Centrally Finite and Centrally Infinite Division Rings

Q: Why insist on the centre rather than an arbitrary subfield?

Because the centre is canonical and because scalars from it behave: for ∈ Z(D) the left and right actions on D agree, so needs no side. Over a noncentral subfield K the left and right dimensions are genuinely different invariants, and Cohn and Schofield showed they can disagree.

Q: Is a centrally finite division ring the same thing as a finite-dimensional division algebra?

Only if the base field is the centre. A division algebra can be finite-dimensional over a base field F that is strictly smaller than Z(D) — C over R is the trivial instance. The word *centrally* is doing real work: it fixes the base field to be the whole centre, which is what makes the dimension a perfect square.

Q: Can a centrally infinite division ring be algebraic over its centre?

Yes. Köthe's 1931 construction, using infinite tensor products, produces division algebras all of whose elements are algebraic over the centre while the total dimension is infinite. So *algebraic over the centre* is strictly weaker than *centrally finite*, and the Kurosh-type questions this raises are much harder.

Q: Does every centrally finite division algebra have a Galois maximal subfield?

No. Algebras that do are called crossed products, and Amitsur constructed central division algebras of suitable degrees which are not crossed products. In low degree the answer is yes: degree 2 and degree 3 central division algebras are cyclic, hence crossed products.

Q: How do I actually compute the centre of a division ring given by generators and relations?

There is no general procedure. In the constructions of this section it is done by hand: expand a general element in a normal form, impose commutation with the generators, and read off the constraints. For k((x;)) this takes a paragraph; for a free field it is a theorem, not a computation.

Q: Why does the degree, rather than the dimension, get a name?

Because it is the multiplicative invariant. Degrees multiply under tensor product of division algebras when the classes are coprime, maximal subfields have degree exactly n, and the index of a Brauer class is a degree. Dimensions are squares of degrees, so nothing is lost and the bookkeeping is cleaner.

Executive Summary

The centre of a division ring $D$ is a field $F = Z (D)$ , so $D$ is always an algebra over a field and the number $\dim_{F} D$ is always defined. Lam's dichotomy $(14.1)$ takes that number seriously: $D$ is centrally finite when $\dim_{F} D < \infty$ and centrally infinite otherwise.

The division is not cosmetic. Centrally finite division rings are finite-dimensional central simple algebras: they are algebraic over $F$ , satisfy both chain conditions, have square dimension $n^{2}$ over $F$ , and are classified — as far as they are classified at all — by the Brauer group of $F$ . Centrally infinite division rings have none of that structure, and are built by completion, ordering or free constructions rather than by Galois descent.

$Z (D)$ Always a field

$n^{2}$ Dimension if finite

$4$

\dim_{ℝ} ℍ

1899First centrally infinite example

Overview

Fields are division rings, and from the point of view of noncommutative ring theory they are the uninteresting ones. To organise the interesting ones we need an invariant that measures how far a division ring is from being commutative. The centre supplies it: $Z (D)$ is the largest commutative piece of $D$ that everything else respects, and $\dim_{Z (D)} D$ measures the size of $D$ relative to it.

Z (D) = {z \in D : z d = d z for all d \in D}

(14.1a)

A commutative subring; the point of the first proposition below is that it is closed under inversion, hence a field.

A field has $\dim_{Z (D)} D = 1$ . At the other extreme, Hilbert's twisted Laurent series ring of 1899 has centre $ℚ$ and is infinite-dimensional over it. In between sit the real quaternions with $\dim_{ℝ} ℍ = 4$ , and Dickson's cyclic algebras with $\dim_{F} D = s^{2}$ for every $s$ for which a cyclic extension of degree $s$ is available.

The two classical construction techniques in this part of the theory sit on opposite sides of the line: cyclic algebras, treated on Cyclic Algebras, produce centrally finite examples; the Mal'cev–Neumann series construction produces centrally infinite ones. Twisted Laurent series, described on Twisted Laurent Series Division Rings, straddle the line and are the cleanest place to watch the dichotomy in action.

Learning Objectives

Prove that $Z (D)$ is a field for every division ring $D$ .
State $(14.1)$ and decide the dichotomy for a given $D$ .
Show that a centrally finite $D$ is algebraic over $F = Z (D)$ and is left and right artinian.
Use the square-dimension theorem to define the degree $n$ of $D$ and relate it to maximal subfields.
Classify the standard examples: $ℍ$ , $(K / F, σ, a)$ , $k ((x; σ))$ , Mal'cev–Neumann series, free fields.
Explain why $[D : K]_{ℓ} = [D : K]_{r}$ for subfields is safe in the centrally finite case and false in general.

Definitions

Definition(14.1)Centrally finite, centrally infinite

Let $D$ be a division ring with centre $F = Z (D)$ . $D$ is centrally finite if $\dim_{F} D$ is finite, and centrally infinite otherwise.

Since $F$ is central, left and right $F$ -dimensions of $D$ agree, so no side needs to be specified — an exception rather than the rule in this subject.

$Z (D)$: The centre of $D$ ; a field whenever $D$ is a division ring.
$F$ -algebra: A ring $A$ with a ring map $F \to Z (A)$ . For a division ring $D$ the tautological choice $F = Z (D)$ is always available.
Central simple $F$ -algebra: A simple ring $A$ with $Z (A) = F$ and $\dim_{F} A < \infty$ . Centrally finite division rings are exactly the central simple algebras that happen to be division rings.
Degree: $\deg D = n$ where $\dim_{F} D = n^{2}$ ; well defined by the square-dimension theorem below.
$C_{D} (S)$: The centraliser ${d \in D : d s = s d for all s \in S}$ of a subset $S \subseteq D$ . Note $Z (D) = C_{D} (D)$ .
Maximal subfield: A subfield of $D$ maximal among commutative subrings of $D$ ; it necessarily contains $Z (D)$ .

Throughout, division ring means an associative ring with $1 \neq 0$ in which every nonzero element is invertible. Skew field and division algebra are used interchangeably in the literature; we reserve division algebra for the situation where a base field has been fixed.

Core Concepts

Why the centre must be a field

Centrality is preserved by inversion. If $z \neq 0$ commutes with everything, then conjugating the identity $z d = d z$ by $z^{- 1}$ on both sides gives $d z^{- 1} = z^{- 1} d$ , so $z^{- 1}$ is central too. Nothing here needs finiteness, and nothing here works for a general ring: the centre of $M_{2} (ℤ)$ is $ℤ$ , not a field.

Finite dimension forces algebraicity, and much else

Once $\dim_{F} D$ is finite, every $d \in D$ satisfies a polynomial over $F$ for the cheapest possible reason: the powers $1, d, d^{2}, \dots$ cannot stay independent. The commutative subalgebra $F [d]$ is then a finite-dimensional domain, hence a field, so $D$ is a union of finite field extensions of $F$ . Left ideals are $F$ -subspaces, so both chain conditions hold automatically and $D$ is artinian — which is why Wedderburn–Artin theory applies to $D$ and to every algebra built from it.

\dim_{F} D < \infty

D

algebraic over

F

D

left and right artinian

D

central simple over

F

The last arrow is immediate rather than deep: a division ring is simple, and $Z (D) = F$ holds by construction. What is deep is the converse direction of the theory — that finite-dimensional central simple algebras are classified up to Morita equivalence by the Brauer group $Br (F)$ , whose elements are represented by exactly these centrally finite division rings.

Degree, index and the square-dimension theorem

Extending scalars to an algebraic closure turns any central simple $F$ -algebra into a matrix algebra, so $\dim_{F} D = \dim_{\bar{F}} (D \otimes_{F} \bar{F}) = n^{2}$ for some $n$ . That integer is the degree; when $D$ is a division ring it also equals the index of its Brauer class. Degrees, not dimensions, are the currency of the subject: $ℍ$ has degree $2$ , Dickson's nine-dimensional example has degree $3$ .

Key Results

Proposition(14.1a)The centre of a division ring is a field

Let $D$ be a division ring. Then $F = Z (D)$ is a subfield of $D$ , and $D$ is an $F$ -algebra in which left and right $F$ -dimensions coincide.

Proof

$Z (D)$ is closed under addition and multiplication and contains $0$ and $1$ , so it is a commutative subring. Let $z \in Z (D)$ with $z \neq 0$ ; since $D$ is a division ring, $z^{- 1}$ exists in $D$ . For any $d \in D$ , multiply $z d = d z$ on the left and on the right by $z^{- 1}$ to get $d z^{- 1} = z^{- 1} d$ . Hence $z^{- 1} \in Z (D)$ and $Z (D)$ is a field.

For the last clause: scalar multiplication by $F$ on either side gives the same operation, because $λ d = d λ$ for $λ \in F$ . So an $F$ -basis on the left is an $F$ -basis on the right.

Proposition—Consequences of central finiteness

Let $D$ be a centrally finite division ring with centre $F$ and $\dim_{F} D = m$ . Then:

every $d \in D$ is algebraic over $F$ , and $F [d] = F (d)$ is a subfield of $D$ with $[F (d) : F] \leq m$ ;
$D$ is left and right artinian and left and right noetherian;
every subfield of $D$ containing $F$ is a finite extension of $F$ , and maximal subfields exist;
$D$ is a simple $F$ -algebra with $Z (D) = F$ , i.e. a central simple $F$ -algebra.

Proof

(1) The $m + 1$ elements $1, d, \dots, d^{m}$ lie in an $m$ -dimensional $F$ -vector space, so some nontrivial $F$ -linear relation $\sum_{i = 0}^{m} λ_{i} d^{i} = 0$ holds; as $F$ is central this says $d$ is a root of a nonzero polynomial in $F [X]$ . Consequently $F [d]$ is a commutative $F$ -algebra of dimension at most $m$ . It is a subring of a division ring, hence a domain, and a finite-dimensional commutative domain over a field is a field: multiplication by a nonzero element is an injective $F$ -linear endomorphism of a finite-dimensional space, hence surjective, so inverses exist. Thus $F [d] = F (d)$ is a subfield with $[F (d) : F] \leq m$ .

(2) Every left ideal and every right ideal of $D$ is in particular an $F$ -subspace, and $F$ -subspaces of an $m$ -dimensional space satisfy both chain conditions. (For a division ring the point is vacuous — the only one-sided ideals are $0$ and $D$ — but the same argument applies verbatim to $M_{r} (D)$ and to any $D$ -algebra of finite $F$ -dimension, which is where it is used.)

(3) A subfield $K$ with $F \subseteq K \subseteq D$ is an $F$ -subspace of $D$ , so $[K : F] \leq m$ . A chain of subfields therefore has bounded degrees and Zorn's Lemma (or simply maximality of the degree) produces maximal ones.

(4) A division ring has no two-sided ideals other than $0$ and itself, so $D$ is simple; $Z (D) = F$ is the definition of $F$ . Together with $\dim_{F} D < \infty$ this is precisely the definition of a central simple $F$ -algebra.

Theorem—Square dimension and maximal subfields

Let $D$ be a centrally finite division ring with centre $F$ . Then $\dim_{F} D = n^{2}$ for a unique positive integer $n = \deg D$ , and every maximal subfield $K$ of $D$ satisfies $[K : F] = n$ and splits $D$ , in the sense that $D \otimes_{F} K ≅ M_{n} (K)$ .

*This belongs to the theory of central simple algebras rather than to $§ 14$ ; it is proved by extending scalars to a splitting field, where $D$ becomes a matrix algebra. The maximal-subfield half is developed on* Maximal Subfields of Division Rings. *The cyclic algebras of $§ 14$ realise it explicitly: $K$ is visibly a maximal subfield of $(K / F, σ, a)$ of degree $s$ , and the algebra has dimension $s^{2}$ .*

Corollary—Wedderburn's little theorem, restated

Every finite division ring is a field. Hence a finite division ring is centrally finite of degree $1$ , and there are no noncommutative examples of the dichotomy in the finite world at all — every noncommutative division ring is infinite. (Lam proves this as $(13.1)$ ; see Wedderburn's Little Theorem for the argument.)

Theorem—Frobenius

The only finite-dimensional associative division algebras over $ℝ$ are $ℝ$ , $ℂ$ and $ℍ$ . Of these, $ℝ$ and $ℍ$ are central over $ℝ$ , of degrees $1$ and $2$ ; $ℂ$ has centre $ℂ$ . Consequently $ℍ$ is, up to isomorphism, the only noncommutative centrally finite division ring with centre $ℝ$ .

Proposition(14.2)The dichotomy for twisted Laurent series

Let $k$ be a field, $σ \in Aut (k)$ , and $D = k ((x; σ))$ the twisted Laurent series division ring with $x a = σ (a) x$ . Then $D$ is centrally finite if and only if $σ$ has finite order $s$ , in which case $\dim_{Z (D)} D = s^{2}$ and $Z (D) = k_{0} ((x^{s}))$ with $k_{0}$ the fixed field of $σ$ . If $σ$ has infinite order, $Z (D) = k_{0}$ and $D$ is centrally infinite.

Proved on Twisted Laurent Series Division Rings. *It is the cheapest source of examples on both sides of $(14.1)$ : one automorphism, two completely different answers.*

Proof Techniques and Method

How these arguments work, and which move to reuse.

Move 1

Conjugate the centrality relation

To show a set is closed under inversion, multiply the defining identity by the inverse on both sides. This one line proves $Z (D)$ is a field and reappears whenever centralisers are shown to be division subrings.

Move 2

Count dimensions to force algebraicity

$m + 1$ powers in an $m$ -dimensional space must be dependent. Every finiteness consequence on this page starts from that pigeonhole, not from any structure theory.

Move 3

Finite domain over a field is a field

Multiplication by a nonzero element is injective and $F$ -linear, hence surjective. This upgrades $F [d]$ to $F (d)$ without any explicit inverse formula.

To prove a given $D$ is centrally infinite, the standard route is to compute $Z (D)$ explicitly and exhibit an infinite $Z (D)$ -independent family — usually the powers of a single element, as in the Hilbert example where ${x^{i}}_{i \in ℤ}$ is independent over $ℚ$ . To prove it is centrally finite, exhibit a finite spanning set over the computed centre; a cyclic-algebra presentation does this in one line.

Worked Example

The real quaternions: centrally finite of degree 2

Let $ℍ = ℝ \oplus ℝ i \oplus ℝ j \oplus ℝ k$ with $i^{2} = j^{2} = - 1$ and $i j = - j i = k$ . A quaternion $q = t + x i + y j + z k$ is central iff it commutes with $i$ and with $j$ . Commuting with $i$ : $i q - q i = 2 y i j + 2 z i k$ , wait — compute directly. $i q = t i + x i^{2} + y i j + z i k = - x + t i + y k - z j$ and $q i = t i + x i^{2} + y j i + z k i = - x + t i - y k + z j$ . So $i q = q i$ forces $y = z = 0$ . Commuting with $j$ then forces $x = 0$ .

Z (ℍ) = ℝ, \dim_{ℝ} ℍ = 4 = 2^{2}, \deg ℍ = 2 .

(E.1)

Square dimension, as the general theorem predicts.

A maximal subfield is $ℝ (i) ≅ ℂ$ , of degree $2 = \deg ℍ$ over the centre — again as predicted. And indeed $ℍ \otimes_{ℝ} ℂ ≅ M_{2} (ℂ)$ .

Hilbert's series ring: centrally infinite

Take $k = ℚ (t)$ and let $σ$ be the $ℚ$ -automorphism with $σ (t) = 2 t$ . Then $σ^{n} (t) = 2^{n} t$ , so $σ$ has infinite order. Its fixed field is $ℚ$ : a rational function $f$ with $f (2 t) = f (t)$ has a divisor on the projective line invariant under scaling by $2$ , and every point other than $0$ and $\infty$ has infinite orbit, so $f = c t^{m}$ ; then $f (2 t) = f (t)$ forces $2^{m} = 1$ , i.e. $m = 0$ .

D = ℚ (t) ((x; σ)), x \cdot a (t) = a (2 t) \cdot x, Z (D) = ℚ,

(E.2)

and $\dim_{ℚ} D$ is infinite — already the family ${t^{i}}_{i \geq 0} \subseteq k \subseteq D$ is $ℚ$ -independent. Historically this is the first known centrally infinite division ring, produced by Hilbert in 1899 to separate axioms of ordered geometry.

Frameworks and Models

A working taxonomy of division rings by their relationship to the centre.

Division rings $D$ , $F = Z (D)$
- Centrally finite: $\dim_{F} D = n^{2} < \infty$
  - $n = 1$ : $D = F$ is a field
  - $n = 2$ : quaternion algebras (char $\neq 2$ )
  - $n$ arbitrary, cyclic: $D ≅ (K / F, σ, a)$
  - $n$ arbitrary, crossed product but not cyclic
  - not a crossed product at all (Amitsur, 1972)
- Centrally infinite
  - algebraic over $F$ (Köthe's infinite tensor products)
  - not algebraic: $k ((x; σ))$ , $ord (σ) = \infty$
  - Mal'cev–Neumann series over an ordered group
  - free fields: the division ring generated by a free algebra

Division rings

Z (D)

is a field; nothing else is guaranteed

Algebraic over the centreevery element satisfies a polynomial over

F

Centrally finite

\dim_{F} D = n^{2}

; central simple over

F

Crossed products

D

contains a maximal subfield Galois over

F

Cyclic algebrasthat Galois group is cyclic:

D = (K / F, σ, a)

Quaternion algebrasdegree

2

;

D = (a, b)_{F}

Each containment above is strict. In particular Amitsur's 1972 examples are centrally finite division algebras that are not crossed products, so the innermost bands do not exhaust the centrally finite world.

Comparison and Classification

Classical division rings and their position in the dichotomy
Division ring $D$	Centre $F$	$\dim_{F} D$	Verdict
Any field $k$	$k$	$1$	centrally finite, degree $1$
$ℍ$ (real quaternions)	$ℝ$	$4$	centrally finite, degree $2$
$(a, b)_{F}$ , char $F \neq 2$ , division	$F$	$4$	centrally finite, degree $2$
$(K / F, σ, a)$ division, $[K : F] = s$	$F$	$s^{2}$	centrally finite, degree $s$
$k ((x; σ))$ , $ord (σ) = s$	$k_{0} ((x^{s}))$	$s^{2}$	centrally finite, degree $s$
$k ((x; σ))$ , $ord (σ) = \infty$	$k_{0}$	infinite	centrally infinite
$ℚ (t) ((x; σ))$ , $σ (t) = 2 t$	$ℚ$	infinite	centrally infinite (Hilbert, 1899)
Mal'cev–Neumann $R ((G, ω))$ , $ω$ injective	fixed subfield of $R$	infinite	centrally infinite
Division ring generated by a free algebra	base field	infinite	centrally infinite

Which tools are available on each side
	Centrally finite	Centrally infinite
Wedderburn–Artin applies to $D$ -algebras	yes	partial
$D$ algebraic over $Z (D)$	yes	partial
Dimension over $Z (D)$ is a perfect square	yes	no
Represents a class in $Br (Z (D))$	yes	no
Maximal subfields all of the same degree	yes	no
Reduced norm and reduced trace defined	yes	no
$[D : K]_{ℓ} = [D : K]_{r}$ for division subrings $K$	yes	partial
Built by Galois-theoretic data	yes	no
Built by completion, ordering or freeness	partial	yes

Which tools are available on each side

Applications and Industry Use

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

Number theory

Brauer groups of global fields

Centrally finite division algebras over a number field $F$ are exactly the objects classified by local invariants in the Albert–Brauer–Hasse–Noether theorem; over such $F$ every one of them is cyclic.

Cryptography

Supersingular isogeny problems

The endomorphism algebra of a supersingular elliptic curve over ${\bar{𝔽}}_{p}$ is the centrally finite quaternion algebra over $ℚ$ ramified exactly at $p$ and $\infty$ ; the Deuring correspondence turns isogeny problems into arithmetic in its orders.

Coding and signal processing

Space-time codes

Cyclic division algebras of degree $n$ supply fully diverse space-time block codes for $n$ -antenna wireless links: nonzero codeword differences are invertible precisely because the algebra has no zero divisors.

Symbolic computation

What a CAS can actually do

Centrally finite means finite structure constants, so a computer algebra system can represent $D$ exactly, decide splitting, and compute reduced norms. Centrally infinite division rings admit no such uniform finite presentation.

Analysis and geometry

Quaternionic structures

The centrally finite algebra $ℍ$ underlies unit-quaternion rotation representations, $SU (2)$ -spinors, and hyperkähler geometry — the only noncommutative option Frobenius leaves over $ℝ$ .

Ring theory itself

A supply of counterexamples

Centrally infinite division rings are the standard source of pathologies: infinite-dimensional simple algebras, non-Ore phenomena, and the failure of naive dimension arguments.

Design Considerations

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

Fix the base field first. A division ring has one centre, but many possible base fields $F \subseteq Z (D)$ . Statements such as degree, split, Brauer class are relative to that choice; central means $F = Z (D)$ exactly, and results routinely fail if $F$ is smaller.
Decide whether you need finiteness or only algebraicity. Chain conditions and reduced norms need finite dimension. Many arguments about torsion or algebraic elements need only algebraicity, and then Köthe-type examples remain in scope.
Choose the construction to match the target. Want a prescribed degree and a prescribed centre? Use a cyclic algebra. Want a division ring containing a given awkward ring? Use Mal'cev–Neumann or an Ore localisation, and accept a centrally infinite answer.
Track sides even when they agree. Over the centre, left and right dimensions coincide; over a noncentral division subring they need not. Writing $[D : K]$ without a subscript is safe only when $K \subseteq Z (D)$ or $D$ is centrally finite.

Failure Modes and Common Mistakes

Do not read centrally infinite as not algebraic: the two are different conditions, and the first does not imply the second.
Do not assume the centre is easy to compute. For twisted Laurent series it is a short calculation; for Mal'cev–Neumann rings over a general ordered group Lam does not compute it at all, treating only the case of injective $ω$ .
Do not confuse the degree $n$ with the dimension $n^{2}$ . Sources differ, and *dimension $9$ * versus *degree $3$ * has caused real confusion in the literature on space-time codes.
Do not expect finite division rings to give examples: by Wedderburn's little theorem there are no noncommutative ones.

Historical Notes and Lessons Learned

1843Hamilton's quaternionsThe first noncommutative division ring, and the first centrally finite one: $\dim_{ℝ} ℍ = 4$ .
1878Frobenius's theorem $ℝ$ , $ℂ$ and $ℍ$ exhaust the finite-dimensional associative division algebras over $ℝ$ — the first classification of a class of centrally finite objects.
1899Hilbert's twisted seriesConstructed while studying the independence of the axioms of ordered geometry, and the first example of a centrally infinite division ring.
1905Wedderburn's little theoremEvery finite division ring is commutative, so the dichotomy has no finite noncommutative instances.
1906Dickson's cyclic algebrasAbstracting the finite-order case of Hilbert's construction produced a systematic supply of centrally finite division algebras of every available degree.
1931–32Brauer group and the arithmetic caseKöthe builds centrally infinite division algebras that are algebraic over their centre; Albert, Brauer, Hasse and Noether prove that over a number field every centrally finite division algebra is cyclic.
1948–49Mal'cev and NeumannOrdered-group Laurent series give centrally infinite division rings containing prescribed rings, including free algebras.
1972Amitsur's non-crossed productsCentrally finite does not imply crossed product; the classification of division algebras of high degree remains open.

The methodological lesson is that the two halves of the dichotomy were discovered by different techniques and are still studied by different techniques. Finite dimension over the centre invites Galois theory and cohomology; infinite dimension invites orderings, valuations and completions. Almost no theorem transfers between the halves.

Quick Reference

Definition

D

centrally finite

iff \dim_{Z (D)} D < \infty

Centre

Z (D)

is a field for every division ring

D

Dimension

\dim_{F} D = n^{2}

;

n = \deg D

Maximal subfieldsall of degree

n

over

F

; each splits

D

Implicationcentrally finite

\Rightarrow

algebraic over

F

(not conversely)

Test case

k ((x; σ))

: finite

iff ord (σ) < \infty

Finite ringsnone are noncommutative (Wedderburn)

Over $ℝ$

ℍ

is the only noncommutative example (Frobenius)

Deciding the dichotomy in practice
If $D$ is given as…	Compute	Conclusion
$(K / F, σ, a)$ with $[K : F] = s$	nothing — $\dim_{F} D = s^{2}$	centrally finite of degree $s$ when $D$ is a division ring
$k ((x; σ))$	$ord (σ)$	finite order $s$ gives degree $s$ ; infinite order gives centrally infinite
$R ((G, ω))$ with $ω$ injective, $G \neq 1$	fixed subring $R^{ω}$	centrally infinite
a finite ring	nothing	a field, degree $1$
the Ore quotient ring of a Weyl algebra	centre of the Weyl algebra	centrally infinite in characteristic $0$
an $ℝ$ -algebra of finite dimension	Frobenius	$ℝ$ , $ℂ$ or $ℍ$

Frequently Asked Questions

Why insist on the centre rather than an arbitrary subfield?

Because the centre is canonical and because scalars from it behave: for $λ \in Z (D)$ the left and right actions on $D$ agree, so $\dim$ needs no side. Over a noncentral subfield $K$ the left and right dimensions are genuinely different invariants, and Cohn and Schofield showed they can disagree.

Is a centrally finite division ring the same thing as a finite-dimensional division algebra?

Only if the base field is the centre. A division algebra can be finite-dimensional over a base field $F$ that is strictly smaller than $Z (D)$ — $ℂ$ over $ℝ$ is the trivial instance. The word centrally is doing real work: it fixes the base field to be the whole centre, which is what makes the dimension a perfect square.

Can a centrally infinite division ring be algebraic over its centre?

Yes. Köthe's 1931 construction, using infinite tensor products, produces division algebras all of whose elements are algebraic over the centre while the total dimension is infinite. So algebraic over the centre is strictly weaker than centrally finite, and the Kurosh-type questions this raises are much harder.

Does every centrally finite division algebra have a Galois maximal subfield?

No. Algebras that do are called crossed products, and Amitsur constructed central division algebras of suitable degrees which are not crossed products. In low degree the answer is yes: degree $2$ and degree $3$ central division algebras are cyclic, hence crossed products.

How do I actually compute the centre of a division ring given by generators and relations?

There is no general procedure. In the constructions of this section it is done by hand: expand a general element in a normal form, impose commutation with the generators, and read off the constraints. For $k ((x; σ))$ this takes a paragraph; for a free field it is a theorem, not a computation.

Why does the degree, rather than the dimension, get a name?

Because it is the multiplicative invariant. Degrees multiply under tensor product of division algebras when the classes are coprime, maximal subfields have degree exactly $n$ , and the index of a Brauer class is a degree. Dimensions are squares of degrees, so nothing is lost and the bookkeeping is cleaner.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §14, especially (14.1)–(14.2) (pp. 227–229).
T. Y. Lam, A First Course in Noncommutative Rings, §13 (Wedderburn's little theorem) and §15 (maximal subfields and scalar extension).
N. Jacobson, Basic Algebra II, 2nd edition, W. H. Freeman, 1989, Chapter 4 (finite-dimensional central simple algebras).
A. A. Albert, Structure of Algebras, American Mathematical Society Colloquium Publications 24, 1939.
P. Gille and T. Szamuely, Central Simple Algebras and Galois Cohomology, Cambridge Studies in Advanced Mathematics 101, Cambridge University Press, 2006.
P. M. Cohn, Skew Fields: Theory of General Division Rings, Encyclopedia of Mathematics and its Applications 57, Cambridge University Press, 1995.

AI Suggested Questions

Give a complete proof that a finite-dimensional central simple algebra has square dimension over its centre.
Construct a centrally infinite division ring that is algebraic over its centre, following Köthe's infinite tensor product method.
How does the index of a division algebra relate to the exponent of its class in the Brauer group, and when are they equal?
Work through Amitsur's construction of a central division algebra that is not a crossed product.
What is known about Artin's question on left versus right dimension over a division subring, and what did Schofield actually construct?
Explain how cyclic division algebras of degree $n$ are used to build fully diverse space-time block codes.
For which fields $F$ is every centrally finite division algebra with centre $F$ necessarily cyclic?

Centrally Finite Division Rings

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Why the centre must be a field

Finite dimension forces algebraicity, and much else

Degree, index and the square-dimension theorem

Key Results

Proof Techniques and Method

Conjugate the centrality relation

Count dimensions to force algebraicity

Finite domain over a field is a field

Worked Example

The real quaternions: centrally finite of degree 2

Hilbert's series ring: centrally infinite

Frameworks and Models

Comparison and Classification

Applications and Industry Use

Brauer groups of global fields

Supersingular isogeny problems

Space-time codes

What a CAS can actually do

Quaternionic structures

A supply of counterexamples

Design Considerations

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

Why the centre must be a field

Finite dimension forces algebraicity, and much else

Degree, index and the square-dimension theorem

06Key Results

07Proof Techniques and Method

Conjugate the centrality relation

Count dimensions to force algebraicity

Finite domain over a field is a field

08Worked Example

The real quaternions: centrally finite of degree 2

Hilbert's series ring: centrally infinite

09Frameworks and Models

10Comparison and Classification

11Applications and Industry Use

Brauer groups of global fields

Supersingular isogeny problems

Space-time codes

What a CAS can actually do

Quaternionic structures

A supply of counterexamples

12Design Considerations

13Failure Modes and Common Mistakes

14Historical Notes and Lessons Learned

15Quick Reference

16Frequently Asked Questions

17References

18AI 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

Applications and Industry Use

Design Considerations

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions