Why is (G) preferable to the centre?

Because it is usually much bigger and is still tractable. Infinite groups routinely have trivial centre — free groups and the finitary symmetric group, for instance — while (G) captures every element that commutes with a finite-index subgroup. When G is torsion-free, (G) is abelian, so nothing is lost in tractability.

Is the truncation _ multiplicative?

No, and this is the central technical point. Elements supported outside can multiply to something supported inside it, so (') and ()(') differ in general. What holds is the bimodule identity over k, plus Passman's proposition, which recovers multiplicativity in the one situation the theory needs.

Does (6.28) need any hypothesis on k or G?

None at all. The proposition holds for an arbitrary ring k and an arbitrary group G. Hypotheses enter only when the conclusion is used: to make k a domain one needs k a domain and torsion-free abelian, and the latter comes from torsion-freeness of G via (6.24).

What is ^+(G) good for?

It isolates the torsion of the FC part. Dietzmann's lemma shows it is a subgroup, equal to the union of all finite normal subgroups of G, and (G)/^+(G) is torsion-free abelian. Most structural characterisations of group rings — primeness, semiprimeness — are stated in terms of ^+(G) rather than (G).

FC Groups, the Δ-Subgroup and Trace Methods

Q: Where does Neumann's lemma actually enter?

At the end of the proof of (6.28). Conjugation constraints put the finite-index subgroup C inside a finite union of cosets of the centralisers C_G(v_i), with each v_i outside and hence with centraliser of infinite index. Neumann's lemma says such a covering is impossible, which is the contradiction.

Q: Why does (6.24) reduce to two-generator subgroups?

Commutativity is a statement about pairs. Since subgroups of FC groups are FC, one may replace G by the subgroup generated by the two elements in question, and that subgroup has centre of finite index because it is generated by two elements each with a finite-index centraliser. The finite-index hypothesis of (6.22) is then available.

Executive Summary

Group-ring arguments about zero divisors run into a wall: the support of a product can collapse in ways that are impossible to control element by element. Passman's $Δ$ -method removes the wall by projecting onto the part of $G$ that behaves almost centrally. The subgroup $Δ (G)$ of elements with finitely many conjugates is normal, characteristic, and — when $G$ is torsion-free — abelian.

Three results carry the method. A group with centre of finite index $n$ satisfies $(a b)^{n} = a^{n} b^{n}$ ; a torsion-free FC group is therefore abelian; and Neumann's lemma says a group covered by finitely many cosets of subgroups has one of those subgroups of finite index. Together they prove Passman's proposition that $γ k G γ' = 0$ forces $π (γ) π (γ') = 0$ in $k Δ$ , and hence that a reduced group ring of a torsion-free group over a domain is a domain.

$Δ (G)$ FC centre

$n$ -abelianConsequence of

[G : Z (G)] = n

(6.25)Neumann covering lemma

R $\Rightarrow$ DWhat the method delivers

Overview

Let $G$ be any group. Conjugation partitions $G$ into classes, and the elements whose class is finite form a subgroup:

Δ (G) = {g \in G : [G : C_{G} (g)] < \infty},

(6.23)

The finite conjugate subgroup, also called the FC centre. The index of the centraliser counts the conjugates of $g$ .

That this is a subgroup is a two-line check: for $a, b \in Δ (G)$ and $x \in G$ , $x^{- 1} (a b) x = (x^{- 1} a x) (x^{- 1} b x)$ can take only finitely many values, and inverses are handled the same way. Since the defining condition is invariant under every automorphism, $Δ (G)$ is characteristic, hence normal; it is an FC group, and $Δ (Δ (G)) = Δ (G)$ .

The point of the construction is that $Δ (G)$ sits between $Z (G)$ and $G$ and is much larger than the centre in general, while retaining enough finiteness to be analysed. For the zero-divisor problem $G$ is torsion-free, and then $Δ (G)$ is abelian — a group ring over which is completely understood.

Learning Objectives

Verify that $Δ (G)$ is a characteristic normal FC subgroup of $G$ .
Prove the $n$ -abelian identity for a group with centre of index $n$ .
Prove that a torsion-free FC group is abelian.
Prove Neumann's covering lemma by induction on the number of subgroups.
Show that $π_{Δ}$ is a $k Δ$ -bimodule map detecting nonzero right ideals.
Assemble these into Passman's proposition and the implication that reduced implies domain.

Definitions

Definition(6.23d)FC groups and the

Δ

-subgroup

A group $H$ is an FC group (finite conjugate group) if every element of $H$ has only finitely many conjugates in $H$ . For an arbitrary group $G$ , $Δ (G)$ denotes the set of $g \in G$ with finitely many $G$ -conjugates. Subgroups and quotients of FC groups are FC, and $Δ (G)$ is a normal FC subgroup of $G$ containing $Z (G)$ .

$C_{G} (g)$: The centraliser ${x \in G : x g = g x}$ . The number of conjugates of $g$ equals $[G : C_{G} (g)]$ .
$Δ^{+} (G)$: The set of torsion elements of $Δ (G)$ ; a characteristic subgroup, equal to the union of all finite normal subgroups of $G$ .
$n$ -abelian: The identity $(a b)^{n} = a^{n} b^{n}$ holds identically. It is much weaker than commutativity but strong enough to force commutators to have exponent dividing $n$ in favourable cases.
$π = π_{Δ}$: The truncation $k G \to k Δ$ deleting all terms whose group element lies outside $Δ = Δ (G)$ . It is $k$ -linear and a $(k Δ, k Δ)$ -bimodule map, but not a ring homomorphism.
Right transversal: A set of representatives $g_{i}$ with $G$ the disjoint union of the cosets $Δ g_{i}$ ; it makes $k G$ a free left $k Δ$ -module on the $g_{i}$ .

Here $k$ is an arbitrary ring unless stated otherwise; the $Δ$ -machinery needs no hypothesis on the coefficients until the very last step.

Core Concepts

Almost central elements

An element of $Δ (G)$ is almost central: it commutes with a subgroup of finite index. Almost centrality is exactly the property that survives averaging arguments, because a finite conjugacy class can be summed over. This is why $Δ (G)$ and not $Z (G)$ is the right target — the centre of an infinite group is frequently trivial, while $Δ (G)$ can be large.

The truncation map

Write $Δ = Δ (G)$ . Choosing a right transversal ${g_{i}}$ for $Δ$ in $G$ gives a decomposition of $k G$ as a free left $k Δ$ -module,

k G = ⨁_{i \in I} k Δ \cdot g_{i},

(6.27a)

Each $γ \in k G$ has unique coordinates $α_{i} \in k Δ$ ; the truncation $π$ is the coordinate at $g_{i} = 1$ .

The coordinate functions are computed by $α_{i} = π (γ g_{i}^{- 1})$ , which is why right ideals are detected: multiplying on the right stays inside a right ideal, so if some coordinate of some element is nonzero then $π$ of something in the ideal is nonzero.

Coverings by cosets

Neumann's lemma is the combinatorial engine. If a group is the union of finitely many right cosets $H_{i} a_{i j}$ , then one of the $H_{i}$ has finite index. Contrapositively, a finite family of infinite-index subgroups can never cover $G$ by finitely many cosets — and in Passman's proof, every $v_{i} \notin Δ$ has $C_{G} (v_{i})$ of infinite index, so the covering that the argument produces is impossible.

v \notin Δ (G)

[G : C_{G} (v)] = \infty

no finite coset coveringcontradiction

Why torsion-freeness makes $Δ$ abelian

A finitely generated FC group has centre of finite index, so it is $n$ -abelian for some $n$ ; the $n$ -th power map then kills every commutator, and in a torsion-free group killing an $n$ -th power kills the element. Hence a torsion-free FC group is abelian, and $k Δ$ becomes the group ring of a torsion-free abelian group — a domain, by the ordered-group theorem.

Key Results

Lemma(6.22)Centre of finite index

Let $G$ be a group whose centre $H = Z (G)$ has finite index $n$ in $G$ . Then $G$ is $n$ -abelian: $(a b)^{n} = a^{n} b^{n}$ for all $a, b \in G$ .

Proof

Sketch, via the transfer. For a subgroup $H$ of finite index $n$ there is a canonical homomorphism $V : G \to H / H'$ , the transfer. Here $H = Z (G)$ is abelian, so $H' = {1}$ and $V$ maps into $H$ itself. Evaluating the transfer when $H$ is central shows that $V (a) = a^{n}$ for every $a \in G$ — the coset representatives contribute trivially because they commute with everything in $H$ . Since $V$ is a homomorphism,

(a b)^{n} = V (a b) = V (a) V (b) = a^{n} b^{n} \in H . ■

(6.22a)

Remark

The same hypothesis yields Schur's theorem: if $[G : Z (G)]$ is finite then the commutator subgroup $G'$ is finite. Both statements say that a group which is almost abelian is abelian up to bounded error.

Corollary(6.24)Torsion-free FC groups

Every torsion-free FC group $G$ is abelian.

Proof

Take $x, y \in G$ . Every subgroup of an FC group is FC, so we may replace $G$ by $⟨ x, y ⟩$ and assume $G$ is generated by $x$ and $y$ . Then $Z (G) = C_{G} (x) \cap C_{G} (y)$ , an intersection of two subgroups of finite index, so $[G : Z (G)] = n < \infty$ . By $(6.22)$ , $G$ is $n$ -abelian, and applying the identity to the pair $x^{- 1}$ , $y^{- 1} x y$ gives

(x^{- 1} y^{- 1} x y)^{n} = (x^{- 1})^{n} (y^{- 1} x y)^{n} = x^{- n} y^{- 1} x^{n} y = x^{- n} x^{n} = 1,

(6.24a)

The last step uses $x^{n} \in Z (G)$ , which holds because the transfer takes values in the centre.

Since $G$ is torsion-free, $x^{- 1} y^{- 1} x y = 1$ , that is, $x y = y x$ . $■$

Lemma(6.25)B. H. Neumann's covering lemma

Let $H_{1}, \dots, H_{m}$ be subgroups of a group $G$ and suppose there are finitely many elements $a_{i j} \in G$ $(1 \leq i \leq m, 1 \leq j \leq n)$ with

G = ⋃_{i = 1}^{m} ⋃_{j = 1}^{n} H_{i} a_{i j} .

(6.25a)

Then $[G : H_{i}] < \infty$ for at least one $i$ .

Proof

Induct on $m$ . For $m = 1$ the group is a union of $n$ right cosets of $H_{1}$ , so $[G : H_{1}] \leq n$ .

Let $m \geq 2$ and suppose $[G : H_{1}]$ is infinite. Only finitely many right cosets of $H_{1}$ occur among the $H_{1} a_{1 j}$ , so there is a right coset $H_{1} b$ disjoint from all of them; consequently $H_{1} b \subseteq ⋃_{i \geq 2} ⋃_{j} H_{i} a_{i j}$ . Right multiplication by $b^{- 1} a_{1 j}$ is a bijection carrying $H_{1} b$ onto $H_{1} a_{1 j}$ , so

H_{1} a_{1 j} \subseteq ⋃_{i \geq 2} ⋃_{j'} H_{i} a_{i j'} b^{- 1} a_{1 j} (1 \leq j \leq n) .

(6.25b)

Substituting into $(6.25 a)$ exhibits $G$ as a union of finitely many right cosets of $H_{2}, \dots, H_{m}$ , and the inductive hypothesis gives some $H_{i}$ of finite index. $■$

Proposition(6.27)Properties of the truncation

Let $k$ be any ring, $Δ$ a subgroup of $G$ , and $π = π_{Δ} : k G \to k Δ$ the truncation of $(6.26)$ . Then:

$π (α γ β) = α π (γ) β$ for all $α, β \in k Δ$ and $γ \in k G$ ; that is, $π$ is a homomorphism of $(k Δ, k Δ)$ -bimodules;
if $𝔄$ is a right ideal of $k G$ then $𝔄 \subseteq π (𝔄) \cdot k G$ ; in particular $𝔄 \neq 0$ implies $π (𝔄) \neq 0$ .

Proof

(1) is immediate from the definition: left or right multiplication by an element of $k Δ$ permutes the cosets $Δ g$ and fixes $Δ$ setwise, so truncation commutes with it.

(2) Choose a right transversal ${g_{i}}_{i \in I}$ with $G = ⋃_{i} Δ g_{i}$ disjointly, so that $k G = ⨁_{i} k Δ \cdot g_{i}$ as left $k Δ$ -modules. Given $γ \in 𝔄$ , write $γ = \sum_{i} α_{i} g_{i}$ with $α_{i} \in k Δ$ , almost all zero. Multiplying on the right by $g_{j}^{- 1}$ and truncating gives $α_{j} = π (γ g_{j}^{- 1})$ , and $γ g_{j}^{- 1} \in 𝔄$ because $𝔄$ is a right ideal. Hence every coordinate $α_{j}$ lies in $π (𝔄)$ and $γ \in π (𝔄) \cdot k G$ . $■$

Proposition(6.28)Passman

Let $k$ be any ring and $G$ any group; put $Δ = Δ (G)$ and $π = π_{Δ}$ . If $γ, γ' \in k G$ satisfy $γ k G γ' = 0$ , then $π (γ) π (γ') = 0$ in $k Δ$ .

Proof

By $(6.27) (1)$ it is enough to prove $π (γ) γ' = 0$ , since applying $π$ then gives $0 = π (π (γ) γ') = π (γ) π (γ')$ .

Split $γ = γ_{0} + γ_{1}$ with $γ_{0} = π (γ) = a_{1} u_{1} + \dots + a_{r} u_{r}$ supported in $Δ$ and $γ_{1} = b_{1} v_{1} + \dots + b_{m} v_{m}$ supported off $Δ$ , and write $γ' = c_{1} w_{1} + \dots + c_{n} w_{n}$ . Set $C = ⋂_{ℓ} C_{G} (u_{ℓ})$ , a subgroup of finite index because each $u_{ℓ} \in Δ$ .

Suppose $γ_{0} γ' \neq 0$ , and fix $g \in G$ occurring with nonzero coefficient in $γ_{0} γ'$ . For each pair $(i, j)$ such that $v_{i}$ is conjugate to $g w_{j}^{- 1}$ , fix $g_{i j} \in G$ with $g_{i j}^{- 1} v_{i} g_{i j} = g w_{j}^{- 1}$ ; otherwise set $g_{i j} = 1$ .

For $x \in C$ the hypothesis gives $γ x γ' = 0$ , hence $x^{- 1} γ x γ' = 0$ . Since $x$ commutes with every $u_{ℓ}$ we have $x^{- 1} γ_{0} x = γ_{0}$ , so

γ_{0} γ' = x^{- 1} γ_{0} x γ' = - x^{- 1} γ_{1} x γ' .

(6.28a)

The element $g$ occurs with nonzero coefficient on the left, hence on the right, so $g = x^{- 1} v_{i} x w_{j}$ for some $i, j$ . Then $x^{- 1} v_{i} x = g w_{j}^{- 1} = g_{i j}^{- 1} v_{i} g_{i j}$ , which rearranges to $x g_{i j}^{- 1} \in C_{G} (v_{i})$ , that is, $x \in C_{G} (v_{i}) g_{i j}$ . As $x \in C$ was arbitrary,

C \subseteq ⋃_{i, j} C_{G} (v_{i}) g_{i j} .

(6.28b)

Because $[G : C] < \infty$ , translating $(6.28 b)$ by coset representatives of $C$ covers all of $G$ by finitely many right cosets of $C_{G} (v_{1}), \dots, C_{G} (v_{m})$ . Neumann's lemma $(6.25)$ then forces some $C_{G} (v_{i})$ to have finite index — that is, $v_{i} \in Δ$ , contradicting the choice of $γ_{1}$ . Hence $γ_{0} γ' = 0$ . $■$

Theorem(6.28c)Reduced implies domain

Let $k$ be a domain and $G$ a torsion-free group. If $k G$ is reduced, then $k G$ is a domain. This is the implication $R \Rightarrow D$ of $(6.20)$ .

Proof

Suppose $γ γ' = 0$ with $γ, γ' \neq 0$ . For every $r \in k G$ , $(γ' r γ)^{2} = γ' r (γ γ') r γ = 0$ , so reducedness gives $γ' r γ = 0$ ; that is, $γ' k G γ = 0$ .

By $(6.27) (2)$ applied to the nonzero right ideal $γ' k G$ there is $r$ with $π (γ' r) \neq 0$ , and by the left-handed mirror image of the same statement there is $r'$ with $π (r' γ) \neq 0$ . Since $(γ' r) k G (r' γ) \subseteq γ' k G γ = 0$ , Proposition $(6.28)$ gives $π (γ' r) π (r' γ) = 0$ in $k Δ$ .

But $Δ = Δ (G)$ is FC and torsion-free, hence abelian by $(6.24)$ ; a torsion-free abelian group is orderable, so $k Δ$ is a domain. Two nonzero elements of $k Δ$ cannot multiply to zero — contradiction. Hence no such $γ, γ'$ exist. $■$

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Conjugate by a finite-index subgroup

Elements of $Δ$ are fixed by conjugation by $C = ⋂ C_{G} (u_{ℓ})$ , a subgroup of finite index. Averaging or comparing over $C$ isolates the $Δ$ -part of an element and leaves the rest to be contradicted.

Move 2

Manufacture a coset covering

Every constraint of the form $x^{- 1} v x = w$ says $x$ lies in a coset of $C_{G} (v)$ . Collecting such constraints over all $x$ in a finite-index subgroup produces a covering, and Neumann's lemma turns the covering into a finite-index conclusion.

Move 3

Push down, then compute

Reduce a statement about $k G$ to the same statement in $k Δ$ , where the group is abelian and every question has a known answer. The truncation is not a ring map, so the reduction must be engineered — that is exactly what $(6.28)$ provides.

The pattern generalises far beyond this page. Passman's structure theory of group rings — primeness, semiprimeness, primitivity, the description of central idempotents — is organised around $Δ (G)$ and $Δ^{+} (G)$ , with the same two-step architecture: prove a truncation statement, then analyse the FC group that remains.

Worked Example

The infinite dihedral group

Let $G = ⟨ a, b ∣ b^{2} = 1, b a b^{- 1} = a^{- 1} ⟩$ . Conjugating $a^{n}$ gives only $a^{n}$ and $a^{- n}$ , so every $a^{n}$ lies in $Δ (G)$ . Conjugating the reflection $a^{n} b$ by $a^{m}$ gives $a^{n + 2 m} b$ , infinitely many elements, so no reflection lies in $Δ (G)$ . Hence

Δ (D_{\infty}) = ⟨ a ⟩ ≅ ℤ .

(E.1)

A characteristic subgroup of index $2$ , torsion-free and abelian — as $(6.24)$ requires of any torsion-free FC group.

Two extreme cases

For a free group $F$ of rank at least $2$ , the centraliser of any nontrivial element is an infinite cyclic subgroup of infinite index, so $Δ (F) = {1}$ : the truncation collapses to the coefficient ring and the $Δ$ -method gives nothing beyond what $k$ itself supplies. At the other extreme, if $G = H \times A$ with $H$ finite and $A$ abelian, then every conjugacy class has at most $| H |$ elements, so $G$ is FC and $Δ (G) = G$ .

A torsion-free nilpotent check

Let $G$ be the discrete Heisenberg group of upper unitriangular $3 \times 3$ integer matrices. It is torsion-free and nonabelian, so by $(6.24)$ it cannot be FC — and indeed every noncentral element has an infinite conjugacy class, its centraliser being of infinite index. Therefore

Δ (G) = Z (G) ≅ ℤ, k Δ (G) ≅ k [z, z^{- 1}],

(E.2)

A Laurent polynomial ring over $k$ : a domain whenever $k$ is, which is what the proof of $(6.28 c)$ consumes.

Neumann's lemma in the smallest case

Take $G = ℤ$ and $H_{1} = \dots = H_{m} = {0}$ . Cosets of the trivial subgroup are single points, so finitely many of them cannot cover $ℤ$ — consistent with the lemma, since $[ℤ : {0}]$ is infinite. Replacing one $H_{i}$ by $2 ℤ$ makes a covering possible with two cosets, and that $H_{i}$ does have finite index.

Process and Workflow

The shape of the R implies D argument, step by step.

Assume a zero divisorSuppose

γ γ' = 0

with both factors nonzero in

k G

Upgrade to an annihilationReducedness turns

(γ' r γ)^{2} = 0

into

γ' r γ = 0

, so

γ' k G γ = 0

Find surviving truncations

(6.27) (2)

produces

r

and

r'

with

π (γ' r) \neq 0

and

π (r' γ) \neq 0

Push the annihilation down

(6.28)

transports the vanishing into

k Δ

π (γ' r) π (r' γ) = 0

Contradict in the abelian case

Δ

is torsion-free FC, hence abelian and orderable, so

k Δ

is a domain — the product of two nonzero elements cannot vanish.

Each step is where a hypothesis is spent: reducedness at step 2, the group structure of $Δ$ at step 5, and Neumann's lemma inside step 4. Removing torsion-freeness breaks step 5 immediately, since $Δ$ could then contain a finite normal subgroup and $k Δ$ would have zero divisors.

Comparison and Classification

$Δ (G)$ for standard groups
Group $G$	$Δ (G)$	$Δ^{+} (G)$
Abelian $G$	$G$	torsion subgroup of $G$
Finite $G$	$G$	$G$
Infinite dihedral $D_{\infty}$	$⟨ a ⟩ ≅ ℤ$	${1}$
Free group of rank $\geq 2$	${1}$	${1}$
Discrete Heisenberg group	$Z (G) ≅ ℤ$	${1}$
$H \times A$ , $H$ finite, $A$ torsion-free abelian	$G$	$H$
Finitary symmetric group on an infinite set	${1}$	${1}$

What each hypothesis buys
	$Δ$ abelian	$k Δ$ a domain	R $\Rightarrow$ D available
$G$ torsion-free, $k$ a domain	yes	yes	yes
$G$ torsion-free, $k$ arbitrary	yes	no	no
$G$ with a finite normal subgroup	no	no	no
$G$ free, $k$ a domain	yes	yes	yes

What each hypothesis buys

In the second row $Δ$ is still abelian, but nothing prevents $k$ itself from having zero divisors, and the final step of the argument fails.

Relationship Map

The relevant subgroups nest, and each layer carries a different amount of finiteness.

$G$ arbitrary group

$Δ (G)$ elements with finitely many conjugates; normal, characteristic, FC

$Δ^{+} (G)$ torsion elements of

Δ (G)

; the union of all finite normal subgroups of

G

${1}$ the case

G

torsion-free, where

Δ (G)

is torsion-free abelian

By a theorem of B. H. Neumann, $Δ (G) / Δ^{+} (G)$ is always torsion-free abelian — it is torsion-free by construction and FC as a quotient of an FC group, so $(6.24)$ applies. Dietzmann's lemma, that a group generated by finitely many torsion elements each with finitely many conjugates is finite, is what makes $Δ^{+} (G)$ a subgroup at all.

Z (G)

Δ (G)

G

Downstream, this page supplies the hard arrow of The Group Ring Problems and depends on the orderability results of Ordered Groups, Trivial Units and Domains for the final contradiction.

Applications and Industry Use

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

Structure of group rings

Primeness and semiprimeness

Connell characterised primeness of a group algebra over a field by the absence of nontrivial finite normal subgroups, and Passman characterised semiprimeness in characteristic $p$ by the absence of finite normal subgroups of order divisible by $p$ . Both proofs are $Δ$ -arguments.

Zero-divisor problem

The reduction machine

Every known reduction of the zero-divisor conjecture to a smaller class of groups passes through $Δ$ : one splits off the FC part, handles it by abelian theory, and studies the quotient.

Infinite group theory

FC groups as a class

FC groups form a natural class between abelian and arbitrary groups, closed under subgroups, quotients and finite direct products. The structure theory — a torsion-free FC group is abelian, and in general $Δ / Δ^{+}$ is torsion-free abelian — is used well outside ring theory.

Computation

Centraliser indices

For a group given by a polycyclic or finite presentation with a solvable conjugacy problem, membership in $Δ (G)$ reduces to computing centraliser indices. GAP and Magma supply the centraliser and index primitives; for polycyclic groups the FC centre is computable.

This material is internal to algebra. Its value is as a machine: it converts questions about a general group ring into questions about an abelian group ring, which is where the answers live.

Standards and Notation

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

Preferred

Δ (G)

for the FC centre,

Δ^{+} (G)

for its torsion part

Variants

Δ_{G}

, and FC centre or finite conjugate subgroup in the group theory literature

FCwritten f.c. in older sources including Lam; capitalised FC is now standard

Truncation

π_{Δ}

π

; sometimes called the projection or trace map onto

k Δ

Centraliser

C_{G} (g)

;

𝒞_{G} (g)

and

Z_{G} (g)

also occur

Implementationscentraliser and index computations in GAP and Magma; polycyclic group packages expose the FC centre

Failure Modes and Common Mistakes

Do not assume $Δ (G)$ is finitely generated or even countable; it is only FC.
Do not expect $Δ (G / N)$ to be the image of $Δ (G)$ ; the FC condition is not preserved by taking preimages, and quotients can create new finite conjugacy classes.
Do not use $(6.27) (2)$ for left ideals without mirroring the proof: the transversal must be chosen on the correct side.
Do not forget that $(6.28)$ needs the full annihilation $γ k G γ' = 0$ , not merely $γ γ' = 0$ . Producing the stronger statement is what reducedness is for.

Historical Notes and Lessons Learned

1937Dietzmann's lemmaA group generated by finitely many torsion elements each with finitely many conjugates is finite — the fact that makes $Δ^{+} (G)$ a subgroup.
1951-54B. H. Neumann on FC groupsNeumann develops the theory of groups with finite conjugacy classes and proves the covering lemma $(6.25)$ , in the course of studying groups covered by cosets.
1960sPassman's $Δ$ -methodPassman makes $Δ (G)$ the central tool of group ring theory, obtaining semiprimitivity, primeness and zero-divisor results by systematic truncation.
1963ConnellConnell's paper on group rings characterises primeness in terms of finite normal subgroups, one of the first structural payoffs of the method.
1977ConsolidationThe Algebraic Structure of Group Rings presents the $Δ$ -theory in its mature form and remains the standard reference.

The methodological lesson is the value of finding the right subgroup. Neither $Z (G)$ nor $G$ is usable — one is too small, the other too general. $Δ (G)$ is chosen so that a finiteness property (finitely many conjugates) converts into an algebraic property (abelian, once torsion is excluded), and the whole theory is the exploitation of that conversion.

Quick Reference

$Δ (G)$

{g : [G : C_{G} (g)] < \infty}

, normal and characteristic

FC groupevery conjugacy class finite

$(6.22)$

[G : Z (G)] = n \Rightarrow (a b)^{n} = a^{n} b^{n}

$(6.24)$ torsion-free

+

\Rightarrow

abelian

$(6.25)$ finite coset covering

\Rightarrow

some

H_{i}

of finite index

$(6.27)$

π

is a

k Δ

-bimodule map; nonzero right ideals have nonzero image

$(6.28)$

γ k G γ' = 0 \Rightarrow π (γ) π (γ') = 0

Payoff

k

a domain,

G

torsion-free,

k G

reduced

\Rightarrow

k G

a domain

Hypotheses of the main results
Result	On $k$	On $G$
$(6.22)$	not used	$[G : Z (G)] = n < \infty$
$(6.24)$	not used	torsion-free FC
$(6.25)$	not used	arbitrary
$(6.27)$	any ring	arbitrary, $Δ$ any subgroup
$(6.28)$	any ring	arbitrary, $Δ = Δ (G)$
R $\Rightarrow$ D	domain	torsion-free

Frequently Asked Questions

Why is

Δ (G)

preferable to the centre?

Because it is usually much bigger and is still tractable. Infinite groups routinely have trivial centre — free groups and the finitary symmetric group, for instance — while $Δ (G)$ captures every element that commutes with a finite-index subgroup. When $G$ is torsion-free, $Δ (G)$ is abelian, so nothing is lost in tractability.

Is the truncation

π_{Δ}

multiplicative?

No, and this is the central technical point. Elements supported outside $Δ$ can multiply to something supported inside it, so $π (γ γ')$ and $π (γ) π (γ')$ differ in general. What holds is the bimodule identity over $k Δ$ , plus Passman's proposition, which recovers multiplicativity in the one situation the theory needs.

Where does Neumann's lemma actually enter?

At the end of the proof of $(6.28)$ . Conjugation constraints put the finite-index subgroup $C$ inside a finite union of cosets of the centralisers $C_{G} (v_{i})$ , with each $v_{i}$ outside $Δ$ and hence with centraliser of infinite index. Neumann's lemma says such a covering is impossible, which is the contradiction.

Does

(6.28)

need any hypothesis on

k

G

None at all. The proposition holds for an arbitrary ring $k$ and an arbitrary group $G$ . Hypotheses enter only when the conclusion is used: to make $k Δ$ a domain one needs $k$ a domain and $Δ$ torsion-free abelian, and the latter comes from torsion-freeness of $G$ via $(6.24)$ .

Why does

(6.24)

reduce to two-generator subgroups?

Commutativity is a statement about pairs. Since subgroups of FC groups are FC, one may replace $G$ by the subgroup generated by the two elements in question, and that subgroup has centre of finite index because it is generated by two elements each with a finite-index centraliser. The finite-index hypothesis of $(6.22)$ is then available.

What is

Δ^{+} (G)

good for?

It isolates the torsion of the FC part. Dietzmann's lemma shows it is a subgroup, equal to the union of all finite normal subgroups of $G$ , and $Δ (G) / Δ^{+} (G)$ is torsion-free abelian. Most structural characterisations of group rings — primeness, semiprimeness — are stated in terms of $Δ^{+} (G)$ rather than $Δ (G)$ .

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §6, results (6.22)-(6.28) (pp. 96-100).
D. S. Passman, The Algebraic Structure of Group Rings, Wiley-Interscience, New York, 1977, Chapters 4 and 12.
B. H. Neumann, “Groups covered by permutable subsets”, Journal of the London Mathematical Society 29 (1954).
I. G. Connell, “On the group ring”, Canadian Journal of Mathematics 15 (1963).
D. J. S. Robinson, A Course in the Theory of Groups, 2nd edition, Graduate Texts in Mathematics 80, Springer-Verlag, 1996, Chapter 14 (FC groups and the transfer).
D. S. Passman, Infinite Group Rings, Pure and Applied Mathematics 6, Marcel Dekker, New York, 1971.

AI Suggested Questions

Give the full construction of the transfer homomorphism and verify that it is $g \mapsto g^{n}$ for a central subgroup of index $n$ .
Prove Dietzmann's lemma and deduce that $Δ^{+} (G)$ is a subgroup.
State Connell's criterion for a group algebra to be prime and locate the $Δ$ -argument in its proof.
Construct an FC group with infinite commutator subgroup.
How is $Δ (G)$ used in Passman's characterisation of semiprime group algebras in characteristic $p$ ?
Compute $Δ (G)$ for a torsion-free polycyclic group and describe an algorithm that does it.
What replaces the $Δ$ -method for groups with trivial $Δ$ , such as free groups?

FC Groups and Δ-Methods

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Almost central elements

The truncation map

Coverings by cosets

Why torsion-freeness makes $Δ$ abelian

Key Results

Proof Techniques and Method

Conjugate by a finite-index subgroup

Manufacture a coset covering

Push down, then compute

Worked Example

The infinite dihedral group

Two extreme cases

A torsion-free nilpotent check

Neumann's lemma in the smallest case

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Primeness and semiprimeness

The reduction machine

FC groups as a class

Centraliser indices

Standards and Notation

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

Almost central elements

The truncation map

Coverings by cosets

Why torsion-freeness makes Δ abelian

06Key Results

07Proof Techniques and Method

Conjugate by a finite-index subgroup

Manufacture a coset covering

Push down, then compute

08Worked Example

The infinite dihedral group

Two extreme cases

A torsion-free nilpotent check

Neumann's lemma in the smallest case

09Process and Workflow

10Comparison and Classification

11Relationship Map

12Applications and Industry Use

Primeness and semiprimeness

The reduction machine

FC groups as a class

Centraliser indices

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

Why torsion-freeness makes $Δ$ abelian

Key Results

Proof Techniques and Method

Worked Example

Process and Workflow

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