WorksMathematics
Subset-system derived topologies and coreflective categories of -spaces
Abstract
Given a subset system on -spaces, we associate with each topology a closure-testing derived topology and study the fixed points of this operator. The -determined topologies above a given topology are closed under intersections and hence determine a least fixed-point hull. We prove that this hull is obtained by transfinite iteration, is functorial, and defines a right adjoint to the inclusion of -determined spaces into . Consequently, the corresponding full subcategory is coreflective and complete. We isolate restriction-stability and open-locality hypotheses under which determinedness is inherited by closed and open subspaces, and we establish a general non-reflectivity criterion. If is contained in the directed-subset system, the category of -determined spaces is Cartesian closed; its exponentials are obtained by coreflecting the pointwise convergence topology on continuous maps. The construction subsumes familiar monotone- and irreducible-determined cases while separating the formal consequences of the subset-system axioms from the additional hypotheses needed for heredity and exponentials. It therefore provides a uniform categorical mechanism for producing complete Cartesian closed categories of -spaces with explicit higher-order function spaces, as required in denotational semantics.
Introduction
Domain theory provides order-theoretic models of computation in which directed suprema describe approximation and the Scott topology records observable information; see, for example, [1, 5, 6]. From the viewpoint of denotational semantics, a category of semantic spaces is particularly useful when it supports products and higher-order function spaces; Cartesian closed categories provide exactly this structure through exponentials. A parallel line of work develops the order-theoretic core of domain theory directly on -spaces. Monotone determined spaces are a basic example: their topology is detected by directed subsets, and they form a Cartesian closed category extending the Scott-topological setting; see [4, 7, 14]. The question pursued here is whether this categorical infrastructure can be generated uniformly from a functorially chosen class of distinguished subsets, rather than constructed separately for each convergence or approximation notion.
Many constructions in this area describe a topology by specifying which distinguished subsets are allowed to detect open sets. Replacing directed subsets by irreducible subsets leads to irreducible determined spaces. In [13], the corresponding derived topology, its transfinite idempotent hull, hereditary properties and categorical structure were studied for the irreducible subset system. The purpose of the present paper is not to replace the irreducible system there by a formal parameter. Instead, we isolate the fixed-point mechanism behind these constructions and determine what follows for an arbitrary subset system
The notion of subset system, inspired by the uniform closure constructions of Wright, Wagner and Thatcher [10], was used in [11] to organize several sobriety-type notions. Chen and Xu [3] also studied topological spaces determined by subset systems through an induced-operator framework. Here the subset system is used in a different way: from a -topology we form a closure-testing operator . This operator is monotone and inflationary on the poset of -topologies on the fixed underlying set, but its idempotence is not part of the subset-system axioms. The central object is therefore its least fixed point above .
A second recent uniform approach is due to Chen, Kou and Lyu [2]. Their construction starts from convergence classes and C-generated spaces; in particular, it recovers monotone determined spaces as a Cartesian closed coreflective category and treats directed, chain and monotone sequential convergence classes in a common framework. The present construction organizes the data differently. It begins with a closure-testing operator on the lattice of -topologies: the distinguished family depends functorially on the current space, candidate open sets are tested against closures of -sets, and the one-step operator may require transfinite iteration before reaching a fixed point. Thus the two approaches overlap in the monotone-determined case, but the fixed-point completion, its ordinal length and the associated heredity questions are intrinsic to the closure-testing operator studied here.
Our main results form three categorical levels. First, the -determined topologies above a given -topology are closed under intersections. Hence there is a least fixed-point hull , obtained by transfinite iteration of ; the assignment is a functorial closure operator on the poset of -topologies on a fixed set. Second, this hull defines a coreflector: the full category HD of -determined spaces is coreflective in and therefore complete. We also obtain retract stability, a general non-reflectivity criterion, and restriction and localization hypotheses that ensure closed- and open-heredity. Third, if , then HD is Cartesian closed. The exponential is obtained by applying the coreflector to the pointwise convergence topology on the set of continuous maps. Thus the same fixed-point mechanism produces the completion, the adjunction and explicit higher-order function spaces, which is the categorical structure most directly relevant to denotational semantics.
The paper is organized around these structural points. Section 2 introduces subset-system derived topologies and records the basic order-theoretic relations among them. Section 3 develops the fixed-point completion, its transfinite description and functoriality. Section 4 turns the completion into a coreflector and derives completeness, retraction stability and non-reflectivity criteria. Section 5 treats subspaces under explicit stability assumptions on . Section 6 constructs exponentials and establishes Cartesian closedness when . The final section records questions suggested by the general framework.
Subset-system derived topologies
We recall only the notation needed below. For a -space , the specialization order is denoted by ; thus iff . We write for the Alexandroff topology of upper sets of this order. A nonempty subset is directed if every finite subset of has an upper bound in . The family of directed subsets of is denoted by , and denotes the family of nonempty irreducible subsets of .
Definition 2.1 A covariant functor is a subset system if, for every -space ,
where , and for every continuous map one has
If for every , we write .
Standard examples are the systems and of singletons, nonempty finite subsets, nonempty chains, directed subsets, Rudin subsets, well-filtered determined subsets, irreducible subsets and nonempty subsets, respectively. The familiar inclusions include
see [12].
In what follows, the capital letter always denotes a subset system .
Definition 2.2. Let be a -space. A subset is -open relative to if
The family of all such sets is denoted by . The topology generated by as a subbasis is denoted by and is called the -derived topology. The space is -determined if
The full subcategory of formed by the -determined spaces is denoted by .
Since every -open set satisfies the defining condition, . Moreover every member of is an upper set for the specialization order because .
Proposition 2.3. Let be a -topology on .
. In particular, and have the same specialization order.
If , then and .
If are -topologies on the same set, then .
and .
The family is a topology; hence .
If , then is a topology and therefore .
Proof. For (1), only the last inclusion needs comment. If and , then ; since , we obtain . Thus is upper. Since both and induce the same specialization order, every topology between them does as well.
For (2), increasing the family of test sets strengthens the defining condition. For (3), the identity is continuous, so , while .
For (4), the closure of a finite set is the finite union of the closures of its points; hence the singleton and finite tests both yield exactly the upper sets. The assertion for is the usual closure characterization of open sets.
For (5), arbitrary unions cause no difficulty. If and a chain has closure meeting , choose and . The two points are comparable, and are upper; hence one of belongs to .
The same argument proves (6): if and has closure meeting , choose and . A common upper bound belongs to because both sets are upper.
Corollary 2.4. If , every -determined space is -determined. Thus
as full subcategories of .
Proof. If is -determined, then
so . Since , Proposition 2.3(2) gives
Hence , and therefore .
The fixed-point condition also has a useful closed-set formulation.
Proposition 2.5 A -space is -determined if and only if every subset satisfying
is closed.
Proof. Take complements in Definition 2.2.
The general construction contains several familiar cases. For one recovers the monotone determined topology of [4, 7, 14]; for one recovers the irreducible derived topology of [13]. At the two extremes, - and -determined spaces are precisely the Alexandroff -spaces, while every -space is -determined. Notice already that categorical statements cannot be uniform in without additional hypotheses.
Fixed-point completion and transfinite iteration
The operator is monotone and inflationary by Proposition 2.3, but it need not be idempotent for an arbitrary subset system. We therefore pass to its least fixed point above a given topology.
Lemma 3.1 Let be a family of -determined topologies on a set , all containing a fixed -topology . Then is again an -determined -topology containing .
Proof. The intersection is because it contains the -topology . By monotonicity,
for every . Hence , and equality follows.
Definition 3.2 For a -topology on , define
This family is nonempty because is -determined and (note that is equipped with the specialization order).
Theorem 3.3 For every -space , we have the following conclusions:
is the least -determined topology containing ;
(2) , where is equipped with the order ;
(3) and have the same specialization order;
(4) the assignment is a closure operator on the poset of -topologies on : it is inflationary, monotone and idempotent.
Proof. Parts (1) and (2) follow from Lemma 3.1. Part (3) follows from the two inclusions . Monotonicity in (4) follows because the class of -determined topologies above a larger initial topology is smaller; idempotence follows from (1).
We next describe the hull by iteration. Put and define, for ordinals ,
where the join is the topology generated by .
Theorem 3.4 There is a least ordinal such that
For this ordinal,
Proof. By transfinite induction,
for every , and the chain is increasing. Put . If the iteration did not stabilize before , then for every we could choose
Whenever , monotonicity gives , whereas ; hence the sets are pairwise distinct. This would give an injection , a contradiction. Thus the iteration stabilizes.
Let be the first stationary stage. Then is -determined and contains , so by minimality. The reverse inclusion follows from the displayed chain.
Remark 3.5 The distinction between and is essential in the general theory. For several standard systems the first step is already idempotent (for example , , and ), while for the transfinite construction gives the fixed-point hull considered in [13]. A useful general problem is to characterize those subset systems for which is idempotent on every -topology.
The construction is functorial.
Proposition 3.6 If is continuous, then the same underlying map is continuous
and
Proof. Let and . If , then
and . Hence and therefore . This proves the first assertion. Iterating the argument transfinitely, including the evident limit step, gives continuity at every stage of Theorem 3.4 and hence the second assertion.
For maps whose domain is already a fixed point, continuity has a particularly simple test.
Proposition 3.7 Let be -determined, let be a -space, and let be a map. Then the following are equivalent:
is continuous;
for every ;
is continuous.
Proof. The implication (1) (2) is standard, and (1) (3) follows from Proposition 3.6 because . The implication (3) (1) follows by composing with the identity .
For completeness, assume (2). If and have , then . Since is open, . Hence , proving (1).
The coreflector and categorical consequences
We use standard categorical terminology as in, for example, [8]. Let be the inclusion. Define
Proposition 3.6 makes a functor.
Theorem 4.1. The functor is right adjoint to the inclusion . Equivalently, is coreflective in , with coreflection map
Proof. For and , Proposition 3.7 gives, naturally in both variables,
with the same underlying set maps on the two sides.
This formulation makes the limit theory immediate.
Corollary 4.2. The category is complete. If a diagram in has limit in , then its limit in is
In particular, for -determined spaces ,
Proof. The category is complete, and the right adjoint preserves limits.
Proposition 4.3. Every retract of an -determined space is -determined.
Proof. Let and be continuous with , and suppose is -determined. If , Proposition 3.6 gives . Hence
Thus .
Reflectivity behaves differently. We use the criterion of Shen, Xi and Zhao [9]: if a full subcategory of is closed under homeomorphism and is not contained in , then reflectivity of implies that every sober space belongs to .
Proposition 4.4. For every subset system , the category contains non- spaces. Consequently, if , then is not reflective in .
Proof. First note that is closed under homeomorphisms. Indeed, let be a homeomorphism and suppose that is -determined. Proposition 3.6 gives a continuous map
Hence implies , and the original homeomorphism then gives . Thus , while the reverse inclusion is automatic, so is -determined.
Every Alexandroff -space is -determined by Proposition 2.3(1). In particular, the Sierpiński space belongs to HD and is not . The reflectivity criterion of Shen, Xi and Zhao [9] now applies and proves the second statement.
Theorem 4.5 If is an irreducible subset system, i.e. , then HD is not reflective in .
Proof. Let be any non-discrete Hausdorff space. It is sober, while . Since , we have , and therefore
which is the discrete topology because is . Thus is not -determined, so . Proposition 4.4 applies.
Corollary 4.6 The categories of -, -, - and -determined spaces are not reflective in . The same holds for -determined spaces because the - and -determined spaces coincide.
Remark 4.7 The hypothesis in Theorem 4.5 cannot simply be omitted. For , Proposition 2.3(4) gives for every -topology, so HD ; in particular this category is reflective as well as coreflective. This also shows why a blanket non-reflectivity assertion for arbitrary subset systems is false.
Subspaces: stability hypotheses on the subset system
For the irreducible system , irreducibility of a subset is unchanged when it is regarded inside a subspace containing it; this fact underlies the hereditary results in [13]. An arbitrary subset system need not have this property, and functoriality of only gives one direction. We therefore make the required hypotheses explicit.
Definition 5.1 A subset system is called
restriction-stable if, whenever is a nonempty subspace of and , the implication
holds;
open-local if, whenever is open in and with , one has
The reverse implication in (1) is automatic from the subset-system axioms because the inclusion is continuous. The standard systems , , , and are restriction-stable and open-local. For , for example, a nonempty intersection of a directed set with an open upper set is again directed; for , a nonempty intersection of an irreducible set with an open set is irreducible in the open subspace.
Theorem 5.2 If is restriction-stable, then every nonempty closed subspace of an -determined space is -determined.
Proof. Let be -determined, let be nonempty and closed, and let be -open relative to the subspace topology. Put . We claim that .
Suppose and . If there is nothing to prove. Otherwise . Restriction-stability gives , while is closed, so . Hence
and the defining property of yields , a contradiction. Thus is -open in and therefore open because is -determined. Consequently is open in .
Theorem 5.3 If is open-local, then every nonempty open subspace of an -determined space is -determined.
Proof. Let be -determined, be nonempty and open, and let be -open relative to . We show directly that is -open in .
Take with . Then , and since is open, . Put . By open-locality, . Moreover
and is closed, so
Intersecting with gives , equivalently . Hence , and therefore . Thus is -open in , so it is open. In particular it is open in .
Corollary 5.4 For , the property of being -determined is both closed-hereditary and open-hereditary.
For , Corollary 5.4 recovers the corresponding results of [13]; the point here is that the general argument identifies precisely the properties of the subset system that are used.
Cartesian closedness below the directed subset system
Assume throughout this section that
For , write for their categorical product in , namely the -determined hull of their product in as in Corollary 4.2.
For -spaces , let be the set of continuous maps from to . Equip with the topology of pointwise convergence, whose subbasic open sets are
The pointwise topology is . Indeed, if , choose with and use the property of to find an open set that contains exactly one of and ; the corresponding subbasic set separates and . Moreover, its specialization order is the pointwise order: if and only if for every . Define
Thus is automatically an object of .
Lemma 6.1 For , the evaluation map
is continuous.
Proof. We use Proposition 3.7. Let
and let . The two projections are continuous, hence
The topology of is finer than the pointwise topology, so also belongs to the pointwise closure of .
Let contain . Since is continuous, is an open neighborhood of . Choose with . Then . Because lies in the pointwise closure of , the subbasic neighborhood meets ; choose such that .
Now , so is directed in the specialization order. Choose above both and . By Theorem 3.3(3), the categorical product has the same specialization order as the ordinary product, and the specialization order on is the pointwise order. Since every continuous map is order preserving,
Because is an upper set, . Thus every open neighborhood of meets , and hence
Proposition 3.7 now yields continuity of .
Theorem 6.2 If , then is Cartesian closed. For , the object defined above is an exponential of by .
Proof. Let . Given a continuous map
define by . For each fixed , the section is the pairing of the constant map with and is therefore continuous by the universal property of the categorical product. Thus is continuous. Similarly, for each fixed , the section is continuous, so is continuous. Hence is continuous by the defining initial property of the pointwise topology. Since is -determined, Theorem 4.1 (or Proposition 3.7) lifts this uniquely to a continuous map
Conversely, if is continuous, then
is continuous by Lemma 6.1. These two constructions are inverse to one another on underlying maps. Therefore there is a natural bijection
which is precisely the exponential universal property.
Corollary 6.3 The categories of -, - and -determined spaces are Cartesian closed.
Proof. The - and -determined spaces coincide by Proposition 2.3(4), while . Apply Theorem 6.2.
Concluding remarks and questions
The fixed-point formulation separates three levels of the construction: the one-step topology , its idempotent hull , and the categorical coreflector . This makes clear which conclusions are formal consequences of the subset-system axioms and which require additional structure on .
From the standpoint of programming semantics, the main categorical consequence is Theorem 6.2: for , the same fixed-point mechanism that produces the coreflective category also supplies its exponentials. The paper does not commit to a particular programming language or type system; rather, it isolates a reusable topological and categorical mechanism that can be tested in future against concrete domain-theoretic models and higher-order type interpretations.
The following questions arise naturally.
Question 7.1 Characterize the subset systems for which is idempotent on every -topology, equivalently, for which always holds.
Question 7.2 Can the restriction-stability and open-locality assumptions in § 5 be replaced by weaker intrinsic conditions that are still necessary and sufficient for closed- and open-heredity of -determined spaces?
Question 7.3 Characterize the subset systems for which is reflective in . Theorem 4.5 gives a broad negative class, while gives the opposite extreme.
Question 7.4 Can the hypothesis in Theorem 6.2 be weakened? More generally, which properties of guarantee continuity of the evaluation map for the coreflected pointwise function space?
Statements and Declarations
Funding. This research was supported by the National Natural Science Foundation of China (Nos. 12471070 and 12071199).
Author contributions. Xiaoquan Xu: Conceptualization, Investigation, methodology, formal analysis, writing-review and editing. Xinjie Huang: Conceptualization, investigation, formal analysis, writing-original draft. Yi Yang: Investigation, formal analysis, writing-review and editing All authors have reviewed and approved the final manuscript.
Competing interests. The authors have no relevant financial or non-financial interests to disclose.
Data availability. No datasets were generated or analysed in this theoretical study.
Use of AI-assisted tools. During the preparation of this work, the authors used ChatGPT (OpenAI) to assist with organization, language and readability. The authors reviewed and edited the resulting text, verified the mathematical content and references, and take full responsibility for the final manuscript.
References
- [1]Abramsky, S., Jung, A.: Domain theory. In: Abramsky, S., Gabbay, D.M., Maibaum, T.S.E. (eds.) Handbook of Logic in Computer Science, vol. 3, pp. 1–168. Clarendon Press, Oxford (1994)
- [2]Chen, Y., Kou, H., Lyu, Z.: Monotone determined spaces via C-generated spaces. Topology Appl. 393, 109973 (2026). https://doi.org/10.1016/j.topol.2026.109973
- [3]Chen, Y., Xu, X.: On topological spaces determined by subset systems. Appl. Math. J. Chinese Univ. 34, 487–491 (2019). https://doi.org/10.3785/amjcu-a.2019.04.0487
- [4]Erné, M.: Infinite distributive laws versus local connectedness and compactness properties. Topology Appl. 156, 2054–2069 (2009). https://doi.org/10.1016/j.topol.2009.03.029
- [5]Gierz, G., Hofmann, K.H., Keimel, K., Lawson, J.D., Mislove, M., Scott, D.S.: Continuous Lattices and Domains. Encyclopedia of Mathematics and its Applications, vol. 93. Cambridge University Press, Cambridge (2003). https://doi.org/10.1017/CBO9780511542725
- [6]Goubault-Larrecq, J.: Non-Hausdorff Topology and Domain Theory. New Mathematical Monographs, vol. 22. Cambridge University Press, Cambridge (2013). https://doi.org/10.1017/CBO9781139524438
- [7]Luo, S., Xu, X.: On monotone determined spaces. Electron. Notes Theor. Comput. Sci. 333, 63–72 (2017). https://doi.org/10.1016/j.entcs.2017.08.006
- [8]Mac Lane, S.: Categories for the Working Mathematician, 2nd edn. Springer, New York (1997). https://doi.org/10.1007/978-1-4757-4721-8
- [9]Shen, C., Xi, X., Zhao, D.: The reflectivity of some categories of T₀ spaces in domain theory. Rocky Mountain J. Math. 54, 1149–1166 (2024). https://doi.org/10.1216/rmj.2024.54.1149
- [10]Wright, J.B., Wagner, E.G., Thatcher, J.W.: A uniform approach to inductive posets and inductive closure. Theoret. Comput. Sci. 7, 57–77 (1978). https://doi.org/10.1016/0304-3975(78)90040-3
- [11]Xu, X.: On H-sober spaces and H-sobrifications of T₀ spaces. Topology Appl. 289, 107548 (2021). https://doi.org/10.1016/j.topol.2020.107548
- [12]Xu, X., Shen, C., Xi, X., Zhao, D.: On T₀ spaces determined by well-filtered spaces. Topology Appl. 282, 107323 (2020). https://doi.org/10.1016/j.topol.2020.107323
- [13]Xu, X., Yang, Y., Luo, S.: On irreducible determined T₀-spaces. Topology Appl. 382, 109762 (2026). https://doi.org/10.1016/j.topol.2026.109762
- [14]Yu, Y., Kou, H.: Directed spaces defined through T₀ spaces with specialization order. J. Sichuan Univ. Nat. Sci. Ed. 52, 217–222 (2015) (in Chinese)