---
title: Finite-Closed Topology
url: https://www.emergentmind.com/topics/finite-closed-topology
type: topic
---

# Finite-Closed Topology

Searching arXiv for papers directly relevant to the different established uses of “finite-closed topology.”
“Finite-closed topology” is not a single universally fixed term across mathematics. In the arXiv literature, it refers at least to: the finite topology on a vector space, defined by testing openness or closedness on every finite-dimensional subspace [1801.09085]; Michael’s lower semifinite topology on hyperspaces of closed sets, identified there with the finite-closed topology [0811.3276]; and a broader finite-closure-space setting in which closure systems, finite closure spaces, and their topological resolutions are studied categorically and combinatorially [2310.16636], [2111.12801]. A further, distinct usage appears in the theory of topological spaces with no infinite discrete subspace, where “finite-closed” is used descriptively for spaces in which closed sets decompose into finitely many irreducible closed subsets [1712.08739]. These usages share a common methodological theme: finitary control of closure or openness, but they arise in different categories and should not be conflated.

## 1. Finite topology on vector spaces

Let \(V\) be a real or complex vector space. The finite topology of \(V\) consists of all subsets \(U\) for which the intersection \(U \cap F\) is closed in \(F\) for every finite-dimensional linear subspace of \(V\) [1801.09085]. Equivalently, a set \(A \subseteq V\) is finite closed if \(A \cap F\) is closed in \(F\) for every finite-dimensional subspace \(F \subseteq V\) [1801.09085].

The corresponding open-set formulation is standard in the same paper: a subset \(U \subseteq V\) is finite open if, for every finite-dimensional linear subspace \(F \subseteq V\), the set \(U \cap F\) is open in \(F\), and the collection of all such finite open subsets forms a topology, denoted \(\tau_f(V)\) [1801.09085]. Formally,
$$
\tau_f(V) := \{ U \subseteq V : \forall F \subset V,\, \dim F < \infty \implies U \cap F \text{ is open in } F\}.
$$

This topology is closely related to extremal linear topologies. If \(V\) has countable dimension, then Kakutani and Klee showed
$$
\tau_f(V) = \tau_{\max}(V) = \tau_{\max}^{LC}(V)
$$
where \(\tau_{\max}(V)\) is the greatest vector space topology and \(\tau_{\max}^{LC}(V)\) is the greatest locally convex vector space topology [1801.09085]. In the uncountable-dimensional case, the situation changes sharply: \(\tau_f(V)\) is not a vector space topology, and
$$
\tau_{\max}^{LC}(V) \subsetneq \tau_{\max}(V) \subsetneq \tau_f(V).
$$
This establishes that the finite topology is maximal only in the countable-dimensional regime [1801.09085].

A further characterization uses norms. Let \(\mathcal{N}(V)\) be the set of all norms on \(V\), and for each norm \(N\), let \(\mathcal{O}_N\) denote its open sets. Define
$$
\tau_N(V) = \bigcup_{N \in \mathcal{N}(V)} \mathcal{O}_N.
$$
If \(V\) has countable dimension, then \(\tau_N(V) = \tau_f(V)\); otherwise, \(\tau_N(V)\) is not a topology, and in particular \(\tau_N(V) \neq \tau_f(V)\) [1801.09085]. Thus, in countable dimension, the finite topology is simply the union of all normed space topologies, while in uncountable dimension even that union fails to define a topology.

## 2. Domination of families of norms and the finite topology

The vector-space finite topology is tied in [1801.09085] to a domination problem for families of norms. A norm \(N\) dominates a norm \(N'\) if \(N' \leq \alpha N\) for some \(\alpha > 0\), and a family \((N_i)_{i \in I}\) is dominated if there is a norm \(N\) such that each \(N_i \leq \alpha_i N\) for some \(\alpha_i > 0\) [1801.09085]. This control problem is used to explain when finite-open sets can be realized as open in a single norm topology.

The main structural result is Theorem 5 of [1801.09085]: every \(I\)-indexed family of norms on \(V\) is dominated if and only if one of the following holds: \(V\) is finite-dimensional; \(I\) is finite; \(V\) is countable-dimensional and \(|I| < \mathfrak{b}\); or \(I\) is countable and \(\dim V < \mathfrak{b}\), where \(\mathfrak{b}\) is the bounding number [1801.09085]. The paper defines \(\mathfrak{b}\) as the minimum cardinality of a set of functions \(f:\omega \rightarrow \omega\) which is unbounded with respect to eventual domination [1801.09085].

This cardinal-theoretic criterion explains why countable-dimensional spaces behave particularly well. In that case, every countable family of norms is dominated, and this is used to show that every finite open set is open in some norm topology [1801.09085]. By contrast, when \(V\) or the index set is sufficiently large, undominated families of norms exist, and the union-of-norm-topologies description breaks down [1801.09085]. A plausible implication is that the finite topology sits at an interface between linear-topological structure and set-theoretic asymptotics.

## 3. Finite-closed topology in hyperspace theory

In hyperspace theory, “finite-closed topology” has a different, classical meaning. For a topological space \(X\), let \(F(X)\) denote the set of all closed subsets of \(X\). The paper “Hyperspaces of Closed Limit Sets” identifies Michael’s lower semifinite topology \(\tau_w\) with the finite-closed topology on \(F(X)\) [0811.3276].

The basis elements are
$$
U(\varnothing, \Phi) = \{ A \in F(X) \mid \forall O \in \Phi: A \cap O \neq \varnothing \},
$$
where \(\Phi\) is a finite family of open subsets of \(X\) [0811.3276]. Thus the topology is generated by finitely many intersection conditions with open sets. In the same paper, Fell’s topology \(\tau_s\) on \(F(X)\) has basis
$$
U(C, \Phi) = \{ A \in F(X)\mid A \cap C = \varnothing,\, \forall O \in \Phi : A \cap O \neq \varnothing \},
$$
with \(C\) compact and \(\Phi\) finite [0811.3276]. The difference is that Fell’s topology adds avoidance of compact sets, making it finer than \(\tau_w\).

Several standard properties are recorded in [0811.3276]. The Fell topology is compact on \(F(X)\); if \(X\) is locally compact, it is also Hausdorff, and if second countable, metrizable. The lower semifinite topology is weaker than Fell’s topology, is always \(T_0\), and is second countable when \(X\) is second countable [0811.3276]. The paper then restricts both topologies to the family \(L(X)\) of closed limit sets and to the subfamily \(\mathrm{ML}(X)\) of maximal limit sets.

A key result is that the two topologies coincide on maximal limit sets:
$$
(\mathrm{ML}(X), \tau_s) = (\mathrm{ML}(X), \tau_w)
$$
[0811.3276]. Also, the closure in \(\tau_w\) of the set of points embedded as singletons is exactly the set \(L(X)\) of closed limit sets [0811.3276]. In this usage, “finite-closed” does not concern finite-dimensional subspaces or finite closure spaces; it refers specifically to a hyperspace topology generated by finitely many hit conditions.

## 4. Closure spaces and finite closure spaces

A third major context is the theory of closure spaces. A closure space is a pair \((A,\mathcal{C})\), in which \(A\) is a set and \(\mathcal{C}\) a set of subsets of \(A\) closed under arbitrary intersections [2310.16636]. Equivalently, one can specify a closure operator \(X \mapsto \bar{X}\) satisfying extensivity, idempotency, and monotonicity [2310.16636]. Every topological space is a closure space, but not every closure space is topological, because closure spaces require closure under arbitrary intersections of closed sets, not arbitrary unions [2310.16636].

The paper “Strict monadic topology II: descent for closure spaces” studies morphisms \(f:A\to B\) such that preimages of closed sets are closed, and characterizes descent morphisms in the category \(\mathbf{CLS}\) of closure spaces [2310.16636]. For a morphism \(p:E\to B\), the following are equivalent: \(p\) is a pullback-stable regular epimorphism; for all \(X\subseteq B\), \(X\subseteq p(p^{-1}(X))\); for all \(X\subseteq B\), \(X=p(p^{-1}(X))\); and for all \(X\subseteq B\), the subset \(p(p^{-1}(X))\) is closed [2310.16636]. The paper’s main categorical results include that every descent morphism in the category of finite closure spaces is an effective descent morphism, and that every surjective closed map and every surjective open map of closure spaces is an effective descent morphism [2310.16636].

Finite closure spaces are also linked to finite topological spaces by a canonical resolution. For every finite closure space \(X\), one can define a finite topological space \(\operatorname{Top} X\) together with a natural projection \(\pi:\operatorname{Top} X\to X\) [2111.12801]. If \(\mathcal{M}(x)\) denotes the set of minimal neighborhoods of \(x\), then
$$
\operatorname{Top} X := \{ (x, M) \mid x \in X, M \in \mathcal{M}(x) \},
$$
with order
$$
xM \leq yN \quad :\Longleftrightarrow \quad N \subset M
$$
[2111.12801]. The projection \(\pi(xM)=x\) is both continuous and open, and if \(X\) is already a topological space, then \(\operatorname{Top} X \cong X\) [2111.12801]. This construction is designed to allow the techniques of topological combinatorics to be applied to finite closure spaces [2111.12801].

This suggests a broad interpretation of “finite-closed” as finitary closure structure, but the papers keep the notions distinct: closure spaces, finite closure spaces, and finite topologies are treated as related but not identical objects [2310.16636], [2111.12801].

## 5. Regular closed sets and finite closure-space lattices

The lattice-theoretic study of closure spaces provides another strand of the subject. For a closure space \((P,\varphi)\) with \(\varphi(\emptyset)=\emptyset\), the closures of open subsets of \(P\), called the regular closed subsets, form an ortholattice \(\Reg(P,\varphi)\), extending the poset \(\Clop(P,\varphi)\) of all clopen subsets [1307.1480]. Here the interior operator is
$$
\mathrm{int}(x) = P \setminus \varphi(P \setminus x),
$$
and a subset is regular closed when \(x=\varphi(\mathrm{int}(x))\) [1307.1480].

The paper proves several structural results for finite closure spaces [1307.1480]. In any finite convex geometry, \(\Reg(P,\varphi)\) is pseudocomplemented. For finite closure spaces of semilattice type, \(\Reg(P,\varphi)\) satisfies an infinite collection of stronger and stronger quasi-identities, weaker than both meet- and join-semidistributivity, although it may fail semidistributivity. If \(\Reg(P,\varphi)\) is semidistributive, then it is a bounded homomorphic image of a free lattice [1307.1480].

One sharp criterion concerns the clopen poset. The paper states that \(\Clop(P,\varphi)\) is a lattice if and only if every regular closed set is clopen [1307.1480]. In many important cases, \(\Reg(P,\varphi)\) is the Dedekind–MacNeille completion of \(\Clop(P,\varphi)\), including the semilattice setting and certain graph-theoretic constructions such as block graphs and cycles [1307.1480]. The extended permutohedron \(R(G)\) on a graph \(G\) and the extended permutohedron \(\Reg(S)\) on a join-semilattice \(S\) are both defined as lattices of regular closed sets of suitable closure spaces [1307.1480].

Within an encyclopedia treatment of finite-closed topology, these results are relevant because they show how closure notions on finite carriers generate robust lattice structures even when the ambient setting is not an ordinary topology.

## 6. FAC spaces and the descriptive use of “finite-closed”

A different use of the term appears in the theory of spaces with no infinite discrete subspace. A topological space \(T=(E,\mathcal{F})\) is called an FAC space if it contains no infinite discrete subset [1712.08739]. The paper proves that the following are equivalent: no infinite subset of \(E\) is discrete; every closed set is a finite union of irreducible closed subsets; and every closed set contains a dense subset on which the induced topology is Noetherian [1712.08739].

The same paper gives further equivalent conditions: no infinite subspace is both sober and \(T_1\); no infinite subspace is a KC-space; no infinite subspace is Hausdorff; and no subspace contains an infinite relatively Hausdorff subset [1712.08739]. It also establishes a min-max property: for every closed set \(C\) in an FAC space, the maximum cardinality of relatively Hausdorff subsets of \(C\), the maximum number of pairwise disjoint non-empty open subsets of \(C\), the least number of irreducible closed subsets needed to cover \(C\), and the least number of hyperconnected subspaces covering \(C\) are finite and equal [1712.08739].

The paper remarks that FAC spaces are sometimes called “finite-closed” or “finite-union-of-irreducibles” spaces, capturing that all closed sets break into finitely many irreducible atoms [1712.08739]. This terminology is descriptive rather than definitional. It does not denote the finite topology on a vector space or the lower semifinite topology on a hyperspace; instead it names a structural property of the lattice of closed sets.

## 7. Related finite-topological constructions and terminological boundaries

Several adjacent arXiv works clarify the broader landscape in which finite-closed terminology appears. “Finite topologies for finite geometries” develops finite topological spaces from finite abstract simplicial complexes and finite graphs using stars \(U(x)=\{y\mid x\subset y\}\) as a basis, with closed sets given by subcomplexes [2301.03156]. All such spaces are Alexandroff, every point has a smallest neighborhood, and classical formulas such as the Lefschetz fixed point theorem are shown to hold for continuous maps on finite topological spaces [2301.03156]. This paper concerns finite topologies rather than the specific finite topology of [1801.09085] or the finite-closed hyperspace topology of [0811.3276].

“Introduction To Typed Topological Space” addresses a standard pathology of finite topologies: in classical finite topological spaces, especially those with the \(T_1\) property, every singleton is both open and closed [1804.04627]. The paper introduces type assignments from a finite bounded distributive lattice and defines closure and neighborhood systems relative to chains of types, with
$$
A^c = \{ x \in X : \forall U \in J_c(x),\, U \cap A \neq \emptyset \}
$$
for the \(c\)-closure of \(A\) [1804.04627]. This is another finitary closure formalism, but not one called finite-closed topology in the cited work.

A further source of possible confusion is the algebraic use of “finite topology” on module duals. For a right \(R\)-module \(M\), the dual \(M^*=\mathrm{Hom}_R(M,R)\) carries the finite topology whose neighborhoods of \(0\) are
$$
F^- := \{ f \in \mathrm{Hom}_R(M,N) \mid f(x) = 0,\, \forall x \in F \}
$$
for finitely generated submodules \(F\leq M\) [1109.3003]. The closed submodules of \(M^*\) are precisely those of the form \(X^-\), and the resulting annihilator correspondence is an anti-isomorphism of lattices for all modules if and only if \(R\) is a PF ring [1109.3003]. This is again a “finite topology,” but not the hyperspace finite-closed topology.

The terminological boundary is therefore essential. “Finite-closed topology” may denote Michael’s lower semifinite hyperspace topology [0811.3276]; it may denote the finite topology on a vector space when phrased in terms of finite closed sets [1801.09085]; or it may be used descriptively for topological spaces whose closed sets decompose finitely [1712.08739]. A plausible implication is that the term functions more as a family resemblance than as a single canonical concept: in each context, finite data determine closure behavior, but the ambient category, objects, and closure operators differ substantially.

Source: https://www.emergentmind.com/topics/finite-closed-topology