Well-Ordered Subsets of Ordered Groups

Q: Is the order type of ST controlled by those of S and T?

Yes, but only by inequalities. If S and T have order types and , the order type of S T is at most the natural sum and that of ST at most the natural (Hessenberg) product of and . Equality fails in general because different products st can coincide.

Q: Why does inverting a series need more than (14.17)?

Because the geometric series 1++^2+… has support inside _n≥ 1S^n, an infinite union. (14.17) gives each S^n separately but says nothing about the union, and infinite unions of WO sets are generally not WO. The extra hypothesis S ⊆ P and a separate argument are required.

Executive Summary

To multiply two formal series indexed by a group, the coefficient of a given group element must be a finite sum. In the classical Laurent series ring the index group is $ℤ$ and finiteness comes from supports being bounded below. For an arbitrary ordered group the right condition is that supports be well-ordered: every nonempty subset has a least element.

Three facts are needed, and they are exactly what this page establishes. Well-ordered subsets of a totally ordered set are characterised by the descending chain condition, and equivalently by the existence of nondecreasing subsequences. In an ordered group, the product $S T$ of two well-ordered subsets is again well-ordered. And each element of $S T$ has only finitely many factorisations $s t$ with $s \in S$ , $t \in T$ .

$3$ Equivalent characterisations

$S \cup T$ , $S T$ Closed operations

finiteFactorisations of each

u \in S T

$S^{- 1}$ Not closed

Overview

Let $(G, \leq)$ be a group with a total order satisfying $x \leq y \Rightarrow a x b \leq a y b$ for all $a, b \in G$ . Write $P = {g \in G : g > 1}$ for the positive cone. A subset $S \subseteq G$ is well-ordered (WO) if every nonempty subset of $S$ has a least element.

Given a ring $R$ and formal sums $α = \sum_{g} a_{g} g$ , the product is defined by convolution:

α β = \sum_{u \in G} (\sum_{g h = u} a_{g} ω_{g} (b_{h})) u,

(14.20)

The inner sum ranges over all factorisations $u = g h$ with $g \in supp (α)$ and $h \in supp (β)$ . For this to define an element of $R$ it must be finite; for the answer to be a legitimate series, its support must again be WO.

When $G = ℤ$ with its usual order, a subset is WO exactly when it is bounded below, and everything here is the familiar bookkeeping behind Laurent series. The content is that the same bookkeeping survives for an arbitrary — possibly nonabelian — ordered group, which is what The Mal'cev–Neumann Construction of Laurent Series Rings needs.

Learning Objectives

Define ordered group, positive cone, and well-ordered subset.
Prove the equivalence of well-ordering, the descending chain condition, and nondecreasing subsequences.
Prove $S \cup T$ is WO whenever $S$ and $T$ are.
Prove $S T$ is WO in an ordered group, and that factorisations are finite in number.
Exhibit WO and non-WO subsets of $ℝ$ and of $ℤ^{2}$ with the lexicographic order.
Explain precisely which step of the convolution formula each lemma licenses.

Definitions

DefinitionOrdered group and positive cone

An ordered group is a group $G$ together with a total order $\leq$ such that $x \leq y$ implies $a x b \leq a y b$ for all $a, b \in G$ . Equivalently, the set $P = {g : g > 1}$ satisfies $P P \subseteq P$ , $G = P ⊔ {1} ⊔ P^{- 1}$ , and $g P g^{- 1} = P$ for every $g \in G$ . $P$ is the positive cone of the ordering.

An ordered group is necessarily torsion-free: if $g > 1$ then $g^{n} > 1$ for all $n \geq 1$ . Free groups, torsion-free abelian groups and torsion-free nilpotent groups all admit orderings.

Definition(14.16a)Well-ordered subset

Let $(G, \leq)$ be a totally ordered set. A subset $S \subseteq G$ is well-ordered, written WO, if every nonempty subset of $S$ has a least element. Equivalently $(S, \leq)$ is order-isomorphic to an ordinal, its order type.

$P$: The positive cone ${g \in G : g > 1}$ . Note $1 \notin P$ .
$supp (α)$: For a formal sum $α = \sum_{g} a_{g} g$ , the set ${g \in G : a_{g} \neq 0}$ .
$S T$: The product set ${s t : s \in S, t \in T}$ , computed in the group.
$S^{n}$: ${s_{1} s_{2} \dots s_{n} : s_{i} \in S}$ , the $n$ -fold product set.
DCC: The descending chain condition: no strictly decreasing infinite sequence exists in the set.
Order type: The unique ordinal order-isomorphic to a given well-ordered set.

Well-ordering here is a property of a subset of an already totally ordered ambient set, not the assertion that some order can be found. The ambient order is fixed once and for all.

Core Concepts

What well-ordering buys and what it costs

Well-ordering is a one-sided finiteness condition. It says nothing about how large a set is — $ℤ_{\geq 0}$ is WO and infinite — but it forbids infinite descent. That asymmetry is exactly right for series: a series may have infinitely many terms, but only finitely many below any given point.

Every subset of a WO set is WO.
Every finite subset is WO; the empty set is WO.
A WO subset of $ℤ$ is precisely a subset bounded below.
In $ℝ$ , ${1 - 1 / n : n \geq 1}$ is WO of order type $ω$ , while ${1 / n : n \geq 1}$ is not WO — it has no least element.

Where each lemma is spent

The convolution $(14.20)$ makes three demands, and the results below supply exactly those three.

Each coefficient is a finite sumSupplied by the finiteness of factorisations in

(14.17)

The product has WO supportSupplied by

supp (α β) \subseteq supp (α) supp (β)

together with closure of WO under products.

The sum has WO supportSupplied by

supp (α + β) \subseteq supp (α) \cup supp (β)

and closure under finite unions.

The two failures

Well-ordering is not preserved by inversion: in $ℤ$ , $S = {0, 1, 2, \dots}$ is WO but $S^{- 1} = {0, - 1, - 2, \dots}$ is not. Nor is it preserved by infinite unions: each $S_{n} = {- n} \subseteq ℤ$ is WO, but $⋃_{n} S_{n} = {- 1, - 2, - 3, \dots}$ is not.

Key Results

Lemma(14.16)Three characterisations of well-ordering

Let $(G, \leq)$ be a totally ordered set and $S \subseteq G$ . The following are equivalent:

$S$ is well-ordered;
$S$ satisfies DCC: every sequence $s_{1} \geq s_{2} \geq s_{3} \geq \dots$ in $S$ is eventually constant;
every sequence $s_{1}, s_{2}, s_{3}, \dots$ in $S$ has a nondecreasing subsequence $s_{n (1)} \leq s_{n (2)} \leq s_{n (3)} \leq \dots$ with $n (1) < n (2) < n (3) < \dots$ .

Proof

**(3) $\Rightarrow$ (2).** A strictly decreasing sequence has no nondecreasing subsequence with two distinct terms, so (3) rules strictly decreasing sequences out; a nonincreasing sequence with no strict decrease from some point on is eventually constant.

**(2) $\Rightarrow$ (1).** Suppose some nonempty $T \subseteq S$ has no least element. Pick $t_{1} \in T$ ; since $t_{1}$ is not least, some $t_{2} \in T$ has $t_{2} < t_{1}$ ; iterating produces a strictly decreasing sequence in $S$ , contradicting DCC. (This step uses the axiom of dependent choice.)

**(1) $\Rightarrow$ (3).** Let $s_{1}, s_{2}, \dots$ be a sequence in a WO set $S$ . Choose $n (1)$ with $s_{n (1)} = \min {s_{i} : i \geq 1}$ , which exists because ${s_{i} : i \geq 1}$ is a nonempty subset of $S$ . Having chosen $n (1) < \dots < n (r)$ , choose $n (r + 1) > n (r)$ with $s_{n (r + 1)} = \min {s_{i} : i > n (r)}$ . Each minimum is taken over a subset of the previous index range's tail, so $s_{n (r)} \leq s_{n (r + 1)}$ , and the resulting subsequence is nondecreasing.

Lemma(14.17)Closure under union and product

Let $(G, \leq)$ be a totally ordered set and let $S, T \subseteq G$ be WO. Then $S \cup T$ is WO. If in addition $(G, \leq)$ is an ordered group, then

$U : = S T = {s t : s \in S, t \in T}$ is WO;
for each $u \in U$ there are only finitely many pairs $(s, t) \in S \times T$ with $u = s t$ .

Proof

Union. Let $\emptyset \neq A \subseteq S \cup T$ . At least one of $A \cap S$ , $A \cap T$ is nonempty; each nonempty one has a least element by hypothesis, and the smaller of those (or the only one) is least in $A$ since the ambient order is total.

(1) Product. Suppose $U$ is not WO. By $(14.16)$ there is a strictly decreasing sequence $s_{1} t_{1} > s_{2} t_{2} > \dots$ with $s_{i} \in S$ , $t_{i} \in T$ . Since $S$ is WO, $(14.16) (3)$ lets us pass to a subsequence along which $s_{1} \leq s_{2} \leq \dots$ ; the sequence $s_{i} t_{i}$ remains strictly decreasing. Now if $t_{i} \leq t_{i + 1}$ for some $i$ , then compatibility of the order with multiplication gives

s_{i} t_{i} \leq s_{i + 1} t_{i} \leq s_{i + 1} t_{i + 1},

contradicting $s_{i} t_{i} > s_{i + 1} t_{i + 1}$ . Hence $t_{1} > t_{2} > t_{3} > \dots$ strictly, contradicting that $T$ is WO. So $U$ is WO.

(2) Finitely many factorisations. Fix $u \in U$ and suppose $u = s_{i} t_{i}$ for infinitely many distinct pairs $(s_{i}, t_{i})$ , $i = 1, 2, \dots$ . By $(14.16) (3)$ pass to a subsequence with $s_{1} \leq s_{2} \leq \dots$ . In an ordered group $s_{i} \leq s_{i + 1}$ implies $s_{i + 1}^{- 1} \leq s_{i}^{- 1}$ , so from $t_{i} = s_{i}^{- 1} u$ we get $t_{1} \geq t_{2} \geq \dots$ . Since $T$ is WO, $(14.16) (2)$ makes this sequence eventually constant, say $t_{i} = t$ for all $i \geq i_{0}$ ; but then $s_{i} = u t^{- 1}$ is also constant for $i \geq i_{0}$ , so the pairs $(s_{i}, t_{i})$ are not distinct — a contradiction.

Corollary(14.17a)Finite products and powers

By induction, if $S_{1}, \dots, S_{n}$ are WO subsets of an ordered group then $S_{1} S_{2} \dots S_{n}$ is WO and every element of it has only finitely many factorisations as $s_{1} s_{2} \dots s_{n}$ with $s_{i} \in S_{i}$ . In particular $S^{n}$ is WO for every $n \geq 1$ and every WO $S$ .

Counterexample(14.17b)Infinite unions and inverses

Both restrictions in $(14.17)$ are necessary. In $G = ℤ$ : the sets $S_{n} = {- n}$ are WO but $⋃_{n \geq 1} S_{n} = {- 1, - 2, \dots}$ is not; and $S = ℤ_{\geq 0}$ is WO but $S^{- 1} = ℤ_{\leq 0}$ is not. Consequently $⋃_{n \geq 1} S^{n}$ need not be WO for a general WO set $S$ — an extra hypothesis, namely $S \subseteq P$ , is required.

Corollary(14.18)The support condition is closed under the ring operations

Fix a ring $R$ , an ordered group $(G, \leq)$ and a homomorphism $ω : G \to Aut (R)$ . On the set of formal sums with WO support,

supp (α + β) \subseteq supp (α) \cup supp (β), supp (α β) \subseteq supp (α) \cdot supp (β),

so by $(14.17)$ both results again have WO support, and by $(14.17) (2)$ each coefficient of $α β$ in $(14.20)$ is a finite sum in $R$ . Hence addition and multiplication are well defined on

A = R ((G, ω)) = {textstyle \sum_{g \in G} a_{g} g : supp (α) \subseteq G is WO} .

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Refuting well-ordering by a bad sequence

To prove a set is WO, assume it is not and extract a strictly decreasing sequence. Sequences are far easier to manipulate than arbitrary subsets, and $(14.16)$ is what licenses the translation.

Move 2

Pass to a monotone subsequence first

Given a bad sequence in $S T$ , tidy the $S$ -parts into a nondecreasing sequence before arguing about the $T$ -parts. This converts a two-variable problem into a one-variable one.

Move 3

Invert to trade monotonicity

In an ordered group, $s \leq s'$ gives $s'^{- 1} \leq s^{- 1}$ . Applying this to $t_{i} = s_{i}^{- 1} u$ turns a nondecreasing sequence in $S$ into a nonincreasing sequence in $T$ , where DCC finishes the argument.

Move 1 is the reason $(14.16)$ is stated before anything else: without the equivalence of well-ordering with DCC, every proof would have to manipulate arbitrary nonempty subsets, and the compatibility of the group operation with the order would be far harder to use. The pattern — replace a chain condition on subsets by a chain condition on sequences — is the same one used to relate noetherian modules to ascending chains.

Note also which hypotheses are spent where. $(14.16)$ needs only a total order on a set. The union statement in $(14.17)$ needs only that. Everything about products needs the group structure and its two-sided compatibility with the order; a merely left-invariant order would not support the inversion step in Move 3.

Worked Example

Subsets of $ℝ$ under addition

$(ℝ, +, \leq)$ is an ordered abelian group. Take

S = {1 - \frac{1}{n} : n \geq 1} = {0, \frac{12}{,} \frac{23}{,} \frac{34}{,} \dots}, T = {2 - \frac{1}{m} : m \geq 1} .

(E.1)

Both are WO of order type $ω$ : any nonempty subset has a least element because the sequences are strictly increasing. By $(14.17)$ the sum set $S + T = {3 - 1 / n - 1 / m}$ is WO. Its order type is larger than $ω$ — for each fixed $n$ the values accumulate from below at $3 - 1 / n$ , and those limits themselves accumulate at $3$ — but well-ordering survives, which is all the construction needs.

Finiteness of factorisations is visible here: for a fixed value $3 - 1 / n - 1 / m$ the quantity $1 / n + 1 / m$ is a fixed positive rational $q$ , and $n$ must satisfy $1 / q < n \leq 2 / q$ , leaving only finitely many possibilities for the pair $(n, m)$ .

A lexicographic ordered group

Order $G = ℤ^{2}$ lexicographically: $(a, b) < (c, d)$ iff $a < c$ , or $a = c$ and $b < d$ . This is a total order compatible with addition, so $G$ is an ordered abelian group with positive cone $P = {(a, b) : a > 0} \cup {(0, b) : b > 0}$ .

Well-ordering in $ℤ^{2}$ with the lexicographic order
Subset	WO?	Reason
${(0, n) : n \geq 0}$	yes, type $ω$	increasing sequence with least element $(0, 0)$
${(1, n) : n \in ℤ}$	no	$(1, 0) > (1, - 1) > (1, - 2) > \dots$ strictly decreasing
${(0, n) : n \geq 0} \cup {(1, n) : n \geq 0}$	yes, type $ω \cdot 2$	finite union of WO sets
${(n, 0) : n \geq 0}$	yes, type $ω$	least element $(0, 0)$ ; the first coordinate dominates
${(- n, 0) : n \geq 0}$	no	strictly decreasing
${(0, n) : n \geq 0} + {(1, n) : n \geq 0}$	yes	product of WO sets, by $(14.17)$

The second row shows why an arbitrary bounded-below-looking set can fail: every element of ${(1, n)}$ exceeds every element of ${(0, n) : n \geq 0}$ , so the set is bounded below in the ambient order, yet it is not WO. In $ℤ$ the two conditions coincide; in a general ordered group they do not.

Comparison and Classification

Closure properties of well-ordered subsets
	Totally ordered set	Ordered group	$G = ℤ$
Subsets of a WO set	yes	yes	yes
Finite unions	yes	yes	yes
Finite intersections	yes	yes	yes
Products $S T$	n/a	yes	yes
Finitely many factorisations in $S T$	n/a	yes	yes
Inverses $S^{- 1}$	n/a	no	no
Arbitrary unions	no	no	no
$⋃_{n \geq 1} S^{n}$ for $S \subseteq P$	n/a	yes	yes

Closure properties of well-ordered subsets

The last row is the deep case: it is a separate theorem, proved on the Mal'cev-Neumann page, and it fails without the hypothesis that S lies in the positive cone.

Well-ordering compared with neighbouring finiteness conditions
Condition on $S$	Meaning	Relation to WO
Finite	finitely many elements	strictly stronger
Well-ordered	every nonempty subset has a least element	—
DCC	no strictly decreasing infinite sequence	equivalent, by $(14.16)$
Bounded below	some $g$ with $g \leq s$ for all $s \in S$	implied by WO; equivalent only for $G = ℤ$ and similar discrete orders
Well-quasi-ordered	no infinite antichain and no infinite strictly decreasing sequence	for a total order the antichain condition is vacuous, so it coincides with WO
Noetherian (ACC)	no strictly increasing infinite sequence	independent: $ℤ_{\geq 0}$ has DCC not ACC, $ℤ_{\leq 0}$ has ACC not DCC

Relationship Map

Totally ordered set $(G, \leq)$

(14.16)

available: WO

\Leftrightarrow

DCC

\Leftrightarrow

nondecreasing subsequences

Ordered group $(G, \leq)$

(14.17)

available: products of WO sets are WO with finite factorisations

Positive cone $P$ the deeper lemma on

⋃_{n} S^{n}

for

S \subseteq P

becomes available

Mal'cev–Neumann ring $R ((G, ω))$ geometric series converge;

R

a division ring makes

A

a division ring

(14.16)

(14.17)

well-defined convolution

(14.18)

Mal'cev–Neumann ring

The same combinatorics reappears wherever formal sums are indexed by an ordered structure: Hahn series over an ordered abelian group, Novikov rings in symplectic topology, and the field of Levi-Civita and other non-archimedean number systems. Each is a special case of the support condition established here.

Applications and Industry Use

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

Algebra

Existence of Mal'cev–Neumann division rings

The entire construction of series division rings over ordered groups rests on these two lemmas; without closure under products the multiplication is not even defined.

Valuation theory

Hahn series and value groups

Hahn's embedding theorem realises any ordered abelian group as the value group of a series field whose supports are WO. Real closed fields and surreal numbers are built this way.

Symplectic topology

Novikov rings

Floer homology is defined over Novikov rings, whose elements are formal series with supports satisfying exactly a well-ordering or finiteness-below condition on the energy grading.

Computer algebra

Lazy and transseries arithmetic

Systems that manipulate generalised power series with real exponents enforce a well-ordered support so that each coefficient of a product is computed by a finite sum.

Termination analysis

Ranking functions

Proofs that a program terminates map states into a well-ordered set so that each step strictly decreases; the DCC formulation on this page is the same principle.

Combinatorics on words

Well-quasi-ordering

Higman's and Kruskal's theorems are the antichain-tolerant generalisation of $(14.16)$ ; for total orders the two notions coincide.

The honest summary is that this page is infrastructure. Its results are never the object of study, but every construction of a series ring over a group larger than $ℤ$ silently invokes them.

Standards and Notation

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

This collectionWO for well-ordered;

P

for the positive cone;

supp

for support

Common variantsWell-ordered versus reverse well-ordered, depending on whether series are indexed by increasing or decreasing exponents

Ordered groupSome authors say fully ordered or totally ordered group, reserving ordered group for partial orders

One-sided ordersLeft-orderable is strictly weaker than bi-orderable; the results here need bi-invariance

Order typesWritten as ordinals

ω

ω \cdot 2

ω^{2}

; natural (Hessenberg) sum and product bound the types of unions and products

GAPOrderings on free and polycyclic groups via package-level support

SagePuiseuxSeriesRing and Hahn-type series constructions enforce support conditions

Computational Notes

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

Representation. A general WO subset of an ordered group is not finitely describable, so implementations restrict to supports that are finitely generated as $⋃_{n} S^{n}$ for a finite $S \subseteq P$ , or to $ℤ$ -graded supports where WO means bounded below.
Truncated arithmetic. Because supports are WO, one can compute all coefficients below a chosen threshold in finite time; this is the operational meaning of the theory and is what lazy series libraries implement.
Cost of a product coefficient. Computing the coefficient at $u$ requires enumerating the factorisations $u = s t$ ; $(14.17) (2)$ guarantees the enumeration terminates, but gives no bound on how many there are, so worst-case cost is not controlled by the theory alone.
Deciding well-ordering is not effective for an arbitrary given subset: it is a statement about all nonempty subsets, and even the DCC formulation quantifies over all sequences. Algorithms rely on structural guarantees, not on testing.
Order comparison in a lexicographically ordered $ℤ^{n}$ costs $O (n)$ and is the primitive on which every support manipulation is built.

Failure Modes and Common Mistakes

Do not confuse well-ordered with well-quasi-ordered; they agree for total orders only.
Do not assume the order type of $S T$ is the sum or product of the order types of $S$ and $T$ — only bounds hold, via natural ordinal arithmetic.
Do not use DCC for subsets when you mean sequences without invoking $(14.16)$ ; the equivalence needs dependent choice and is a genuine step.
Do not assume a WO set has a largest element or is bounded above; $ℤ_{\geq 0}$ is neither.

Quick Reference

Ordered grouptotal order with

x \leq y \Rightarrow a x b \leq a y b

Positive cone

P = {g \in G : g > 1}

WOevery nonempty subset has a least element

EquivalentDCC; every sequence has a nondecreasing subsequence

Closed undersubsets, finite unions and intersections, finite products

Not closed underinverses, infinite unions

Factorisationseach

u \in S T

has finitely many

u = s t

In $ℤ$ WO

\Leftrightarrow

bounded below

Statements and their hypotheses
Statement	Hypotheses	Reference
WO $\Leftrightarrow$ DCC $\Leftrightarrow$ nondecreasing subsequences	$(G, \leq)$ a totally ordered set	(14.16)
$S \cup T$ is WO	$(G, \leq)$ a totally ordered set; $S, T$ WO	(14.17)
$S T$ is WO	$(G, \leq)$ an ordered group; $S, T$ WO	(14.17)
each $u \in S T$ has finitely many factorisations	same	(14.17)
$S_{1} \dots S_{n}$ WO with finite factorisations	same, by induction	(14.17a)
convolution is well defined on WO-supported sums	$R$ a ring, $ω : G \to Aut (R)$	(14.18)

Frequently Asked Questions

Why not simply require supports to be bounded below?

Because in a densely ordered group that condition is too weak for the convolution to be finite. In $ℝ$ the set ${1 / n}$ is bounded below yet has no least element, and one can then build sums in which the coefficient of a single group element is an infinite sum. Well-ordering is the correct strengthening, and in $ℤ$ the two coincide.

Where exactly is the group structure used?

Only in the statements about products. The characterisation $(14.16)$ and the closure under finite unions need nothing but a total order on a set. Closure under $S T$ needs multiplication compatible with the order on both sides, and the factorisation-finiteness argument additionally uses that inversion reverses the order.

Does the proof of

(14.16)

need the axiom of choice?

The implication from DCC to well-ordering builds a strictly decreasing sequence by repeated selection, which is the axiom of dependent choice. The other implications are choice-free. In practice this is never an obstacle, but it is worth knowing which direction carries the set-theoretic weight.

Is the order type of

S T

controlled by those of

S

and

T

Yes, but only by inequalities. If $S$ and $T$ have order types $α$ and $β$ , the order type of $S \cup T$ is at most the natural sum and that of $S T$ at most the natural (Hessenberg) product of $α$ and $β$ . Equality fails in general because different products $s t$ can coincide.

Why does inverting a series need more than

(14.17)

Because the geometric series $1 + α + α^{2} + \dots$ has support inside $⋃_{n \geq 1} S^{n}$ , an infinite union. $(14.17)$ gives each $S^{n}$ separately but says nothing about the union, and infinite unions of WO sets are generally not WO. The extra hypothesis $S \subseteq P$ and a separate argument are required.

Which groups actually carry such an order?

Torsion-freeness is necessary. Torsion-free abelian groups, free groups, and torsion-free nilpotent groups are all bi-orderable, and free groups being orderable is exactly what lets a free ring be embedded in a division ring. Not every torsion-free group is orderable, and left-orderability is a strictly weaker condition.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §14, (14.16)–(14.18), pp. 240–243.
B. H. Neumann, “On ordered division rings”, Transactions of the American Mathematical Society 66 (1949), 202–252.
A. I. Mal'cev, “On the embedding of group algebras in division algebras”, Doklady Akademii Nauk SSSR 60 (1948), 1499–1501.
L. Fuchs, Partially Ordered Algebraic Systems, Pergamon Press, 1963, Chapters II and VIII.
D. S. Passman, The Algebraic Structure of Group Rings, Wiley-Interscience, 1977, Chapter 13 (ordered groups and crossed products).
P. W. Carruth, “Arithmetic of ordinals with applications to the theory of ordered abelian groups”, Bulletin of the American Mathematical Society 48 (1942), 262–271.

AI Suggested Questions

Prove that the order type of a product of two well-ordered subsets is bounded by the natural product of their order types.
Give an example of a torsion-free group that admits no bi-invariant total order.
Show that a free group of rank $2$ is bi-orderable, and describe one such order explicitly.
How do Higman's lemma and well-quasi-ordering generalise the results on this page to partial orders?
What breaks in the Mal'cev–Neumann construction if the group order is only left-invariant?
Compare well-ordered supports with the finiteness conditions used to define Novikov rings.
Give an explicit ordered group and well-ordered sets $S, T$ in which some element of $S T$ has exactly $n$ factorisations, for each $n$ .

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

What well-ordering buys and what it costs

Where each lemma is spent

The two failures

06Key Results

07Proof Techniques and Method

Refuting well-ordering by a bad sequence

Pass to a monotone subsequence first

Invert to trade monotonicity

08Worked Example

Subsets of ℝ under addition

A lexicographic ordered group

09Comparison and Classification

10Relationship Map

11Applications and Industry Use

Existence of Mal'cev–Neumann division rings

Hahn series and value groups

Novikov rings

Lazy and transseries arithmetic

Ranking functions

Well-quasi-ordering

12Standards and Notation

13Computational Notes

14Failure Modes and Common Mistakes

15Quick Reference

16Frequently Asked Questions

17Related KEVOS Topics

18References

19AI Suggested Questions

Continue learning