Skew Group Rings and Crossed Products

Q: Is k * G ever a division ring?

For G finite and k a field with G acting faithfully, never once |G| > 1: the ring is M_n(F) with n = |G| > 1. Twisting can change this — cyclic algebras with a suitable non-norm parameter are division rings. For infinite G the untwisted case can also be a division ring, for example k((x;)) arises from an infinite cyclic action after completing.

Q: How does one compute the centre of k * G?

An element _g a_g u_g is central iff it commutes with every b ∈ k and with every u_h. The first condition gives a_g b = _g(b) a_g for each g, which kills every g whose _g is not inner on the support of a_g. The second condition imposes a twisted conjugacy invariance. For k a field with G finite and faithful, only g = 1 survives and the centre is the fixed field k^G.

Q: Does the radical of k * G relate simply to the radical of k?

Not in general. One always has that rad(k) (k*G) is contained in a nil ideal when G is finite, but equality of rad(k*G) with rad(k)(k*G) fails; the modular group ring F_pG with |G| = p has zero coefficient radical and nonzero radical. Determining the radical of a group ring is a research subject in its own right.

Executive Summary

A group ring $k G$ records a group inside a ring but keeps the two ingredients apart: group elements commute with coefficients. The skew group ring $k * G$ removes that separation. If $G$ acts on $k$ by ring automorphisms $σ_{g}$ , one adjoins symbols $u_{g}$ subject to $u_{g} b = σ_{g} (b) u_{g}$ , so that inside $k * G$ the action becomes conjugation: $σ_{g} (b) = u_{g} b u_{g}^{- 1}$ .

Allowing the symbols to multiply with a twist, $u_{g} u_{h} = τ (g, h) u_{g h}$ with $τ (g, h)$ a unit of $k$ , gives the crossed product. Two rings with the same $k$ , the same $G$ and the same action can then be genuinely different: over $ℂ$ with the Galois group of $ℂ / ℝ$ , the untwisted choice gives $M_{2} (ℝ)$ and the twist $τ = - 1$ gives Hamilton's quaternions.

$u_{g} b = σ_{g} (b) u_{g}$ Defining relation

$| G |$ Rank as a free

k

-module

$H^{2} (G, K^{\times})$ Classifies the twists

(1.11)Lam's numbering

Overview

Let $k$ be a ring and let $G$ be a group acting on $k$ as a group of ring automorphisms, that is, a group homomorphism $G \to Aut (k)$ written $g \mapsto σ_{g}$ . The skew group ring is the free left $k$ -module on a copy of $G$ ,

k * G = ⨁_{g \in G} k u_{g}, (a u_{g}) (b u_{h}) = a σ_{g} (b) u_{g h},

(1.11)

Associativity is exactly the statement that $g \mapsto σ_{g}$ is a homomorphism.

Taking the trivial action recovers the ordinary group ring $k G$ described in Group Rings and Semigroup Rings. Taking $G$ infinite cyclic recovers the skew Laurent polynomial ring $k [x, x^{- 1}; σ]$ of Skew Polynomial Rings and Hilbert's Twist. So the construction is a common generalisation of two families already in the collection, and it is a genuinely useful one: even ordinary group rings decompose as skew group rings when the group is a semidirect product.

The crossed product loosens one further constraint. If the symbols are allowed to multiply up to a unit of $k$ , then $g \mapsto σ_{g}$ need only be a homomorphism modulo inner automorphisms. The bookkeeping is the theory of factor sets, and over a Galois extension of fields it is exactly group cohomology in degree two.

Learning Objectives

Write down the multiplication in $k * G$ and verify associativity from the action axioms.
State the two conditions on $(σ, τ)$ that make a crossed product associative.
Prove $k * ⟨ σ ⟩ ≅ k [x, x^{- 1}; σ]$ for $⟨ σ ⟩$ infinite cyclic.
Prove $k G ≅ (k T) * H$ when $G = T ⋊ H$ .
Prove that $K * G ≅ M_{n} (F)$ for a Galois extension $K / F$ with group $G$ of order $n$ .
State Maschke's theorem for crossed products with its invertibility hypothesis and identify where it is used.

Definitions

Definition(1.11)Skew group ring

Let $k$ be a ring and $G$ a group together with a homomorphism $G \to Aut (k)$ , $g \mapsto σ_{g}$ . The skew group ring $k * G$ is the free left $k$ -module with basis ${u_{g} : g \in G}$ , with multiplication defined on the basis by $(a u_{g}) (b u_{h}) = a σ_{g} (b) u_{g h}$ and extended additively. Its identity is $1 u_{1}$ , and $k \to k * G$ , $a \mapsto a u_{1}$ , is an injective ring homomorphism.

Definition—Crossed product

Let $k$ be a ring, $G$ a group, $σ : G \to Aut (k)$ any map, and $τ : G \times G \to U (k)$ any map. The crossed product $k *_{σ}^{τ} G$ is the free left $k$ -module on ${u_{g} : g \in G}$ with

u_{g} b = σ_{g} (b) u_{g} (b \in k), u_{g} u_{h} = τ (g, h) u_{g h} .

This multiplication is associative if and only if, for all $a \in k$ and $g, h, l \in G$ ,

σ_{g} (σ_{h} (a)) = τ (g, h) σ_{g h} (a) τ (g, h)^{- 1}, τ (g, h) τ (g h, l) = σ_{g} (τ (h, l)) τ (g, h l) .

(C)

The first condition says $σ$ is a homomorphism only up to inner automorphisms; the second is the twisted 2-cocycle identity.

$k * G$: Skew group ring: the crossed product with $τ \equiv 1$ . Then $(C)$ reduces to $σ$ being a group homomorphism.
$k^{τ} G$: Twisted group ring: the crossed product with $σ \equiv id$ . Requires $τ$ to take values in the centre and to be an ordinary 2-cocycle.
$k G$: Ordinary group ring: both $σ$ and $τ$ trivial.
$k^{G}$: The fixed ring ${a \in k : σ_{g} (a) = a for all g}$ , a subring of $k$ and the natural base for the whole construction.
Support: For $r = \sum_{g} a_{g} u_{g}$ , the finite set ${g : a_{g} \neq 0}$ . Arguments about crossed products are almost always inductions on support size.

The elements $u_{g}$ are units of the crossed product, with $u_{g}^{- 1} = σ_{g}^{- 1} (τ (g, g^{- 1})^{- 1}) u_{g^{- 1}}$ up to the normalisation $τ (1, 1) = 1$ , which can always be arranged.

Core Concepts

Making an outer automorphism inner

An automorphism $σ$ of $k$ is inner if $σ (b) = c b c^{- 1}$ for a fixed unit $c \in U (k)$ . Most interesting automorphisms are not: complex conjugation on $ℂ$ is outer, since $ℂ$ is commutative and every inner automorphism of a commutative ring is the identity. The skew group ring supplies the missing conjugator by force. Inside $k * G$ one has $u_{g} b u_{g}^{- 1} = σ_{g} (b)$ , so the action of $G$ has become a group of inner automorphisms of the larger ring.

This is the same manoeuvre as the skew polynomial ring, which makes an endomorphism into a one-sided conjugation, and it explains why the two constructions coincide for $G$ infinite cyclic.

Grading, and how to recognise a crossed product

Every crossed product is $G$ -graded: $R = ⨁_{g} R_{g}$ with $R_{g} = k u_{g}$ and $R_{g} R_{h} = R_{g h}$ . The homogeneous component of degree $g$ contains the unit $u_{g}$ . That property characterises the construction.

The factor set is only defined up to a coboundary

Rescaling the basis, $u_{g} \to c_{g} u_{g}$ with $c_{g} \in U (k)$ , leaves the ring unchanged but replaces $τ (g, h)$ by $c_{g} σ_{g} (c_{h}) τ (g, h) c_{g h}^{- 1}$ . Over a Galois extension $K / F$ with abelian coefficient field this is exactly the coboundary relation, so the isomorphism class of the crossed product depends on the class of $τ$ in $H^{2} (G, K^{\times})$ and not on $τ$ itself.

k G

k * G

k *_{σ}^{τ} G

G

-graded with units in each degree

Why the fixed ring matters

The centre of $k * G$ contains $k^{G} \cap Z (k)$ , and for $k$ a field with $G$ finite and faithful the centre is exactly the fixed field $k^{G}$ . The crossed product is then a $k^{G}$ -algebra of dimension $| G |^{2}$ , which is why crossed products over Galois extensions are candidates for central simple algebras of a prescribed degree.

Key Results

Proposition(1.11a)Infinite cyclic groups give skew Laurent polynomials

Let $k$ be a ring and let $G = ⟨ σ ⟩$ be an infinite cyclic group acting on $k$ , the generator acting by the automorphism $σ \in Aut (k)$ . Then $k * G ≅ k [x, x^{- 1}; σ]$ , the skew Laurent polynomial ring, by an isomorphism fixing $k$ and sending $u_{σ} \mapsto x$ .

Proof

Both rings are free left $k$ -modules: $k * G$ on ${u_{σ^{n}} : n \in ℤ}$ and $k [x, x^{- 1}; σ]$ on ${x^{n} : n \in ℤ}$ . Define the $k$ -linear bijection $φ (a u_{σ^{n}}) = a x^{n}$ . It respects multiplication because both sides obey the same commutation rule: in $k * G$ we have $u_{σ^{n}} b = σ^{n} (b) u_{σ^{n}}$ , and in the skew Laurent ring $x^{n} b = σ^{n} (b) x^{n}$ . Hence

φ ((a u_{σ^{m}}) (b u_{σ^{n}})) = φ (a σ^{m} (b) u_{σ^{m + n}}) = a σ^{m} (b) x^{m + n} = (a x^{m}) (b x^{n}),

and $φ (1) = 1$ . Note that $σ$ must be an automorphism, not merely an endomorphism, for negative powers to make sense — the same hypothesis the skew Laurent construction requires. $□$

Proposition(1.11b)Group rings of semidirect products

Let $k$ be a ring and let $G = T ⋊ H$ be a semidirect product of a normal subgroup $T$ with a complement $H$ . Let $H$ act on the group ring $k T$ by extending the conjugation action of $H$ on $T$ linearly, so that $h \cdot \sum_{t} a_{t} t = \sum_{t} a_{t} (h t h^{- 1})$ . Then $k G ≅ (k T) * H$ , the skew group ring for that action.

Proof

Every element of $G$ is uniquely $t h$ with $t \in T$ , $h \in H$ , because $T \cap H = 1$ and $T H = G$ . Hence $k G = ⨁_{h \in H} (k T) h$ as a left $k T$ -module, free on ${h : h \in H}$ . Inside $k G$ ,

h (\sum_{t \in T} a_{t} t) = (\sum_{t \in T} a_{t} h t h^{- 1}) h = (h \cdot textstyle \sum_{t} a_{t} t) h,

which is precisely the skew group ring relation $u_{h} b = σ_{h} (b) u_{h}$ with $b \in k T$ . Since $h h' \in H$ with no twist, the factor set is trivial, and the $k T$ -linear map sending $h$ to $u_{h}$ is a ring isomorphism $k G \to (k T) * H$ . $□$

Theorem—Galois skew group rings are matrix rings

Let $K$ be a field and $G$ a finite group of order $n$ acting faithfully on $K$ by field automorphisms, and let $F = K^{G}$ be the fixed field. Then $[K : F] = n$ , the extension $K / F$ is Galois with group $G$ , and the skew group ring satisfies

K * G ≅ {End}_{F} (K) ≅ M_{n} (F) .

(G)

In particular $K * G$ is a simple artinian $F$ -algebra of dimension $n^{2}$ with centre $F$ , and it is never a division ring for $n > 1$ .

Proof

That $K / F$ is Galois with group $G$ and $[K : F] = n$ is Artin's theorem on fixed fields. Define $φ : K * G \to {End}_{F} (K)$ on the $K$ -basis by $φ (a u_{g}) (y) = a g (y)$ for $y \in K$ ; each $φ (a u_{g})$ is $F$ -linear because $g$ fixes $F$ pointwise.

$φ$ is a ring homomorphism: for $y \in K$ ,

φ (a u_{g}) φ (b u_{h}) (y) = a g (b h (y)) = a g (b) (g h) (y) = φ (a g (b) u_{g h}) (y) = φ ((a u_{g}) (b u_{h})) (y) .

$φ$ is injective: if $\sum_{g} a_{g} u_{g}$ lies in the kernel then $\sum_{g} a_{g} g (y) = 0$ for every $y \in K$ , and Dedekind's lemma on the linear independence of distinct field automorphisms over $K$ forces every $a_{g} = 0$ .

Finally $\dim_{F} (K * G) = n \cdot [K : F] = n^{2} = \dim_{F} {End}_{F} (K)$ , so the injection is onto. Choosing an $F$ -basis of $K$ identifies ${End}_{F} (K)$ with $M_{n} (F)$ . $□$

Theorem—Maschke's theorem for crossed products

Let $R = k *_{σ}^{τ} G$ be a crossed product with $G$ finite of order $n$ , and suppose $n \cdot 1$ is invertible in $k$ . Let $V$ be a left $R$ -module and $W \subseteq V$ an $R$ -submodule that is a direct summand of $V$ as a $k$ -module. Then $W$ is a direct summand of $V$ as an $R$ -module. Consequently, if $k$ is a semisimple ring and $n \cdot 1 \in U (k)$ , then $R$ is semisimple.

Proof

Let $π : V \to W$ be a $k$ -linear projection with $π |_{W} = id$ . Each $u_{g}$ is a unit of $R$ , so we may average:

π' (v) = n^{- 1} \sum_{g \in G} u_{g} π (u_{g}^{- 1} v) (v \in V) .

The scalar $n^{- 1}$ is central in $R$ , being a value of the prime subring. For $w \in W$ each $u_{g}^{- 1} w \in W$ , so $π (u_{g}^{- 1} w) = u_{g}^{- 1} w$ and $π' (w) = n^{- 1} \sum_{g} w = w$ ; also $π' (V) \subseteq W$ because $W$ is an $R$ -submodule.

$π'$ is $R$ -linear. For $b \in k$ , using $u_{g}^{- 1} b = σ_{g}^{- 1} (b') u_{g}^{- 1}$ for the appropriate $b' \in k$ and the $k$ -linearity of $π$ , one gets $π' (b v) = b π' (v)$ . For a basis unit $u_{h}$ , substitute $w_{g} = u_{h}^{- 1} u_{g}$ , which runs over a set of homogeneous units indexed bijectively by $G$ as $g$ does; then $\sum_{g} u_{g} π (u_{g}^{- 1} u_{h} v) = u_{h} \sum_{g} w_{g} π (w_{g}^{- 1} v)$ , so $π' (u_{h} v) = u_{h} π' (v)$ .

Thus $π'$ is an $R$ -linear projection onto $W$ and $V = W \oplus \ker π'$ . For the consequence: if $k$ is semisimple, every $k$ -submodule is a $k$ -direct summand, so every $R$ -submodule of every $R$ -module is an $R$ -direct summand, which is one of the equivalent definitions of semisimplicity. $□$

Remark—Where the hypothesis bites

The invertibility of $n \cdot 1$ cannot be dropped. Over $k = 𝔽_{p}$ with $G$ of order $p$ acting trivially, $R = 𝔽_{p} G ≅ 𝔽_{p} [t] / (t^{p} - 1) = 𝔽_{p} [t] / ((t - 1)^{p})$ is local with nonzero nilpotent radical, hence not semisimple. Modular representation theory is the study of exactly this failure.

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Freeness plus a commutation rule

To identify two crossed products, exhibit a $k$ -linear bijection on the distinguished bases and check the single relation $u_{g} b = σ_{g} (b) u_{g}$ . Multiplicativity on general elements then follows by bilinearity.

Move 2

Average over the group

Every $u_{g}$ is a unit, so $v \mapsto n^{- 1} \sum_{g} u_{g} π (u_{g}^{- 1} v)$ converts a $k$ -linear map into an $R$ -linear one. This is Maschke's argument and needs only $n \cdot 1 \in U (k)$ .

Move 3

Independence of characters

Distinct automorphisms of a field are linearly independent over it. This is what makes the representation $K * G \to {End}_{F} (K)$ injective, and it is the standard route from a group action to a faithful matrix model.

A fourth technique, used constantly in the infinite case, is induction on support size: to prove a nonzero ideal of $k * G$ meets $k$ , choose an element of minimal support and multiply it by suitable $u_{g}$ and coefficients to shorten the support, deriving a contradiction unless the support has one element. This is exactly the argument used for simplicity of skew Laurent rings.

Worked Example

The same action, two factor sets, two rings

Take $K = ℂ$ , $F = ℝ$ , and $G = {1, σ}$ with $σ$ complex conjugation. Both rings below are $4$ -dimensional $ℝ$ -algebras with basis ${1, i, u, i u}$ and the relation $u z = \bar{z} u$ for $z \in ℂ$ . They differ only in the value of $u^{2} = τ (σ, σ)$ .

Crossed products of $ℂ$ by $Gal (ℂ / ℝ)$
Factor set	Relation	Ring	Reason
$τ (σ, σ) = 1$	$u^{2} = 1$	$M_{2} (ℝ)$	$e = \frac{1}{2} (1 + u)$ is a nontrivial idempotent
$τ (σ, σ) = - 1$	$u^{2} = - 1$	$ℍ$	set $u = j$ ; every nonzero element is invertible

Check the idempotent in the untwisted case: $e^{2} = \frac{1}{4} (1 + 2 u + u^{2}) = \frac{1}{4} (2 + 2 u) = \frac{1}{2} (1 + u) = e$ , and $e \neq 0, 1$ . So $ℂ * G$ has a nontrivial idempotent and cannot be a division ring; it is $4$ -dimensional over its centre $ℝ$ , hence isomorphic to $M_{2} (ℝ)$ — in agreement with the Galois theorem $(G)$ .

Verify the centre in that case directly. If $a + b u$ is central with $a, b \in ℂ$ , commuting with $z = i$ gives $b \bar{i} = i b$ , so $- b i = b i$ and $b = 0$ ; commuting with $u$ gives $a = \bar{a}$ , so $a \in ℝ$ . Hence $Z (ℂ * G) = ℝ$ as predicted.

The twisted case is Hamilton's algebra

With $u^{2} = - 1$ , write $j = u$ and $k = i j$ . Then $j i = \bar{i} j = - i j$ , so $i j = - j i$ , and $j^{2} = - 1$ , $i^{2} = - 1$ . These are exactly Hamilton's relations: the crossed product is $ℍ$ . The two rings represent the two classes of $Br (ℂ / ℝ) ≅ H^{2} (ℤ / 2 ℤ, ℂ^{\times}) ≅ ℝ^{\times} / N_{ℂ / ℝ} (ℂ^{\times}) ≅ ℤ / 2 ℤ$ .

An infinite example

Take $k = ℚ (t)$ and $G = ⟨ σ ⟩ ≅ ℤ$ with $σ (t) = 2 t$ . Then $k * G ≅ ℚ (t) [x, x^{- 1}; σ]$ , in which $x t x^{- 1} = 2 t$ . No nonzero power of $σ$ is the identity, and this is exactly the condition making the ring simple — see Simplicity of Skew Laurent Rings.

Frameworks and Models

The constructions form a lattice determined by which of the two twists is present.

$G$ -graded rings — $R = ⨁_{g} R_{g}$ with $R_{g} R_{h} \subseteq R_{g h}$
- strongly graded — $R_{g} R_{h} = R_{g h}$
  - crossed products — each $R_{g}$ contains a unit of $R$
    - skew group rings $k * G$ — trivial factor set
    - twisted group rings $k^{τ} G$ — trivial action
    - group rings $k G$ — both trivial
    - cyclic algebras $(K / F, σ, a)$ — $G$ cyclic, $K / F$ Galois

Degenerate case

Trivial group

$k * {1} = k$ . Nothing is gained; the construction is only interesting when the action has elements acting nontrivially.

Degenerate case

Inner action

If every $σ_{g}$ is inner, say $σ_{g} =$ conjugation by $c_{g}$ , then rescaling $u_{g} \to c_{g}^{- 1} u_{g}$ turns $k * G$ into a twisted group ring. Only the outer part of the action produces something genuinely new.

Leading case

Cyclic algebras

$G = ⟨ σ ⟩$ of order $n$ , $K / F$ cyclic Galois. Then $τ$ is determined by the single element $a = u^{n} \in F^{\times}$ , and the algebra $(K / F, σ, a)$ is central simple of degree $n$ over $F$ .

Limit of the method

Noncrossed products

Not every central division algebra is a crossed product. Amitsur constructed the first examples in 1972, closing a question open since Noether's lectures.

Comparison and Classification

Which construction has which feature
	$k G$	$k * G$	$k^{τ} G$	$k *_{σ}^{τ} G$
Free left $k$ -module of rank $\| G \|$	yes	yes	yes	yes
$G$ embeds as a subgroup of the units	yes	yes	no	no
$k$ is central when $k$ is commutative	yes	no	yes	no
Maschke applies when $\| G \|^{- 1} \in k$	yes	yes	yes	yes
Determined by cohomological data	no	no	yes	yes

Which construction has which feature

Worked identifications
Data	Crossed product	Identification
$G = ⟨ σ ⟩ ≅ ℤ$ acting on $k$	$k * G$	$k [x, x^{- 1}; σ]$ , skew Laurent polynomials
$G = T ⋊ H$ , trivial action on $k$	$(k T) * H$	$k G$ , an ordinary group ring
$K / F$ Galois with group $G$ , $\| G \| = n$	$K * G$	${End}_{F} (K) ≅ M_{n} (F)$
$ℂ / ℝ$ , $τ (σ, σ) = - 1$	$ℂ *^{τ} G$	$ℍ$ , the real quaternions
$K / F$ cyclic of degree $n$ , $u^{n} = a \in F^{\times}$	$(K / F, σ, a)$	cyclic algebra, central simple of degree $n$

Relationship Map

The construction sits at the junction of three pages already in the collection.

action of

G

k

k * G

G

acts by inner automorphismsstructure of

k * G

from that of

k

and

G

$k * G$ — specialises to
- $G$ infinite cyclic
  - skew Laurent polynomial ring — Simplicity of Skew Laurent Rings
  - and, dropping negative powers, $k [x; σ]$ — Skew Polynomial Rings and Hilbert's Twist
- $G$ acting trivially
  - ordinary group ring — Group Rings and Semigroup Rings
- $k$ a Galois extension field
  - matrix algebra over the fixed field, one instance of the Wedderburn–Artin classification
  - with a twist, cyclic algebras — Cyclic Algebras

In the other direction, crossed products consume the general theory: Maschke's theorem uses semisimplicity, the Galois identification is a Wedderburn–Artin statement, and the radical of a crossed product is a research topic in its own right.

Applications and Industry Use

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

Division algebras

Brauer groups and number theory

Crossed products over Galois extensions realise every class of $Br (K / F)$ , and the isomorphism $Br (K / F) ≅ H^{2} (G, K^{\times})$ turns the classification of division algebras into a cohomology computation. Over number fields this is the Brauer–Hasse–Noether theorem.

Coding theory

Space–time block codes

Cyclic division algebras — crossed products over cyclic Galois extensions — supply the algebraic backbone of full-rate, full-diversity space–time codes for multiple-antenna wireless channels. The non-norm condition guaranteeing that the algebra is a division ring is exactly what guarantees full diversity.

Operator algebras

Crossed product C*-algebras

The analytic analogue, a group acting on a C*-algebra, produces the crossed product C*-algebras at the centre of noncommutative geometry; the algebraic construction on this page is its purely ring-theoretic skeleton.

Internal to algebra

Reduction machinery

The identity $k G ≅ (k T) * H$ for $G = T ⋊ H$ is a workhorse: it reduces questions about a group ring to questions about a smaller group ring plus an action, and underlies much of Passman's structure theory for infinite group rings.

The honest summary: outside coding theory and operator algebras the applications are internal. Crossed products are the standard device for building an algebra with prescribed centre, prescribed dimension and prescribed splitting behaviour — they are a construction kit rather than an object of independent interest.

Standards and Notation

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

Skew group ring

k * G

(Lam, Passman); some authors write

k ⋊ G

k # G

Crossed product

k *_{σ}^{τ} G

, or

k * G

with the twist left implicit

Basis elements

u_{g}

\bar{g}

; Lam writes group elements directly and lets the context supply the twist

Cyclic algebra

(K / F, σ, a)

, older sources

(a, K / F, σ)

; the quaternion case is

(a, b)_{F}

Fixed ring

k^{G}

; the trace map is

tr (a) = \sum_{g} σ_{g} (a)

Action side

σ_{g} (a)

^{g} a

; left actions are standard, but right-action conventions occur in the group ring literature

GAPGroupRing for the untwisted case; twisted group algebras via TwistedGroupAlgebra in the Wedderga package

Magma / SageCyclicAlgebra and Brauer group machinery in Magma; Sage supports group algebras and quaternion algebras

Failure Modes and Common Mistakes

A crossed product with $k$ a field and $G$ finite need not be simple unless the action is faithful; a kernel of the action produces a group ring factor, which is semisimple only under Maschke's hypothesis.
Two factor sets differing by a coboundary give isomorphic rings, but non-cohomologous factor sets can still give isomorphic rings when the isomorphism does not preserve the grading. The cohomology class classifies graded isomorphism.
$k * G$ is free as a left $k$ -module, but the induced right $k$ -module structure is twisted; do not assume left and right ranks are computed the same way.
Not every central simple algebra is a crossed product — Amitsur's 1972 examples. Do not use crossed product and *central simple algebra of degree $n$ * interchangeably.

Historical Notes and Lessons Learned

1843Hamilton's quaternionsHamilton constructs $ℍ$ directly. In retrospect it is the crossed product of $ℂ$ by $Gal (ℂ / ℝ)$ with factor set $- 1$ .
1906Dickson's cyclic algebrasDickson introduces algebras generated by a cyclic field extension and one extra element $u$ with $u^{n}$ scalar, the first systematic family of crossed products.
1929–1933Noether's crossed productsEmmy Noether develops the general crossed product with factor sets in her Göttingen lectures; the cohomological interpretation of the Brauer group follows.
1932Brauer–Hasse–NoetherEvery central division algebra over an algebraic number field is a cyclic algebra — the high point of the crossed product method.
1972Amitsur's noncrossed productsAmitsur constructs central division algebras that are not crossed products over any Galois extension, showing the method is not universal.
1977–1989Infinite crossed productsPassman systematises the structure theory of crossed products over infinite groups, with the $X$ -inner automorphisms controlling primeness and simplicity.

The methodological lesson is the same one that recurs throughout structure theory: parameterise the possible twists rather than the objects. Once factor sets are seen to be cocycles, the classification problem becomes a computation in a cohomology group, and non-obvious finiteness results follow for free.

Quick Reference

Skew group ring

k * G = ⨁_{g} k u_{g}

(a u_{g}) (b u_{h}) = a σ_{g} (b) u_{g h}

Key relation

u_{g} b u_{g}^{- 1} = σ_{g} (b)

Crossed product

u_{g} u_{h} = τ (g, h) u_{g h}

τ (g, h) \in U (k)

Associativity

σ_{g} σ_{h} = τ (g, h) σ_{g h} τ (g, h)^{- 1}

and the 2-cocycle identity

Cyclic case

G ≅ ℤ

gives

k [x, x^{- 1}; σ]

Semidirect products

G = T ⋊ H \Rightarrow k G ≅ (k T) * H

Galois case

K / F

Galois,

| G | = n \Rightarrow K * G ≅ M_{n} (F)

Maschke

| G | = n < \infty

n \cdot 1 \in U (k)

k

semisimple

\Rightarrow k * G

semisimple

Recognising the four members of the family
Ring	Action $σ$	Factor set $τ$	$G$ inside the units?
$k G$	trivial	trivial	yes
$k * G$	nontrivial	trivial	yes
$k^{τ} G$	trivial	nontrivial	no
$k *_{σ}^{τ} G$	nontrivial	nontrivial	no

Frequently Asked Questions

What is the difference between a skew group ring and a semidirect product of groups?

They are different levels of structure that interact. A semidirect product $G = T ⋊ H$ is a group; a skew group ring is a ring. The link is the isomorphism $k G ≅ (k T) * H$ : the group-level semidirect product becomes a ring-level skew group ring after applying the group ring functor. The action of $H$ on $T$ becomes an action of $H$ on the ring $k T$ .

Why can the same action give two non-isomorphic rings?

Because the multiplication of the basis elements is extra data. The action fixes how $u_{g}$ commutes past coefficients but says nothing about what $u_{g} u_{h}$ equals. Choosing $u^{2} = 1$ over $ℂ / ℝ$ gives $M_{2} (ℝ)$ , choosing $u^{2} = - 1$ gives $ℍ$ . The possible choices modulo harmless rescalings form the cohomology group $H^{2} (G, K^{\times})$ .

k * G

ever a division ring?

For $G$ finite and $k$ a field with $G$ acting faithfully, never once $| G | > 1$ : the ring is $M_{n} (F)$ with $n = | G | > 1$ . Twisting can change this — cyclic algebras with a suitable non-norm parameter are division rings. For infinite $G$ the untwisted case can also be a division ring, for example $k ((x; σ))$ arises from an infinite cyclic action after completing.

How does one compute the centre of

k * G

An element $\sum_{g} a_{g} u_{g}$ is central iff it commutes with every $b \in k$ and with every $u_{h}$ . The first condition gives $a_{g} b = σ_{g} (b) a_{g}$ for each $g$ , which kills every $g$ whose $σ_{g}$ is not inner on the support of $a_{g}$ . The second condition imposes a twisted conjugacy invariance. For $k$ a field with $G$ finite and faithful, only $g = 1$ survives and the centre is the fixed field $k^{G}$ .

Why is the recognition criterion for crossed products stated with units rather than with a basis?

Because a basis is a choice and the units are intrinsic. Saying that each graded component $R_{g}$ contains a unit of $R$ is a property of the graded ring; choosing one unit per degree then produces $σ$ and $τ$ , and a different choice changes $τ$ by a coboundary. The criterion is therefore checkable without constructing anything.

Does the radical of

k * G

relate simply to the radical of

k

Not in general. One always has that $rad (k) \cdot (k * G)$ is contained in a nil ideal when $G$ is finite, but equality of $rad (k * G)$ with $rad (k) (k * G)$ fails; the modular group ring $𝔽_{p} G$ with $| G | = p$ has zero coefficient radical and nonzero radical. Determining the radical of a group ring is a research subject in its own right.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §1, example (1.11) (pp. 14–15).
D. S. Passman, Infinite Crossed Products, Pure and Applied Mathematics 135, Academic Press, 1989.
D. S. Passman, The Algebraic Structure of Group Rings, Wiley-Interscience, 1977.
R. S. Pierce, Associative Algebras, Graduate Texts in Mathematics 88, Springer-Verlag, 1982, Chapters 12–15 (crossed products and the Brauer group).
S. A. Amitsur, “On central division algebras”, Israel Journal of Mathematics 12 (1972), 408–420.
C. Nastasescu and F. van Oystaeyen, Methods of Graded Rings, Lecture Notes in Mathematics 1836, Springer-Verlag, 2004.

AI Suggested Questions

Derive the coboundary formula for rescaling the basis of a crossed product and verify it defines the same ring.
Prove the recognition criterion: a $G$ -graded ring is a crossed product iff each component contains a unit.
Compute $H^{2} (ℤ / n ℤ, K^{\times})$ for a cyclic Galois extension and read off the cyclic algebras.
When is a cyclic algebra $(K / F, σ, a)$ a division ring rather than a matrix ring?
Sketch Amitsur's construction of a central division algebra that is not a crossed product.
What is an $X$ -inner automorphism, and how does it control primeness of $k * G$ for infinite $G$ ?
Compare the algebraic crossed product with the crossed product C*-algebra of a group acting on a C*-algebra.

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Making an outer automorphism inner

Grading, and how to recognise a crossed product

The factor set is only defined up to a coboundary

Why the fixed ring matters

06Key Results

07Proof Techniques and Method

Freeness plus a commutation rule

Average over the group

Independence of characters

08Worked Example

The same action, two factor sets, two rings

The twisted case is Hamilton's algebra

An infinite example

09Frameworks and Models

Trivial group

Inner action

Cyclic algebras

Noncrossed products

10Comparison and Classification

11Relationship Map

12Applications and Industry Use

Brauer groups and number theory

Space–time block codes

Crossed product C*-algebras

Reduction machinery

13Standards and Notation

14Failure Modes and Common Mistakes

15Historical Notes and Lessons Learned

16Quick Reference

17Frequently Asked Questions

18Related KEVOS Topics

19References

20AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

Worked Example

Frameworks and Models

Comparison and Classification

Relationship Map

Applications and Industry Use

Standards and Notation

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions