Constructing Ordered Division Rings

Q: Why order by the least element of the support rather than the greatest?

Because supports are well-ordered, the least element always exists while the greatest generally does not — a series may run upward forever. Ordering by the lowest term is therefore the only option available, and it is also the one that matches the valuation-theoretic picture, where low order means large.

Q: Does the ordering on A restrict to the given ordering on R?

Yes. A nonzero element of R is a series supported at the identity of G, so its leading coefficient is itself; it lies in P exactly when it lies in P_0. Hence P R = P_0, and (A,P) is an ordered extension of (R,P_0).

Q: Is the ordering produced by (18.5) the only one on A?

Not in general, but it is the unique one compatible with the natural valuation. Exercise 18.2 makes this precise in the untwisted case: P is the unique ordering with P R = P_0, with G ⊆ P, and with 1 + m ⊆ P where m is the maximal ideal of the valuation ring.

Q: Why must a noncommutative ordered division ring be non-archimedean?

Because (17.21) says an archimedean ordered ring is commutative and order-isomorphic to a subring of R. So every noncommutative example has infinitely large or infinitely small elements, and the series construction supplies them in the most transparent way possible.

Q: What happens if the ordered group G is noncommutative?

Nothing breaks. The proof of (18.5) uses only that G is totally ordered with two-sided compatible multiplication, so free groups — orderable by (6.19) — are legitimate inputs, and they produce ordered division rings containing free subgroups inside the positive cone.

Q: How does this construction interact with the level of a division ring?

Directly. Choosing the twist (y) = cy with c = -(1 + c_1^2 + … + c_r^2) destroys formal reality in a controlled way, and Scharlau and Tschimmel showed the resulting level is exactly r+1. The same machine therefore produces both the formally real examples and the counterexamples.

Executive Summary

$(18.2)$ decides whether a division ring can be ordered but produces no examples. $(18.5)$ produces them in bulk. Take an ordered ring $(R, P_{0})$ , an ordered group $(G, \leq)$ and a twist $ω : G \to Aut (R)$ whose image consists of order-preserving automorphisms; then the Mal'cev–Neumann series ring $A = R ((G, ω))$ carries a canonical ordering, and when $R$ is a division ring so is $A$ .

The cone is defined by the sign of the leading coefficient: the coefficient sitting at the least element of the support. Well-orderedness of supports, the property that makes the Mal'cev–Neumann multiplication converge, is also what makes leading coefficient meaningful. The two constructions fit together with nothing left over.

$3$ axiomsAll verified on leading terms

$ω_{g} (P_{0}) = P_{0}$ The only hypothesis on the twist

Order-reversingEmbedding of

G

into

A

1903Hilbert's original example

Overview

Recall the construction of $§ 14$ . Let $R$ be a ring, $(G, \leq)$ a multiplicative ordered group and $ω : G \to Aut (R)$ a group homomorphism, writing $ω_{g}$ for the image of $g$ . The Mal'cev–Neumann series ring is

A = R ((G, ω)) = {α = \sum_{g \in G} a_{g} g : supp (α) = {g : a_{g} \neq 0} is well-ordered},

(14.18)

Addition is coefficientwise; multiplication uses the twist law $g r = ω_{g} (r) g$ and converges because a product of well-ordered sets is well-ordered and each coefficient receives only finitely many contributions.

The key theorem of $§ 14$ is $(14.21)$ : if $R$ is a division ring then so is $A$ . The theorem of this page adds an order. The recipe is the obvious one — look at the bottom of the series — and the only surprise is how little needs to be assumed.

Two consequences make the construction indispensable. Any ordered group embeds into the positive cone of an ordered division ring $(18.6)$ , so the ordered groups place no restriction on the theory; and taking $G$ infinite cyclic recovers Hilbert's 1903 example of a noncommutative ordered division ring, the historical origin of the whole subject.

Learning Objectives

State the cone $P$ of $(18.5)$ precisely and verify $(17.1)$ – $(17.3)$ for it.
Explain why a sum of two elements of $P$ cannot cancel at the least support element.
Show that $supp (α β)$ has least element $g_{0} h_{0}$ with coefficient $a_{g_{0}} ω_{g_{0}} (b_{h_{0}})$ .
Deduce $(18.6)$ : every ordered group embeds order-reversingly into the positive cone of an ordered division ring.
Reconstruct Hilbert's example $ℚ ((y)) ((x; σ))$ with $σ (y) = 2 y$ and verify it is noncommutative.
Compute the commutator $x y x^{- 1} y^{- 1}$ in that ring and exhibit it as a square-product.

Definitions

Definition—Leading coefficient and the cone

For $0 \neq α = \sum_{g} a_{g} g \in A$ , the support is nonempty and well-ordered, so it has a least element $g_{0}$ ; the leading coefficient of $α$ is $a_{g_{0}} \neq 0$ . Given an ordering $P_{0}$ of $R$ , set

P = {α \in A ∖ {0} : a_{g_{0}} \in P_{0}, g_{0} = \min supp (α)} .

(18.5a)

$(G, \leq)$: A multiplicative group with a total order satisfying $g \leq h \Rightarrow u g v \leq u h v$ for all $u, v \in G$ .
$ω : G \to Aut (R)$: The twist. Multiplication in $A$ obeys $g r = ω_{g} (r) g$ ; $ω$ trivial gives the untwisted series ring $R ((G))$ .
$supp (α)$: The set of $g \in G$ with $a_{g} \neq 0$ ; required to be well-ordered, i.e. every nonempty subset has a least element.
$R ((x; σ))$: The twisted Laurent series division ring: the case $G = ⟨ x ⟩$ infinite cyclic, $ω_{x} = σ$ , with $x r = σ (r) x$ .
Order-preserving: $σ (P_{0}) = P_{0}$ ; equivalently $σ$ is an automorphism of the ordered ring $(R, \leq)$ .

Well-ordered support is not a technicality: an infinite union of well-ordered sets need not be well-ordered, and $(14.22)$ — the crux of the Mal'cev–Neumann theorem — is precisely the statement that keeps inverses inside $A$ .

Core Concepts

Sums do not cancel at the bottom

Let $α, β \in P$ with least support elements $g_{0}, h_{0}$ . If $g_{0} \neq h_{0}$ , say $g_{0} < h_{0}$ , then $\min supp (α + β) = g_{0}$ and its coefficient is $a_{g_{0}} \in P_{0}$ . If $g_{0} = h_{0}$ , the candidate coefficient is $a_{g_{0}} + b_{g_{0}}$ , a sum of two elements of $P_{0}$ ; since $P_{0} + P_{0} \subseteq P_{0}$ and $0 \notin P_{0}$ , that sum is again positive and in particular nonzero. So the bottom of the support does not move, and $α + β \in P$ .

Products multiply their bottoms

In an ordered group, $g \geq g_{0}$ and $h \geq h_{0}$ imply $g h \geq g_{0} h \geq g_{0} h_{0}$ , with equality only when $g = g_{0}$ and $h = h_{0}$ . So $g_{0} h_{0}$ is the least element of $supp (α β)$ and receives exactly one contribution.

α β = a_{g_{0}} ω_{g_{0}} (b_{h_{0}}) g_{0} h_{0} + (terms supported above g_{0} h_{0}) .

(18.5b)

The twist appears because $g_{0}$ must be moved past $b_{h_{0}}$ : $(a_{g_{0}} g_{0}) (b_{h_{0}} h_{0}) = a_{g_{0}} ω_{g_{0}} (b_{h_{0}}) g_{0} h_{0}$ .

The hypothesis $ω_{g} (P_{0}) = P_{0}$ enters here and only here: it guarantees $ω_{g_{0}} (b_{h_{0}}) \in P_{0}$ , so the leading coefficient of $α β$ lies in $P_{0} \cdot P_{0} \subseteq P_{0}$ .

Why the embedding of $G$ reverses order

Identify $g \in G$ with the series $1 \cdot g$ . Its leading coefficient is $1 \in P_{0}$ , so $G \subseteq P$ : every group element is a positive element of $A$ . Now suppose $g > 1$ in $G$ . The series $1 - g$ has support ${1, g}$ with least element the identity of $G$ , whose coefficient is $1 \in P_{0}$ ; hence $1 - g \in P$ , that is $g < 1$ in $A$ .

g > 1

(G, \leq)

1 - g

has leading coefficient

1

1 - g \in P

g < 1

(A, P)

Key Results

Proposition(18.5)Ordering a Mal'cev–Neumann series ring

Let $(R, P_{0})$ be an ordered ring, $(G, \leq)$ a multiplicative ordered group, and $ω : G \to Aut (R)$ a homomorphism such that $ω_{g} (P_{0}) = P_{0}$ for every $g \in G$ . Let $A = R ((G, ω))$ and let $P$ be the set of nonzero $α \in A$ whose coefficient at $g_{0} = \min supp (α)$ lies in $P_{0}$ . Then $P$ is an ordering on $A$ . If moreover $(R, P_{0})$ is an ordered division ring, then $(A, P)$ is an ordered division ring.

Proof

** $(17.3)$ .** If $α \neq 0$ , its support is nonempty and well-ordered, so $g_{0} = \min supp (α)$ exists and $a_{g_{0}} \neq 0$ . Since $P_{0} \cup (- P_{0}) = R ∖ {0}$ , either $a_{g_{0}} \in P_{0}$ or $- a_{g_{0}} \in P_{0}$ , i.e. $α \in P$ or $- α \in P$ . The two cannot happen simultaneously because $P_{0} \cap (- P_{0}) = \emptyset$ .

** $(17.1)$ .** Let $α, β \in P$ with bottoms $g_{0}, h_{0}$ . If $g_{0} < h_{0}$ then $\min supp (α + β) = g_{0}$ with coefficient $a_{g_{0}} \in P_{0}$ ; symmetrically if $h_{0} < g_{0}$ . If $g_{0} = h_{0}$ then the coefficient there is $a_{g_{0}} + b_{g_{0}} \in P_{0} + P_{0} \subseteq P_{0}$ , which is nonzero since $0 \notin P_{0}$ ; so the bottom is unchanged and $α + β \in P$ .

** $(17.2)$ .** By the ordered-group computation above, $\min supp (α β) = g_{0} h_{0}$ , and the coefficient there is $a_{g_{0}} ω_{g_{0}} (b_{h_{0}})$ . Since $b_{h_{0}} \in P_{0}$ and $ω_{g_{0}} (P_{0}) = P_{0}$ , we get $ω_{g_{0}} (b_{h_{0}}) \in P_{0}$ , hence the leading coefficient lies in $P_{0} \cdot P_{0} \subseteq P_{0}$ and is in particular nonzero. Thus $α β \in P$ — and incidentally $α β \neq 0$ , which re-proves that $A$ is a domain.

For the last assertion: if $R$ is a division ring then $A$ is a division ring by $(14.21)$ , and $P$ is an ordering on it by the above.

Remark—The hypothesis cannot simply be dropped

If some $ω_{g}$ fails to preserve $P_{0}$ , the set $P$ of $(18.5 a)$ is still closed under addition and still satisfies $(17.3)$ , but $(17.2)$ can fail: with $α = g$ and $β = b$ a positive scalar, the product has leading coefficient $ω_{g} (b)$ , which may be negative. Note the hypothesis is about each $ω_{g}$ individually; it does not follow from $ω$ being a homomorphism.

Corollary(18.6)Every ordered group sits in an ordered division ring

Any ordered group $(G, \leq)$ can be embedded, in an order-reversing way, as a subgroup of the multiplicative group of positive elements of an ordered division ring $A$ . If $G$ is commutative, $A$ may be chosen to be an ordered field.

Proof

Take $R = ℚ$ with its usual ordering $P_{0} = ℚ_{> 0}$ and $ω$ the trivial homomorphism; the hypothesis of $(18.5)$ holds vacuously since $ω_{g} = id$ . Then $A = ℚ ((G))$ is a division ring by $(14.21)$ and $P$ is an ordering by $(18.5)$ . Identifying $g$ with $1 \cdot g$ embeds $G$ as a subgroup of $A^{*}$ lying inside $P$ , and the computation above shows $g > 1$ in $G$ forces $g < 1$ in $A$ , so the embedding reverses order. If $G$ is abelian and $ω$ is trivial then $A$ is commutative, hence an ordered field.

Construction—Hilbert's noncommutative ordered division ring

Let $(R, P_{0})$ be an ordered field, let $G = ⟨ x ⟩$ be infinite cyclic with positive cone ${x^{n} : n \geq 1}$ , and let $σ$ be an order-preserving automorphism of $(R, P_{0})$ . Setting $ω_{x} = σ$ gives $A = R ((x; σ))$ , the twisted Laurent series division ring with $x r = σ (r) x$ , and $(18.5)$ furnishes an ordering $P$ of $A$ extending $P_{0}$ . If $σ \neq id$ then $A$ is a noncommutative ordered division ring.

Hilbert's own example is the concrete case $R = ℚ ((y))$ ordered by the sign of the lowest $y$ -coefficient, with $σ$ induced by $y \mapsto 2 y$ .

Proof Techniques and Method

How these proofs work, and which move to reuse.

Everything in $(18.5)$ is decided by leading terms; nothing about the tails matters. That is the reusable idea, and it recurs whenever a valuation-like filtration is present.

Move 1

Order by the leading term

Any structure with a well-defined bottom coefficient inherits an ordering from its coefficient ring. The only checks are: sums do not cancel at the bottom, and bottoms multiply.

Move 2

Make the twist invisible on signs

Assume $ω_{g} (P_{0}) = P_{0}$ . The twist then permutes the coefficients without moving them across zero, so every sign computation is as in the untwisted case.

Move 3

Inherit the division-ring property

The order and the ring structure are proved independently: $(14.21)$ supplies inverses, $(18.5)$ supplies the cone. Neither proof uses the other.

The resulting ordering is non-archimedean by design: $x$ is positive but smaller than every positive rational, so $A$ has infinitely small elements. By $(17.21)$ an archimedean ordered ring is commutative and embeds in $ℝ$ , so every noncommutative ordered division ring must look like this — infinitesimals are not an accident of the construction, they are compulsory.

Worked Example

Hilbert's division ring, explicitly

Let $R = ℚ ((y))$ , the field of formal Laurent series, ordered by

P_{0} = {\sum_{i \geq n} a_{i} y^{i} : n \in ℤ, a_{n} > 0 in ℚ},

(E.1)

The sign of a Laurent series is the sign of its lowest coefficient. Thus $y > 0$ and $1 - y > 0$ , so $0 < y < 1$ ; in fact $0 < y < 1 / n$ for every $n$ , and $y$ is infinitely small.

Let $σ \in Aut (R)$ be induced by $y \mapsto 2 y$ , fixing $ℚ$ . It is order-preserving: $σ$ sends $\sum_{i \geq n} a_{i} y^{i}$ to $\sum_{i \geq n} a_{i} 2^{i} y^{i}$ , whose lowest coefficient $a_{n} 2^{n}$ has the sign of $a_{n}$ . So $σ (P_{0}) = P_{0}$ and $(18.5)$ applies to $A = ℚ ((y)) ((x; σ))$ with $x r = σ (r) x$ .

Noncommutativity, in one line

x y = σ (y) x = 2 y x \neq y x .

(E.2)

Comparing the two infinitesimals

Is $x$ larger or smaller than $y$ ? Expand $x - y$ as a series in $x$ with coefficients in $ℚ ((y))$ : it is $(- y) x^{0} + 1 \cdot x^{1}$ , whose lowest $x$ -exponent is $0$ with coefficient $- y$ . Now $- y \notin P_{0}$ , so $x - y \notin P$ and therefore $x < y$ . The same argument gives $x < y^{n}$ for every $n \geq 1$ : the variable $x$ is infinitely small even relative to $y$ .

A commutator that is a square-product but not a square

From $(E . 2)$ , $x y x^{- 1} = 2 y$ , hence

x y x^{- 1} y^{- 1} = 2 = x^{2} (x^{- 1} y)^{2} (y^{- 1})^{2},

(E.3)

The second equality is the commutator identity of $(18.1)$ ; it exhibits $2$ as a product of three squares of $A$ .

Yet $2$ is not a square in $A$ . If $α^{2} = 2$ with $α$ of lowest $x$ -exponent $m$ and lowest coefficient $a \in ℚ ((y))$ , then $α^{2}$ has lowest exponent $2 m$ with coefficient $a σ^{m} (a)$ ; comparing with $2 = 2 x^{0}$ gives $m = 0$ and $a^{2} = 2$ in $ℚ ((y))$ . Repeating the argument one level down, the lowest $y$ -coefficient $c$ of $a$ would satisfy $c^{2} = 2$ in $ℚ$ — impossible.

Process and Workflow

Choose the coefficient ringAn ordered ring

(R, P_{0})

—

ℚ

ℝ

for the plain construction, an already-ordered series field to iterate.

Choose the ordered group

(G, \leq)

determines the value group of the resulting non-archimedean order. Infinite cyclic gives Laurent series; free groups give noncommutative

G

Choose the twist

ω : G \to Aut (R)

with every

ω_{g}

order-preserving. Trivial twist gives a commutative example; any nontrivial one gives a noncommutative division ring.

Form $A$ and its cone

A = R ((G, ω))

is a division ring by

(14.21)

; the leading-coefficient set

P

is an ordering by

(18.5)

Read off the properties

G \subseteq P

, order-reversed;

A

is non-archimedean;

A

is noncommutative iff

ω

is nontrivial or

G

is.

What do you want the example to exhibit?

An ordered field with prescribed value groupTake

G

abelian and

ω

trivial:

A = ℚ ((G))

is an ordered field whose archimedean classes reproduce

G

A noncommutative ordered division ringTake

G = ⟨ x ⟩

and

σ \neq id

order-preserving on

R

. Hilbert's example. It is necessarily centrally infinite by

(18.11)

A division ring where $- 1$ is a square-product but not a sum of squaresTake

R = k ((y))

over a formally real

k

, and

σ (y) = c y

with

c

a suitable negative element. This is

(18.7)

— no longer formally real.

An ordered ring that is archimedeanNot available from this construction, and not available at all in the noncommutative world:

(17.21)

forces archimedean ordered rings to be commutative subrings of

ℝ

Comparison and Classification

Inputs and what they produce
$R$ , $P_{0}$	$G$	$ω$	$A = R ((G, ω))$
$ℚ$ , usual	$ℤ$	trivial	$ℚ ((y))$ : ordered field, non-archimedean
$ℚ$ , usual	any abelian ordered $G$	trivial	ordered field with value group $G$
$ℚ$ , usual	free group of rank $2$	trivial	noncommutative ordered division ring containing a free group in its positive cone
$ℚ ((y))$ , leading sign	$ℤ$	$y \mapsto 2 y$	Hilbert's example: noncommutative, formally real, centrally infinite
$ℝ ((y))$ , leading sign	$ℤ$	$y \mapsto - y$	not order-preserving; $(18.5)$ does not apply, and indeed $- 1$ becomes a square-product

Properties of the output
	Division ring	Ordered	Commutative	Archimedean
$R ((G, ω))$ , $R$ a division ring, $ω$ order-preserving	yes	yes	no	no
$R ((G))$ , $G$ abelian, $R$ an ordered field	yes	yes	yes	no
$R ((G, ω))$ , $ω$ not order-preserving	yes	partial	no	no
A subfield of $ℝ$	yes	yes	yes	yes

Properties of the output

The third row says only that $(18.5)$ gives no cone; such a ring may or may not be orderable by other means, and $(18.7)$ shows it can fail to be formally real.

Relationship Map

The construction stands at the junction of three earlier threads: ordered groups, well-ordered supports, and the Mal'cev–Neumann division ring theorem.

$(18.5)$ Ordered Mal'cev–Neumann series — the source of all standard examples
- consumes
  - $(14.18)$ well-ordered support
  - $(14.21)$ $A$ is a division ring
  - $(17.1)$ – $(17.3)$ ordering axioms
  - an ordered group $(G, \leq)$
- produces
  - $(18.6)$ every ordered group embeds in an ordered division ring
  - Hilbert's 1903 example
  - the formally real examples used to test $(18.10)$
  - the non-formally-real examples of $(18.7)$
- is constrained by
  - $(17.21)$ archimedean ordered rings are commutative
  - $(18.11)$ formally real centrally finite division rings are fields

The last constraint deserves emphasis. Once $σ \neq id$ is order-preserving, Exercise 17.13 shows $σ$ has infinite order, so by $(14.2)$ the centre of $R ((x; σ))$ is a subfield of $R$ and the division ring is centrally infinite — consistent with $(18.11)$ , which forbids formally real centrally finite noncommutative examples.

Applications and Industry Use

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

Embedding theorems

Free algebras in division rings

The same construction with $G$ a free group gives the Mal'cev–Neumann–Moufang theorem $(14.25)$ : the free ring over any division ring embeds in a division ring. The ordering is a by-product that makes the embedding concrete.

Valuation theory

Value groups realised

$(18.6)$ shows every ordered group is the value group of a valued, ordered field. Series rings are the standard witnesses in the theory of ordered fields and real places.

Model theory

Non-archimedean models

Ordered series fields are the algebraic counterparts of non-standard models of the ordered field axioms, and they supply explicit infinitesimals without an ultrapower.

Symbolic computation

Skew series arithmetic

Truncated twisted Laurent series are implemented in computer algebra for skew polynomial arithmetic — Ore domains, linear differential and difference operators — where the leading term drives every algorithm.

The honest summary is that this is infrastructure inside algebra. Ordered division rings are not used to model physical systems; they exist to demarcate what the axioms of an ordered ring can and cannot force, and Hilbert built the first one for exactly that purpose in the foundations of geometry.

Failure Modes and Common Mistakes

Do not expect $P$ to restrict to a familiar order on subrings other than $R$ : an element $g > 1$ of $G$ becomes infinitely small in $A$ — indeed $n g < 1$ for every integer $n$ — while $g < 1$ in $G$ becomes infinitely large.
Do not conclude that $A$ is noncommutative merely because $ω$ is nontrivial as a map; you need some $ω_{g} \neq id$ , and then $A$ contains $x$ and $r$ with $x r \neq r x$ .
Do not try to make the example archimedean. $(17.21)$ says archimedean forces commutativity and an embedding into $ℝ$ , so the infinitesimals are unavoidable.
Do not confuse $R ((G, ω))$ with the twisted group ring $R [G, ω]$ of finite sums; the latter is a domain but generally not a division ring, and it is the former that $(14.21)$ concerns.

Historical Notes and Lessons Learned

1899–1903Hilbert's GrundlagenStudying which geometric axioms force commutativity of the coordinate division ring, Hilbert constructs an ordered noncommutative division ring from twisted Laurent series — the first example of the subject.
1927Artin–SchreierFormally real fields are characterised, and the commutative side of the theory takes its modern shape.
1948–49Mal'cev and NeumannIndependently, both replace Hilbert's single variable by an arbitrary ordered group, using well-ordered supports to make the multiplication converge; the free ring embedding theorem follows.
1952The orderability criterion for division ringsSzele and Pickert extend Artin–Schreier to division rings using square-products; Johnson's preordering argument covers arbitrary rings.
1983Levels of division ringsScharlau and Tschimmel use exactly this series construction, with a carefully chosen twist, to realise every positive integer as the level of a division ring.

The lesson is that the interesting examples in this subject are all completions. Polynomial-style constructions produce domains, not division rings; only after passing to series with well-ordered support do inverses appear, and only then does a leading-coefficient order make sense. The two facts have the same source.

Quick Reference

Inputs

(R, P_{0})

ordered,

(G, \leq)

an ordered group,

ω : G \to Aut (R)

Hypothesis

ω_{g} (P_{0}) = P_{0}

for all

g \in G

The ring

A = R ((G, ω))

, series with well-ordered support,

g r = ω_{g} (r) g

The cone

P = {α \neq 0 : a_{g_{0}} \in P_{0}, g_{0} = \min supp (α)}

Product rule

\min supp (α β) = g_{0} h_{0}

, coefficient

a_{g_{0}} ω_{g_{0}} (b_{h_{0}})

Division ring

R

a division ring

\Rightarrow

A

a division ring, by

(14.21)

Embedding

G \subseteq P

, order-reversed:

g > 1

G

gives

g < 1

A

Hilbert

ℚ ((y)) ((x; σ))

σ (y) = 2 y

— noncommutative, formally real

Axiom-by-axiom summary of the proof
Axiom	What is checked	Hypothesis used
$(17.1)$ $P + P \subseteq P$	no cancellation at the least support element	$P_{0} + P_{0} \subseteq P_{0}$ , $0 \notin P_{0}$
$(17.2)$ $P \cdot P \subseteq P$	leading coefficient $a_{g_{0}} ω_{g_{0}} (b_{h_{0}})$	$ω_{g} (P_{0}) = P_{0}$ and $P_{0} \cdot P_{0} \subseteq P_{0}$
$(17.3)$ $P \cup (- P)$	existence of $\min supp (α)$	supports are well-ordered
Inverses	$α^{- 1}$ has well-ordered support	$(14.21)$ , via $(14.22)$

Frequently Asked Questions

Why order by the least element of the support rather than the greatest?

Because supports are well-ordered, the least element always exists while the greatest generally does not — a series may run upward forever. Ordering by the lowest term is therefore the only option available, and it is also the one that matches the valuation-theoretic picture, where low order means large.

Does the ordering on

A

restrict to the given ordering on

R

Yes. A nonzero element of $R$ is a series supported at the identity of $G$ , so its leading coefficient is itself; it lies in $P$ exactly when it lies in $P_{0}$ . Hence $P \cap R = P_{0}$ , and $(A, P)$ is an ordered extension of $(R, P_{0})$ .

Is the ordering produced by

(18.5)

the only one on

A

Not in general, but it is the unique one compatible with the natural valuation. Exercise 18.2 makes this precise in the untwisted case: $P$ is the unique ordering with $P \cap R = P_{0}$ , with $G \subseteq P$ , and with $1 + 𝔪 \subseteq P$ where $𝔪$ is the maximal ideal of the valuation ring.

Why must a noncommutative ordered division ring be non-archimedean?

Because $(17.21)$ says an archimedean ordered ring is commutative and order-isomorphic to a subring of $ℝ$ . So every noncommutative example has infinitely large or infinitely small elements, and the series construction supplies them in the most transparent way possible.

What happens if the ordered group

G

is noncommutative?

Nothing breaks. The proof of $(18.5)$ uses only that $G$ is totally ordered with two-sided compatible multiplication, so free groups — orderable by $(6.19)$ — are legitimate inputs, and they produce ordered division rings containing free subgroups inside the positive cone.

How does this construction interact with the level of a division ring?

Directly. Choosing the twist $σ (y) = c y$ with $c = - (1 + c_{1}^{2} + \dots + c_{r}^{2})$ destroys formal reality in a controlled way, and Scharlau and Tschimmel showed the resulting level is exactly $r + 1$ . The same machine therefore produces both the formally real examples and the counterexamples.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §18, pp. 286–289, and §14, pp. 240–248.
D. Hilbert, Grundlagen der Geometrie, 2nd edition, B. G. Teubner, Leipzig, 1903.
A. I. Mal’cev, “On the embedding of group algebras in division algebras”, Doklady Akademii Nauk SSSR 60 (1948).
B. H. Neumann, “On ordered division rings”, Transactions of the American Mathematical Society 66 (1949), 202–252.
L. Fuchs, Partially Ordered Algebraic Systems, Pergamon Press, Oxford, 1963.
P. M. Cohn, Skew Fields: Theory of General Division Rings, Encyclopedia of Mathematics and its Applications 57, Cambridge University Press, 1995, Chapter 2.

AI Suggested Questions

Work through the proof of $(14.22)$ that a union of powers of a well-ordered subset of the positive cone is well-ordered.
Show that the ordering of $(18.5)$ is compatible with the natural Krull valuation on $ℚ ((G))$ and is uniquely determined by that compatibility.
Compute the centre of $ℚ ((y)) ((x; σ))$ with $σ (y) = 2 y$ and confirm it is $ℚ$ .
Which ordered groups arise as the group of archimedean classes of an ordered field?
Give an ordered division ring whose positive cone contains a free group of rank $2$ , and describe its ordering explicitly.
How would one order a skew polynomial ring $R [x; σ]$ directly, without passing to series?
What changes in $(18.5)$ if $ω_{g}$ merely permutes the orderings of $R$ rather than fixing $P_{0}$ ?

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Sums do not cancel at the bottom

Products multiply their bottoms

Why the embedding of G reverses order

06Key Results

07Proof Techniques and Method

Order by the leading term

Make the twist invisible on signs

Inherit the division-ring property

08Worked Example

Hilbert's division ring, explicitly

Noncommutativity, in one line

Comparing the two infinitesimals

A commutator that is a square-product but not a square

09Process and Workflow

10Comparison and Classification

11Relationship Map

12Applications and Industry Use

Free algebras in division rings

Value groups realised

Non-archimedean models

Skew series arithmetic

13Failure Modes and Common Mistakes

14Historical Notes and Lessons Learned

15Quick Reference

16Frequently Asked Questions

17Related KEVOS Topics

18References

19AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Why the embedding of $G$ reverses order

Key Results

Proof Techniques and Method

Worked Example

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions