The Structure of kG modulo Its Radical

Q: Why is kG automatically artinian?

Because _k kG = |G| is finite, and a finite-dimensional algebra over a field satisfies both chain conditions on one-sided ideals — descending chains of k-subspaces must stabilise. This is the single hypothesis that makes the whole theorem available, and it is exactly what fails for infinite G.

Q: Does (8.1) tell me what rad kG is?

Only its dimension, and only once everything else in the equation is known. Identifying the radical as an ideal requires separate arguments: Maschke in the coprime case, the theory of normal p-subgroups in the modular case, and Wallace's formula when a Sylow p-subgroup is normal.

Q: How do I know when I have found all the irreducibles?

When the count closes. If the known irreducibles contribute n_i^2 _k D_i = |G| - _k rad kG and you know the radical dimension independently, the list is complete. When char k |G| the radical is zero, so a total of exactly |G| certifies completeness.

Q: Can two of the D_i be different division rings?

Yes. For G = Q_8 over Q four of the D_i equal Q and the fifth is the rational quaternion algebra. There is no reason for the endomorphism rings of different simple modules to agree, and their variation is exactly what a splitting field removes.

Q: What is the relation between (8.1) and the character-theoretic formula |G| = _i(1)^2?

They are the same statement in the semisimple split case, where _i(1) = _k M_i = n_i. Outside that case the character formula is either false or meaningless, while (8.1)(4) remains correct — which is the reason to state the theorem in module-theoretic terms.

Q: Is the decomposition in (8.1)(1) unique?

Yes, in the strong sense supplied by the uniqueness part of Wedderburn–Artin: the number r, the multiset of matrix sizes n_i and the isomorphism classes of the D_i are determined by the ring. The individual idempotents cutting out the components are not unique, but the central primitive idempotents are.

Executive Summary

For a finite group $G$ and a field $k$ , the algebra $k G$ is finite-dimensional, hence artinian, hence its radical is nilpotent and the quotient $k G / rad k G$ is semisimple. Wedderburn–Artin then forces that quotient to be a finite product of matrix rings over division algebras. Theorem $(8.1)$ records the four consequences that get used.

The fourth of them, $| G | = \dim_{k} (rad k G) + \sum_{i = 1}^{r} n_{i}^{2} \dim_{k} D_{i}$ , is a hard arithmetic constraint on a small number of positive integers. It is routinely strong enough to determine $\dim_{k} rad k G$ , the number $r$ of irreducibles, or the fact that $k$ is a splitting field, from a couple of representations found by inspection.

$r$ Simple components

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

$\sum n_{i}^{2} \dim_{k} D_{i}$ Accounts for

| G | - \dim rad

$n_{i} M_{i}$ Multiplicity in the regular module

Overview

Every structural question about the $k$ -representations of a finite group $G$ passes through one exact sequence:

0 ⟶ rad k G ⟶ k G ⟶ k G / rad k G ⟶ 0 .

(8.1a)

The radical carries no simple modules; the quotient carries all of them.

The simple left $k G$ -modules are exactly the simple left modules of the quotient, because $rad k G$ annihilates every simple module by definition of the radical. So the classification problem is a problem about a semisimple ring, and there Wedderburn–Artin is a complete answer. What remains genuinely hard is the left-hand term: computing $rad k G$ is the modular theory.

Two special cases collapse the statement usefully. If $char k ∤ | G |$ then $rad k G = 0$ by Maschke's theorem and the first term disappears. If $k$ is in addition a splitting field for $G$ then every $D_{i} = k$ and $n_{i} = \dim_{k} M_{i}$ , so the identity becomes $| G | = \sum_{i} (\dim_{k} M_{i})^{2}$ — the classical formula.

Learning Objectives

State $(8.1)$ in full, with the definitions of $r$ , $M_{i}$ , $D_{i}$ and $n_{i}$ .
Explain why $k G$ being finite-dimensional is what licenses the whole argument.
Prove the four conclusions from Wedderburn–Artin and Schur's Lemma.
Recognise $\dim_{k} M_{i} = n_{i} \dim_{k} D_{i}$ as a statement about $M_{i}$ as a $D_{i}$ -vector space.
Specialise to abelian $G$ , to the semisimple case, and to a splitting field.
Use the dimension formula to pin down $rad (𝔽_{2} S_{3})$ exactly.

Definitions

Definition(7.1)The standing notation

Let $k$ be a field and $G$ a finite group, and set $R = k G$ . Let $M_{1}, \dots, M_{r}$ be a complete set of representatives for the isomorphism classes of simple left $R$ -modules — a finite set, because $R$ is left artinian. Put

D_{i} = End (_{R} M_{i}), n_{i} = \dim_{D_{i}} M_{i} .

By Schur's Lemma each $D_{i}$ is a division ring, and it is finite-dimensional over $k$ because $M_{i}$ is. Following Lam, elements of $D_{i}$ are composed as right operators on $M_{i}$ , so that $M_{i}$ is a right $D_{i}$ -vector space and $End (M_{i})_{D_{i}} ≅ M_{n_{i}} (D_{i})$ without any opposite ring appearing.

$rad k G$: The Jacobson radical of the group algebra: the intersection of the annihilators of the simple left $k G$ -modules. Nilpotent, because $k G$ is artinian.
Simple component: A factor $M_{n_{i}} (D_{i})$ of the semisimple quotient. It is a two-sided ideal of $k G / rad k G$ and a simple artinian ring.
$r$: The number of isomorphism classes of simple left $k G$ -modules, equivalently the number of simple components.
$n_{i}$: The multiplicity of $M_{i}$ in the left regular module modulo the radical; equal to $\dim_{D_{i}} M_{i}$ .
Splitting field: A field $k$ over which every simple $k G$ -module has $D_{i} = k$ ; then $n_{i} = \dim_{k} M_{i}$ .

Throughout this page G is finite and k is a field of arbitrary characteristic. All modules are finite-dimensional left modules.

Key Results

Theorem(8.1)Structure of kG modulo its radical

Let $G$ be a finite group and $k$ a field, and adopt the notation of $(7.1)$ above. Then:

$k G / rad k G ≅ M_{n_{1}} (D_{1}) \times \dots \times M_{n_{r}} (D_{r})$ as $k$ -algebras;
as a left $k G$ -module, $k G / rad k G ≅ n_{1} M_{1} \oplus \dots \oplus n_{r} M_{r}$ ;
$\dim_{k} M_{i} = n_{i} \dim_{k} D_{i}$ for each $i$ ;
$| G | = \dim_{k} (rad k G) + \sum_{i = 1}^{r} n_{i}^{2} \dim_{k} D_{i}$ .

Proof

Setup. $\dim_{k} k G = | G | < \infty$ , so $k G$ is a finite-dimensional $k$ -algebra and therefore left and right artinian. Hence $rad k G$ is nilpotent and $\bar{R} : = k G / rad k G$ is a semisimple ring. Since $rad k G$ annihilates every simple left $k G$ -module, the simple left $k G$ -modules are precisely the simple left $\bar{R}$ -modules, and $End (_{k G} M_{i}) = End (_{\bar{R}} M_{i})$ .

(1). By Wedderburn–Artin, $\bar{R} ≅ B_{1} \times \dots \times B_{r}$ with each $B_{i}$ simple artinian, and the number of factors equals the number of isomorphism classes of simple $\bar{R}$ -modules. Each $B_{i}$ has a unique simple left module, which we may take to be $M_{i}$ ; and $B_{i} ≅ End (M_{i})_{D_{i}} ≅ M_{n_{i}} (D_{i})$ where $D_{i} = End (_{\bar{R}} M_{i})$ and $n_{i} = \dim_{D_{i}} M_{i}$ . This is exactly the uniqueness statement of Wedderburn–Artin. The isomorphism is $k$ -linear because $k$ is central in $k G$ and the decomposition is by central idempotents.

(2). As a left module over itself, $\bar{R} = ⨁_{i} B_{i}$ , and $M_{n_{i}} (D_{i})$ decomposes as a left module over itself into its $n_{i}$ columns, each isomorphic to $M_{i}$ . The factors $B_{j}$ with $j \neq i$ act as zero on $M_{i}$ , so the summands are $k G$ -submodules with the stated isomorphism types. Hence $\bar{R} ≅ ⨁_{i} n_{i} M_{i}$ .

(3). $M_{i}$ is a right $D_{i}$ -vector space of dimension $n_{i}$ by definition of $n_{i}$ , so $M_{i} ≅ D_{i}^{n_{i}}$ as a right $D_{i}$ -module. Since $k$ maps into the centre of $D_{i}$ and the $k$ -structures are compatible, counting $k$ -dimensions gives $\dim_{k} M_{i} = n_{i} \dim_{k} D_{i}$ .

(4). Take $k$ -dimensions in $(8.1 a)$ : $\dim_{k} k G = \dim_{k} rad k G + \dim_{k} \bar{R}$ . By (1), $\dim_{k} \bar{R} = \sum_{i} \dim_{k} M_{n_{i}} (D_{i}) = \sum_{i} n_{i}^{2} \dim_{k} D_{i}$ . Finally $\dim_{k} k G = | G |$ because $G$ is a $k$ -basis of $k G$ .

Corollary(8.1abelian)The abelian case

If $G$ is a finite abelian group and $k$ any field, then all $n_{i} = 1$ , every $D_{i}$ is a finite field extension of $k$ , and

k G / rad k G ≅ D_{1} \times \dots \times D_{r} .

Viewed as ideals of $k G / rad k G$ , the $D_{i}$ afford the $r$ distinct irreducible $k$ -representations of $G$ .

Proof

$G$ abelian makes $k G$ commutative, hence so is the quotient $k G / rad k G ≅ \prod_{i} M_{n_{i}} (D_{i})$ . A matrix ring $M_{n} (D)$ is commutative only when $n = 1$ and $D$ is commutative, so $n_{i} = 1$ and each $D_{i}$ is a field. Each $D_{i}$ is a finite-dimensional commutative $k$ -algebra that is a division ring, hence a finite field extension of $k$ .

Corollary(8.1split)Semisimple and split cases

Let $G$ be finite and $k$ a field.

If $char k ∤ | G |$ then $rad k G = 0$ and $| G | = \sum_{i = 1}^{r} n_{i}^{2} \dim_{k} D_{i}$ .
If in addition $k$ is a splitting field for $G$ , then every $D_{i} = k$ , so $n_{i} = \dim_{k} M_{i}$ and $| G | = \sum_{i = 1}^{r} (\dim_{k} M_{i})^{2}$ .
If $k$ is a splitting field but $char k$ divides $| G |$ , then $| G | = \dim_{k} (rad k G) + \sum_{i = 1}^{r} (\dim_{k} M_{i})^{2}$ .

Remark—What the theorem does not do

$(8.1)$ describes the semisimple quotient completely and says nothing at all about $rad k G$ beyond its dimension. Determining the radical requires separate input: Maschke's theorem in the coprime case, the results on normal $p$ -subgroups in the modular case, and, in general, no closed formula at all.

Proof Techniques and Method

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

Move 1

Quotient by the radical first

Nothing about simple modules is lost, and the quotient is semisimple, where a complete structure theorem exists. This is the standard opening for any finite-dimensional algebra.

Move 2

Count dimensions on both sides

An isomorphism of finite-dimensional algebras is worth an equation between integers. Because the integers $n_{i}$ , $\dim_{k} D_{i}$ and $r$ are small and positive, that equation often has few solutions.

Move 3

Find representations by hand, then close the count

Exhibit whatever irreducibles are visible — trivial, sign, permutation, an embedding into a division algebra — subtract their contribution from $| G |$ , and see what room is left.

Move 3 is the workhorse. In the $𝔽_{2} S_{3}$ example below, two representations found by inspection contribute $1 + 4 = 5$ , leaving exactly $1$ for the radical — which then has to be the span of the sum of all group elements, because that is a square-zero ideal of dimension $1$ .

Worked Example

$G = S_{3}$ over $ℚ$ : everything is visible

$| G | = 6$ and $char ℚ = 0$ , so $rad ℚ S_{3} = 0$ by Maschke. Three irreducibles are immediate: the trivial module $M_{1}$ , the sign module $M_{1}'$ , and the $2$ -dimensional module

M_{2} = (k e_{1} \oplus k e_{2} \oplus k e_{3}) / k (e_{1} + e_{2} + e_{3}),

(E.1)

$S_{3}$ permutes the $e_{i}$ ; the quotient is $2$ -dimensional and irreducible over $ℚ$ .

All three have $D_{i} = ℚ$ , so $n_{i} = \dim_{ℚ} M_{i} = 1, 1, 2$ and

6 = 0 + 1^{2} \cdot 1 + 1^{2} \cdot 1 + 2^{2} \cdot 1, ℚ S_{3} ≅ ℚ \times ℚ \times M_{2} (ℚ) .

(E.2)

The count is exact with $r = 3$ , so there are no further irreducibles and $ℚ$ is already a splitting field for $S_{3}$ .

$G = C_{3}$ over $ℚ$ : a division ring larger than $k$

Here $ℚ C_{3} ≅ ℚ [x] / (x^{3} - 1) ≅ ℚ \times ℚ (ω)$ with $ω$ a primitive cube root of unity. So $r = 2$ , $D_{1} = ℚ$ , $D_{2} = ℚ (ω)$ , and $n_{1} = n_{2} = 1$ . Check $(8.1) (3)$ : $\dim_{ℚ} M_{2} = 2 = n_{2} \dim_{ℚ} D_{2} = 1 \cdot 2$ . Check $(8.1) (4)$ : $3 = 0 + 1 \cdot 1 + 1 \cdot 2$ . Note $\dim_{k} M_{i}$ and $n_{i}$ differ here — $ℚ$ is not a splitting field.

$G = S_{3}$ over $𝔽_{2}$ : the count determines the radical

Now $char k = 2$ divides $6$ , so $rad k G \neq 0$ . Over $𝔽_{2}$ the sign representation coincides with the trivial one, so $M_{1} = M_{1}'$ is the only $1$ -dimensional module in sight. The module $M_{2}$ of $(E . 1)$ is still $2$ -dimensional; in $𝔽_{2}^{3} / 𝔽_{2} (1, 1, 1)$ the three nonzero vectors are the images ${\bar{e}}_{1}, {\bar{e}}_{2}, {\bar{e}}_{3}$ , which $S_{3}$ permutes transitively, so no $1$ -dimensional submodule is stable and $M_{2}$ is simple. One checks $End (_{k G} M_{2}) = k$ , so $n_{2} = 2$ .

6 = \dim_{k} (rad k G) + 1^{2} \cdot 1 + 2^{2} \cdot 1 + \dots

(E.3)

Two irreducibles already account for $5$ . Any further simple module would contribute at least $1$ , forcing $\dim_{k} rad k G \leq 0$ — impossible in the modular case. Hence $r = 2$ and $\dim_{k} rad k G = 1$ .

To identify the radical, put $σ = \sum_{g \in G} g$ . Then $g σ = σ g = σ$ for every $g$ , so $k σ$ is a two-sided ideal, and $σ^{2} = | G | σ = 6 σ = 0$ in characteristic $2$ . A square-zero ideal lies in the radical, and $\dim_{k} k σ = 1 = \dim_{k} rad k G$ , so

rad (𝔽_{2} S_{3}) = 𝔽_{2} \cdot σ, 𝔽_{2} S_{3} / rad ≅ 𝔽_{2} \times M_{2} (𝔽_{2}) .

(E.4)

Process and Workflow

Write down $| G |$ This is

\dim_{k} k G

and the total budget for the count.

Decide whether the radical is zeroMaschke:

rad k G = 0

iff

char k ∤ | G |

. Otherwise expect a nonzero radical and keep the first term.

List the irreducibles you can seeTrivial; sign if

char k \neq 2

; the deleted permutation module; any embedding of

G

into a division algebra.

Compute each $D_{i}$

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

. If it is

k

then

n_{i} = \dim_{k} M_{i}

; otherwise divide.

Close the countSubstitute into

| G | = \dim_{k} rad k G + \sum_{i} n_{i}^{2} \dim_{k} D_{i}

and see what the remaining budget permits.

Identify the radicalIf the budget forces a small dimension, look for an explicit nilpotent ideal of that dimension — often the span of

\sum_{g} g

, or the kernel of

k G \to k [G / H]

The count leaves an unexplained remainder. What now?

Remainder is $0$ You have found all the irreducibles and

rad k G = 0

. Confirm that

char k ∤ | G |

, or you have made an error.

Remainder is small and positiveEither an unnoticed irreducible or the radical. Compare with

\dim_{k} rad k G

obtained independently — for example from Wallace's formula when a Sylow

p

-subgroup is normal.

Remainder is largeLook for missing irreducibles first: an unsplit division algebra

D_{i}

contributes

n_{i}^{2} \dim_{k} D_{i}

, which grows fast. Consider extending

k

to a splitting field and comparing.

Comparison and Classification

Worked decompositions of kG modulo the radical
$G$	$k$	$k G / rad k G$	$\dim_{k} rad k G$	$r$
$C_{3}$	$ℚ$	$ℚ \times ℚ (ω)$	$0$	$2$
$C_{3}$	$𝔽_{3}$	$𝔽_{3}$	$2$	$1$
$S_{3}$	$ℚ$	$ℚ \times ℚ \times M_{2} (ℚ)$	$0$	$3$
$S_{3}$	$𝔽_{2}$	$𝔽_{2} \times M_{2} (𝔽_{2})$	$1$	$2$
$S_{3}$	$𝔽_{3}$	$𝔽_{3} \times 𝔽_{3}$	$4$	$2$
$Q_{8}$	$ℚ$	$ℚ^{4} \times ℍ_{ℚ}$	$0$	$5$
$Q_{8}$	$ℚ (\sqrt{- 1})$	$K^{4} \times M_{2} (K)$	$0$	$5$

Which conclusion of (8.1) is available under which hypotheses
	$G$ finite, $k$ any	$char k ∤ \| G \|$	$k$ splitting	$G$ abelian
$k G / rad$ is a product of matrix rings over division rings	yes	yes	yes	yes
$rad k G = 0$	no	yes	no	no
all $D_{i} = k$	no	no	yes	no
all $n_{i} = 1$	no	no	no	yes
$\| G \| = \sum_{i} (\dim_{k} M_{i})^{2}$	no	no	partial	no
$r$ equals the number of conjugacy classes	no	partial	partial	no

Which conclusion of (8.1) is available under which hypotheses

Relationship Map

The theorem is a specialisation, not a new argument. Its content is that group algebras are finite-dimensional, so every general result about finite-dimensional algebras applies.

Rings with identity

rad R

defined

Left artinian rings

rad R

nilpotent;

R / rad R

semisimple (Hopkins–Levitzki)

Finite-dimensional $k$ -algebrasall data finite-dimensional over

k

; dimension counting available

$k G$ , $G$ finite

\dim_{k} = | G |

;

G

acts by units; the count of

(8.1) (4)

$char k ∤ | G |$

rad k G = 0

by Maschke:

k G

itself is semisimple

Hopkins–Levitzki

k G / rad k G

semisimpleWedderburn–Artin

\prod_{i} M_{n_{i}} (D_{i})

(8.1) (4)

Applications and Industry Use

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

Computer algebra

Algebra decomposition as a primitive

GAP and Magma implement Wedderburn decomposition of $k G$ directly: compute the radical, then split the semisimple quotient into simple components with their matrix sizes and division algebras. The output is precisely the data $r$ , $n_{i}$ , $D_{i}$ .

Coding theory

Idempotents and group codes

The primitive central idempotents of $k G$ generate the minimal group codes. Knowing $k G / rad k G$ tells you how many such codes exist and their dimensions; when $char k ∣ | G |$ the codes are no longer generated by idempotents, which is why coding theory usually assumes the semisimple case.

Number theory

Schur indices and Brauer groups

When $D_{i} \neq k$ its class in the Brauer group of $k$ is the obstruction, measured by the Schur index. The classification of the $D_{i}$ that arise from finite groups over number fields is a genuine research subject.

Harmonic analysis

Fourier inversion

In the semisimple case the isomorphism of $(8.1) (1)$ is the Fourier transform on $G$ , and $| G | = \sum n_{i}^{2}$ over a splitting field is the Plancherel identity for the group.

Honestly stated: this theorem is the engine, not the product. It is quoted, not admired — but almost nothing else in the modular theory can be proved without it.

Design Considerations

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

Which base field? Enlarging $k$ to a splitting field makes all $D_{i} = k$ and simplifies every formula, at the cost of leaving the field you care about. Descent problems — is the representation realisable over $k$ ? — are then measured by the Schur index.
**Right operators for $D_{i}$ .** Composing endomorphisms as right operators avoids opposite rings in $End (M_{i})_{D_{i}} ≅ M_{n_{i}} (D_{i})$ . If you prefer left operators, every occurrence of $D_{i}$ in $(8.1)$ becomes $D_{i}^{op}$ ; the dimensions are unchanged, which is why the numerical formulas are convention-independent.
Radical first or last? Quotienting by the radical immediately is almost always right; it is the only step that loses information, and it is what makes the rest mechanical. Keep a record of $\dim_{k} rad k G$ so the count can be closed.
Dimension counting versus character counting. In the semisimple case characters give the same information more quickly. In the modular case characters over $k$ are unreliable and the dimension count of $(8.1) (4)$ is the honest tool.

Standards and Notation

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

Semisimple quotient

k G / rad k G

; also written

\bar{k G}

k G / J (k G)

Matrix ring

M_{n} (D)

here;

{Mat}_{n} (D)

and

D^{n \times n}

elsewhere

Simple modules

M_{1}, \dots, M_{r}

(Lam);

S_{λ}

L (λ)

in the highest-weight literature

Multiplicity

n_{i} = \dim_{D_{i}} M_{i}

; equals

\dim_{k} M_{i}

only over a splitting field

Number of irreducibles

r

here;

l (G)

for the modular count in the Brauer-character literature

GAPWedderburnDecomposition (Wedderga package), RadicalOfAlgebra

MagmaWedderburnDecomposition, JacobsonRadical, AbsolutelyIrreducibleModules

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.

Computing the data of $(8.1)$ for a finite-dimensional algebra of dimension $n = | G |$ over a field $k$ splits into two very different stages.

Radical. In characteristic $0$ the radical is the kernel of the trace form, one $n \times n$ nullspace computation, $O (n^{3})$ field operations. In characteristic $p$ the trace form is inadequate and the Friedl–Rónyai algorithm is used; still polynomial time, but with a chain of higher trace conditions.
Splitting the semisimple quotient. Over a finite field this is effective and fast. Over $ℚ$ it requires factoring polynomials and identifying division algebras by their local invariants, which is where the real cost lies.
Simple modules directly. Often cheaper to find the $M_{i}$ by the MeatAxe on explicit matrix modules than to decompose $k G$ itself; the numbers $n_{i}$ and $\dim_{k} D_{i}$ then come out of the MeatAxe's absolute-irreducibility test.
Scale. The regular module has dimension $| G |$ , so direct computation in $k G$ is impractical beyond a few thousand elements. Condensation methods work in a much smaller algebra $e k G e$ and recover the multiplicities afterwards.

Failure Modes and Common Mistakes

Do not assume the $D_{i}$ are commutative. $Q_{8}$ over $ℚ$ produces the rational quaternion division algebra as a $D_{i}$ .
Do not read $(8.1) (2)$ as a decomposition of $k G$ itself. It decomposes $k G / rad k G$ ; the regular module $k G$ is generally not a direct sum of simples in the modular case.
Do not forget that $r$ counts isomorphism classes. Two simple modules with the same dimension and the same character over a non-splitting field can still be non-isomorphic.

Quick Reference

Hypotheses

G

finite,

k

a field,

char k

arbitrary

(1) Algebra

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

(2) Module

k G / rad k G ≅ n_{1} M_{1} \oplus \dots \oplus n_{r} M_{r}

(3) Dimension

\dim_{k} M_{i} = n_{i} \dim_{k} D_{i}

(4) Count

| G | = \dim_{k} rad k G + \sum_{i} n_{i}^{2} \dim_{k} D_{i}

Abelian $G$ all

n_{i} = 1

, all

D_{i}

finite field extensions of

k

Maschke case

char k ∤ | G | \Rightarrow rad k G = 0

Split case

D_{i} = k \Rightarrow n_{i} = \dim_{k} M_{i}

Which term vanishes when
Hypothesis	Effect on the count	Resulting identity
$char k ∤ \| G \|$	$rad k G = 0$	$\| G \| = \sum_{i} n_{i}^{2} \dim_{k} D_{i}$
$k$ a splitting field	all $D_{i} = k$	$\| G \| = \dim_{k} rad k G + \sum_{i} n_{i}^{2}$
both	radical zero and $D_{i} = k$	$\| G \| = \sum_{i} (\dim_{k} M_{i})^{2}$
$G$ abelian	all $n_{i} = 1$	$\| G \| = \dim_{k} rad k G + \sum_{i} \dim_{k} D_{i}$
$G$ a $p$ -group, $char k = p$	$r = 1$ , $M_{1} = k$ , $D_{1} = k$	$\| G \| = \dim_{k} rad k G + 1$

Frequently Asked Questions

Why is

k G

automatically artinian?

Because $\dim_{k} k G = | G |$ is finite, and a finite-dimensional algebra over a field satisfies both chain conditions on one-sided ideals — descending chains of $k$ -subspaces must stabilise. This is the single hypothesis that makes the whole theorem available, and it is exactly what fails for infinite $G$ .

Does

(8.1)

tell me what

rad k G

is?

Only its dimension, and only once everything else in the equation is known. Identifying the radical as an ideal requires separate arguments: Maschke in the coprime case, the theory of normal $p$ -subgroups in the modular case, and Wallace's formula when a Sylow $p$ -subgroup is normal.

How do I know when I have found all the irreducibles?

When the count closes. If the known irreducibles contribute $\sum n_{i}^{2} \dim_{k} D_{i} = | G | - \dim_{k} rad k G$ and you know the radical dimension independently, the list is complete. When $char k ∤ | G |$ the radical is zero, so a total of exactly $| G |$ certifies completeness.

Can two of the

D_{i}

be different division rings?

Yes. For $G = Q_{8}$ over $ℚ$ four of the $D_{i}$ equal $ℚ$ and the fifth is the rational quaternion algebra. There is no reason for the endomorphism rings of different simple modules to agree, and their variation is exactly what a splitting field removes.

What is the relation between

(8.1)

and the character-theoretic formula

| G | = \sum χ_{i} (1)^{2}

They are the same statement in the semisimple split case, where $χ_{i} (1) = \dim_{k} M_{i} = n_{i}$ . Outside that case the character formula is either false or meaningless, while $(8.1) (4)$ remains correct — which is the reason to state the theorem in module-theoretic terms.

Is the decomposition in

(8.1) (1)

unique?

Yes, in the strong sense supplied by the uniqueness part of Wedderburn–Artin: the number $r$ , the multiset of matrix sizes $n_{i}$ and the isomorphism classes of the $D_{i}$ are determined by the ring. The individual idempotents cutting out the components are not unique, but the central primitive idempotents are.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §8, Theorem (8.1) (pp. 125–126); §7 (7.1)–(7.2) for the underlying notation.
C. W. Curtis and I. Reiner, Representation Theory of Finite Groups and Associative Algebras, Wiley-Interscience, 1962, Chapters IV–V.
D. S. Passman, The Algebraic Structure of Group Rings, Wiley-Interscience, 1977, Chapter 1.
N. Jacobson, Structure of Rings, American Mathematical Society Colloquium Publications 37, revised edition, 1964, Chapter IV.
H. Nagao and Y. Tsushima, Representations of Finite Groups, Academic Press, 1989, Chapter 1.
L. H. Rowen, Ring Theory, Volume I, Academic Press, 1988, Chapter 2.

AI Suggested Questions

Compute the Wedderburn decomposition of $ℚ D_{8}$ and compare it with $ℚ Q_{8}$ to see how the group algebra distinguishes the two groups of order $8$ .
For which finite groups $G$ is $ℚ G$ a product of fields, and what does that say about $G$ ?
How does one compute $\dim_{k} rad k G$ when no Sylow $p$ -subgroup is normal?
Explain the role of the Schur index in controlling the division algebras $D_{i}$ over a number field.
Derive the Plancherel formula for a finite group from the isomorphism in $(8.1) (1)$ .
What can be said about $k G / rad k G$ when $G$ is infinite but $k G$ is still artinian, and which groups have that property?
Work out $(8.1)$ for $G = A_{4}$ over $ℚ$ , $ℚ (ω)$ and $𝔽_{2}$ , and reconcile the three answers.

kG Modulo Its Radical

Executive Summary

Overview

Learning Objectives

Definitions

Key Results

Proof Techniques and Method

Quotient by the radical first

Count dimensions on both sides

Find representations by hand, then close the count

Worked Example

$G = S_{3}$ over $ℚ$ : everything is visible

$G = C_{3}$ over $ℚ$ : a division ring larger than $k$

$G = S_{3}$ over $𝔽_{2}$ : the count determines the radical

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Algebra decomposition as a primitive

Idempotents and group codes

Schur indices and Brauer groups

Fourier inversion

Design Considerations

Standards and Notation

Computational Notes

Failure Modes and Common Mistakes

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions

Continue learning

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Key Results

06Proof Techniques and Method

Quotient by the radical first

Count dimensions on both sides

Find representations by hand, then close the count

07Worked Example

G=S3 over ℚ: everything is visible

G=C3 over ℚ: a division ring larger than k

G=S3 over 𝔽2: the count determines the radical

08Process and Workflow

09Comparison and Classification

10Relationship Map

11Applications and Industry Use

Algebra decomposition as a primitive

Idempotents and group codes

Schur indices and Brauer groups

Fourier inversion

12Design Considerations

13Standards and Notation

14Computational Notes

15Failure Modes and Common Mistakes

16Quick Reference

17Frequently Asked Questions

18Related KEVOS Topics

19References

20AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Key Results

Proof Techniques and Method

Worked Example

$G = S_{3}$ over $ℚ$ : everything is visible

$G = C_{3}$ over $ℚ$ : a division ring larger than $k$

$G = S_{3}$ over $𝔽_{2}$ : the count determines the radical

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Design Considerations

Standards and Notation

Computational Notes

Failure Modes and Common Mistakes

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions