Why is no hypothesis needed on for D to be a division ring?

Because invertibility is proved by a valuation argument that never inspects . One factors out the lowest term, reducing to inverting 1 - with of positive order, and the geometric series _r ≥ 0^r is well defined purely because ord(^r) ≥ r. The automorphism only affects which element the coefficients turn out to be, not whether the sums make sense.

Is k a maximal subfield of D?

Only when has infinite order. If || = s 1.

Why does an automorphism of infinite order force _k_0 k to be infinite?

If _k_0 k were finite then k/k_0 would be a finite extension, and the group of k_0-automorphisms of a finite extension is finite; but sits inside that group and is infinite. So _k_0 k = ∈fty, and since k ⊆ D and Z(D) = k_0 in this case, D is centrally infinite.

How does this generalise beyond Z-indexed exponents?

Replace the exponent group Z by any ordered group G, replace "support bounded below" by "support well-ordered", and replace the single automorphism by a homomorphism G → Aut(R). That is the Mal'cev–Neumann construction, and the geometric series argument survives verbatim once one knows that products of well-ordered sets are well-ordered.

Twisted Laurent Series Division Rings

Q: What is the relation to the skew polynomial ring k[x;]?

k[x;] is the subring of series with finite support and no negative exponents. It is a principal left ideal domain, and k((x;)) is a completion of its Ore quotient division ring with respect to the x-adic filtration. Passing to series is what makes invertibility trivial: in k[x;] one has to build the Ore quotient ring by hand.

Q: Does the same construction work if is only an endomorphism?

No. If is not surjective the rule x a = (a) x still defines an associative multiplication on the polynomial ring k[x;], but negative powers of x cannot be introduced consistently, so there is no Laurent series ring of this shape. Surjectivity is exactly what lets one move x^-1 past a coefficient using ^-1.

Executive Summary

Fields are abundant; noncommutative division rings are not. The first systematic supply came from Hilbert in 1899: take formal Laurent series over a field $k$ , but insist that the variable does not commute with the coefficients — instead $x a = σ (a) x$ for a fixed automorphism $σ$ of $k$ . The result $D = k ((x; σ))$ is a division ring for every choice of $σ$ , and it is noncommutative as soon as $σ \neq id$ .

The one substantive computation on this page is the centre. It splits cleanly on the order of $σ$ : infinite order gives $Z (D) = k_{0}$ , the fixed field, and $D$ is centrally infinite; finite order $s$ gives $Z (D) = k_{0} ((x^{s}))$ and $\dim_{Z (D)} D = s^{2}$ . The finite-order case is the prototype from which Dickson abstracted the cyclic algebra.

1899Hilbert's example

$x a = σ (a) x$ The twist

$s^{2}$

\dim_{Z (D)} D

when

| σ | = s

$\infty$ …when

| σ | = \infty

Overview

Let $R$ be a division ring and $σ \in Aut (R)$ . The underlying set of $D = R ((x; σ))$ consists of formal sums $\sum_{i \geq n} a_{i} x^{i}$ with $n \in ℤ$ and $a_{i} \in R$ — that is, functions $ℤ \to R$ whose support is bounded below. Addition is coefficientwise. Multiplication is forced by distributivity together with the single rule $x a = σ (a) x$ , iterated to $x^{i} a = σ^{i} (a) x^{i}$ .

(\sum_{i} a_{i} x^{i}) (\sum_{j} b_{j} x^{j}) = \sum_{n} (\sum_{i + j = n} a_{i} σ^{i} (b_{j})) x^{n}

(14.T)

Each inner sum is finite because both supports are bounded below — this is the only convergence issue, and it is combinatorial, not analytic.

Two features make the construction work. First, supports bounded below are closed under the addition of index sets, so the product is again a legitimate series. Second, a series with nonzero lowest coefficient can be normalised to $1 - α$ with $α$ of strictly positive order, and $1 - α$ is then invertible by a geometric series that terminates in each degree. Nothing about $σ$ is used in either step, which is why the theorem carries no hypothesis on $σ$ at all.

This page is the source of both halves of the classification introduced in Centrally Finite Division Rings. Hilbert's own choice of $σ$ had infinite order and produced the first centrally infinite division ring; the finite-order case reappears, abstracted, as the theory of Cyclic Algebras.

Learning Objectives

Write down the twist rule and verify that multiplication of twisted Laurent series is well defined.
Prove that $R ((x; σ))$ is a division ring whenever $R$ is, for every $σ \in Aut (R)$ .
Compute $Z (k ((x; σ)))$ in the two cases $| σ | = \infty$ and $| σ | = s < \infty$ .
Deduce the criterion for central finiteness and the value $\dim_{Z (D)} D = s^{2}$ .
Exhibit $D$ as a left $K$ -space with basis $1, x, \dots, x^{s - 1}$ over $K = k ((x^{s}))$ .
Reconstruct Hilbert's example over $ℚ (t)$ and a degree- $4$ example over $ℝ ((t))$ .

Definitions

Definition(14.1)Centrally finite, centrally infinite

The centre $Z (D)$ of a division ring $D$ is a field, so $D$ is an algebra over $Z (D)$ . Call $D$ centrally finite if $\dim_{Z (D)} D < \infty$ , and centrally infinite otherwise.

ConstructionThe twisted Laurent series ring

Let $R$ be a ring and $σ \in Aut (R)$ . Set

R ((x; σ)) = {\sum_{i \geq n} a_{i} x^{i} : n \in ℤ, a_{i} \in R}

with coefficientwise addition and the multiplication $(14 . T)$ , equivalently the associative extension of $x a = σ (a) x$ for $a \in R$ . The subring of series with finitely many nonzero terms and no negative exponents is the skew polynomial ring $R [x; σ]$ .

$ord (f)$: The least $i$ with $a_{i} \neq 0$ , for $f = \sum a_{i} x^{i} \neq 0$ . One sets $ord (0) = \infty$ .
$k_{0}$: The fixed field ${a \in k : σ (a) = a}$ of $σ$ acting on $k$ .
$| σ |$: The order of $σ$ in the group $Aut (k)$ .
$K$: When $| σ | = s < \infty$ : the commutative Laurent series field $k ((x^{s}))$ , obtained because $x^{s}$ is central over $k$ .
$F$: The field $k_{0} ((x^{s}))$ , which the main theorem identifies with $Z (D)$ .

Lam writes $k ((x, σ))$ with a comma; this collection uses the semicolon $k ((x; σ))$ throughout, to match the skew polynomial notation $k [x; σ]$ .

Core Concepts

Why the multiplication converges

Let $f, g$ be nonzero with $ord (f) = m$ and $ord (g) = n$ . The coefficient of $x^{N}$ in $f g$ is a sum over pairs $(i, j)$ with $i + j = N$ , $i \geq m$ , $j \geq n$ . There are at most $N - m - n + 1$ such pairs, so every coefficient is a finite sum, and the support of $f g$ is contained in ${N : N \geq m + n}$ . Thus $ord (f g) \geq ord (f) + ord (g)$ , with equality when $R$ has no zero divisors, since the bottom coefficient of $f g$ is $a_{m} σ^{m} (b_{n})$ .

ord (f g) = ord (f) + ord (g), ord (f + g) \geq \min {ord (f), ord (g)}

(V)

For $R$ a division ring, $ord$ is a discrete valuation on $D$ with value group $ℤ$ and residue division ring $R$ .

Why every nonzero series is invertible

Given $0 \neq f = \sum_{i \geq m} a_{i} x^{i}$ with $a_{m} \neq 0$ , the element $a_{m} x^{m}$ is a unit, with inverse $σ^{- m} (a_{m}^{- 1}) x^{- m}$ . Dividing on the left normalises the bottom coefficient to $1$ :

(a_{m} x^{m})^{- 1} f = 1 - α, ord (α) \geq 1 .

(N)

Since $ord (α^{r}) \geq r$ , the series $1 + α + α^{2} + \dots$ has only finitely many terms contributing to each power of $x$ , so it defines an element of $D$ ; it is a two-sided inverse of $1 - α$ . Hence $f$ is a unit.

f \neq 0

f = a_{m} x^{m} (1 - α)

(1 - α)^{- 1} = \sum_{r \geq 0} α^{r}

f \in U (D)

How the twist creates the centre condition

Commuting a series past a scalar $a \in k$ compares $a_{i} σ^{i} (a)$ with $a a_{i}$ in degree $i$ . So a central series can only have a nonzero coefficient in degree $i$ when $σ^{i}$ fixes all of $k$ , i.e. when $| σ |$ divides $i$ . Commuting past $x$ then forces every surviving coefficient into $k_{0}$ . Those two constraints are the whole of the centre computation.

Key Results

Proposition§1Hilbert's twisted Laurent series ring is a division ring

Let $R$ be a division ring and $σ \in Aut (R)$ . Then $D = R ((x; σ))$ is a division ring, and $ord : D ∖ {0} \to ℤ$ is a discrete valuation with $ord (f g) = ord (f) + ord (g)$ .

Proof

Associativity and distributivity are a direct check on the formula $(14 . T)$ , using $σ^{i} σ^{j} = σ^{i + j}$ . For invertibility, take $0 \neq f$ of order $m$ with bottom coefficient $a_{m}$ . Then $(a_{m} x^{m})^{- 1} f = 1 - α$ with $ord (α) \geq 1$ as in $(N)$ . Because $ord (α^{r}) \geq r$ , for each fixed $N$ only the terms $α^{0}, \dots, α^{N}$ can contribute to the coefficient of $x^{N}$ ; so $γ = \sum_{r \geq 0} α^{r}$ is a well-defined element of $D$ and $(1 - α) γ = γ (1 - α) = 1$ . Therefore $f^{- 1} = γ^{- 1} σ^{- m} (a_{m}^{- 1}) x^{- m}$ exists. The order formula was proved above.

Proposition(14.2)The centre of a twisted Laurent series ring

Let $k$ be a field, $σ \in Aut (k)$ , $D = k ((x; σ))$ , and let $k_{0} = {a \in k : σ (a) = a}$ be the fixed field. Then

Z (D) = {\begin{cases} k_{0}, & if σ has infinite order, \\ [2 p t] k_{0} ((x^{s})), & if σ has finite order s . \end{cases}

In particular $D$ is centrally finite if and only if $σ$ has finite order, and in that case $\dim_{Z (D)} D = s^{2}$ .

Proof

The scalar test. Let $f = \sum_{i} a_{i} x^{i} \in Z (D)$ and $a \in k$ . Then $f a = \sum_{i} a_{i} σ^{i} (a) x^{i}$ while $a f = \sum_{i} a a_{i} x^{i}$ . Comparing the coefficient of $x^{i}$ gives $a_{i} σ^{i} (a) = a a_{i}$ for all $a \in k$ . Hence for every index $i$ with $a_{i} \neq 0$ we get $σ^{i} = {id}_{k}$ .

**The $x$ test.** From $f x = x f$ we get $\sum_{i} a_{i} x^{i + 1} = \sum_{i} σ (a_{i}) x^{i + 1}$ , so $σ (a_{i}) = a_{i}$ , i.e. $a_{i} \in k_{0}$ , for every $i$ .

**Case $| σ | = \infty$ .** The scalar test allows $a_{i} \neq 0$ only for $i = 0$ , so $f = a_{0}$ , and the $x$ test gives $a_{0} \in k_{0}$ . Conversely every element of $k_{0}$ commutes with all scalars and with $x$ , hence is central. So $Z (D) = k_{0}$ . Since $k \subseteq D$ and $\dim_{k_{0}} k$ is infinite — the fixed field of an automorphism of infinite order has infinite index, because a finite extension $k / k_{0}$ would force $Gal$ -type finiteness on $⟨ σ ⟩ \subseteq Aut (k / k_{0})$ — $D$ is centrally infinite.

**Case $| σ | = s < \infty$ .** The scalar test now says $a_{i} \neq 0$ forces $s ∣ i$ , and the $x$ test puts every $a_{i}$ in $k_{0}$ ; hence $f \in k_{0} ((x^{s}))$ . Conversely, for $a \in k_{0}$ the monomial $a x^{j s}$ commutes with every $b \in k$ (because $σ^{j s} = id$ and $σ (a) = a$ ) and with $x$ ; so $Z (D) = k_{0} ((x^{s})) = : F$ .

The dimension count. Since $σ^{s} = id$ , the element $x^{s}$ is central over $k$ , so $K : = k ((x^{s}))$ is an ordinary commutative Laurent series field over $k$ , and $F = k_{0} ((x^{s})) \subseteq K$ . Artin's theorem applied to the finite automorphism group $⟨ σ ⟩$ of order $s$ gives $\dim_{k_{0}} k = s$ , and extending coefficients to Laurent series in the central variable $x^{s}$ gives $\dim_{F} K = s$ . Splitting a series by the residue of its exponent modulo $s$ yields the internal direct sum

D = K \cdot 1 \oplus K \cdot x \oplus \dots \oplus K \cdot x^{s - 1},

(14.3)

$D$ as a left $K$ -vector space of dimension $s$ .

so $\dim_{K} D = s$ . Transitivity of dimension gives $\dim_{F} D = \dim_{F} K \cdot \dim_{K} D = s \cdot s = s^{2}$ .

Corollary(14.2)Existence of centrally infinite division rings

If $k$ admits an automorphism of infinite order then $k ((x; σ))$ is a centrally infinite division ring. Such $k$ exist: $ℚ (t)$ with $σ (t) = 2 t$ is the classical choice.

Remark(14.4)The multiplication in the basis

1, x, \dots, x^{s - 1}

Write $\bar{σ}$ for the automorphism of $K = k ((x^{s}))$ which acts as $σ$ on $k$ and fixes $x^{s}$ . Then $K / F$ is Galois with $Gal (K / F) = ⟨ \bar{σ} ⟩$ cyclic of order $s$ , and the multiplication in $(14.3)$ is determined by

x \cdot c = \bar{σ} (c) \cdot x (c \in K), x^{s} \in F .

(14.4)

Exactly the defining relations of the cyclic algebra $(K / F, \bar{σ}, x^{s})$ .

Thus $D ≅ (K / F, \bar{σ}, x^{s})$ . Reading these two relations as a definition rather than a computed consequence is precisely Dickson's step, taken up in Cyclic Algebras.

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Order as a valuation

Everything about invertibility is controlled by $ord$ . Normalise the bottom term to $1$ , then invert a geometric series. The same move proves the Mal'cev–Neumann theorem, with $ℤ$ replaced by an ordered group.

Move 2

Test centrality degree by degree

Commute against a general scalar to constrain degrees; commute against the variable to constrain coefficients. Two tests, and the centre falls out.

Move 3

Transitivity of dimension

Compute $\dim_{F} K$ by Galois theory and $\dim_{K} D$ by an explicit basis, then multiply. The pattern $s \cdot s = s^{2}$ recurs for every cyclic algebra.

Move 1 is worth isolating because it shows how little is needed: no finiteness, no chain conditions, no hypothesis on $σ$ . All that is used is that the exponent set of a nonzero element has a least element and that products add orders. Replacing $ℤ$ by an arbitrary ordered group and "bounded below" by "well-ordered" gives the general construction described in The Mal'cev–Neumann Construction of Laurent Series Rings.

Choose $k$ and $σ$ Any field and any automorphism will do; the construction imposes no compatibility condition.

Determine $| σ |$ This single number decides central finiteness. Look for an element of

k

with an infinite

σ

-orbit.

Compute the fixed field $k_{0}$ For finite order

s

, Artin gives

\dim_{k_{0}} k = s

automatically.

Read off the centre

k_{0}

| σ | = \infty

; otherwise

k_{0} ((x^{s}))

, with

\dim_{Z (D)} D = s^{2}

Present as a cyclic algebraIn the finite case, set

K = k ((x^{s}))

and read

(14.4)

as the defining relations.

Worked Example

Hilbert's original example (centrally infinite)

Take $k = ℚ (t)$ and let $σ$ be the $ℚ$ -automorphism with $σ (t) = 2 t$ . Elements of $D = k ((x; σ))$ are series $\sum_{i \geq n} a_{i} (t) x^{i}$ with $a_{i} \in ℚ (t)$ , multiplied using $x \cdot a (t) = a (2 t) x$ .

Since $σ^{m} (t) = 2^{m} t$ and $2^{m} \neq 1$ for $m \neq 0$ , the automorphism $σ$ has infinite order. Its fixed field is $k_{0} = ℚ$ : a rational function $a (t)$ with $a (2 t) = a (t)$ is constant, because a nonconstant $a$ takes a given value at only finitely many points while $t \mapsto 2 t$ has infinite orbits. Hence $Z (D) = ℚ$ and $D$ is centrally infinite — indeed $ℚ (t) \subseteq D$ already has infinite dimension over $ℚ$ .

A centrally finite example: $σ$ of order $2$

Take $k = ℂ$ and $σ =$ complex conjugation, so $s = 2$ and $k_{0} = ℝ$ . Then $D = ℂ ((x; σ))$ has $K = ℂ ((x^{2}))$ , $F = Z (D) = ℝ ((x^{2}))$ , and $\dim_{F} D = 4$ . Setting $t = x^{2}$ and $j = x$ :

D = K \oplus K x, i^{2} = - 1, j^{2} = t, j i = - i j,

(E.1)

The generalised quaternion algebra $(\frac{- 1, t}{ℝ ((t))})$ , of dimension $4$ over $F = ℝ ((t))$ .

Consistency check against the splitting criterion for cyclic algebras: $D$ is a division algebra if and only if $t \notin N_{K / F} (K^{\times})$ . For $f = c t^{n} + (higher) \in K^{\times}$ with $0 \neq c \in ℂ$ , the norm is $N (f) = f \bar{f} = | c |^{2} t^{2 n} + (higher)$ . Every norm therefore has even order and a positive real leading coefficient, whereas $t$ has order $1$ . So $t$ is not a norm, $D$ is a division algebra — as it must be, since we built it as a Laurent series ring.

This example is the smallest case where the two descriptions meet: the series construction and the cyclic-algebra construction produce literally the same ring, and each verifies the other.

Comparison and Classification

The dichotomy governed by the order of $σ$
Feature	$\| σ \| = \infty$	$\| σ \| = s < \infty$
Centre $Z (D)$	$k_{0}$ (the fixed field)	$k_{0} ((x^{s}))$
$\dim_{Z (D)} D$	infinite	$s^{2}$
Maximal subfield containing $k$	$k$ itself, because the scalar test gives $C_{D} (k) = k$	$K = k ((x^{s}))$
Cyclic algebra structure	none	$D ≅ (K / F, \bar{σ}, x^{s})$
Classification	centrally infinite	centrally finite
Historical role	Hilbert 1899	prototype for Dickson 1906

Which hypotheses each conclusion actually needs
	$R$ a division ring	$R$ commutative	$σ$ of finite order	$σ \neq id$
$D$ is a ring	no	no	no	no
$D$ is a division ring	yes	no	no	no
$ord$ is a valuation	yes	no	no	no
$Z (D)$ computed by $(14.2)$	yes	yes	no	no
$D$ centrally finite	yes	yes	yes	no
$D$ noncommutative	no	no	no	yes

Which hypotheses each conclusion actually needs

You need a division ring with prescribed behaviour — which construction?

Centrally infinite, explicit elementsUse

k ((x; σ))

with

| σ | = \infty

. Elements are series; arithmetic is mechanical.

Centrally finite of degree $s$ Use a cyclic algebra

(K / F, σ, a)

and a norm condition. The series ring only realises the special case

a = x^{s}

Division ring containing a free ringNeither suffices: pass to the Mal'cev–Neumann construction over an ordered free group.

Ordered division ringTake

R

ordered and

G

ordered in the Mal'cev–Neumann construction; the

ℤ

-graded case here is the first instance.

Relationship Map

The twisted series ring sits between the polynomial constructions of §1 and the general ordered-group construction at the end of §14.

k [x; σ]

k (x; σ)

(Ore quotients)

k ((x; σ))

k ((G, w))

All division ringsno finiteness assumed

Twisted Laurent series $R ((x; σ))$ always a division ring; carries a

ℤ

-valued valuation

$| σ | < \infty$ centrally finite,

\dim_{Z (D)} D = s^{2}

Cyclic algebra $(K / F, \bar{σ}, x^{s})$ the same ring, presented by generators and relations

$| σ | = \infty$ centrally infinite; Hilbert's example

$D = k ((x; σ))$ — structural consequences
- always
  - a division ring, for every $σ$
  - a discretely valued ring with residue field $k$
  - contains $k [x; σ]$ and its Ore quotient ring
- when $| σ | = s$
  - $K = k ((x^{s}))$ is a maximal subfield
  - $K / F$ is cyclic of degree $s$
  - $D ≅ (K / F, \bar{σ}, x^{s})$ has degree $s$
- never
  - artinian simple with finite centre index when $| σ | = \infty$
  - commutative unless $σ = id$

Applications and Industry Use

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

Valuation theory

Model discretely valued division rings

$k ((x; σ))$ is the standard local model of a complete discretely valued division ring with residue field $k$ and totally ramified twist; the order function is the valuation.

Coding theory

Skew cyclic codes

Codes are constructed as left ideals in quotients of $𝔽_{q} [x; σ]$ with $σ$ a Frobenius power. The series ring is the completion in which the division algorithm and root-finding arguments are carried out.

Control theory

Linear time-varying systems

Transfer-function calculus for time-varying and delay systems is done in twisted series rings, where the shift operator satisfies exactly the Hilbert twist relation with respect to the coefficient field.

Symbolic computation

Ore algebras in CAS

Maple's OreTools, Sage's OrePolynomialRing and Magma's twisted polynomial rings implement $k [x; σ]$ ; formal solution algorithms work in the associated twisted Laurent series ring.

Division algebra theory

Source of examples

Almost every early counterexample about noncommutative division rings — non-conjugate maximal subfields, infinite-dimensional centres, valued but non-commutative fields — is realised here.

Space–time coding

Cyclic division algebras for MIMO

Full-rate full-diversity space–time block codes are built from cyclic division algebras; this page supplies the smallest family in which the defining relations can be inspected directly.

The honest summary is that this construction is a generator of examples. Its downstream engineering value is indirect but real: skew polynomial and skew series rings are the algebraic setting for time-varying linear systems and for skew-cyclic codes, and both start from the twist relation on this page.

Standards and Notation

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

This collection

k ((x; σ))

, with

k [x; σ]

for the polynomial subring

Lam's notation

k ((x, σ))

— comma rather than semicolon

Ore's notation

k [x; σ, δ]

for the general skew polynomial ring;

δ = 0

here

Twist conventionLeft coefficients with

x a = σ (a) x

. The mirror convention

a x = x σ (a)

appears in the literature and swaps

σ

for

σ^{- 1}

SageR['x', sigma] via OrePolynomialRing; Laurent series via completion

Magma / GAPTwistedPolynomials in Magma; skew polynomial support in GAP is package-level

MarkupPresentation MathML per ISO/IEC 40314; operator and set symbols per ISO 80000-2

Computational Notes

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

Arithmetic in $D$ is exact and truncatable: to know a product to precision $N$ it suffices to know each factor to precision $N$ minus the other's order.

Multiplication. Computing the coefficients of $f g$ up to $x^{N}$ costs $O (N^{2})$ ring operations plus $O (N^{2})$ applications of powers of $σ$ ; caching $σ^{i} (b_{j})$ removes the repeated automorphism evaluations.
Inversion. Newton iteration on $1 - α$ doubles the known precision each step, giving $O (N^{2})$ overall with the naive product and quasi-linear cost with fast multiplication — the twist does not obstruct Newton's method because $σ$ is applied coefficientwise.
Deciding central finiteness. This reduces to deciding the order of $σ$ in $Aut (k)$ . For $k$ finite this is immediate; for $k$ a rational function field it is a question about the induced action on generators; for general $k$ there is no algorithm.
Truncation is not equality. Two series agreeing to precision $N$ need not be equal, so no finite computation certifies an identity in $D$ . Certificates must come from the algebraic relations, not from numerics.

Failure Modes and Common Mistakes

Do not assume $x$ is central. It commutes with $k$ only through $σ$ , and $x^{s}$ is central over $k$ only when $σ^{s} = id$ .
Do not write $σ$ on the wrong side: $x a = σ (a) x$ , so $x^{i} a = σ^{i} (a) x^{i}$ , and the product formula carries $σ^{i}$ on the second factor's coefficient.
Do not conclude that $k$ is a maximal subfield in the finite-order case — it is properly contained in $K = k ((x^{s}))$ .
Do not expect uniqueness: non-isomorphic maximal subfields coexist inside $D$ , as Lam's exercises for §14 make explicit.

Historical Notes and Lessons Learned

1843Hamilton's quaternionsThe first noncommutative division ring, of dimension $4$ over its centre $ℝ$ — centrally finite.
1899Hilbert's twisted seriesIn the Grundlagen der Geometrie, seeking to prove the independence of the axiom of Pappus from the other incidence axioms, Hilbert constructs $ℚ (t) ((x; σ))$ with $σ (t) = 2 t$ : the first centrally infinite division ring.
1906Dickson's cyclic algebrasAnalysing the finite-order case, Dickson isolates the relations $(14.4)$ and defines cyclic algebras over an arbitrary cyclic extension.
1907Hahn seriesHahn embeds ordered abelian groups into series groups, replacing $ℤ$ by an arbitrary ordered abelian group in the commutative, untwisted case.
1933Ore's theoryOre develops the systematic theory of skew polynomial rings $k [x; σ, δ]$ and their quotient rings, placing Hilbert's example in a general framework.
1948–49Mal'cev and NeumannThe twist and Hahn's ordered-group idea are combined, and the noncommutative ordered-group case is settled.

The methodological lesson: Hilbert did not look for a division ring, he looked for a counterexample in geometry and needed coordinates that failed to commute. The construction has outlived its original purpose by more than a century because the twist relation — not the series — is the durable idea.

Quick Reference

Object

D = k ((x; σ)) = {\sum_{i \geq n} a_{i} x^{i}}

Twist

x a = σ (a) x

, hence

x^{i} a = σ^{i} (a) x^{i}

Alwaysa division ring, for any

σ \in Aut (k)

Valuation

ord (f g) = ord (f) + ord (g)

Centre, $| σ | = \infty$

Z (D) = k_{0}

; centrally infinite

Centre, $| σ | = s$

Z (D) = k_{0} ((x^{s}))

;

\dim_{Z (D)} D = s^{2}

Maximal subfield

K = k ((x^{s}))

when

| σ | = s

Cyclic form

D ≅ (K / F, \bar{σ}, x^{s})

F = Z (D)

Results at a glance
Statement	Hypotheses	Reference
$R ((x; σ))$ is a division ring	$R$ a division ring; $σ \in Aut (R)$ arbitrary	§1
$Z (D) = k_{0}$	$k$ a field, $\| σ \| = \infty$	(14.2)
$Z (D) = k_{0} ((x^{s}))$	$k$ a field, $\| σ \| = s < \infty$	(14.2)
$\dim_{Z (D)} D = s^{2}$	$k$ a field, $\| σ \| = s < \infty$	(14.2)
$D = K \oplus K x \oplus \dots \oplus K x^{s - 1}$	$K = k ((x^{s}))$ , $\| σ \| = s$	(14.3)
$D ≅ (K / F, \bar{σ}, x^{s})$	$F = k_{0} ((x^{s})) = Z (D)$	(14.4)

Frequently Asked Questions

Why is no hypothesis needed on

σ

for

D

to be a division ring?

Because invertibility is proved by a valuation argument that never inspects $σ$ . One factors out the lowest term, reducing to inverting $1 - α$ with $α$ of positive order, and the geometric series $\sum_{r \geq 0} α^{r}$ is well defined purely because $ord (α^{r}) \geq r$ . The automorphism only affects which element the coefficients turn out to be, not whether the sums make sense.

k

a maximal subfield of

D

Only when $σ$ has infinite order. If $| σ | = s < \infty$ then $x^{s}$ commutes with $k$ , so $K = k ((x^{s}))$ is a commutative subfield strictly containing $k$ ; it is maximal, of degree $s$ over the centre. Lam's Exercise 6 for §14 shows $k_{0} ((x))$ is a second maximal subfield, not isomorphic to $K$ over the centre when $s > 1$ .

What is the relation to the skew polynomial ring

k [x; σ]

$k [x; σ]$ is the subring of series with finite support and no negative exponents. It is a principal left ideal domain, and $k ((x; σ))$ is a completion of its Ore quotient division ring with respect to the $x$ -adic filtration. Passing to series is what makes invertibility trivial: in $k [x; σ]$ one has to build the Ore quotient ring by hand.

Does the same construction work if

σ

is only an endomorphism?

No. If $σ$ is not surjective the rule $x a = σ (a) x$ still defines an associative multiplication on the polynomial ring $k [x; σ]$ , but negative powers of $x$ cannot be introduced consistently, so there is no Laurent series ring of this shape. Surjectivity is exactly what lets one move $x^{- 1}$ past a coefficient using $σ^{- 1}$ .

Why does an automorphism of infinite order force

\dim_{k_{0}} k

to be infinite?

If $\dim_{k_{0}} k$ were finite then $k / k_{0}$ would be a finite extension, and the group of $k_{0}$ -automorphisms of a finite extension is finite; but $⟨ σ ⟩$ sits inside that group and is infinite. So $\dim_{k_{0}} k = \infty$ , and since $k \subseteq D$ and $Z (D) = k_{0}$ in this case, $D$ is centrally infinite.

How does this generalise beyond

ℤ

-indexed exponents?

Replace the exponent group $ℤ$ by any ordered group $G$ , replace "support bounded below" by "support well-ordered", and replace the single automorphism by a homomorphism $G \to Aut (R)$ . That is the Mal'cev–Neumann construction, and the geometric series argument survives verbatim once one knows that products of well-ordered sets are well-ordered.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §14, especially (14.1)–(14.4), pp. 227–231; the construction itself appears in §1.
D. Hilbert, Grundlagen der Geometrie, Teubner, Leipzig, 1899; the noncommutative coordinate system used for the independence of the Pappus axiom.
N. Jacobson, Structure of Rings, American Mathematical Society Colloquium Publications 37, revised edition, 1964, Chapter VII.
O. Ore, “Theory of non-commutative polynomials”, Annals of Mathematics 34 (1933), 480–508.
P. M. Cohn, Skew Fields: Theory of General Division Rings, Encyclopedia of Mathematics and its Applications 57, Cambridge University Press, 1995.
L. H. Rowen, Ring Theory, Volume I, Academic Press, 1988, Chapter 1 (skew polynomial and skew series constructions).

AI Suggested Questions

Work out the centre of $R ((x; σ))$ when $R$ is a noncommutative division ring rather than a field.
Give an explicit isomorphism between $ℂ ((x; conjugation))$ and a generalised quaternion algebra over $ℝ ((t))$ , with the quaternion basis written down.
Which discretely valued division rings arise as twisted Laurent series rings, and what is the obstruction in general?
Compare the twisted Laurent series ring with the Ore quotient division ring of $k [x; σ]$ — are they ever equal?
How is the Brauer class of $(K / F, \bar{σ}, x^{s})$ computed in terms of $σ$ and $s$ ?
Show that a free algebra on two generators embeds in some twisted Laurent series division ring, or explain why it cannot.
What replaces the order function when $ℤ$ is replaced by a nonabelian ordered group?

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Why the multiplication converges

Why every nonzero series is invertible

How the twist creates the centre condition

06Key Results

07Proof Techniques and Method

Order as a valuation

Test centrality degree by degree

Transitivity of dimension

08Worked Example

Hilbert's original example (centrally infinite)

A centrally finite example: σ of order 2

09Comparison and Classification

10Relationship Map

11Applications and Industry Use

Model discretely valued division rings

Skew cyclic codes

Linear time-varying systems

Ore algebras in CAS

Source of examples

Cyclic division algebras for MIMO

12Standards and Notation

13Computational Notes

14Failure Modes and Common Mistakes

15Historical Notes and Lessons Learned

16Quick Reference

17Frequently Asked Questions

18Related KEVOS Topics

19References

20AI Suggested Questions

Continue learning