Maschke’s Theorem and Semisimplicity of Group Rings

Q: Where exactly does the proof break when char k divides |G|?

At the division. The sum _ f(^-1v) is still defined and still G-equivariant, but its restriction to W is multiplication by |G|, which is 0 rather than invertible. The construction produces the zero map instead of a retraction.

Q: Is the invertibility of |G| really necessary, or just convenient?

Necessary. The group sum G satisfies G^2 = |G|G, so if |G| 1_k = 0 then kGG is a nonzero ideal of square zero and rad(kG) ≠ 0. Over a general ring the full necessity argument goes through von Neumann regularity and the lemma on the annihilator of 1-.

Q: Does semisimplicity of kG tell you the number of simple modules?

It tells you there are finitely many and that kG is their matrix-ring product. The count is the number of conjugacy classes only when k is a splitting field of characteristic zero. Over Q the count can be smaller because Galois-conjugate complex representations fuse into a single rational component.

Q: How does the relative version help with infinite groups?

It replaces |G| by an index [G:H]. If G is infinite but has a subgroup H of finite invertible index, semisimplicity of a module over kH still transfers to kG. This does not make kG semisimple — that is impossible for infinite G — but it does transfer semisimplicity of individual modules.

Executive Summary

Maschke's theorem is the dividing line of group representation theory. For $G$ finite and $k$ a coefficient ring, $k G$ is semisimple if and only if two conditions hold: $k$ is itself semisimple, and $| G | \cdot 1_{k}$ is a unit in $k$ . Over a field this reads: $k G$ is semisimple exactly when $char k$ does not divide $| G |$ .

The sufficiency is a single construction — average a $k$ -linear projection over the group — and it is the origin of the Reynolds operator throughout invariant theory. The necessity is subtler and is where the augmentation map earns its keep: pushing a von Neumann regular identity through $ε$ forces each prime dividing $| G |$ to be invertible.

1898Maschke

$| G |^{- 1}$ The whole hypothesis

$\Leftrightarrow$ Both directions hold

$\sum n_{i}^{2} = | G |$ Complex decomposition

Overview

Emmy Noether's reformulation makes the setting precise: a $k$ -representation of $G$ is the same thing as a left $k G$ -module, and equivalence of representations is isomorphism of modules. Complete reducibility of representations therefore becomes semisimplicity of the ring $k G$ , and the whole Wedderburn–Artin structure theory becomes available.

The theorem below is stated for an arbitrary coefficient ring, not merely a field, because both hypotheses are then visible separately. Over a field the first condition is automatic and only the arithmetic condition survives.

k G semisimple iff k semisimple and | G | \cdot 1_{k} \in U (k) .

(6.1)

$G$ finite. The finiteness of $G$ is essential: for infinite $G$ no coefficient ring makes $k G$ semisimple.

The complementary result, that $k G$ is never semisimple for infinite $G$ , is proved in Group Rings of Infinite Groups Are Never Semisimple. Together the two statements settle semisimplicity completely, and the residual question — what survives when semisimplicity fails — becomes the J-semisimplicity problem.

Learning Objectives

State $(6.1)$ with both hypotheses on $k$ and on $| G |$ , and explain why neither can be dropped.
Build the averaging operator and check $k$ -linearity, $G$ -equivariance and idempotence on the submodule.
Prove the annihilator lemma $(6.2)$ for an element of prime order.
Run the necessity argument: augmentation applied to a von Neumann regular identity.
State the field version and the complex Wedderburn decomposition with the sum-of-squares identity.
Compute $ℚ S_{3}$ and $𝔽_{3} S_{3}$ and identify the radical in the modular case.

Definitions

$k G$: The group ring: free as a left $k$ -module on $G$ , with multiplication extending the group law. Here $k$ is any ring with identity, not necessarily commutative.
$| G | \cdot 1_{k}$: The sum of $| G |$ copies of $1_{k}$ . It is central in $k$ for any ring $k$ , so if it is invertible its inverse is central too.
$\hat{G}$: The group sum $\sum_{g \in G} g \in k G$ . It is central, $g \hat{G} = \hat{G}$ for all $g$ , and ${\hat{G}}^{2} = | G | \hat{G}$ .
$\hat{σ}$: For $σ$ of order $p$ , the element $1 + σ + \dots + σ^{p - 1}$ . It satisfies $\hat{σ} (1 - σ) = (1 - σ) \hat{σ} = 0$ and $ε (\hat{σ}) = p \cdot 1_{k}$ .
Semisimple module: A module that is a sum of simple submodules; equivalently, one in which every submodule is a direct summand.

Modules are left modules and unital. A ring is semisimple when it is semisimple as a left module over itself — a condition that is left-right symmetric, unlike most conditions in this collection.

Core Concepts

Averaging: turning a $k$ -splitting into a $k G$ -splitting

Let $W \subseteq V$ be $k G$ -modules. Because $k$ is semisimple, $W$ is a direct summand of $V$ as a $k$ -module, so there is a $k$ -linear retraction $f : V \to W$ with $f |_{W} = id$ . This $f$ has no reason to respect the $G$ -action. The repair is to symmetrise it:

g (v) = | G |^{- 1} \sum_{σ \in G} σ f (σ^{- 1} v), v \in V .

(6.1a)

The averaging, or Reynolds, operator. It requires exactly one thing: that $| G |$ be invertible in $k$ .

The three verifications are short. The image lies in $W$ because $f$ lands in $W$ and $W$ is $G$ -stable. The restriction to $W$ is the identity because each summand contributes $σ σ^{- 1} w = w$ . And equivariance is a reindexing: replacing $σ$ by $τ σ$ converts $g (τ v)$ into $τ g (v)$ .

Why the converse needs the augmentation

Suppose instead that $k G$ is semisimple. Two consequences are immediate. First, $k$ is a homomorphic image of $k G$ under $ε$ , and quotients of semisimple rings are semisimple. Second, semisimple rings are von Neumann regular, so every element $a$ admits $x$ with $a x a = a$ .

Apply regularity to $a = 1 - σ$ for $σ$ of prime order $p$ , given by Cauchy's theorem whenever $p$ divides $| G |$ . The resulting identity $[1 - (1 - σ) x] (1 - σ) = 0$ says that $1 - (1 - σ) x$ lies in the right annihilator of $1 - σ$ , and the lemma identifies that annihilator as $k G \hat{σ}$ . Augmenting turns this into an equation in $k$ with $p$ on one side.

What replaces the theorem in the modular case

When $char k = p$ divides $| G |$ , the group sum satisfies ${\hat{G}}^{2} = | G | \hat{G} = 0$ while $\hat{G} \neq 0$ , so $k G \hat{G}$ is a nonzero ideal of square zero and $rad (k G) \neq 0$ . Everything then happens inside $k G / rad (k G)$ and in the lifting problem back to $k G$ — the subject of modular representation theory.

Key Results

Lemma(6.2)Right annihilator of

1 - σ

Let $k$ be any ring, $G$ any group, $R = k G$ , and let $σ \in G$ have finite order $p$ . Put $\hat{σ} = 1 + σ + \dots + σ^{p - 1}$ . Then for $r \in R$ ,

r (1 - σ) = 0 iff r \in R \hat{σ} .

(6.2a)

Proof

**( $\Leftarrow$ )** $\hat{σ} (1 - σ) = \hat{σ} - \hat{σ} σ = \hat{σ} - \hat{σ} = 0$ , because right multiplication by $σ$ permutes the terms of $\hat{σ}$ cyclically. Hence $β \hat{σ} (1 - σ) = 0$ for every $β \in R$ .

**( $\Rightarrow$ )** Write $r = \sum_{μ \in G} r_{μ} μ$ . The condition $r (1 - σ) = 0$ says $r σ = r$ , and the coefficient of $μ$ in $r σ$ is $r_{μ σ^{- 1}}$ . So $r_{μ σ^{- 1}} = r_{μ}$ for every $μ$ : the coefficient function is constant on the right cosets $μ ⟨ σ ⟩$ , each of which has exactly $p$ elements.

Choose representatives $τ_{1}, \dots, τ_{s}$ of the cosets meeting the support of $r$ . Grouping the sum by cosets gives

r = \sum_{j = 1}^{s} r_{τ_{j}} (τ_{j} + τ_{j} σ + \dots + τ_{j} σ^{p - 1}) = (\sum_{j = 1}^{s} r_{τ_{j}} τ_{j}) \hat{σ} \in R \hat{σ} .

(6.2b)

Theorem(6.1)Maschke's theorem

Let $k$ be any ring with identity and let $G$ be a finite group. Then the group ring $R = k G$ is semisimple if and only if $k$ is semisimple and $| G | \cdot 1_{k}$ is a unit in $k$ .

Proof

Sufficiency. Assume $k$ is semisimple and $| G | \cdot 1_{k} \in U (k)$ ; write $| G |^{- 1}$ for its inverse, which is central because $| G | \cdot 1_{k}$ is. Let $V$ be any left $R$ -module and $W \subseteq V$ an $R$ -submodule. Since $k$ is semisimple, $W$ is a $k$ -direct summand of $V$ , so there is a $k$ -linear $f : V \to W$ with $f |_{W} = {id}_{W}$ .

Define $g : V \to W$ by $(6.1 a)$ . It is $k$ -linear because $f$ is and because $k$ commutes with $G$ inside $R$ . For $w \in W$ we have $σ^{- 1} w \in W$ , so $f (σ^{- 1} w) = σ^{- 1} w$ and $g (w) = | G |^{- 1} \sum_{σ} w = w$ . For $τ \in G$ and $v \in V$ , substituting $σ = τ ρ$ gives

g (τ v) = | G |^{- 1} \sum_{σ \in G} σ f (σ^{- 1} τ v) = | G |^{- 1} \sum_{ρ \in G} τ ρ f (ρ^{- 1} v) = τ g (v),

(6.1b)

so $g$ is $R$ -linear. Thus $g$ is an $R$ -module retraction onto $W$ and $V = W \oplus \ker g$ . Every submodule of every $R$ -module is a direct summand, which is one of the standard characterisations of a semisimple ring.

Necessity. Assume $R = k G$ is semisimple. The augmentation $ε : k G \to k$ is a surjective ring homomorphism, and a homomorphic image of a semisimple ring is semisimple; hence $k$ is semisimple.

Now let $p$ be any prime dividing $| G |$ . By Cauchy's theorem there is $σ \in G$ of order $p$ . Semisimple rings are von Neumann regular, so there is $x \in R$ with $(1 - σ) x (1 - σ) = 1 - σ$ , that is $[1 - (1 - σ) x] (1 - σ) = 0$ . By $(6.2)$ there is $β \in R$ with

1 - (1 - σ) x = β \hat{σ}, \hat{σ} = 1 + σ + \dots + σ^{p - 1} .

(6.1c)

Apply $ε$ . On the left, $ε (1 - σ) = 0$ kills the second term and we get $1$ . On the right, $ε (\hat{σ}) = p \cdot 1_{k}$ , so we get $ε (β) (p \cdot 1_{k})$ . Hence $ε (β) (p \cdot 1_{k}) = 1$ , and since $p \cdot 1_{k}$ is central this makes $p \cdot 1_{k}$ a unit of $k$ .

Every prime divisor of $| G |$ is therefore invertible in $k$ ; writing $| G | = \prod_{i} p_{i}^{e_{i}}$ and multiplying the corresponding units gives $| G | \cdot 1_{k} \in U (k)$ .

Corollary—The field case

Let $k$ be a field and $G$ a finite group. Then $k G$ is semisimple if and only if $char k ∤ | G |$ — in particular whenever $char k = 0$ . Equivalently, every $k G$ -module is a direct sum of simple modules, and every finite-dimensional $k$ -representation of $G$ is completely reducible.

Corollary—Wedderburn form over

ℂ

For $G$ finite, $ℂ G$ is semisimple, and since $ℂ$ is algebraically closed every division algebra finite-dimensional over $ℂ$ equals $ℂ$ . Wedderburn–Artin therefore gives

ℂ G ≅ \prod_{i = 1}^{r} M_{n_{i}} (ℂ), \sum_{i = 1}^{r} n_{i}^{2} = | G |,

(6.1d)

with $r$ equal to the number of conjugacy classes of $G$ , because the class sums form a $ℂ$ -basis of the centre.

For a general field $k$ of characteristic zero the same argument gives $k G ≅ \prod_{i} M_{n_{i}} (D_{i})$ with each $D_{i}$ a division algebra finite-dimensional over $k$ ; the $D_{i}$ need not be commutative, as $ℚ Q_{8}$ shows, where the quaternion algebra appears.

Proposition—Relative version

Let $k$ be a ring, $G$ a group and $H \leq G$ a subgroup of finite index $i$ with $i \cdot 1_{k} \in U (k)$ . If a $k G$ -module $V$ is semisimple as a $k H$ -module, then $V$ is semisimple as a $k G$ -module.

Proof

Let $W \subseteq V$ be a $k G$ -submodule. It is in particular a $k H$ -submodule, so by hypothesis there is a $k H$ -linear retraction $f : V \to W$ . Fix a left transversal $t_{1}, \dots, t_{i}$ of $H$ in $G$ and set $g (v) = i^{- 1} \sum_{j} t_{j} f (t_{j}^{- 1} v)$ .

This is independent of the choice of representatives: replacing $t_{j}$ by $t_{j} h$ with $h \in H$ gives $t_{j} h f (h^{- 1} t_{j}^{- 1} v) = t_{j} f (t_{j}^{- 1} v)$ by $k H$ -linearity of $f$ . Equivariance under $τ \in G$ follows because ${τ t_{j}}$ is again a transversal, and $g |_{W} = id$ as before. So $W$ is a $k G$ -direct summand, and $V$ is a semisimple $k G$ -module.

Taking $H = {1}$ and $k$ semisimple recovers the sufficiency half of $(6.1)$ .

Proof Techniques and Method

The reusable moves behind the two halves of the proof.

Move 1

Symmetrise, then divide

Given any $k$ -linear gadget, sum its $G$ -conjugates and divide by $| G |$ . The result is $G$ -equivariant and agrees with the original on $G$ -fixed data. This is the Reynolds operator and it recurs in invariant theory and in the construction of invariant inner products.

Move 2

Push an identity through $ε$

Any equation in $k G$ becomes an equation in $k$ under augmentation, and elements like $\hat{σ}$ that look invisible in $k G$ produce the integer $p$ in $k$ . This is the only way arithmetic enters the necessity proof.

Move 3

Use regularity instead of idempotents

Semisimple gives von Neumann regular, and regularity supplies an equation involving a chosen element rather than an abstract decomposition. Choosing $a = 1 - σ$ targets exactly the torsion of $G$ .

Move 1 is robust: it needs only invertibility of an index, which is why the relative version goes through verbatim for a subgroup of invertible finite index. Move 2 is fragile in a useful way: it is precisely the failure of $p$ to be invertible that manufactures the radical.

Worked Example

$ℚ S_{3}$ : the semisimple side

Let $G = S_{3}$ , of order $6$ , and $k = ℚ$ . Since $char ℚ = 0$ , Maschke applies and $ℚ S_{3}$ is semisimple. The group has three conjugacy classes — identity, transpositions, three-cycles — and three irreducible representations over $ℚ$ : trivial, sign, and the two-dimensional standard representation on ${(a, b, c) \in ℚ^{3} : a + b + c = 0}$ .

All three are realised over $ℚ$ and each has endomorphism ring $ℚ$ , so

ℚ S_{3} ≅ ℚ \times ℚ \times M_{2} (ℚ), 1 + 1 + 4 = 6 = | S_{3} | .

(E.1)

The dimension count is the identity $\sum n_{i}^{2} = | G |$ in the smallest interesting case.

$𝔽_{3} S_{3}$ : the modular side

Now take $k = 𝔽_{3}$ . Since $3$ divides $6$ , Maschke fails and $rad (𝔽_{3} S_{3}) \neq 0$ . The Sylow $3$ -subgroup $P = A_{3} = ⟨ c ⟩$ is normal, and the relative augmentation ideal $Δ (G, P) = k G Δ_{k} (P)$ is nilpotent because $Δ_{𝔽_{3}} (P)^{3} = 0$ and $P$ is normal. Moreover $k G / Δ (G, P) ≅ 𝔽_{3} [G / P] = 𝔽_{3} C_{2}$ , which is semisimple by Maschke since $2$ is invertible in $𝔽_{3}$ . A nilpotent ideal with semisimple quotient is the radical, so

rad (𝔽_{3} S_{3}) = Δ (S_{3}, A_{3}), 𝔽_{3} S_{3} / rad ≅ 𝔽_{3} C_{2} ≅ 𝔽_{3} \times 𝔽_{3} .

(E.2)

Dimensions: $\dim rad = 6 - 2 = 4$ . Since $Δ (G, P)^{n} = k G Δ_{k} (P)^{n}$ and $Δ_{𝔽_{3}} (P)$ has powers of dimension $2, 1, 0$ , we get $\dim {rad}^{2} = \dim k G (1 + c + c^{2}) = 2$ and ${rad}^{3} = 0$ .

Radical filtration of $𝔽_{3} S_{3}$
Layer	$𝔽_{3}$ -dimension	Description
$𝔽_{3} S_{3}$	$6$	the whole algebra
$rad$	$4$	$Δ (S_{3}, A_{3})$ , nilpotent of index $3$
${rad}^{2}$	$2$	spanned by $\hat{A_{3}}$ and $τ \hat{A_{3}}$ for a transposition $τ$
${rad}^{3}$	$0$	vanishes

Process and Workflow

Check finiteness of $G$ If

G

is infinite, stop:

k G

is not semisimple for any nonzero

k

Check $k$ Is

k

semisimple? A field, a finite product of matrix rings over division rings — yes.

ℤ

ℤ / 4

k [[x]]

— no.

Check the arithmeticIs

| G | \cdot 1_{k}

a unit? Over a field this is: does

char k

divide

| G |

Semisimple: decomposeApply Wedderburn–Artin. Over an algebraically closed field of characteristic

0

the factors are matrix rings and the dimensions satisfy

\sum n_{i}^{2} = | G |

Not semisimple: localise the failureCompute

rad (k G)

, pass to the quotient, and study lifting. For a normal Sylow

p

-subgroup

P

the radical is

Δ (G, P)

$k$ is a field, $G$ is finite — what does $char k$ do?

Characteristic $0$

k G

is semisimple. Ordinary representation theory: characters determine modules, and

ℂ G

splits as a product of matrix algebras.

Characteristic $p$ , $p ∤ | G |$ Still semisimple. The theory is essentially the ordinary one, though the field may not be a splitting field.

Characteristic $p$ , $G$ a $p$ -groupMaximally non-semisimple:

k G

is local,

rad (k G) = Δ_{k} (G)

is nilpotent and there is one simple module.

Characteristic $p$ , $p$ divides $| G |$ , $G$ not a $p$ -groupGenuinely modular. The radical is nonzero and strictly inside

Δ_{k} (G)

; Brauer theory, blocks and defect groups govern the structure.

Comparison and Classification

Maschke's theorem across coefficient rings, $G$ finite and nontrivial
$k$	$k$ semisimple?	$\| G \| \cdot 1_{k}$ a unit?	$k G$ semisimple?
$ℂ$ , $ℝ$ , $ℚ$	yes	yes	yes
$𝔽_{p}$ with $p ∤ \| G \|$	yes	yes	yes
$𝔽_{p}$ with $p$ dividing $\| G \|$	yes	no	no
$ℤ$	no	only if $\| G \| = 1$	no
$ℤ [1 / \| G \|]$	no	yes	no
$M_{n} (D)$ , $D$ a division ring of characteristic $0$	yes	yes	yes

Which hypothesis each half of the proof consumes
	$G$ finite	$k$ semisimple	$\| G \|$ invertible	von Neumann regularity
Averaging operator is defined	yes	no	yes	no
A $k$ -linear retraction exists	no	yes	no	no
$k$ is semisimple (necessity)	no	no	no	no
$p \cdot 1_{k}$ is a unit (necessity)	yes	no	no	yes

Which hypothesis each half of the proof consumes

The necessity of $k$ being semisimple uses only that $ε$ is a surjective ring map, so it holds for infinite $G$ as well — vacuously, since $k G$ is then never semisimple.

Applications and Industry Use

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

Signal processing

Fast transforms from group algebra splittings

The isomorphism $ℂ C_{n} ≅ ℂ^{n}$ is the discrete Fourier transform; the general $ℂ G ≅ \prod M_{n_{i}} (ℂ)$ is the non-abelian Fourier transform underlying fast convolution algorithms on finite groups.

Coding theory

Group codes and idempotent generators

When $| G |$ is invertible in $𝔽_{q}$ the group algebra splits into a product, and every group code is generated by an idempotent computable from the splitting. When $char 𝔽_{q}$ divides $| G |$ the code is not generated by an idempotent, which is exactly the modular case.

Physics and chemistry

Symmetry-adapted bases

Projection operators onto irreducible components are averaging operators of the Maschke type. Molecular vibration analysis and crystal-field splitting both compute with them; characteristic zero guarantees they exist.

Invariant theory

The Reynolds operator

Averaging over a finite group produces a projection onto invariants, giving finite generation of invariant rings for finite groups in characteristic zero. Modular invariant theory is hard precisely because this projection disappears.

The theorem is also the reason character theory works. Over $ℂ$ semisimplicity makes a module determined up to isomorphism by its character; in characteristic $p$ dividing $| G |$ , Brauer characters recover only the composition factors.

Computational Notes

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

Deciding semisimplicity is trivial once $| G |$ and $char k$ are known — it is a divisibility test, not an algebraic computation.
Producing the Wedderburn decomposition is the real work. Over $ℚ$ or a number field the components are matrix rings over division algebras, and the standard tool is GAP's Wedderga package, which computes the decomposition from strong Shoda pairs rather than from characters.
Over an algebraically closed field of characteristic $0$ the decomposition is read off from the character table: the block sizes are the irreducible degrees, and $\sum n_{i}^{2} = | G |$ is the check.
In the modular case the analogous task is computing $rad (k G)$ and the simple modules; the Meataxe algorithm, which splits a module by finding a singular element of the acting algebra, is the workhorse and is implemented in GAP, Magma and the standalone C Meataxe.
Dimension is the practical constraint: $\dim_{k} k G = | G |$ , so a direct matrix approach costs $O (| G |^{3})$ field operations and becomes infeasible well before groups of interest to representation theorists.

Failure Modes and Common Mistakes

Do not conclude from failure of Maschke that $k G$ has large radical. For $G$ finite with $p ∣ | G |$ the radical is nonzero, but it may be small — its dimension is $| G |$ minus the dimension of the semisimple quotient.
The converse direction needs $G$ finite in an essential way: Cauchy's theorem is applied to produce an element of order $p$ .
Do not assume the number of simple modules is the number of conjugacy classes in characteristic $p$ ; it is the number of $p$ -regular classes, and for $𝔽_{3} S_{3}$ that is $2$ rather than $3$ .
The relative version needs the index to be invertible, not the order of $G$ ; this is what makes it useful for infinite groups with a finite-index subgroup.

Historical Notes and Lessons Learned

1896–97Frobenius invents charactersFrobenius factorises the group determinant and creates the character theory of finite groups, working entirely with complex representations.
1898Maschke's theoremMaschke proves complete reducibility for complex representations of a finite group, the statement that now bears his name.
Early 1900sDickson's extensionDickson observes that the argument needs only that the characteristic of the base field does not divide the group order, and that the conclusion genuinely fails otherwise.
1920sNoether's reformulationNoether recasts representations as modules over the group ring, so that complete reducibility becomes semisimplicity and Wedderburn's structure theory applies directly. Many results of Frobenius and Schur are re-derived ring-theoretically.
1935 onwardsBrauer and the modular theoryBrauer develops the representation theory of finite groups over fields of characteristic dividing the group order — blocks, defect groups, decomposition matrices — the systematic study of what Maschke's theorem excludes.
1945Jacobson's radicalWith semisimplicity settled for finite groups, the Jacobson radical supplies a usable weaker notion for infinite groups, and the J-semisimplicity problem for group algebras opens.

The methodological lesson is that the theorem's content is an averaging construction, and averaging is available whenever an index is invertible. Every later generalisation — relative projectivity, Higman's criterion, cohomological vanishing for finite groups with invertible order — is a refinement of the same division by $| G |$ .

Quick Reference

Statement $(6.1)$

G

finite:

k G

semisimple

iff

k

semisimple and

| G | \cdot 1_{k} \in U (k)

Field version

k G

semisimple

iff char k ∤ | G |

Averaging

g (v) = | G |^{- 1} \sum_{σ} σ f (σ^{- 1} v)

Lemma $(6.2)$

r (1 - σ) = 0 iff r \in k G \hat{σ}

σ

of order

p

Necessity enginevon Neumann regularity at

1 - σ

, then apply

ε

Complex form

ℂ G ≅ \prod M_{n_{i}} (ℂ)

\sum n_{i}^{2} = | G |

r =

number of classes

Relative form

[G : H] = i

invertible and

V

semisimple over

k H

\Rightarrow

V

semisimple over

k G

Failure witness

{\hat{G}}^{2} = | G | \hat{G} = 0

when

char k

divides

| G |

What each hypothesis buys
Hypothesis	Used for	What fails without it
$G$ finite	the sum in the averaging operator; Cauchy's theorem	no averaging, and semisimplicity fails outright
$k$ semisimple	existence of a $k$ -linear retraction $f$	no starting projection to average
$\| G \| \cdot 1_{k} \in U (k)$	the division by $\| G \|$	$\hat{G}$ becomes a square-zero element and $rad (k G) \neq 0$
$σ$ of prime order	the annihilator lemma $(6.2)$	the annihilator is no longer a single principal left ideal of this shape

Frequently Asked Questions

Why does Lam state Maschke's theorem for an arbitrary ring

k

rather than a field?

Because the two hypotheses then separate cleanly. Over a field, semisimplicity of $k$ is automatic and the theorem looks like a single arithmetic condition; over a general ring one sees that the averaging argument needs a $k$ -linear retraction to exist — which is what semisimplicity of $k$ provides — and separately needs $| G |$ to be invertible. The general statement also covers matrix rings over division rings as coefficients.

Where exactly does the proof break when

char k

divides

| G |

At the division. The sum $\sum_{σ} σ f (σ^{- 1} v)$ is still defined and still $G$ -equivariant, but its restriction to $W$ is multiplication by $| G |$ , which is $0$ rather than invertible. The construction produces the zero map instead of a retraction.

Is the invertibility of

| G |

really necessary, or just convenient?

Necessary. The group sum $\hat{G}$ satisfies ${\hat{G}}^{2} = | G | \hat{G}$ , so if $| G | \cdot 1_{k} = 0$ then $k G \hat{G}$ is a nonzero ideal of square zero and $rad (k G) \neq 0$ . Over a general ring the full necessity argument goes through von Neumann regularity and the lemma on the annihilator of $1 - σ$ .

What is the analogue of Maschke's theorem for compact groups?

Averaging with respect to normalised Haar measure replaces the finite sum, and every finite-dimensional continuous representation of a compact group over $ℂ$ is completely reducible. The structural analogue of the Rickart argument for infinite discrete groups uses the same measure-theoretic idea in the Banach algebra setting.

Does semisimplicity of

k G

tell you the number of simple modules?

It tells you there are finitely many and that $k G$ is their matrix-ring product. The count is the number of conjugacy classes only when $k$ is a splitting field of characteristic zero. Over $ℚ$ the count can be smaller because Galois-conjugate complex representations fuse into a single rational component.

How does the relative version help with infinite groups?

It replaces $| G |$ by an index $[G : H]$ . If $G$ is infinite but has a subgroup $H$ of finite invertible index, semisimplicity of a module over $k H$ still transfers to $k G$ . This does not make $k G$ semisimple — that is impossible for infinite $G$ — but it does transfer semisimplicity of individual modules.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §6, (6.1)–(6.2) (pp. 82–85).
C. W. Curtis and I. Reiner, Representation Theory of Finite Groups and Associative Algebras, Wiley-Interscience, 1962, Chapter IV.
D. S. Passman, The Algebraic Structure of Group Rings, Wiley-Interscience, 1977, Chapter 2.
J.-P. Serre, Linear Representations of Finite Groups, Graduate Texts in Mathematics 42, Springer-Verlag, 1977, Part I.
W. Feit, The Representation Theory of Finite Groups, North-Holland, 1982, Chapter I.
H. Maschke, “Beweis des Satzes, dass diejenigen endlichen linearen Substitutionsgruppen, in welchen einige durchgehends verschwindende Coefficienten auftreten, intransitiv sind”, Mathematische Annalen 52 (1899).

AI Suggested Questions

Decompose $ℚ Q_{8}$ explicitly and identify the quaternion division algebra component.
State Higman's criterion for relative projectivity and explain how it generalises the averaging argument.
Compute the Cartan matrix and decomposition matrix of $𝔽_{3} S_{3}$ .
How does Brauer's splitting field theorem bound the field extension needed to split $k G$ ?
Give the Haar measure version of Maschke's theorem for compact topological groups and identify where compactness is used.
For which finite groups $G$ and primes $p$ is $𝔽_{p} G$ a local ring?
Explain why modular invariant theory loses finite generation results that hold in characteristic zero.

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Averaging: turning a k-splitting into a kG-splitting

Why the converse needs the augmentation

What replaces the theorem in the modular case

06Key Results

07Proof Techniques and Method

Symmetrise, then divide

Push an identity through ε

Use regularity instead of idempotents

08Worked Example

ℚS3: the semisimple side

𝔽3S3: the modular side

09Process and Workflow

10Comparison and Classification

11Applications and Industry Use

Fast transforms from group algebra splittings

Group codes and idempotent generators

Symmetry-adapted bases

The Reynolds operator

12Computational Notes

13Failure Modes and Common Mistakes

14Historical Notes and Lessons Learned

15Quick Reference

16Frequently Asked Questions

17Related KEVOS Topics

18References

19AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Averaging: turning a $k$ -splitting into a $k G$ -splitting

Key Results

Proof Techniques and Method

Push an identity through $ε$

Worked Example

$ℚ S_{3}$ : the semisimple side

$𝔽_{3} S_{3}$ : the modular side

Process and Workflow

Comparison and Classification

Applications and Industry Use

Computational Notes

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions