Complete Domain Relations in Theory

• Complete Domain Relations are concepts where a family of relations suffices to guarantee consistency, continuity, and expressive adequacy across various mathematical frameworks.
• They encompass diverse completeness notions, ranging from maximal acyclic restrictions in social choice to expressive universality in algebraic calculi and relational presentations in domain theory.
• In practice, these relations bridge local conditions to global models, enabling effective application in areas such as recursive domain constructions, topological completions, and machine learning.

“Complete domain relations” does not denote a single universally fixed formal object. In the current literature, the phrase names, or is closely approximated by, several technically distinct completeness phenomena: maximal acyclic restrictions of preference domains in social choice, complete fragments for domain and codomain comparison in algebraic relation calculi, relational presentations of continuous domains, topological and metric-like completeness notions for domain-shaped spaces, and expressive completeness of relations on finite domains. A plausible unifying theme is that each setting asks when a family of relations is rich enough to determine admissible behavior without loss of consistency, continuity, or expressive power (Liversidge, 2020, Zhang et al., 2024, Wu et al., 2022, Brecht et al., 2019, Koshkin, 17 Jun 2026, Amorim, 2022).

1. Terminological scope

Across the cited literature, the term is used in several non-equivalent senses. In some settings, “complete” refers to an inclusion-maximal family of admissible linear orders; in others, it refers to completeness of an inequational fragment, completeness of a space under $G_\delta$- or Smyth-style criteria, completeness of relational presentations of continuous domains, or primitive-positive expressive universality of a relational language.

Context	Relation object	Completeness notion
Social choice	admissible linear orders	maximal acyclic Condorcet domain
TopKAT / Kleene-with-domain	domain or codomain comparison formulas	completeness for restricted inequational fragments
Continuous-domain presentations	CF-approximable relations, abstract bases	equivalence with Scott continuity or ideal completion
Finite-domain relations	pp-definable relations	pp-complete or expressively universal

In preference aggregation, a maximal Condorcet domain is a “complete restriction” in the sense that no further linear order can be added without creating a Condorcet triple (Liversidge, 2020). In TopKAT, completeness concerns formulas of the form $t_1\top \ge t_2\top$ or $\top t_1 \ge \top t_2$, but only when $t_1,t_2$ are KAT terms rather than arbitrary TopKAT terms (Zhang et al., 2024). In domain theory, CF-approximable relations are “complete enough” to recover Scott continuous maps exactly, while in Pitts’ framework relational properties of domains live in complete lattices and satisfy admissibility conditions strong enough to support recursive definitions (Wu et al., 2022, Amorim, 2022). On finite domains, the graph of a Sheffer function composes all relations, yielding a single ternary pp-complete relation (Koshkin, 17 Jun 2026).

2. Complete admissible restrictions in preference aggregation

In Condorcet-domain theory, completeness is attached to a restriction on preferences rather than to a binary relation on alternatives. Let $A$ be a finite set of alternatives and $\mathcal L(A)$ the set of all linear orders on $A$. A Condorcet domain is a subset $D\subseteq \mathcal L(A)$ such that every profile with preferences in $D$ has an acyclic majority relation. Equivalently, $D$ contains no Condorcet triple

$t_1\top \ge t_2\top$0

A domain is maximal if

$t_1\top \ge t_2\top$1

This is the paper’s operative notion of a complete admissible restriction: an inclusion-maximal family of complete linear orders under which majority aggregation remains acyclic (Liversidge, 2020).

A central reformulation identifies each linear order $t_1\top \ge t_2\top$2 with the directed Hamilton path $t_1\top \ge t_2\top$3. Under this translation, a domain becomes a collection of directed Hamilton paths, and the Condorcet property localizes to triples. The key lemma states that a collection $t_1\top \ge t_2\top$4 of Hamilton directed paths defines a Condorcet domain iff for every $t_1\top \ge t_2\top$5 with $t_1\top \ge t_2\top$6, the simplified contraction $t_1\top \ge t_2\top$7 does not contain a double cycle. This reduces global acyclicity to checking every $t_1\top \ge t_2\top$8-subset and makes recursive constructions by contraction and extension possible (Liversidge, 2020).

The paper distinguishes sharply between Black’s and Arrow’s single-peaked domains. Black single-peaked domains are determined by their two extremal orders, but maximal Arrow single-peaked domains are not. On $t_1\top \ge t_2\top$9 elements, the extremal paths

$\top t_1 \ge \top t_2$0

can be completed in two non-isomorphic ways, called the twisted case and the reversed case. This shows that extremal paths do not determine a maximal Arrow single-peaked domain (Liversidge, 2020).

Counting proceeds by contraction, extension, and the pairing notion of self-pairedness. If $\top t_1 \ge \top t_2$1 is the number of isomorphism classes of maximal Arrow single-peaked domains on $\top t_1 \ge \top t_2$2 elements, $\top t_1 \ge \top t_2$3 the number of labeled such domains with a fixed extremal path, and $\top t_1 \ge \top t_2$4 the number of self-paired isomorphism classes, then for $\top t_1 \ge \top t_2$5,

$\top t_1 \ge \top t_2$6

with

$\top t_1 \ge \top t_2$7

The exact values reported are

$\top t_1 \ge \top t_2$8

All of these statements dualize to single-dipped domains, i.e. complete never-top restrictions on every triple (Liversidge, 2020).

3. Complete domain and codomain comparison in algebraic calculi

In algebraic semantics, “complete domain relations” often means completeness for formulas that compare the domain or codomain of relational denotations. In TopKAT, the distinguished top element $\top t_1 \ge \top t_2$9 is interpreted in relational models as the complete relation

$t_1,t_2$0

This makes domain and codomain comparison expressible by ordinary inequalities: $t_1,t_2$1 For KAT terms $t_1,t_2$2, the corresponding formulas are $t_1,t_2$3 and $t_1,t_2$4 (Zhang et al., 2024).

The positive completeness theorem is syntactically sharp. For $t_1,t_2$5,

$t_1,t_2$6

and, from the surrounding text and proof, likewise

$t_1,t_2$7

Thus TopKAT is complete for comparing domains and codomains of terms that are originally KAT terms. The result does not extend to arbitrary TopKAT terms: with

$t_1,t_2$8

the inequality $t_1,t_2$9 is valid in relational TopKAT but not derivable in the base axiomatization. The paper therefore identifies a precise boundary: complete for (co)domain comparison of KAT terms, incomplete for (co)domain comparison of arbitrary TopKAT terms (Zhang et al., 2024).

A related but different completeness result appears for Kleene algebras with domain. In algebras of binary relations, the domain operation is

$A$0

For the signature $A$1, the free algebra is the algebra of reduced pointed labelled finite rooted trees. For the full signature $A$2, the free algebra consists of regular subsets of reduced pointed labelled rooted trees. In the star-free signature $A$3, the axioms of domain semirings provide a finite equational axiomatization of the equational theory of $A$4. For the full star signature, the paper proves decidability of the equational theory, but it does not obtain a finite relational axiomatization (McLean, 2019).

4. Relational presentations of continuous and recursive domains

In domain theory proper, complete domain relations arise as relational presentations of continuous structure. A CF-approximation space is a generalized approximation space $A$5 with $A$6 transitive and $A$7 satisfying a finite consistency condition. A subset $A$8 is CF-closed if every finite $A$9 is contained in some $\mathcal L(A)$0 with $\mathcal L(A)$1 and $\mathcal L(A)$2. The family $\mathcal L(A)$3 of CF-closed sets, ordered by inclusion, is a continuous domain, and every continuous domain is isomorphic to such a family of CF-closed sets (Wu et al., 2022).

The corresponding relational morphisms are CF-approximable relations $\mathcal L(A)$4, defined by five conditions: totality on generators, contravariance in the source approximation, downward closure in the target approximation, an approximation/interpolation condition, and finite directedness of outputs. From $\mathcal L(A)$5 one obtains a Scott continuous map

$\mathcal L(A)$6

and conversely every Scott continuous map $\mathcal L(A)$7 determines a CF-approximable relation $\mathcal L(A)$8. The paper proves

$\mathcal L(A)$9

and hence an equivalence of categories

$A$0

In this precise sense, CF-approximable relations are relational morphisms complete enough to recover Scott continuity exactly (Wu et al., 2022).

Pitts’ framework organizes recursive domain relations differently. A relational structure on a pointed $A$1-category $A$2 is a functor

$A$3

so every object $A$4 carries a complete lattice $A$5 of predicates or relations. If $A$6 is given by an isomorphism

$A$7

with minimal invariant property

$A$8

and if $A$9 is admissible and $D\subseteq \mathcal L(A)$0 acts on $D\subseteq \mathcal L(A)$1, then there exists

$D\subseteq \mathcal L(A)$2

The paper then reinterprets Pitts’ construction through Knaster–Tarski, inverse limits, Banach fixed points on the COFE of uniform relations, and Kleene fixed points (Amorim, 2022).

A predicative and univalent variant appears in the development of dcpos without resizing axioms. The paper explicitly notes that one common predicative strategy is to work with information systems, abstract bases, formal topologies, and approximable relations rather than dcpos and Scott-continuous maps. It chooses the opposite strategy—large but locally small dcpos with small directed suprema—but still develops an abstract-basis presentation. A $D\subseteq \mathcal L(A)$3-abstract basis $D\subseteq \mathcal L(A)$4 yields the rounded ideal completion $D\subseteq \mathcal L(A)$5, and $D\subseteq \mathcal L(A)$6 is a continuous $D\subseteq \mathcal L(A)$7-dcpo with basis $D\subseteq \mathcal L(A)$8. Moreover, a locally small continuous dcpo with basis $D\subseteq \mathcal L(A)$9 is a continuous retract of the algebraic dcpo $D$0 (Jong et al., 2020).

5. Completeness via approximation, topology, distance, and measure

A second cluster of meanings concerns completeness of spaces that arise from domain relations or approximation structure. In “Domains via approximation operators,” an auxiliary relation $D$1 on a poset induces lower and upper approximation operators

$D$2

For an auxiliary relation, interpolation

$D$3

is equivalent to idempotence of these operators: $D$4 on upper sets and $D$5 on lower sets. Most notably, a poset $D$6 is continuous iff

$D$7

equivalently iff

$D$8

The topology $D$9 associated to a pre-approximating relation coincides with the Scott topology when $D$0 on a continuous poset (Zou et al., 2016).

Topological completeness notions are developed further through domain-complete and LCS-complete spaces. A domain-complete space is a $D$1 subset of a continuous dcpo with the Scott subspace topology; an LCS-complete space is a $D$2 subset of a locally compact sober space. The inclusions

$D$3

always hold, and the second implication is strict in general. For countably-based $D$4 spaces,

$D$5

while for metrizable spaces,

$D$6

Every LCS-complete space is sober, Wilker, compactly Choquet-complete, completely Baire, and $D$7-consonant (Brecht et al., 2019).

“Distance domains” replace transitive relations by non-symmetric distances $D$8. The generalized relational notions are $D$9-directed subsets, $t_1\top \ge t_2\top$00-suprema, and $t_1\top \ge t_2\top$01-maxima, together with topological completeness conditions $t_1\top \ge t_2\top$02- and $t_1\top \ge t_2\top$03-completeness. In the hemimetric case, these are the usual Yoneda and Smyth completeness notions. The paper’s main contribution is to characterize a generalized Smyth completeness for arbitrary distances via $t_1\top \ge t_2\top$04-max completeness under interpolation hypotheses (Bice, 2017).

A measure-theoretic completion result appears in the domain-theoretic study of random variables. If $t_1\top \ge t_2\top$05 is a countably-based coherent domain, then the space $t_1\top \ge t_2\top$06 of sub-probability measures, viewed as a valuation domain, is again a countably-based coherent domain with an explicit basis of dyadic simple measures. If $t_1\top \ge t_2\top$07 is a complete chain with $t_1\top \ge t_2\top$08, then $t_1\top \ge t_2\top$09 and $t_1\top \ge t_2\top$10 are continuous lattices. The order on measures is characterized by integration against Scott-continuous test functions: $t_1\top \ge t_2\top$11 These results tie completeness of domains to completeness of associated spaces of valuations and random variables (Mislove, 2016).

6. Expressive completeness and broader categorical or applied senses

On finite domains, “complete domain relations” can mean primitive-positive expressive universality. A set of relations is pp-complete if every finitary relation on a finite domain $t_1\top \ge t_2\top$12 can be expressed by a primitive positive formula using only those relations. The paper proves a new decomposition theorem: $t_1\top \ge t_2\top$13 where $t_1\top \ge t_2\top$14 is an ordinary function. Since every finite-domain function decomposes into $t_1\top \ge t_2\top$15-input operations, every relation decomposes pp-wise into ternary relation graphs. The graph of any Sheffer function is therefore a single ternary relation that composes all relations. On finite domains with $t_1\top \ge t_2\top$16, ternary relations reduce further to binary relations through the explicit formula

$t_1\top \ge t_2\top$17

This yields a finite-domain version of Peirce’s reduction thesis (Koshkin, 17 Jun 2026).

A categorical abstraction of relational completeness appears in double categories of relations. A “double category of relations” is a locally posetal cartesian equipment in which every object is discrete in the sense of a Frobenius law, and with a subobject comprehension scheme it is equivalent to a double category $t_1\top \ge t_2\top$18 for some regular category $t_1\top \ge t_2\top$19. Under this characterization, proarrows are exactly relations $t_1\top \ge t_2\top$20, tabulators recover them as jointly monic spans, and functional completeness means that every proarrow is completely representable by its tabulator (Lambert, 2021).

In machine learning, the same phrase shifts meaning again. In domain generalization, domain relations are pairwise similarities between domains, collected in a matrix

$t_1\top \ge t_2\top$21

D$t_1\top \ge t_2\top$22G learns training-domain-specific predictors and reweights them for a test domain $t_1\top \ge t_2\top$23 by

$t_1\top \ge t_2\top$24

Here “complete” domain relations are the full set of weights needed to connect each unseen test domain to all training domains. The paper proves that relation-based reweighting can achieve stronger out-of-domain generalization than uniform averaging under smoothness assumptions in metadata space (Yao et al., 2023). A different machine-learning use appears in unpaired image translation, where DiscoGAN argues that adversarial realism alone does not recover a complete cross-domain relation and therefore couples two directions with reconstruction constraints to encourage approximate invertibility between domains (Kim et al., 2017).

A plausible conclusion is that the expression “complete domain relations” is best treated as a family resemblance term rather than a single doctrine. In social choice it marks maximal acyclic admissibility; in algebraic semantics, exact completeness of domain-comparison fragments; in domain theory, relations rich enough to recover Scott continuity or recursive invariants; in topology and analysis, completeness conditions induced by approximation structure; and on finite domains, expressive universality under primitive positive composition (Liversidge, 2020, Zhang et al., 2024, Wu et al., 2022, Brecht et al., 2019, Koshkin, 17 Jun 2026, Amorim, 2022).