matrix}) .

(E.3)

Write $A$ and $B$ for these matrices. Then $A^{2} = B^{2} = - I$ , so $a^{4} = 1$ and $b^{2} = a^{2}$ ; and $B A B^{- 1} = A^{- 1}$ , as required.

Frameworks and Models

The fields relevant to a given $G$ sit in a small hierarchy, and it is worth keeping the layers distinct.

$\bar{k}$ — algebraic closurealways splits

G

; infinite over

k

in characteristic

0

Any splitting field $L \supseteq K$ splitting is preserved upward

(7.14)

$K$ — a finite splitting fieldexists by

(8.3)

; Brauer names one

$k$ — the field you started with

End (_{k G} M_{i}) = D_{i}

may exceed

k

$k_{0}$ — the prime field

ℚ

𝔽_{p}

; perfect, which starts the argument

Type A

Already split

$k$ algebraically closed, or $k$ finite and large enough, or accidents such as $ℚ$ for $S_{n}$ . Nothing to do.

Type B

Split by a small explicit extension

Adjoin roots of unity of order $\exp G$ , or just the eigenvalues actually occurring. The quaternion group over $ℚ (\sqrt{- 1})$ is the model case.

Type C

Obstructed by a division algebra

Some $D_{i}$ is noncommutative with nontrivial Schur index. The extension needed is a splitting field of that algebra in the Brauer-group sense, of degree divisible by the index.

Process and Workflow

Compute $\exp G$ Brauer's field

ℚ (ζ_{m})

with

m = \exp G

is a safe first candidate in characteristic

0

List the irreducibles over $k$ Trivial, sign, permutation modules, and any embedding of

G

into a division algebra.

Apply the dimension criterionIf

| G | = \dim_{k} rad k G + \sum_{i} (\dim_{k} M_{i})^{2}

then

k

already splits

G

and you are finished.

Identify the obstructionA shortfall means some

D_{i} \neq k

. Determine that

D_{i}

; its centre and Schur index tell you what to adjoin.

Extend and recountPass to

K

, recompute the simple modules of

K G

, and verify the criterion again over

K

How large an extension do you actually need?

Only roots of unity are missingAdjoin the eigenvalues occurring in the representations. For abelian

G

this suffices and gives

k (ζ_{\exp G})

A commutative $D_{i}$ obstructsAdjoin

D_{i}

itself; the extension needed has degree

\dim_{k} D_{i}

and is generated by one element.

A noncommutative $D_{i}$ obstructsYou need a splitting field for a division algebra of index

m_{i}

; its degree over the centre is a multiple of

m_{i}

. For the rational quaternions, any quadratic imaginary field works.

You only need existenceQuote

(8.3)

: some finite extension works, no construction required.

Comparison and Classification

Does the field split the group?
	$ℚ$	$ℝ$	$ℚ (\sqrt{- 1})$	$ℂ$	$𝔽_{2}$	$𝔽_{3}$
$C_{2}$	yes	yes	yes	yes	yes	yes
$C_{3}$	no	no	no	yes	no	yes
$C_{4}$	no	no	yes	yes	yes	no
$S_{3}$	yes	yes	yes	yes	yes	yes
$Q_{8}$	no	no	yes	yes	yes	yes

Does the field split the group?

Two entries deserve comment. $𝔽_{2}$ splits $C_{3}$ ? No — $𝔽_{2} C_{3} ≅ 𝔽_{2} \times 𝔽_{4}$ , and $𝔽_{4} \neq 𝔽_{2}$ , so the second simple module has $D_{2} = 𝔽_{4}$ . But $𝔽_{3}$ does split $C_{3}$ , trivially: $𝔽_{3} C_{3}$ is local with unique simple module $𝔽_{3}$ . Splitting in characteristic $p$ is often easier when $p$ divides $| G |$ , because the modular irreducibles are fewer and smaller.

How splitting affects each formula
Quantity	General field	Splitting field
$D_{i}$	a division algebra over $k$	$k$
$n_{i}$	$\dim_{D_{i}} M_{i}$	$\dim_{k} M_{i}$
$k G / rad k G$	$\prod_{i} M_{n_{i}} (D_{i})$	$\prod_{i} M_{n_{i}} (k)$
Dimension count	$\| G \| = \dim rad + \sum n_{i}^{2} \dim_{k} D_{i}$	$\| G \| = \dim rad + \sum (\dim_{k} M_{i})^{2}$
Number of irreducibles ( $p = 0$ )	$\leq$ number of conjugacy classes	$=$ number of conjugacy classes
Number of irreducibles ( $p > 0$ )	$\leq$ number of $p$ -regular classes	$=$ number of $p$ -regular classes
Behaviour under $K \supseteq k$	simple modules may split	simple modules stay simple

Relationship Map

k_{0}

perfect

(7.10)

k_{1}

splits

k_{0} G

compositum

K = k \cdot k_{1}

(7.14)

K

splits

G

The logical dependencies of $(8.3)$ are entirely in §7: perfectness gives separability, separability keeps $R^{\bar{k}}$ semisimple, semisimplicity gives matrix units, and finitely many matrix units live in a finite extension. Nothing group-theoretic is used, which is why the theorem holds for any finite-dimensional algebra and $(8.3)$ is a corollary.

$k$ splits $G$
- implies
  - every $D_{i} = k$ and $n_{i} = \dim_{k} M_{i}$
  - $k$ splits every quotient group of $G$
  - every extension of $k$ splits $G$ $(7.14)$
  - the count of irreducibles is maximal among all fields of the same characteristic
- does not imply
  - $k G$ is semisimple — the radical is untouched
  - $k$ splits every subgroup of $G$
  - $k$ is minimal with the property

Applications and Industry Use

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

Computational group theory

Choosing the coefficient field

Before computing irreducible modules, a system must fix a field. GAP and Magma work over $ℚ (ζ_{\exp G})$ or over a finite field chosen large enough to be splitting, precisely on the strength of Brauer's theorem.

Number theory

Schur indices and Brauer groups

The failure of $k$ to split $G$ is recorded by classes in the Brauer group of $k$ . Computing Schur indices of characters is a classical problem connecting representation theory to local class field theory.

Coding theory

Codes over finite fields

Cyclic and abelian group codes over $𝔽_{q}$ decompose completely only when $𝔽_{q}$ contains the relevant roots of unity — that is, when it splits the group. Otherwise the components are extension fields and the code decomposes into fewer, larger pieces.

Physics

Real versus complex representations

The Frobenius–Schur classification into real, complex and quaternionic types is the statement that $ℝ$ fails to split a group in exactly two distinguishable ways. Time-reversal symmetry in quantum mechanics turns on which type occurs.

Stated honestly: splitting fields are a hygiene condition. They are assumed at the start of most treatments so that the formulas are clean, and the work of $(8.3)$ is to show the assumption costs almost nothing.

Design Considerations

Design considerations here means the choices made when modelling a problem with these algebraic structures.

Split first or stay put? Extending to a splitting field makes every formula simpler but changes the object of study. If the question is *is this representation realisable over $ℚ$ ?*, extending destroys the question.
Which splitting field? Brauer's $ℚ (ζ_{m})$ is canonical but often far from minimal — $ℚ$ already splits $S_{n}$ for every $n$ , while $\exp S_{n}$ grows. Minimal splitting fields are not unique and finding one is a genuine computation.
**Characteristic zero or $p$ ?** If the eventual target is modular, choosing a finite splitting field of characteristic $p$ from the start avoids a reduction step. If ordinary character theory is wanted, a cyclotomic field is the right home.
Absolute irreducibility as a design invariant. Recording $End (_{k G} M) = k$ alongside each module is cheap and prevents the common error of extending scalars and finding that a module has silently decomposed.

Standards and Notation

Standards here covers notation, symbol and markup standards, and reference implementations, rather than material or design codes.

Splitting field

K

splits

G

; also *

G

splits over

K

* and *

K

is a splitting field for

k G

Scalar extension

R^{K} = R \otimes_{k} K

(Lam);

R_{K}

K \otimes_{k} R

elsewhere

Roots of unity

ζ_{m}

ζ_{m}

primitive

m

th root;

μ_{m}

for the group of them

Exponent

\exp G

; occasionally

e (G)

Schur index

m_{K} (χ)

for the index of a character

χ

over

K

GAPIrreducibleModules(G, GF(q), 0), MTX.IsAbsolutelyIrreducible

MagmaAbsolutelyIrreducibleModules, SchurIndex

MarkupPresentation MathML per ISO/IEC 40314; symbols per ISO 80000-2

Computational Notes

Computational notes cover algorithms, cost and library behaviour rather than manufacturing process.

Deciding absolute irreducibility is the computational core, and it is cheap relative to everything around it.

For a module of dimension $d$ over $𝔽_{q}$ given by matrices, the MeatAxe's Norton test decides absolute irreducibility in $O (d^{3})$ field operations once a splitting element has been found. This is why finite fields are the preferred working environment.
Over $ℚ$ , deciding whether $ℚ$ splits $G$ amounts to computing the Schur indices of the irreducible characters — a global computation involving local invariants at every ramified place, far more expensive than the finite-field case.
Brauer's theorem gives an a priori field, but $ℚ (ζ_{\exp G})$ can have very large degree. Practical systems compute the character field $ℚ (χ)$ of each character first and adjoin only what is needed.
For a finite field $𝔽_{p}$ with $p ∣ | G |$ , the modular irreducibles are fewer and often smaller than the ordinary ones, so modular computations over a splitting field of characteristic $p$ can be dramatically cheaper — this is the basis of condensation methods.

Failure Modes and Common Mistakes

Do not assume a splitting field for $G$ splits every subgroup of $G$ ; the property descends to quotients, not to subgroups.
Do not assume splitting fields are unique or minimal. Minimal splitting fields need not be unique even up to isomorphism.
Do not read $(8.3)$ as constructive. Its proof locates $k_{1}$ inside $\bar{k}$ without bounding its degree; Brauer's theorem is what supplies an explicit answer.

Historical Notes and Lessons Learned

1896–1900Frobenius over the complex numbersThe original theory is built over $ℂ$ , where the splitting question does not arise; the difficulty is invisible until one asks for representations over $ℚ$ .
1906Schur's indexSchur studies when a complex representation can be realised over a smaller field and introduces the invariant now called the Schur index — the first systematic measure of the failure of splitting.
1929–32Brauer–Noether theoryNoether's module-theoretic reformulation and the developing theory of central simple algebras identify the obstruction as a Brauer-group class, linking splitting fields for groups to splitting fields for algebras.
1945Brauer's cyclotomic theoremBrauer proves that the field of $m$ th roots of unity, $m$ the exponent of $G$ , splits $G$ — a consequence of his induction theorem on characters and the definitive answer in characteristic zero.
1950s onwardModular splitting fieldsThe reduction of Brauer's result to characteristic $p$ via reduction modulo a prime of $ℤ [ζ_{m}]$ becomes standard, and finite splitting fields become the default working environment for modular computation.

The lesson is that the right invariant was not the representation but the algebra of endomorphisms attached to it. Once $End (_{k G} M)$ was recognised as a division algebra with a Brauer class, the question *which fields split $G$ ?* became a question about central simple algebras with an established theory, rather than a collection of accidents.

Quick Reference

Definition

k

splits

G

iff

End (_{k G} M) = k

for every simple

M

Matrix criterion

k G / rad k G ≅ \prod_{i} M_{n_{i}} (k)

Dimension criterion

| G | = \dim_{k} rad k G + \sum_{i} (\dim_{k} M_{i})^{2}

ExistenceSome finite

K \supseteq k

splits

G

—

(8.3)

Brauer

ℚ (ζ_{m})

splits

G

m = \exp G

Upward stability

K

splits

G

and

L \supseteq K

\Rightarrow

L

splits

G

Quotients

k

splits

G

\Rightarrow

k

splits

G / N

for all

N ⊴ G

Abelian caseover a splitting field all irreducibles are

1

-dimensional

Reference computations
$G$	$k$	$k G / rad k G$	Splits?
$Q_{8}$	$ℚ$	$ℚ^{4} \times D$ , $D$ rational quaternions	no
$Q_{8}$	$ℚ (\sqrt{- 1})$	$K^{4} \times M_{2} (K)$	yes
$C_{3}$	$ℚ$	$ℚ \times ℚ (ω)$	no
$C_{3}$	$ℚ (ω)$	$ℚ (ω)^{3}$	yes
$S_{3}$	$ℚ$	$ℚ \times ℚ \times M_{2} (ℚ)$	yes
$S_{3}$	$𝔽_{2}$	$𝔽_{2} \times M_{2} (𝔽_{2})$	yes

Frequently Asked Questions

Is every algebraically closed field a splitting field for every finite group?

Yes. A finite-dimensional division algebra over an algebraically closed field $k$ equals $k$ , because any element generates a finite field extension of $k$ . So all $D_{i} = k$ automatically. The content of $(8.3)$ is that one does not need to go all the way to $\bar{k}$ .

Does splitting depend on the characteristic?

Yes, and not in the direction one might guess. Characteristic $p$ dividing $| G |$ tends to make splitting easier, since the modular irreducibles are fewer and smaller, and Wedderburn's little theorem forbids noncommutative division algebras over finite fields. $𝔽_{2}$ splits both $S_{3}$ and $Q_{8}$ .

k

splits

G

, does

k

split every subgroup of

G

No. The property passes to quotient groups, because $k [G / N]$ is a quotient algebra of $k G$ and its simple modules are among those of $k G$ . Subgroups give subalgebras, not quotients, and simple $k H$ -modules need not appear among the simple $k G$ -modules.

How does one recognise a non-splitting field in practice?

Compute $\sum_{i} (\dim_{k} M_{i})^{2}$ for the irreducibles you have. If it exceeds $| G | - \dim_{k} rad k G$ , some $D_{i}$ is larger than $k$ and $k$ does not split $G$ . If it falls short, either the list is incomplete or, again, a $D_{i}$ is larger than $k$ . Only exact equality certifies splitting.

What is the connection with the Galois-theoretic notion of a splitting field?

They agree for cyclic groups, where $k G ≅ k [x] / (x^{n} - 1)$ and splitting the algebra is factoring the polynomial into linear factors. In general the algebraic notion is about a ring becoming a product of matrix algebras, which for noncommutative $D_{i}$ has no polynomial analogue.

Why does the proof of

(8.3)

go through the prime field?

Because the prime field is perfect, and $(7.10)$ — the existence of a finite splitting field — requires perfectness to guarantee that the algebra remains semisimple after extending to the algebraic closure. An arbitrary $k$ of characteristic $p$ need not be perfect, but its prime field always is, and the compositum step transfers the result back.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §8, (8.2)–(8.3) (pp. 126–129); §7, (7.6)–(7.14) for the algebra-theoretic background.
C. W. Curtis and I. Reiner, Representation Theory of Finite Groups and Associative Algebras, Wiley-Interscience, 1962 — proofs of Brauer's splitting field theorem in both characteristics.
R. Brauer, “On the representation of a group of order g in the field of the g-th roots of unity”, American Journal of Mathematics 67 (1945).
I. M. Isaacs, Character Theory of Finite Groups, Academic Press, 1976, Chapters 9–10 (Schur index and fields of definition).
D. S. Passman, The Algebraic Structure of Group Rings, Wiley-Interscience, 1977.
H. Nagao and Y. Tsushima, Representations of Finite Groups, Academic Press, 1989.

AI Suggested Questions

Compute the minimal splitting field for the dihedral group of order $8$ over $ℚ$ and compare it with the quaternion case.
How is the Schur index of a character computed from local invariants, and what are the possible values for a finite group?
Give an example of a finite group whose minimal splitting fields over $ℚ$ are not unique.
Why does Wedderburn's little theorem make splitting over finite fields purely a question of roots of unity?
State and prove the Frobenius–Schur indicator criterion for a complex irreducible representation to be realisable over $ℝ$ .
How does Brauer's induction theorem lead to the cyclotomic splitting field, and where does the exponent enter?
For which finite groups is $ℚ$ itself a splitting field, and is there a structural characterisation?

Splitting Fields for Groups

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Three ways to test for splitting

Splitting is stable upwards, not downwards

Why the algebraic closure is not the end of the story

Key Results

Proof Techniques and Method

Descend to the prime field

Matrix units are finitely many

Close the dimension count

Worked Example

The quaternion group of order 8

Over $k = ℚ$

Over $K = ℚ (\sqrt{- 1})$

Frameworks and Models

Already split

Split by a small explicit extension

Obstructed by a division algebra

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Choosing the coefficient field

Schur indices and Brauer groups

Codes over finite fields

Real versus complex representations

Design Considerations

Standards and Notation

Computational Notes

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions

Continue learning

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Three ways to test for splitting

Splitting is stable upwards, not downwards

Why the algebraic closure is not the end of the story

06Key Results

07Proof Techniques and Method

Descend to the prime field

Matrix units are finitely many

Close the dimension count

08Worked Example

The quaternion group of order 8

Over k=ℚ

Over K=ℚ(−1)

09Frameworks and Models

Already split

Split by a small explicit extension

Obstructed by a division algebra

10Process and Workflow

11Comparison and Classification

12Relationship Map

13Applications and Industry Use

Choosing the coefficient field

Schur indices and Brauer groups

Codes over finite fields

Real versus complex representations

14Design Considerations

15Standards and Notation

16Computational Notes

17Failure Modes and Common Mistakes

18Historical Notes and Lessons Learned

19Quick Reference

20Frequently Asked Questions

21Related KEVOS Topics

22References

23AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

Worked Example

Over $k = ℚ$

Over $K = ℚ (\sqrt{- 1})$

Frameworks and Models

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Design Considerations

Standards and Notation

Computational Notes

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions