betaFeedback

WorksMathematics

Wilker property, function spaces and hyperspaces

Abstract

The Wilker property requires that a compact set covered by two open sets admit compact covering pieces inside those open sets. We study this property through products, compact-open function spaces and hyperspaces. We construct countable compact Fréchet–Urysohn T1T_1 hereditarily Wilker spaces with non-Wilker products and squares. For spaces whose compact subsets are closed, Wilker-ness is equivalent to every compact subspace being Hausdorff; this yields a positive product theorem when one Wilker factor has closed compact subsets. For an irreducible subset system HH, we characterize preservation of HH-sobriety by compact-open function spaces in terms of HH-supercompactness of compact subsets of the domain, and show that Wilker spaces satisfy this criterion. The Hoare hyperspace of a space is Wilker exactly when binary union on the Scott space of its open-set lattice is jointly continuous for the ordinary product topology. Consequently, the Hoare construction preserves and reflects Wilker-ness on consonant bases, although preservation fails in general. We also identify core-compactness of the base with local compactness of its Hoare hyperspace and prove that the Smyth construction reflects Wilker-ness.

Acknowledgements. This research was supported by the National Natural Science Foundation of China (Nos. 12471070 and 12071199).

Introduction

The compact-cover splitting condition studied in this paper has two classical sources. It appears already in Choquet’s theory of capacities, where a family of sets satisfying the corresponding binary refinement property is called rich [3]. It was later isolated by Wilker as condition (D) in his study of adjoint product and hom functors [17]. We follow the established terminology and call a space XX Wilker if, whenever a compact set KK is contained in the union of two open sets U1,U2U_1, U_2, one can cover KK by compact sets Ki⊆UiK_i \subseteq U_i. Wilker proved, in particular, that every Hausdorff space and every locally compact space has this property.

The condition also occurs in measure-extension theory [10] and in compact-open constructions for multifunction spaces [14]. More recently, every quasi-Polish space was shown to be Wilker [2], every LCS-complete space is Wilker [1], and co-consonance implies Wilker-ness through strong compactness properties [23].

We develop the basic theory of the Wilker property and examine its connections with compact-open function spaces and hyperspaces. The hyperspaces considered here are the Hoare and Smyth powerspaces, equipped with the lower and upper Vietoris topologies, respectively. These constructions connect structural questions in general topology with the order-theoretic study of open-set lattices.

We first study the topology of the class itself. Wilker spaces are hereditary with respect to open, closed and locally closed subspaces, are preserved by retracts, and are preserved and reflected by arbitrary topological sums. Arbitrary heredity, however, fails in a particularly strong way: every T0T_0-space is homeomorphic to a subspace of a locally hypercompact Wilker space. A new separation-theoretic feature emerges when Wilker-ness is combined with the KC property: we prove that

X is KC and Wilker ⟺ every compact subspace of X is Hausdorff.X\text{ is KC and Wilker}\ \Longleftrightarrow\ \text{every compact subspace of }X\text{ is Hausdorff}.

This characterization yields a mixed product theorem. If XX is Wilker and every compact subspace of YY is Hausdorff, then X×YX \times Y is Wilker. Consequently,

X,Y Wilker and one factor KC ⟹ X×Y Wilker,X,Y\text{ Wilker and one factor KC}\ \Longrightarrow\ X \times Y\text{ Wilker},

and in particular the product of a Wilker space with a Hausdorff space is Wilker. The KC hypothesis cannot be omitted. We construct two countable compact Fréchet–Urysohn T1T_1 anti-Hausdorff hereditarily Wilker spaces whose product is not Wilker (Theorem 3.16). A finite sum gives a hereditarily Wilker space with a non-Wilker square; a co-finite-core construction gives an anti-Hausdorff square counterexample as well. Consequently, the full category of Wilker spaces is neither reflective nor epireflective in Top0\mathbf{Top}_0.

The counterexample uses the co-finite modifications studied by Xu and Ji in their preprint Products of co-sober spaces need not be co-sober (2026, arXiv:2610.14434v1) for co-sobriety. Its additional ingredient is that every proper closed subspace of the particular Kovár–van Douwen seed is compact Hausdorff (Lemma 3.14). This makes the singleton modifications hereditarily Wilker. An explicit binary open cover of a compact diagonal then has no compact refinement: the projections of any proposed pieces would express the original anti-Hausdorff KC-space as a union of two proper closed sets. The modifications themselves are credited to the earlier work; the Wilker analysis and the compact-cover obstruction are established here. Together with the mixed product theorem, this locates failure of productivity precisely outside the guarantee supplied by a KC factor.

We next compare Wilker-ness with core-compactness. The rational line is Wilker but not core-compact. In the opposite direction, we show that every subspace of B×JB \times J is Wilker whenever BB is Hausdorff and J=[0,1)J = [0, 1) carries the Scott topology of the reverse usual order. Applied to the classical Hofmann–Lawson construction, this yields a second-countable core-compact Wilker T0T_{0}-space that is not locally compact. Hence

locally compact⟹core-compact and Wilker\text{locally compact} \Longrightarrow\text{core-compact and Wilker}

is a strict implication, while the question whether every core-compact T0T_{0}-space is Wilker remains open.

The Scott-topological consequences make the connection with domain theory more explicit. The Scott space of every quasicontinuous poset is locally compact [13]; hence it is Wilker. For complete lattices one can go further: since their Scott spaces are well-filtered, core-compactness already forces local compactness and therefore Wilker-ness [21, 23]. On the other hand, complete lattices have coherent Scott spaces, and Zhao, He and Wang showed that coherent Wilker spaces are co-sober. Combining this with the recent construction of Ji and Xu of a complete Boolean algebra with a non-co-sober Scott space, given in their preprint Scott spaces of complete Boolean algebras need not be co-sober (2026, arXiv:2608.08484v1), yields a complete Boolean algebra whose Scott space is not Wilker. A parallel phenomenon occurs for upper topologies: the dual of Isbell’s lattice, together with the lower-topology example of Wen and Xu and the coherence theorem of Xu, Yang and Chen, yields a complete lattice whose upper topology is not Wilker. Thus strong order-completeness alone does not force Wilker-ness for either of these standard one-sided order topologies. The unrestricted Boolean, Heyting and meet-continuous Scott-space versions all fail, while the corresponding countable questions for complete Heyting algebras and complete meet-continuous lattices remain open.

The second part of the paper turns from the basic theory to compact-open function spaces. For a T0T_{0}-space XX, let

Ok(X)=(O(X),ϕ(O(X))),\mathcal{O}_{k}(X) = \bigl(\mathcal{O}(X), \phi(\mathcal{O}(X))\bigr),

where ϕ(O(X))\phi(\mathcal{O}(X)) is generated by the sets

Φ(K)={U∈O(X):K⊆U},\Phi(K) = \{U \in\mathcal{O}(X) : K \subseteq U\},

with KK compact in XX. Given an irreducible subset system HH, we call a subset K⊆XK \subseteq X HH-supercompact if every HH-family in Ok(X)\mathcal{O}_{k}(X) whose union covers KK already contains one member covering KK. For a fixed XX, we prove the equivalence of the following four conditions:

  1. Ok(X)\mathcal{O}_{k}(X) is HH-sober;

  1. every compact subset of XX is HH-supercompact;

  1. for every T0T_0-space YY the compact-open topology on C(X,Y)C(X,Y) has property SS with respect to HH;

  1. Ck(X,Y)C_k(X,Y) is HH-sober for every HH-sober space YY.

This reorganizes several compact-open function-space questions from [20] into a single structural statement.

Wilker’s condition is the main bridge to this theory. We prove that every compact subset of a Wilker space is HH-supercompact for every irreducible subset system HH. Consequently, if XX is Wilker, then Ok(X)\mathcal{O}_k(X) is HH-sober for every HH, the compact-open topology on C(X,Y)C(X,Y) has property SS with respect to every HH, and Ck(X,Y)C_k(X,Y) is HH-sober whenever YY is. In particular, Ok(X)\mathcal{O}_k(X) is sober for every Wilker space XX. If XX is also consonant, then ϕ(O(X))=σ(O(X))\phi(\mathcal{O}(X))=\sigma(\mathcal{O}(X)), and the sobriety of the Scott space ΣO(X)\Sigma\mathcal{O}(X) follows immediately.

The third part concerns the Hoare and Smyth hyperspace constructions. For the nonempty Hoare powerspace we prove

X core-compact⟺PH(X) locally compact⟺PH(X) core-compact.X\text{ core-compact} \Longleftrightarrow P_H(X)\text{ locally compact} \Longleftrightarrow P_H(X)\text{ core-compact}.

We further show that every PH(X)P_H(X) is coherent. Since coherent Wilker spaces are co-sober, a known countable zero-dimensional metrizable space MM with non-co-sober PH(M)P_H(M) gives

M Wilker,PH(M) not Wilker.M\text{ Wilker},\quad P_H(M)\text{ not Wilker}.

Thus the Hoare construction does not preserve Wilker-ness in general. A further characterization identifies exactly what the Wilker property of PH(X)P_H(X) measures:

PH(X) Wilker⟺∪:ΣO(X)×ΣO(X)→ΣO(X) is continuous.P_H(X)\text{ Wilker}\Longleftrightarrow\cup:\Sigma\mathcal{O}(X)\times\Sigma\mathcal{O}(X)\to\Sigma\mathcal{O}(X)\text{ is continuous}.

The domain here is the ordinary product of Scott spaces, not the Scott topology of the order product. The proof combines the compact-family correspondence of de Brecht and Kawai [2 Theorem 6.10] with a coherent-space refinement lemma. Wilker-ness of XX itself is characterized by continuity of union on Ok(X)\mathcal{O}_k(X). Hence, for every consonant base,

X Wilker⟺PH(X) Wilker.X\text{ Wilker}\Longleftrightarrow P_H(X)\text{ Wilker}.

This gives a positive solution of the Hoare-reflection problem under consonance; the unrestricted implication remains open. The relation between consonance, Wilker-ness and Scott continuity of union is already recorded in [7 Proposition J]; the additional point here is its connection with the Hoare criterion.

For the Smyth powerspace PS(X)P_S(X), the known local-compactness/core-compactness characterization is complemented by the reflection theorem

PS(X) Wilker⟹X Wilker.P_S(X)\text{ Wilker}\Longrightarrow X\text{ Wilker}.

Thus the two power constructions interact with the Wilker property and with core-compactness in markedly different ways.

The paper is organized as follows. Section 2 fixes notation and recalls the function-space preliminaries. Section 3 develops the basic theory, the KC–Wilker characterization and the mixed product theorem, followed by the product and square counterexamples and the non-reflectivity consequence. Section 4 studies the relation with core-compactness and applications to order topologies. Sections 5 and 6 introduce HH-supercompactness and apply Wilker’s condition to compact-open function spaces. Section 7 treats the Hoare construction, including the Scott-union criterion and consonant reflection; Section 8 treats the Smyth construction. Section 9 singles out two central open problems, Smyth preservation and the core-compactness implication, with a brief discussion of their related directions. The final section summarizes the resulting structural picture.

Preliminaries

Throughout, spaces are T0T_0 unless explicitly stated otherwise. For a space XX, O(X)\mathcal{O}(X) denotes its lattice of open subsets, ordered by inclusion. A subset is saturated if it is an intersection of open sets. We write Q(X)\mathcal{Q}(X) for the set of all nonempty compact saturated subsets of XX.

Definition 2.1 (Choquet–Wilker [3, 17]) A topological space XX is a Wilker space if, whenever K⊆XK \subseteq X is compact and U1,U2∈O(X)U_1,U_2 \in\mathcal{O}(X) satisfy K⊆U1∪U2K \subseteq U_1 \cup U_2, there are compact subsets Ki⊆UiK_i \subseteq U_i such that

K⊆K1∪K2.K \subseteq K_1 \cup K_2.

The compact sets KiK_i are allowed to be empty and need not be subsets of KK. The original condition is binary, but the finite form is immediate.

Lemma 2.2 If XX is Wilker, K⊆XK \subseteq X is compact, and

K⊆U1∪⋯∪UnK \subseteq U_1 \cup\cdots\cup U_n

with each UiU_i open, then there are compact Ki⊆UiK_i \subseteq U_i such that

K⊆K1∪⋯∪Kn.K \subseteq K_1 \cup\cdots\cup K_n.

Proof. Induct on nn, using the binary Wilker property for U1∪(U2∪⋯∪Un)U_1 \cup(U_2 \cup\cdots\cup U_n).

We shall repeatedly use the following elementary fact.

Lemma 2.3 The saturation of a compact subset of a topological space is compact.

Proof. Let KK be compact and let {Ui:i∈I}\{U_i : i \in I\} be an open cover of sat⁡(K)\operatorname{sat}(K). It covers KK, so finitely many members have union UU containing KK. Since UU is open and contains KK, it contains sat⁡(K)\operatorname{sat}(K). Hence those finitely many UiU_i cover sat⁡(K)\operatorname{sat}(K).

Thus the Wilker condition may equivalently be stated using compact saturated sets only.

For later use it is convenient to adjoin the empty compact saturated set and write

Q0(X)=Q(X)∪{∅},\mathcal{Q}_0(X)=\mathcal{Q}(X)\cup\{\emptyset\},

ordered by the Smyth order

K⊑L⟺L⊆K.K\sqsubseteq L\Longleftrightarrow L\subseteq K.

Binary meets in Q0(X)\mathcal{Q}_0(X) are unions.

For a compact subset K⊆XK\subseteq X, put

Φ(K)={U∈O(X):K⊆U}.\Phi(K)=\{U\in\mathcal{O}(X):K\subseteq U\}.

The topology on O(X)\mathcal{O}(X) generated by the family {Φ(K):K⊆X compact}\{\Phi(K):K\subseteq X\text{ compact}\} is denoted by ϕ(O(X))\phi(\mathcal{O}(X)), and

Ok(X)=(O(X),ϕ(O(X))).\mathcal{O}_k(X)=(\mathcal{O}(X),\phi(\mathcal{O}(X))).

As observed in [20],

ν(O(X))⊆ϕ(O(X))⊆σ(O(X)),\nu(\mathcal{O}(X))\subseteq\phi(\mathcal{O}(X))\subseteq\sigma(\mathcal{O}(X)),

and Ok(X)\mathcal{O}_k(X) has inclusion as its specialization order. The latter can also be seen directly, since singletons are compact.

Let H:Top0→SetH:\mathbf{Top}_0\to\mathbf{Set} be an irreducible subset system in the sense of [19]: for every T0T_0-space ZZ,

{{z}:z∈Z}⊆H(Z)⊆Irr⁡(Z),\{\{z\}:z\in Z\}\subseteq H(Z)\subseteq\operatorname{Irr}(Z),

and continuous maps preserve HH-subsets under direct image. A space ZZ is HH-sober if every A∈H(Z)A\in H(Z) has a unique generic point, i.e., there is a unique z∈Zz\in Z such that

cl⁡ZA=cl⁡Z{z}.\operatorname{cl}_Z A=\operatorname{cl}_Z\{z\}.

The systems DD, RDRD, WDWD and RR yield, respectively, dd-spaces, well-filtered spaces (for both RDRD and WDWD), and sober spaces; see [19, 21].

For spaces X,YX,Y, let C(X,Y)\mathcal{C}(X,Y) be the set of continuous maps from XX to YY. The compact-open topology on C(X,Y)\mathcal{C}(X,Y) has subbasic open sets

N(K→V)={f∈C(X,Y):f(K)⊆V},N(K\to V)=\{f\in\mathcal{C}(X,Y):f(K)\subseteq V\},

where K⊆XK\subseteq X is compact and V∈O(Y)V\in\mathcal{O}(Y). The resulting function space is denoted by Ck(X,Y)\mathcal{C}_k(X,Y).

We use property SS from [20]. An order-compatible topology τ\tau on C(X,Y)\mathcal{C}(X,Y) has property SS with respect to HH if, for every F∈H((C(X,Y),τ))\mathcal{F}\in H((\mathcal{C}(X,Y),\tau)) and every g∈C(X,Y)g\in\mathcal{C}(X,Y) satisfying

g−1(V)=⋃f∈Ff−1(V)(V∈O(Y)),g^{-1}(V)=\bigcup_{f\in\mathcal{F}}f^{-1}(V)\quad(V\in\mathcal{O}(Y)),

one has g∈cl⁡τFg\in\operatorname{cl}_{\tau}\mathcal{F}.

We shall use the following result proved in [20 Theorem 36].

Proposition 2.4. Let HH be an irreducible subset system, let X,YX,Y be T0T_{0}-spaces, and let τ\tau be a topology on C(X,Y)C(X,Y) finer than the pointwise convergence topology. If YY is HH-sober, then (C(X,Y),τ)(C(X,Y),\tau) is HH-sober if and only if τ\tau has property SS with respect to HH.

We shall also use the standard homeomorphism

Ck(X,S)⟶Ok(X),f⟼f−1({1}),(1)C_{k}(X,\mathbb{S}) \longrightarrow\mathcal{O}_{k}(X),\quad f \longmapsto f^{-1}(\{1\}), \tag*{(1)}

where S={0,1}\mathbb{S}=\{0,1\} is the Sierpiński space with open sets ∅,{1},S\varnothing,\{1\},\mathbb{S}; see [20].

Basic theory of Wilker spaces

This section develops the structural properties of the Wilker condition used throughout the paper. After elementary permanence properties and the KC characterization, we prove the mixed product theorem and then give counterexamples to unrestricted productivity, already for squares. The same examples yield non-reflectivity.

Heredity, retracts and sums

Proposition 3.1 (Open and closed heredity) Let XX be Wilker.

(1) Every open subspace of XX is Wilker.

(2) Every closed subspace of XX is Wilker.

(3) Consequently, every locally closed subspace of XX is Wilker.

Proof. Let A⊆XA \subseteq X.

If AA is open, every compact K⊆AK \subseteq A is compact in XX, and every open subset of AA is open in XX. The Wilker decomposition in XX therefore already lies in AA.

If AA is closed, write Ui=A∩ViU_i=A\cap V_i with ViV_i open in XX. If K⊆AK\subseteq A is compact and K⊆U1∪U2K\subseteq U_1\cup U_2, apply Wilker-ness of XX to K⊆V1∪V2K\subseteq V_1\cup V_2, obtaining compact Ci⊆ViC_i\subseteq V_i covering KK. Then Ki=Ci∩AK_i=C_i\cap A is compact, Ki⊆UiK_i\subseteq U_i, and K⊆K1∪K2K\subseteq K_1\cup K_2.

The locally closed case follows by composition of the first two cases. □

The preceding result is essentially the strongest general hereditary statement one can expect. In fact, arbitrary heredity fails in a very strong form.

Theorem 3.2 (Failure of arbitrary heredity) Every T0T_{0}-space is homeomorphic to a subspace of a locally hypercompact Wilker space. Consequently, the class of Wilker spaces is not hereditary with respect to arbitrary subspaces.

Proof. Let S={0,1}\mathbb{S}=\{0,1\} be the Sierpiński space and let II be any set. Identify SI\mathbb{S}^{I} with P(I)\mathcal{P}(I) by sending f∈SIf\in\mathbb{S}^{I} to

{i∈I:f(i)=1}.\{i\in I:f(i)=1\}.

Under this identification, a basis for the product topology consists of the sets

↑F={A⊆I:F⊆A},F⊆I finite.\uparrow F=\{A\subseteq I:F\subseteq A\},\quad F\subseteq I\text{ finite}.

Each ↑F\uparrow F is supercompact. Indeed, if an open cover of ↑F\uparrow F is given, one member contains the point FF; since every open subset of SI\mathbb{S}^{I} is an upper set for inclusion, that member contains all of ↑F\uparrow F. Thus SI\mathbb{S}^{I} has a basis of supercompact open sets. In particular, it is locally hypercompact, hence locally compact and therefore Wilker by Wilker’s classical theorem.

Now let XX be a T0T_{0}-space. The canonical map

eX:X⟶SO(X),eX(x)(U)={1,x∈U,0,x∉U,e_{X}:X\longrightarrow\mathbb{S}^{\mathcal{O}(X)},\qquad e_{X}(x)(U)= \begin{cases} 1, & x\in U,\\ 0, & x\notin U, \end{cases}

is a topological embedding. Indeed, it is continuous because the inverse image under the UU-coordinate of the open singleton {1}\{1\} is exactly UU; it is injective because XX is T0T_{0}; and these same coordinate inverse images recover the topology of XX. Hence every T0T_{0}-space embeds into a locally hypercompact Wilker space.

Finally, non-Wilker T0T_{0}-spaces exist. For example, Wilker observed that the one-point compactification Q+\mathbb{Q}^{+} of the rational line is a compact KC non-Hausdorff space and does not satisfy condition (D) [17]. Therefore Q+\mathbb{Q}^{+} embeds into a Wilker space while itself being non-Wilker, so arbitrary heredity fails. □\square

Remark 3.3 The proof shows more than non-heredity. The Sierpiński powers SI\mathbb{S}^{I}, which are naturally identified with the Scott spaces of the algebraic lattices P(I)\mathcal{P}(I), form a universal source of locally hypercompact Wilker spaces for T0T_{0} embeddings. This is one place where the basic theory of Wilker spaces meets a standard domain-theoretic representation.

Proposition 3.4 (Retracts) Every retract of a Wilker space is Wilker.

Proof. Let e:X→Ye:X\to Y and r:Y→Xr:Y\to X be continuous with r∘e=id⁡Xr\circ e=\operatorname{id}_{X}, and suppose YY is Wilker. Let K⊆XK\subseteq X be compact and K⊆U1∪U2K\subseteq U_{1}\cup U_{2} with UiU_{i} open. Then e(K)e(K) is compact and

e(K)⊆r−1(U1)∪r−1(U2).e(K)\subseteq r^{-1}(U_{1})\cup r^{-1}(U_{2}).

Choose compact Ci⊆r−1(Ui)C_{i}\subseteq r^{-1}(U_{i}) covering e(K)e(K). Then Ki=r(Ci)K_{i}=r(C_{i}) is compact, lies in UiU_{i}, and the KiK_{i} cover KK. □\square

Theorem 3.5 (Topological sums) For a family (Xi)i∈I(X_{i})_{i\in I} of spaces, the topological sum ∐i∈IXi\coprod_{i\in I}X_{i} is Wilker if and only if every XiX_{i} is Wilker.

Proof. Each summand is clopen, so the forward implication follows from Proposition 3.1.

Conversely, let X=∐iXiX=\coprod_{i}X_{i} with every XiX_{i} Wilker, and let K⊆XK\subseteq X be compact. Since the summands form an open cover of KK, only finitely many of them meet KK. On each of those finitely many components apply Wilker-ness to K∩XiK\cap X_{i} and the restrictions of the two given open sets. Finite unions of the resulting compact pieces give the required decomposition in XX. □\square

Let Wilk0\mathbf{Wilk}_0 denote the full subcategory of Top0\mathbf{Top}_0 consisting of Wilker spaces. Proposition 3.4 and Theorem 3.5 show that Wilk0\mathbf{Wilk}_0 is closed under retracts and arbitrary coproducts.

Separation properties

Recall that a space is KCKC if every compact subset is closed. Following Wilker [17, p. 279], a space XX is called a KT4KT_4 space if every compact subspace K⊆XK \subseteq X is normal: for any disjoint subsets F1,F2F_1,F_2 that are closed in KK, there are V1,V2∈O(K)V_1,V_2 \in\mathcal{O}(K) such that

F1⊆V1,F2⊆V2,V1∩V2=∅.F_1 \subseteq V_1,\qquad F_2 \subseteq V_2,\qquad V_1 \cap V_2 = \varnothing.

Here normality does not include the T1T_1 axiom. Wilker proved that every KT4KT_4 space satisfies condition (D), and in particular every Hausdorff space is Wilker [17 Theorem 8]. The following elementary characterization shows that, after imposing the KCKC condition, Wilker-ness exactly forces compact subspaces to be Hausdorff.

Theorem 3.6 (KC–Wilker characterization). For a space XX, the following conditions are equivalent.

(1) XX is KCKC and Wilker.

(2) Every compact subspace of XX is Hausdorff.

Proof. (1) ⇒\Rightarrow (2): Assume first that XX is KCKC and Wilker, and let K⊆XK \subseteq X be compact. Since XX is KCKC, KK is closed in XX; hence KK is Wilker by Proposition 3.1. Let x,y∈Kx,y \in K be distinct. Singletons are compact, hence closed in a KCKC space, so

K=(K∖{y})∪(K∖{x})K = (K \setminus\{y\}) \cup(K \setminus\{x\})

is an open cover of the compact space KK. By Wilker-ness there are compact subsets

A⊆K∖{y},B⊆K∖{x}A \subseteq K \setminus\{y\},\qquad B \subseteq K \setminus\{x\}

with K=A∪BK = A \cup B. The sets AA and BB are compact as subsets of XX, hence closed by the KCKC property. Therefore

K∖BandK∖AK \setminus B \quad\text{and}\quad K \setminus A

are disjoint open neighborhoods in KK of xx and yy, respectively. Thus KK is Hausdorff.

(2) ⇒\Rightarrow (1): Suppose that every compact subspace of XX is Hausdorff. We first show that XX is KCKC. Let A⊆XA \subseteq X be compact. If x∈cl⁡XA∖Ax \in\operatorname{cl}_X A \setminus A, then A∪{x}A \cup\{x\} is compact and hence Hausdorff by assumption. Since AA is compact, it is closed in the Hausdorff subspace A∪{x}A \cup\{x\}, contradicting x∈cl⁡XAx \in\operatorname{cl}_X A. Thus every compact subset of XX is closed.

It remains to prove that XX is Wilker. Let K⊆U1∪U2K \subseteq U_1 \cup U_2 be compact, with U1,U2U_1,U_2 open in XX. The subspace KK is compact Hausdorff, hence normal. Applying the finite shrinking lemma in KK to the open cover

K∩U1,K∩U2,K \cap U_1,\qquad K \cap U_2,

we obtain closed subsets FiF_i of KK such that

Fi⊆K∩Ui,K=F1∪F2.F_i \subseteq K \cap U_i,\quad K = F_1 \cup F_2.

Each FiF_i is compact, so these sets give the required Wilker decomposition.

Corollary 3.7 If XX is compact and KCKC, then XX is Wilker if and only if XX is Hausdorff.

Proof. The forward implication follows from Theorem 3.6, since XX itself is a compact subspace. The reverse implication is Wilker’s classical Hausdorff result.

Remark 3.8 Thus Wilker-ness by itself is far from being a separation axiom, but in the presence of KCKC it has a rigid separation-theoretic meaning: it is equivalent to Hausdorffness on every compact subspace. In particular, every compact KCKC non-Hausdorff space is automatically non-Wilker.

Mixed products

The preceding characterization gives a positive product theorem under a separation hypothesis on one factor. The counterexamples in Subsection 3.4 will show that this hypothesis cannot simply be omitted.

Theorem 3.9 (Mixed product theorem) Let XX be a Wilker space, and suppose that every compact subspace of YY is Hausdorff. Then X×YX \times Y is Wilker.

Proof. Let K⊆X×YK \subseteq X \times Y be compact and let U1,U2U_1,U_2 be open in X×YX \times Y with

K⊆U1∪U2.K \subseteq U_1 \cup U_2.

Put C=πY(K)C = \pi_Y(K). Then CC is compact and hence Hausdorff by hypothesis. Moreover, Theorem 3.6 shows that YY is KCKC.

Fix y∈Cy \in C and write

Ky={x∈X:(x,y)∈K},Uiy={x∈X:(x,y)∈Ui}.K_y = \{x \in X : (x,y) \in K\},\quad U_i^y = \{x \in X : (x,y) \in U_i\}.

Since YY is KCKC, {y}\{y\} is closed, so KyK_y is compact. Also

Ky⊆U1y∪U2y.K_y \subseteq U_1^y \cup U_2^y.

By the Wilker property of XX, choose compact sets

Ai(y)⊆Uiy(i=1,2)A_i(y) \subseteq U_i^y \quad(i = 1,2)

such that Ky⊆A1(y)∪A2(y)K_y \subseteq A_1(y) \cup A_2(y). For each ii, compactness of Ai(y)A_i(y) and openness of UiU_i give an open set Vi(y)⊆XV_i(y) \subseteq X and an open neighborhood Wi(y)W_i(y) of yy in YY such that

Ai(y)⊆Vi(y),Vi(y)×Wi(y)⊆Ui.A_i(y) \subseteq V_i(y),\quad V_i(y) \times W_i(y) \subseteq U_i.

If Ai(y)=∅A_i(y)=\varnothing, take Vi(y)=∅V_i(y)=\varnothing and Wi(y)=YW_i(y)=Y. Put

V(y)=V1(y)∪V2(y).V(y)=V_1(y)\cup V_2(y).

Then Ky⊆V(y)K_y \subseteq V(y). Hence

Fy=K∩((X∖V(y))×Y)F_y=K\cap\left((X\setminus V(y))\times Y\right)

is compact and y∉πY(Fy)y\notin\pi_Y(F_y). Since YY is KC, the compact set πY(Fy)\pi_Y(F_y) is closed. Therefore

W0(y)=Y∖πY(Fy)W_0(y)=Y\setminus\pi_Y(F_y)

is an open neighborhood of yy, and

K∩(X×W0(y))⊆V(y)×W0(y).(2)K\cap(X\times W_0(y))\subseteq V(y)\times W_0(y). \tag*{(2)}

The set

Oy=C∩W0(y)∩W1(y)∩W2(y)O_y=C\cap W_0(y)\cap W_1(y)\cap W_2(y)

is an open neighborhood of yy in the compact Hausdorff space CC. Choose an open neighborhood NyN_y of yy in CC such that

Ly:=cl⁡CNy⊆Oy.L_y:=\operatorname{cl}_C N_y\subseteq O_y.

Then LyL_y is compact; since YY is KC, it is also closed in YY. Consequently

K(y)=K∩(X×Ly)K^{(y)}=K\cap(X\times L_y)

is compact, and so is its projection

Py=πX(K(y)).P_y=\pi_X(K^{(y)}).

By (3.1) and Ly⊆W0(y)L_y\subseteq W_0(y),

Py⊆V1(y)∪V2(y).P_y\subseteq V_1(y)\cup V_2(y).

Apply the Wilker property of XX once more to obtain compact sets

Di(y)⊆Vi(y)(i=1,2)D_i(y)\subseteq V_i(y)\quad(i=1,2)

with Py⊆D1(y)∪D2(y)P_y\subseteq D_1(y)\cup D_2(y). Since Ly⊆Wi(y)L_y\subseteq W_i(y),

Di(y)×Ly⊆Ui,D_i(y)\times L_y\subseteq U_i,

and the two compact rectangles Di(y)×LyD_i(y) \times L_y cover K∩(X×Ly)K \cap(X \times L_y). The family (Ny)y∈C(N_y)_{y \in C} covers the compact space CC. Choose a finite subcover

Ny1,…,Nyn.N_{y_1}, \ldots, N_{y_n}.

For i=1,2i=1,2, set

Hi=⋃j=1nDi(yj)×Lyj.H_i = \bigcup_{j=1}^{n} D_i(y_j) \times L_{y_j}.

Each HiH_i is compact, Hi⊆UiH_i \subseteq U_i, and K⊆H1∪H2K \subseteq H_1 \cup H_2. Hence X×YX \times Y is Wilker.

Combining Theorems 3.6 and 3.9 gives a convenient intrinsic form.

Corollary 3.10 If XX and YY are Wilker spaces and at least one of them is KC, then X×YX \times Y is Wilker.

Proof. Suppose, for example, that YY is KC. By Theorem 3.6, every compact subspace of YY is Hausdorff, so Theorem 3.9 applies. The other case follows by symmetry.

Corollary 3.11 (Hausdorff factor) If XX is Wilker and YY is Hausdorff, then X×YX \times Y is Wilker.

Proof. Every compact subspace of a Hausdorff space is Hausdorff, so this is an immediate consequence of Theorem 3.9.

Remark 3.12. Consequently, if the product of two Wilker spaces fails to be Wilker, then both factors must be non-KC, and in particular both must be non-Hausdorff. The next subsection constructs counterexamples of exactly this type.

Remark 3.13. A naive Wallace–Tychonoff argument does not establish productivity in the non-Hausdorff case. Given compact K⊆X×YK \subseteq X \times Y covered by finitely many open rectangles, compactness of the projections and Tychonoff’s theorem do not by themselves give a cover of KK by compact rectangles subordinate to those open rectangles. The proof of Theorem 3.9 succeeds because the second factor makes compact projections closed and supplies finite shrinking inside the compact Hausdorff space πY(K)\pi_Y(K). The counterexample below shows that an unrestricted compact-cover refinement cannot hold.

Failure of product and square stability

We now prove that the separation hypothesis in the mixed product theorem cannot be omitted. The construction uses the co-finite modifications of Xu and Ji (arXiv: 2610.14434v1, Sections 3 and 6), originally used to study products of co-sober spaces. We retain those constructions and analyze their Wilker property; the obstruction to splitting a compact cover in the product is proved directly.

A space is hereditarily Wilker if all of its subspaces are Wilker. A nonempty space is anti-Hausdorff if any two nonempty open sets intersect. We use the particular countably infinite compact Fréchet–Urysohn anti-Hausdorff KC-space (M,τ)(M,\tau) obtained from van Douwen’s construction [4] in the realization described by Kovár [11 Example 2.2 and the following paragraph, p. 6]; see also the product preprint (arXiv:2610.14434v1, Lemma 5.1). In this realization there are a family A\mathcal{A} of infinite subsets of MM and a map ℓ:A→M\ell:\mathcal{A}\to M, with ℓ(A)∉A\ell(A)\notin A, such that a closed subbasis for τ\tau is

S={B∪{ℓ(A)}:B⊆A∈A}.(3)\mathcal{S}=\{B\cup\{\ell(A)\}:B\subseteq A\in\mathcal{A}\}. \tag*{(3)}

Here we use the specific realization with the stated topological properties, not an arbitrary choice of an almost-disjoint family and of the map ℓ\ell. The following additional consequence of the closed subbasis is what is needed for Wilker-ness.

Lemma 3.14. Every proper closed subspace of MM is compact Hausdorff. Consequently, for each a∈Ma\in M, every compact subspace of M∖{a}M\setminus\{a\} is Hausdorff.

Proof. Let S=B∪{ℓ(A)}S=B\cup\{\ell(A)\} be a member of (3). For b∈Bb\in B, the set (B∖{b})∪{ℓ(A)}(B\setminus\{b\})\cup\{\ell(A)\} is closed in MM, so {b}\{b\} is open in SS. Thus every point of SS except possibly ℓ(A)\ell(A) is isolated. Since SS is T1T_1, an isolated point bb and any other point are separated by {b}\{b\} and S∖{b}S\setminus\{b\}. Hence SS is Hausdorff. It is also compact because it is closed in the compact space MM.

Let F⊊MF\subsetneq M be closed and nonempty. Choose a point of M∖FM\setminus F and a basic open neighborhood of it disjoint from FF. The complement of this neighborhood is a finite union S1∪⋯∪SnS_1\cup\cdots\cup S_n of members of (3) containing FF. A finite union YY of closed Hausdorff subspaces SiS_i is Hausdorff: each diagonal ΔSi\Delta_{S_i} is closed in Si2S_i^2, hence in Y2Y^2, and ΔY=⋃i=1nΔSi\Delta_Y=\bigcup_{i=1}^{n}\Delta_{S_i} is closed. Therefore FF is Hausdorff. It is compact by closedness in MM; the empty case is immediate.

Finally, a compact subset KK of M∖{a}M\setminus\{a\} is compact in MM, and hence closed in MM by the KC property. Since it omits aa, it is proper, so the first assertion applies. □\square

Let more generally (M,τ)(M,\tau) be a T1T_1-space and a∈Ma\in M. Following the product preprint (arXiv:2610.14434v1, Section 3), write τcof\tau_{\mathrm{cof}} for the co-finite topology, including the empty set, and put

τa=τcof∪{U∈τ:a∉U},Xa=(M,τa).(4)\tau_a=\tau_{\mathrm{cof}}\cup\{U\in\tau:a\notin U\},\qquad X_a=(M,\tau_a). \tag*{(4)}

This is a compact T1T_1 topology contained in τ\tau. Every subset containing aa is compact, because every neighborhood of aa is co-finite. The restrictions of τa\tau_a and τ\tau to M∖{a}M\setminus\{a\} coincide: for U∈τU\in\tau, the set U∖{a}U\setminus\{a\} is τ\tau-open and belongs to τa\tau_a. In particular, a subset avoiding aa is compact in XaX_a if and only if it is compact in (M,τ)(M,\tau). These are the singleton cases recorded in the product preprint (arXiv:2610.14434v1, Lemma 3.1).

Lemma 3.15. The space XaX_a is Wilker if and only if the original subspace M∖{a}M\setminus\{a\} is Wilker. If every compact subspace of M∖{a}M\setminus\{a\} is Hausdorff, then XaX_a is hereditarily Wilker.

Proof. Necessity of the first assertion follows from open heredity. Conversely, suppose that M∖{a}M\setminus\{a\} is Wilker, and let K⊆U1∪U2K\subseteq U_1\cup U_2, where KK is compact in XaX_a and the $U_i are open in XaX_a. If a∈U1a \in U_1, put

C1=(K∩U1)∪{a},C2=K∖U1.C_1 = (K \cap U_1) \cup\{a\}, \qquad C_2 = K \setminus U_1.

The set C1C_1 is compact because it contains aa, and C2C_2 is finite because U1U_1 is co-finite. They lie in U1,U2U_1,U_2, respectively, and cover KK. The case a∈U2a \in U_2 is symmetric. If neither open set contains aa, apply Wilker-ness in the original open subspace M∖{a}M \setminus\{a\}. Notice that the compact pieces need not be subsets of KK, as in Definition 2.1.

For the hereditary assertion, take S⊆XaS \subseteq X_a. If a∉Sa \notin S, every compact subset of SS is a compact Hausdorff subset of M∖{a}M \setminus\{a\}. Theorem 3.6 makes SS Wilker. If a∈Sa \in S, every relative neighborhood of aa is co-finite in SS, and every subset of SS containing aa is compact. The argument above applies inside SS, using the already established Wilker property of S∖{a}S \setminus\{a\}.

Theorem 3.16 (Failure of productivity) There exist two countably infinite compact Fréchet–Urysohn T1T_1 anti-Hausdorff hereditarily Wilker spaces whose product is not Wilker.

Proof. Take the space (M,τ)(M,\tau) used in Lemma 3.14 and distinct a,b∈Ma,b \in M. By Lemmas 3.14 and 3.15, the modifications Xa,XbX_a,X_b are hereditarily Wilker. They are compact and T1T_1, and they remain anti-Hausdorff because their topologies are coarser than τ\tau. They are also Fréchet–Urysohn, as proved in the product preprint (arXiv:2610.14434v1, Lemma 3.3). Indeed, at the modified point aa sequence of distinct points of any infinite set converges to that point; away from it, closure can be tested in the original open complement, where the Fréchet–Urysohn property applies.

The diagonal map from (M,τ)(M,\tau) to Xa×XbX_a \times X_b is continuous. Thus its image

D={(x,x):x∈M}D = \{(x,x): x \in M\}

is compact. The two open sets

W1=(M∖{a})×M,W2=M×(M∖{b})(5)W_1 = (M \setminus\{a\}) \times M, \qquad W_2 = M \times(M \setminus\{b\}) \tag*{(5)}

cover DD, since a≠ba \ne b.

Suppose that compact sets Ci⊆WiC_i \subseteq W_i covered DD. The projection A=π1(C1)A = \pi_1(C_1) is compact in XaX_a and avoids aa, so it is compact in (M,τ)(M,\tau) and therefore is a proper τ\tau-closed set. Similarly, B=π2(C2)B = \pi_2(C_2) is a proper τ\tau-closed set avoiding bb. Every (x,x)∈D(x,x) \in D belongs to C1C_1 or C2C_2, so M=A∪BM = A \cup B. This contradicts anti-Hausdorffness of MM. Hence the cover (3.4) has no Wilker refinement.

The compact sets CiC_i in the proof are allowed to contain points outside DD. No intersection Ci∩DC_i \cap D is assumed compact. Also, the proof does not infer failure of Wilker-ness merely from the known failure of co-sobriety of the product: it excludes a compact-cover splitting directly by taking projections.

Corollary 3.17 (A non-Wilker square) There is a countable compact Fréchet–Urysohn T1_1 hereditarily Wilker space whose square is not Wilker.

Proof. Let Z=Xa⊔XbZ = X_a \sqcup X_b for the two spaces in Theorem 3.16. A finite topological sum preserves compactness, the Fréchet–Urysohn property and T1T_1. Every subspace of ZZ is a sum of subspaces of the two factors, so Theorem 3.5 gives hereditary Wilker-ness. The product Xa×XbX_a \times X_b is a clopen subspace of Z2Z^2. If Z2Z^2 were Wilker, Proposition 3.1 would contradict Theorem 3.16. □

The square can also be made anti-Hausdorff. We use the co-finite-core construction in the product preprint (arXiv:2610.14434v1, § 6) and record its additional Wilker permanence property.

Proposition 3.18 Let P,QP,Q be disjoint T1T_1 Wilker spaces and RR an infinite set disjoint from them. On Z=P⊔Q⊔RZ = P \sqcup Q \sqcup R put

λ={∅}∪{U∪V∪(R∖F):U∈O(P), V∈O(Q), F⊆R finite}.(6)\lambda= \{\varnothing\} \cup\{U \cup V \cup(R \setminus F) : U \in\mathcal{O}(P),\ V \in\mathcal{O}(Q),\ F \subseteq R\text{ finite}\}. \tag*{(6)}

Then (Z,λ)(Z,\lambda) is Wilker. Consequently, a countable compact Fréchet–Urysohn T1T_1 anti-Hausdorff Wilker space can have a non-Wilker square.

Proof. The basic properties of (3.5) are established in the product preprint (arXiv:2610.14434v1, Lemma 6.1): P,QP,Q are closed subspaces with their original topologies, and K⊆ZK \subseteq Z is compact if and only if K∩PK \cap P and K∩QK \cap Q are compact in their respective spaces. In particular, every subset of RR is compact.

Let K⊆O1∪O2K \subseteq O_1 \cup O_2 be a compact set covered by two open sets of ZZ. Split K∩PK \cap P using compact covering pieces Ai⊆P∩OiA_i \subseteq P \cap O_i, and similarly split K∩QK \cap Q using compact Bi⊆Q∩OiB_i \subseteq Q \cap O_i. Put

E1=K∩R∩O1,E2=(K∩R)∖O1⊆O2.E_1 = K \cap R \cap O_1,\qquad E_2 = (K \cap R) \setminus O_1 \subseteq O_2.

Then Ai∪Bi∪EiA_i \cup B_i \cup E_i are compact subsets of OiO_i covering KK.

For the last assertion, take disjoint copies of Xa,XbX_a,X_b from Theorem 3.16 and countable RR. The construction preserves compactness and the Fréchet–Urysohn property and gives a T1T_1 anti-Hausdorff space (arXiv:2610.14434v1, Lemma 6.1). Its square contains the closed subspace P×QP \times Q, which is not Wilker. Closed heredity completes the proof. □

Categorical consequence

Theorem 3.19 The full subcategory Wilk0\mathrm{Wilk}_0 of Top0\mathrm{Top}_0 is neither reflective nor epireflective. More precisely, the square in Corollary 3.17 has no reflection into Wilker spaces.

Proof. Let ZZ be as in Corollary 3.17, and suppose that η:Z2→C\eta: Z^2 \to C were a reflection into Wilker spaces. Since ZZ is Wilker, its coordinate projections factor uniquely through η\eta: there are continuous pi:C→Zp_i : C \to Z with piη=πip_i\eta= \pi_i. Then r=(p1,p2):C→Z2r = (p_1,p_2) : C \to Z^2 satisfies rη=id⁡Z2r\eta= \operatorname{id}_{Z^2}. The uniqueness clause of the reflection, applied to η\eta, gives ηr=id⁡C\eta r = \operatorname{id}_C, since both sides have composite η\eta with η\eta. Thus η\eta is a homeomorphism, contradicting that CC is Wilker and Z2Z^2 is not. Hence Wilk0\mathrm{Wilk}_0 is not reflective, and therefore cannot be epireflective.

This is the product obstruction to reflectivity, expressed directly by the universal property. Together with Theorem 3.2, it settles the unrestricted heredity and reflection questions negatively. The behavior under quotients and quotient-like maps is not settled by these arguments and remains a separate direction.

Sources of Wilker spaces and the core-compactness problem

Wilker proved that every KT4KT_4 space satisfies condition (D), hence every Hausdorff space is Wilker; he also proved that every locally compact space is Wilker [17]. Later work produced large non-Hausdorff classes. Every quasi-Polish space is Wilker [2], every LCS-complete space is Wilker [1], and every co-consonant T0T_0-space is Wilker by the strong-compactness argument of [23].

Recall that a space XX is Alexandroff if arbitrary intersections of open subsets of XX are open. Equivalently, every point x∈Xx \in X has a least open neighborhood, namely

↑x=⋂{U∈O(X):x∈U},\uparrow x = \bigcap\{U \in\mathcal{O}(X) : x \in U\},

where ↑x\uparrow x is taken in the specialization order. Thus the open sets of an Alexandroff T0T_0-space are exactly the upper sets of its specialization order.

For completeness we record two elementary sources; under our standing T0T_0 convention the zero-dimensional case is already covered by Hausdorffness, while the Alexandroff case is genuinely non-Hausdorff.

Proposition 4.1 Every zero-dimensional space is Wilker. Every Alexandroff space is Wilker.

Proof. Let XX be zero-dimensional, K⊆U1∪U2K \subseteq U_1 \cup U_2 compact. For each x∈Kx \in K choose a clopen neighborhood VxV_x contained in one of the UiU_i. Compactness gives finitely many such VxV_x. Let CiC_i be the union of those assigned to UiU_i and put Ki=K∩CiK_i = K \cap C_i. Each KiK_i is closed in KK, hence compact, lies in UiU_i, and the two sets cover KK.

Now let XX be Alexandroff. By Lemma 2.3 we may assume that KK is compact saturated. Since KK is an upper set and the principal upper sets {↑x:x∈K}\{\uparrow x : x \in K\} form an open cover of KK, compactness yields a finite F⊆KF \subseteq K such that K=↑FK = \uparrow F. Assign each point of FF to one of the UiU_i containing it, and let FiF_i be the corresponding finite subsets. Then Ki=↑FiK_i = \uparrow F_i are compact, lie in UiU_i, and cover KK.

Core-compactness

Recall that XX is core-compact if O(X)\mathcal{O}(X) is a continuous lattice. Every locally compact space is core-compact, and for sober spaces core-compactness is equivalent to local compactness [5]. Hence local compactness gives

locally compact⇒core-compact and Wilker.\text{locally compact} \Rightarrow\text{core-compact and Wilker}.

Neither condition alone characterizes local compactness; the examples below also distinguish their conjunction from local compactness.

The Wilker property does not imply core-compactness: the rational line Q\mathbb{Q} is Hausdorff, hence Wilker, but it is not locally compact; being Hausdorff, it is sober, so it cannot be core-compact.

The opposite implication is one of the central open problems of this paper.

Question 4.2 (Core-compact versus Wilker) Is every core-compact T0T_0-space Wilker?

The classical Hofmann–Lawson example does not provide a negative answer. In fact it yields a rather strong positive test case. We first prove a general lemma.

Let J=[0,1)J = [0,1) equipped with the Scott topology of the reverse usual order. Thus the open sets of JJ are [0,a)[0,a), 0≤a≤10 \le a \le1.

Theorem 4.3 Let BB be a Hausdorff space. Every subspace XX of B×JB \times J is Wilker.

Proof. Let K⊆XK \subseteq X be compact and let U1,U2U_1,U_2 be open in XX with K⊆U1∪U2K \subseteq U_1 \cup U_2. Write Ui=X∩OiU_i = X \cap O_i with OiO_i open in B×JB \times J. Let π:B×J→B\pi: B \times J \to B be projection and put C=π(K)C = \pi(K). Then CC is compact Hausdorff.

For x∈Cx \in C, the fiber

Kx={t∈J:(x,t)∈K}K_x = \{t \in J : (x,t) \in K\}

is nonempty and compact: indeed, KxK_x is homeomorphic to the closed subspace K∩({x}×J)K \cap(\{x\} \times J) of KK, because BB is Hausdorff. Every nonempty compact subset of JJ has a largest element in the usual order. Indeed, if s=sup⁡Kxs = \sup K_x were not attained (or were equal to 11), the increasing family of open initial intervals [0,a)[0,a) with a<sa < s (respectively a<1a < 1) would cover KxK_x without a finite subcover. Write

q(x)=max⁡Kx.q(x) = \max K_x.

The map q:C→[0,1)q : C \to[0,1) is upper semicontinuous. For r∈[0,1)r \in[0,1),

{x∈C:q(x)≥r}=π(K∩(B×[r,1))).\{x \in C : q(x) \ge r\} = \pi\bigl(K \cap(B \times[r,1))\bigr).

Here [r,1)[r,1) is closed in JJ, since its complement [0,r)[0,r) is Scott open. Hence the set inside the projection is compact; its image is compact in the Hausdorff space BB, and therefore closed. Thus qq is upper semicontinuous in the ordinary real-valued sense. For i=1,2i=1,2 and x∈Cx\in C, let

Ai(x)={a∈[0,1]:∃N∈NB(x) such that X∩(N×[0,a))⊆Ui},A_i(x)=\{a\in[0,1]:\exists N\in\mathcal{N}_B(x)\text{ such that }X\cap(N\times[0,a))\subseteq U_i\},
fi(x)=sup⁡Ai(x).f_i(x)=\sup A_i(x).

where NB(x)\mathcal{N}_B(x) denotes the open neighborhoods of xx in BB. Each fif_i is lower semicontinuous: if fi(x)>rf_i(x)>r, one may choose a>ra>r and a witnessing neighborhood NN; then N∩C⊆{fi>r}N\cap C\subseteq\{f_i>r\}.

Put

Wi={x∈C:q(x)<fi(x)}.W_i=\{x\in C:q(x)<f_i(x)\}.

Since qq is upper semicontinuous and fif_i is lower semicontinuous, WiW_i is open in CC. Moreover C=W1∪W2C=W_1\cup W_2. Indeed, for x∈Cx\in C the point (x,q(x))(x,q(x)) belongs to KK, hence to one of the UiU_i. By openness, there are a>q(x)a>q(x) and an open neighborhood NN of xx such that X∩(N×[0,a))⊆UiX\cap(N\times[0,a))\subseteq U_i, so q(x)<fi(x)q(x)<f_i(x).

Since CC is compact Hausdorff, the finite shrinking lemma gives closed sets Fi⊆WiF_i\subseteq W_i with C=F1∪F2C=F_1\cup F_2. Define

Ki=K∩π−1(Fi).K_i=K\cap\pi^{-1}(F_i).

Then each KiK_i is compact and K=K1∪K2K=K_1\cup K_2. We claim Ki⊆UiK_i\subseteq U_i. If (x,t)∈Ki(x,t)\in K_i, then t≤q(x)<fi(x)t\leq q(x)<f_i(x). Choose cc with q(x)<c<fi(x)q(x)<c<f_i(x). By the definition of the supremum, some a>ca>c is witnessed by a neighborhood NN of xx with X∩(N×[0,a))⊆UiX\cap(N\times[0,a))\subseteq U_i. Since t<at<a, we obtain (x,t)∈Ui(x,t)\in U_i.

Hofmann and Lawson constructed a second-countable core-compact T0T_0-space XHLX_{\mathrm{HL}} which is not locally compact; in fact every compact subset of XHLX_{\mathrm{HL}} has empty interior [5, 9]. Their space is a subspace of [0,1]×J[0,1]\times J.

Corollary 4.4 There exists a second-countable core-compact Wilker T0T_0-space that is not locally compact.

Proof. Apply Theorem 4.3 to the Hofmann–Lawson space.

Thus

{locally compact T0-spaces}⊊{core-compact Wilker T0-spaces}.\{\text{locally compact }T_0\text{-spaces}\}\subsetneq\{\text{core-compact Wilker }T_0\text{-spaces}\}.

In particular, even the conjunction of core-compactness and Wilker-ness does not force local compactness.

Scott spaces and domain-theoretic consequences

We now record several consequences for Scott spaces. They provide a direct link between the basic theory developed above and standard classes from domain theory.

Recall that Mao and Xu call a poset PP quasicontinuous if, for every x∈U∈σ(P)x \in U \in\sigma(P), there is a nonempty finite set F⊆PF \subseteq P such that

x∈int⁡σ(P)↑F⊆↑F⊆U.x \in\operatorname{int}_{\sigma(P)} \mathord{\uparrow} F \subseteq\mathord{\uparrow} F \subseteq U.

It is clear that the Scott space of every quasicontinuous poset is locally compact (cf. [13 Proposition 2.6]). Wilker’s classical theorem therefore gives the following immediate consequence.

Corollary 4.5. If PP is a quasicontinuous poset, then its Scott space ΣP\Sigma P is Wilker.

In particular, the conclusion applies to quasicontinuous domains, continuous posets and continuous domains. Notice that directed completeness is not required in Corollary 4.5.

The core-compact Scott-space problem has a complete answer for complete lattices.

Proposition 4.6. Let LL be a complete lattice. If ΣL\Sigma L is core-compact, then ΣL\Sigma L is locally compact, sober and Wilker.

Proof. The Scott space of every complete lattice is well-filtered; see, for example, [18]. Every core-compact well-filtered T0T_0-space is locally compact (and sober) by [21]. Since every locally compact space is Wilker, the conclusion follows. □

Thus the specialization of Question 4.2 to complete lattices is positive. For general Scott spaces, however, a natural domain-theoretic version of the question remains.

Question 4.7 (Core-compact Scott spaces). Let PP be a poset, or even a dcpo. If its Scott space ΣP\Sigma P is core-compact, must ΣP\Sigma P be Wilker?

We next record a useful obstruction. Recall that a T0T_0-space is co-sober if every kk-irreducible compact saturated subset is the saturation of a point. Zhao, He and Wang proved that every Wilker T0T_0-space is weakly co-sober and that, in a coherent space, weak co-sobriety implies co-sobriety [23]. Hence:

Proposition 4.8. Every coherent Wilker T0T_0-space is co-sober.

Since the Scott space of every complete lattice is coherent, we immediately obtain the following complete-lattice form.

Corollary 4.9. Let LL be a complete lattice. If ΣL\Sigma L is Wilker, then ΣL\Sigma L is co-sober.

This turns recent non-co-sobriety examples into Wilker counterexamples without any additional analysis of compact covers. Ji and Xu constructed a complete Boolean algebra BB such that ΣB\Sigma B is not co-sober; in fact their example is also non-sober and well-filtered (arXiv:2608.08484v1). Corollary 4.9 therefore yields:

Corollary 4.10. There exists a complete Boolean algebra whose Scott space is not Wilker. Consequently, there exist a complete Heyting algebra and a complete meet-continuous lattice whose Scott spaces are not Wilker.

Remark 4.11. The counterexample in Corollary 4.10 is much stronger than a merely distributive complete-lattice example: every complete Boolean algebra is a frame, hence a complete Heyting algebra, and in particular a complete meet-continuous lattice. Thus none of these three unrestricted classes has Wilker Scott spaces automatically.

Countability changes the picture at the Boolean level. Ji and Xu proved that every countable complete Boolean algebra is finite (arXiv:2608.08484v1, Proposition 5.4). Hence its Scott topology is finite Alexandroff and therefore Wilker by Proposition 4.1. The next two cases are substantially less clear.

Question 4.12 (Countable Scott–Wilker problems)

  1. Is the Scott space of every countable complete Heyting algebra (equivalently, every countable frame) Wilker?

  2. Is the Scott space of every countable complete meet-continuous lattice Wilker?

The second problem admits an even broader variant in which “complete meet-continuous lattice” is replaced by “countable meet-continuous dcpo”. These countable questions are not consequences of Corollary 4.10, since the Boolean case collapses to the finite case. They therefore isolate a genuinely countable boundary between the positive quasicontinuous theory and the unrestricted Boolean counterexample.

Upper topologies of complete lattices

There is a parallel negative phenomenon for the upper topology. For a poset PP, let ν(P)\nu(P) denote the upper topology generated by the subbasic sets

P∖↓x,x∈P.P \setminus\mathord{\downarrow} x,\quad x \in P.

Equivalently, ν(P)=ω(Pop⁡)\nu(P) = \omega(P^{\operatorname{op}}), where ω\omega denotes the lower topology.

Wen and Xu proved that Isbell’s complete lattice II equipped with its lower topology ω(I)\omega(I) is sober but not co-sober [16]. Let L=Iop⁡L = I^{\operatorname{op}}. Then LL is again a complete lattice and

(L,ν(L))=(I,ω(I)).(L,\nu(L)) = (I,\omega(I)).

as topological spaces. On the other hand, Xu, Yang and Chen proved that the upper topology of every sup-complete poset is coherent (indeed, it satisfies considerably stronger properties) [22]. Hence (L,ν(L))(L,\nu(L)) is coherent. Proposition 4.8 now gives the following.

Corollary 4.13 There exists a complete lattice whose upper topology is not Wilker.

Proof. For the lattice L=IopL = I^{\mathrm{op}} above, (L,ν(L))(L,\nu(L)) is not co-sober but is coherent. If it were Wilker, Proposition 4.8 would make it co-sober, a contradiction.

Remark 4.14 Corollaries 4.10 and 4.13 give two complementary complete-lattice obstructions:

Scott topology | even complete Boolean algebras may fail to be Wilker,

upper topology | even complete lattices may fail to be Wilker.

Thus Wilker-ness is not forced merely by completeness, even for standard order-compatible topologies.

H-supercompactness and compact-open function spaces

We now introduce the graded compactness notion that controls the compact-open topology.

Definition 5.1 Let HH be an irreducible subset system and let XX be a T0T_0-space. A subset K⊆XK \subseteq X is called HH-supercompact if, for every A∈H(Ok(X))\mathcal{A} \in H(\mathcal{O}_k(X)),

K⊆⋃A⟹K⊆U for some U∈A.K \subseteq\bigcup\mathcal{A} \Longrightarrow K \subseteq U \text{ for some } U \in\mathcal{A}.

When H=RH = R, we also say that KK is irreducibly supercompact.

Remark 5.2 If H1≤H2H_1 \leq H_2, then every H2H_2-supercompact subset is an H1H_1-supercompact subset. Hence, for the standard systems,

R-supercompact⟹WD-supercompact⟹RD-supercompactR\text{-supercompact} \Longrightarrow WD\text{-supercompact} \Longrightarrow RD\text{-supercompact}
⟹D-supercompact.\Longrightarrow D\text{-supercompact}.

Ordinary supercompactness implies RR-supercompactness.

The directed level is exactly ordinary compactness.

Proposition 5.3 For a subset KK of a T0T_0-space XX, the following are equivalent:

(1) KK is compact;

(2) KK is DD-supercompact.

Proof. Suppose first that KK is compact, and let D∈D(Ok(X))\mathcal{D} \in D(\mathcal{O}_k(X)) with K⊆⋃DK \subseteq\bigcup\mathcal{D}. Then a finite subfamily U1,…,Un∈DU_1,\ldots,U_n \in\mathcal{D} covers KK. Since D\mathcal{D} is directed under inclusion, there exists U∈DU \in\mathcal{D} with U1∪⋯∪Un⊆UU_1 \cup\cdots\cup U_n \subseteq U. Hence K⊆UK \subseteq U.

Conversely, assume that KK is DD-supercompact and let U⊆O(X)\mathcal{U} \subseteq\mathcal{O}(X) be an open cover of KK. Let D\mathcal{D} be the directed family of all finite unions of members of U\mathcal{U}. Then K⊆⋃DK \subseteq\bigcup\mathcal{D}, so K⊆VK \subseteq V for some V∈DV \in\mathcal{D}. Thus finitely many members of U\mathcal{U} cover KK.

The next observation is the technical link between compact-open function spaces and Ok(X)\mathcal{O}_k(X).

Lemma 5.4 Let X,YX,Y be T0T_0-spaces and let V∈O(Y)V \in\mathcal{O}(Y). Then the map

θV:Ck(X,Y)⟶Ok(X),θV(f)=f−1(V),\theta_V : C_k(X,Y) \longrightarrow\mathcal{O}_k(X), \quad\theta_V(f)=f^{-1}(V),

is continuous.

Proof. For every compact K⊆XK \subseteq X,

θV−1(Φ(K))={f∈C(X,Y):K⊆f−1(V)}=N(K→V),\theta_V^{-1}(\Phi(K))=\{f \in C(X,Y):K \subseteq f^{-1}(V)\}=N(K \to V),

which is open in the compact-open topology.

We now obtain the central equivalence.

Theorem 5.5 Let HH be an irreducible subset system and let XX be a T0T_0-space. Then the following conditions are equivalent:

  1. Ok(X)\mathcal{O}_k(X) is HH-sober;

  2. every compact subset of XX is HH-supercompact;

  3. for every T0T_0-space YY, the compact-open topology on C(X,Y)C(X,Y) has property SS with respect to HH;

  4. for every HH-sober space YY, the function space Ck(X,Y)C_k(X,Y) is HH-sober.

Proof. (1) ⇒\Rightarrow (2). Let K⊆XK \subseteq X be compact and let A∈H(Ok(X))\mathcal{A} \in H(\mathcal{O}_k(X)) satisfy K⊆⋃AK \subseteq\bigcup\mathcal{A}. By the HH-sobriety of Ok(X)\mathcal{O}_k(X), there is a unique UA∈O(X)U_{\mathcal{A}} \in\mathcal{O}(X) such that

cl⁡Ok(X)A=cl⁡Ok(X){UA}.\operatorname{cl}_{\mathcal{O}_k(X)}\mathcal{A}=\operatorname{cl}_{\mathcal{O}_k(X)}\{U_{\mathcal{A}}\}.

Since the specialization order of Ok(X)\mathcal{O}_k(X) is inclusion, every U∈AU \in\mathcal{A} satisfies U⊆UAU \subseteq U_{\mathcal{A}}, whence ⋃A⊆UA\bigcup\mathcal{A} \subseteq U_{\mathcal{A}}.

For the reverse inclusion, let x∈UAx \in U_{\mathcal{A}}. Then UA∈Φ({x})U_{\mathcal{A}} \in\Phi(\{x\}). Since UA∈cl⁡Ok(X)AU_{\mathcal{A}} \in\operatorname{cl}_{\mathcal{O}_k(X)}\mathcal{A}, the open set Φ({x})\Phi(\{x\}) meets A\mathcal{A}. Hence x∈Ux \in U for some U∈AU \in\mathcal{A}. Thus

UA=⋃A.U_{\mathcal{A}}=\bigcup\mathcal{A}.

Now UA∈Φ(K)U_{\mathcal{A}} \in\Phi(K), and again UA∈cl⁡Ok(X)AU_{\mathcal{A}} \in\operatorname{cl}_{\mathcal{O}_k(X)}\mathcal{A} yields Φ(K)∩A≠∅\Phi(K) \cap\mathcal{A} \ne\varnothing. Therefore K⊆UK \subseteq U for some U∈AU \in\mathcal{A}.

(2) ⇒\Rightarrow (1). Let A∈H(Ok(X))\mathcal{A} \in H(\mathcal{O}_k(X)) and put UA=⋃AU_{\mathcal{A}}=\bigcup\mathcal{A}. Since every U∈AU \in\mathcal{A} is contained in UAU_{\mathcal{A}},

cl⁡Ok(X)A⊆cl⁡Ok(X){UA}.\operatorname{cl}_{\mathcal{O}_k(X)}\mathcal{A}\subseteq\operatorname{cl}_{\mathcal{O}_k(X)}\{U_{\mathcal{A}}\}.

We show that UA∈cl⁡Ok(X)AU_{\mathcal{A}} \in\operatorname{cl}_{\mathcal{O}_k(X)}\mathcal{A}. Let Φ(K)\Phi(K) be a basic neighbourhood of UAU_{\mathcal{A}}, where KK is compact. Then

K⊆UA=⋃A.K \subseteq U_{\mathcal{A}}=\bigcup\mathcal{A}.

By (2), K⊆UK \subseteq U for some U∈AU \in\mathcal{A}, so U∈Φ(K)U \in\Phi(K). Hence every basic neighbourhood of UAU_{\mathcal{A}} meets A\mathcal{A}, and therefore

cl⁡Ok(X)A=cl⁡Ok(X){UA}.\operatorname{cl}_{\mathcal{O}_k(X)}\mathcal{A}=\operatorname{cl}_{\mathcal{O}_k(X)}\{U_{\mathcal{A}}\}.

The generic point is unique because Ok(X)\mathcal{O}_k(X) is T0T_0.

(2) ⇒\Rightarrow (3). Let YY be a T0T_0-space, F∈H(Ck(X,Y))\mathcal{F} \in H(C_k(X,Y)), and g∈C(X,Y)g \in C(X,Y) satisfy

g−1(V)=⋃f∈Ff−1(V)(V∈O(Y)).g^{-1}(V)=\bigcup_{f\in\mathcal{F}}f^{-1}(V)\quad(V\in\mathcal{O}(Y)).

Let N(K→V)N(K\to V) be a subbasic compact-open neighbourhood of gg. By Lemma 5.4 and functoriality of HH,

AV:={f−1(V):f∈F}∈H(Ok(X)).\mathcal{A}_V:=\{f^{-1}(V):f\in\mathcal{F}\}\in H(\mathcal{O}_k(X)).

Moreover,

K⊆g−1(V)=⋃AV.K\subseteq g^{-1}(V)=\bigcup\mathcal{A}_V.

By (2), there is f∈Ff\in\mathcal{F} such that K⊆f−1(V)K\subseteq f^{-1}(V), equivalently f∈N(K→V)f\in N(K\to V). Thus every subbasic neighbourhood of gg meets F\mathcal{F}. Since F\mathcal{F} is irreducible, every finite intersection of such neighbourhoods also meets F\mathcal{F}. Hence g∈cl⁡Ck(X,Y)Fg\in\operatorname{cl}_{C_k(X,Y)}\mathcal{F}, proving property SS.

(3) ⇒\Rightarrow (4) follows from Proposition 2.4, since the compact-open topology is finer than the pointwise convergence topology.

(4) ⇒\Rightarrow (1). The Sierpiński space SS is sober, hence HH-sober. Therefore Ck(X,S)C_k(X,S) is HH-sober by (4). By the homeomorphism (2.1), Ok(X)\mathcal{O}_k(X) is HH-sober.

The theorem immediately reorganizes several of the questions from [20].

Corollary 5.6 For a fixed T0T_0-space XX the following hold.

(1) The following are equivalent:

(a) Ok(X)\mathcal{O}_k(X) is sober;

(b) every compact subset of XX is RR-supercompact;

(c) for every T0T_0-space YY, the compact-open topology on C(X,Y)C(X,Y) has property SS with respect to RR;

(d) for every sober space YY, Ck(X,Y)C_k(X,Y) is sober.

(2) The following are equivalent:

(a) Ok(X)\mathcal{O}_k(X) is well-filtered;

(b) every compact subset of XX is WD-supercompact;

(c) every compact subset of XX is RD-supercompact;

(d) for every T0T_0-space YY, the compact-open topology on C(X,Y)C(X,Y) has property SS with respect to WD (equivalently, with respect to RD);

(e) for every well-filtered space YY, Ck(X,Y)C_k(X,Y) is well-filtered.

Consequently, the universal forms of Questions 49, 55, 58 and 60 in [20] are equivalent, and the universal forms of Questions 48, 56, 57 and 59 are equivalent (for either WD or RD).

Proof. Apply Theorem 5.5 with H=RH=R, WDWD and RDRD, and use the characterizations of sober and well-filtered spaces by RR-, WDWD- and RDRD-sobriety from [19, 21].

Corollary 5.7 For every T0T_0-space XX, the following are equivalent:

  1. every compact subset of XX is WD-supercompact;

  2. every compact subset of XX is RD-supercompact.

Remark 5.8 Corollary 5.7 concerns the assertion that all compact subsets have the corresponding property. It does not assert that the WD and RD notions of supercompactness coincide for an individual fixed subset.

There is a close relation with the SI-compactness introduced by Zhao and Ho [24]. They call XX SI-compact if every open cover of XX which is irreducible in the Scott space ΣO(X)\Sigma O(X) contains XX itself. Since ϕ(O(X))⊆σ(O(X))\phi(O(X)) \subseteq\sigma(O(X)), every subset irreducible in the Scott topology is irreducible in ϕ(O(X))\phi(O(X)). Hence R-supercompactness in the sense of Definition 5.1 is at least as strong as the corresponding Scott-irreducible covering condition. If XX is consonant, then ϕ(O(X))=σ(O(X))\phi(O(X)) = \sigma(O(X)), and the two irreducibility tests agree.

Wilker property and H-supercompactness

We now show that Wilker’s condition is a strong bridge from ordinary compactness to the whole HH-supercompactness hierarchy. The Wilker definition and its finite form were given in § 2; we therefore pass directly to the main implication.

Theorem 6.1 Let XX be a Wilker space and let HH be an irreducible subset system. Then every compact subset of XX is HH-supercompact.

Proof. Let K⊆XK \subseteq X be compact and let A∈H(Ok(X))A \in H(O_k(X)) satisfy K⊆⋃AK \subseteq\bigcup A. Since KK is compact, there are U1,…,Un∈AU_1,\ldots,U_n \in A such that

K⊆U1∪⋯∪Un.K \subseteq U_1 \cup\cdots\cup U_n.

By Lemma 2.2, choose compact Ki⊆UiK_i \subseteq U_i such that

K⊆K1∪⋯∪Kn.K \subseteq K_1 \cup\cdots\cup K_n.

For each ii, the basic open set Φ(Ki)\Phi(K_i) of Ok(X)O_k(X) meets AA, because Ui∈A∩Φ(Ki)U_i \in A \cap\Phi(K_i). Since A∈H(Ok(X))⊆Irr⁡(Ok(X))A \in H(O_k(X)) \subseteq\operatorname{Irr}(O_k(X)), irreducibility gives

A∩⋂i=1nΦ(Ki)≠∅.A \cap\bigcap_{i=1}^{n} \Phi(K_i) \ne\varnothing.

Choose UU in this intersection. Then Ki⊆UK_i \subseteq U for every ii, and therefore K⊆UK \subseteq U. Thus KK is HH-supercompact.

Combining Theorems 5.5 and 6.1 yields the main function-space consequence.

Corollary 6.2 Let XX be a Wilker space and let HH be an irreducible subset system. Then we have the following conclusions:

  1. Ok(X)\mathcal{O}_k(X) is HH-sober.

  2. For every T0T_0-space YY, the compact-open topology on C(X,Y)C(X,Y) has property SS with respect to HH.

  3. For every HH-sober space YY, Ck(X,Y)C_k(X,Y) is HH-sober.

Corollary 6.3 Let XX be a Wilker space. Then we have the following conclusions:

  1. Ok(X)\mathcal{O}_k(X) is sober (hence well-filtered).

  2. If YY is sober, then Ck(X,Y)C_k(X,Y) is sober.

  3. If YY is well-filtered, then Ck(X,Y)C_k(X,Y) is well-filtered.

  4. The compact-open topology on C(X,Y)C(X,Y) has property SS with respect to RR, WD and RD, for every T0T_0-space YY.

In particular, Questions 48, 49, 55, 56, 59 and 60 of [20] have affirmative answers whenever the domain space XX is Wilker.

Wilker showed that condition (D) (i.e., the condition of Definition 2.1) holds for some important classes of spaces, including Hausdorff spaces and locally compact spaces [17]. Thus Corollary 6.2 applies to all such domain spaces.

We also obtain a short proof of a recent result on Scott-open set lattices.

Corollary 6.4 If XX is consonant and Wilker, then the Scott space ΣO(X)\Sigma\mathcal{O}(X) is sober.

Proof. By Corollary 6.2, Ok(X)\mathcal{O}_k(X) is sober. Consonance is equivalent to

ϕ(O(X))=σ(O(X)),\phi(\mathcal{O}(X)) = \sigma(\mathcal{O}(X)),

see, for example, [6]. Hence ΣO(X)=Ok(X)\Sigma\mathcal{O}(X) = \mathcal{O}_k(X) is sober. □

Remark 6.5 Corollary 6.4 recovers the consonant-Wilker part of Z. He’s preprint When do weakly first-countable spaces and the Scott topology of open set lattice become sober? (2025, arXiv:2503.12766v2). The present proof separates the two mechanisms: Wilker’s condition supplies the irreducible compact-cover refinement, while consonance identifies the compact-generated topology ϕ(O(X))\phi(\mathcal{O}(X)) with the Scott topology.

The Hoare power construction

Coherence, local compactness and non-preservation

Let PH(X)P_H(X) denote the set of all nonempty closed subsets of XX equipped with the lower Vietoris topology, generated by

◊U={F∈PH(X):F∩U≠∅},U∈O(X).\Diamond U = \{F \in P_H(X) : F \cap U \ne\varnothing\}, \quad U \in\mathcal{O}(X).

Schalk proved that the Hoare hyperspace (with the empty closed set included) is always sober [15]; the nonempty part is the open subspace ◊X\Diamond X, hence it is sober as well.

Proposition 7.1 (Schalk) For every T0T_0-space XX, PH(X)P_H(X) is sober.

The lower Vietoris semilattice also carries a useful compactness property that is independent of the base space.

Proposition 7.2 For every T0T_0-space XX, the Hoare powerspace PH(X)P_H(X) is coherent.

Proof. Put Y=PH(X)Y=P_H(X) and consider the union map

m:Y×Y⟶Y,m(A,B)=A∪B.m:Y\times Y\longrightarrow Y,\quad m(A,B)=A\cup B.

For every U∈O(X)U\in\mathcal{O}(X),

m−1(◊U)=(◊U×Y)∪(Y×◊U).m^{-1}(\Diamond U)=(\Diamond U\times Y)\cup(Y\times\Diamond U).

so mm is continuous. Let K1,K2K_1,K_2 be compact saturated subsets of YY. Since the specialization order of YY is inclusion, both families are upward closed under inclusion. Therefore

m(K1×K2)=K1∩K2.m(K_1\times K_2)=K_1\cap K_2.

Indeed, A∪BA\cup B belongs to both families whenever A∈K1A\in K_1 and B∈K2B\in K_2, while every C∈K1∩K2C\in K_1\cap K_2 equals m(C,C)m(C,C). The intersection is thus a continuous image of a compact space, hence compact; it is saturated as an intersection of saturated sets. Thus YY is coherent.

Combining Proposition 7.2 with Proposition 4.8 yields a useful obstruction.

Corollary 7.3 For every T0T_0-space XX,

PH(X) Wilker⟹PH(X) co-sober.P_H(X)\ \textit{Wilker}\Longrightarrow P_H(X)\ \textit{co-sober}.

For later use we include a direct compact-neighborhood argument in the locally compact case. Schalk proved the corresponding lower-Vietoris local-compactness result [15 Proposition 6.11]; stronger formulations for core-compact bases are also standard [8 Proposition 2.9].

Lemma 7.4 If XX is locally compact, then PH(X)P_H(X) is locally compact.

Proof. Let F∈PH(X)F\in P_H(X) and let

U=◊U1∩⋯∩◊UnU=\Diamond U_1\cap\cdots\cap\Diamond U_n

be a basic neighborhood of FF. Choose xi∈F∩Uix_i \in F \cap U_i. By local compactness, choose open ViV_i and compact saturated QiQ_i such that

xi∈Vi⊆Qi⊆Ui.x_i \in V_i \subseteq Q_i \subseteq U_i.

Then

F∈⋂i◊Vi⊆⋂i◊Qi⊆U.F \in\bigcap_i \Diamond V_i \subseteq\bigcap_i \Diamond Q_i \subseteq\mathcal{U}.

It remains to show that K=⋂i◊QiK = \bigcap_i \Diamond Q_i is compact. Consider

φ:Q1×⋯×Qn⟶PH(X),φ(q1,…,qn)={q1,…,qn}.\varphi: Q_1 \times\cdots\times Q_n \longrightarrow P_H(X), \quad\varphi(q_1,\ldots,q_n) = \{q_1,\ldots,q_n\}.

For open U⊆XU \subseteq X,

φ−1(◊U)=⋃i=1nπi−1(U),\varphi^{-1}(\Diamond U) = \bigcup_{i=1}^{n} \pi_i^{-1}(U),

because an open set meets {q1,…,qn}\{q_1,\ldots,q_n\} if and only if it contains at least one qiq_i. Hence φ\varphi is continuous. Its image is compact. Since the specialization order of PH(X)P_H(X) is inclusion, the saturation of this image consists exactly of the closed sets containing {q1,…,qn}\{q_1,\ldots,q_n\} for some (q1,…,qn)∈Q1×⋯×Qn(q_1,\ldots,q_n) \in Q_1 \times\cdots\times Q_n; equivalently, it is ⋂i◊Qi=K\bigcap_i \Diamond Q_i = K. By Lemma 2.3, KK is compact.

We now relate the Hoare construction to core-compactness. Let XsX^s be the sobrification of XX. Closed subsets of XX and XsX^s correspond naturally: for closed F⊆XF \subseteq X, put

F^={A∈Irr⁡c(X):A⊆F}.\widehat{F} = \{A \in\operatorname{Irr}_c(X) : A \subseteq F\}.

Lemma 7.5 The map F↦F^F \mapsto\widehat{F} is a homeomorphism

PH(X)≅PH(Xs).P_H(X) \cong P_H(X^s).

Proof. The standard sobrification induces an order isomorphism O(X)≅O(Xs)\mathcal{O}(X) \cong\mathcal{O}(X^s) and hence, by complementation, a bijection between the closed subsets of XX and those of XsX^s; under this bijection FF is sent to F^\widehat{F}. It remains only to compare the lower Vietoris topologies. If U∈O(X)U \in\mathcal{O}(X) corresponds to

Us={A∈Irr⁡c(X):A∩U≠∅},U^s = \{A \in\operatorname{Irr}_c(X) : A \cap U \ne\varnothing\},

then

F^∩Us≠∅⟺F∩U≠∅.\widehat{F} \cap U^s \ne\varnothing\quad\Longleftrightarrow\quad F \cap U \ne\varnothing.

Indeed, the forward implication is immediate. Conversely, if x∈F∩Ux \in F \cap U, then {x}⊆F\{x\} \subseteq F is irreducible closed and meets UU. Thus lower Vietoris subbasic opens correspond exactly.

The implication from core-compactness of XX to local compactness of the full lower Vietoris hyperspace is contained in the classical theory recalled above. The following formulation for the nonempty Hoare powerspace also records the converse in a form useful below.

Theorem 7.6 For every T0T_0-space XX, the following are equivalent:

(1) XX is core-compact;

(2) PH(X)P_H(X) is locally compact;

(3) PH(X)P_H(X) is core-compact.

Proof. (1) ⇒\Rightarrow (2). The sobrification XsX^{s} of a core-compact space is locally compact [5]. By Lemma 7.4, PH(Xs)P_H(X^{s}) is locally compact, and by Lemma 7.5 it is homeomorphic to PH(X)P_H(X).

(2) ⇒\Rightarrow (1). Let x∈U∈O(X)x \in U \in\mathcal{O}(X). The canonical map

ηX:X→PH(X),x↦{x}‾,\eta_X : X \to P_H(X),\quad x \mapsto\overline{\{x\}},

is continuous and satisfies ηX−1(◊U)=U\eta_X^{-1}(\Diamond U)=U. By local compactness of PH(X)P_H(X), choose an open V\mathcal{V} and a compact set K\mathcal{K} such that

ηX(x)∈V⊆K⊆◊U.\eta_X(x) \in\mathcal{V} \subseteq\mathcal{K} \subseteq\Diamond U.

Put V=ηX−1(V)V=\eta_X^{-1}(\mathcal{V}). We show V≪UV\ll U in O(X)\mathcal{O}(X). Let (Ui)i∈I(U_i)_{i\in I} be directed with U⊆⋃iUiU\subseteq\bigcup_{i}U_i. Then

K⊆◊U⊆⋃i◊Ui.\mathcal{K}\subseteq\Diamond U\subseteq\bigcup_i\Diamond U_i.

Compactness and directedness yield i0i_0 with K⊆◊Ui0\mathcal{K}\subseteq\Diamond U_{i_0}. Hence

V⊆ηX−1(K)⊆Ui0.V\subseteq\eta_X^{-1}(\mathcal{K})\subseteq U_{i_0}.

Thus every x∈Ux\in U has a neighborhood V≪UV\ll U, so XX is core-compact.

(2) ⇔\Leftrightarrow (3). Local compactness always implies core-compactness. Conversely, PH(X)P_H(X) is sober by Proposition 7.1, and a sober core-compact space is locally compact [5].

Corollary 7.7 The Hoare powerspace construction preserves and reflects core-compactness.

Corollary 7.8 The Hoare powerspace construction preserves local compactness but does not reflect local compactness.

Proof. Preservation is Lemma 7.4. For non-reflection, take the Hofmann–Lawson space XHLX_{HL}. It is not locally compact, but it is core-compact, so PH(XHL)P_H(X_{HL}) is locally compact by Theorem 7.6.

The reason for this asymmetry is transparent: PH(X)P_H(X) carries sobriety automatically. The Hoare construction therefore turns the core-compactness of XX into core-compactness plus sobriety of PH(X)P_H(X), which is exactly local compactness. Combining Theorem 7.6 with Wilker’s classical theorem gives the following large positive region.

Corollary 7.9. If XX is core-compact, then PH(X)P_H(X) is locally compact, sober and Wilker.

This places Question 4.2 in a new light: for every core-compact XX, the Hoare powerspace is already Wilker; the missing issue is whether the Wilker property can be recovered on the base space.

The preservation problem, however, has a strong negative answer. In their preprint A countable metric space whose Hoare power space is not co-sober (2026, arXiv:2610.05868v1), Xu and Ji constructed a countable zero-dimensional metrizable space MM such that PH(M)P_H(M) is not co-sober. Since MM is metrizable, it is Hausdorff and hence Wilker, whereas Corollary 7.3 excludes Wilker-ness of its Hoare powerspace.

Corollary 7.10 (Failure of Hoare preservation) There exists a countable zero-dimensional metrizable Wilker space MM such that PH(M)P_H(M) is not Wilker. Consequently, the Hoare powerspace construction does not preserve Wilker-ness, even within countable zero-dimensional metrizable spaces.

Proof. Take MM as above. If PH(M)P_H(M) were Wilker, then it would be co-sober by Corollary 7.3, contrary to the metric-space preprint (arXiv:2610.05868v1).

The unrestricted reflection direction remains open. The results below give an operation-theoretic reformulation and settle it on consonant bases.

Question 7.11 (Hoare reflection) If PH(X)P_H(X) is Wilker, must XX be Wilker?

A negative answer to Question 4.2 would automatically give a negative answer to Question 7.11: if XX were core-compact but non-Wilker, then PH(X)P_H(X) would be locally compact and hence Wilker by Corollary 7.9.

Although preservation fails in general, it holds on important subclasses. Core-compact bases are covered by Corollary 7.9; moreover, lower powerspaces of quasi-Polish spaces are quasi-Polish, and quasi-Polish spaces are Wilker [2].

Union on the compact-open lattice

We first record an elementary operation-theoretic form of the Wilker condition. In the proof, empty compact saturated sets are allowed.

Proposition 7.12 A space XX is Wilker if and only if

⋃:Ok(X)×Ok(X)⟶Ok(X)\bigcup: \mathcal{O}_k(X) \times\mathcal{O}_k(X) \longrightarrow\mathcal{O}_k(X)

is continuous, where the domain has the ordinary product topology.

Proof. Suppose XX is Wilker and U∪V∈Φ(K)U \cup V \in\Phi(K) for a compact saturated K⊆XK \subseteq X. Choose compact saturated K1⊆UK_1 \subseteq U and K2⊆VK_2 \subseteq V with K⊆K1∪K2K \subseteq K_1 \cup K_2. Then Φ(K1)×Φ(K2)\Phi(K_1) \times\Phi(K_2) is a neighborhood of (U,V)(U,V) mapped into Φ(K)\Phi(K).

Conversely, let K⊆U∪VK \subseteq U \cup V be compact saturated and suppose union is continuous. The sets Φ(C)\Phi(C) form a base for Ok(X)\mathcal{O}_k(X), since finite intersections correspond to finite unions of compact sets. There are therefore compact saturated K1⊆UK_1 \subseteq U and K2⊆VK_2 \subseteq V such that

⋃(Φ(K1)×Φ(K2))⊆Φ(K).\bigcup\left(\Phi(K_1) \times\Phi(K_2)\right) \subseteq\Phi(K).

If x∈K∖(K1∪K2)x \in K \setminus(K_1 \cup K_2), saturation provides open sets Ui′U_i' containing KiK_i and omitting xx. Their union must contain KK, a contradiction. Thus K⊆K1∪K2K \subseteq K_1 \cup K_2, proving the Wilker condition by Lemma 2.3.

A Scott-lattice criterion for Hoare powerspaces

We next show that the corresponding operation on the Scott space ΣO(X)\Sigma\mathcal{O}(X) characterizes Wilker-ness of the Hoare powerspace. One technical step is needed to pass from lower Vietoris subbasic covers to arbitrary open covers.

Lemma 7.13 (A coherent-space refinement lemma) Let YY be coherent, and let S\mathcal{S} be an open subbasis closed under finite unions and containing ∅,Y\varnothing, Y. Suppose that whenever a compact saturated KK satisfies K⊆S∪TK \subseteq S \cup T with S,T∈SS,T \in\mathcal{S}, there are compact saturated sets A⊆SA \subseteq S, B⊆TB \subseteq T covering KK. Then YY is Wilker.

Proof. By Lemma 2.3, it suffices to split compact saturated sets. We may replace the compact pieces supplied by the hypothesis with their intersections with KK, using coherence.

Let K⊆O1∪O2K \subseteq O_1 \cup O_2 with KK compact saturated and OiO_i open. Choose a finite cover of KK by basic opens, each contained in one OiO_i, and group them into two finite unions. Distributivity and finite-union closure of S\mathcal{S} express these unions as

G1=⋂j=1mSj⊆O1,G2=⋂k=1nTk⊆O2,K⊆G1∪G2,G_1 = \bigcap_{j=1}^{m} S_j \subseteq O_1,\qquad G_2 = \bigcap_{k=1}^{n} T_k \subseteq O_2,\qquad K \subseteq G_1 \cup G_2,

where Sj,Tk∈SS_j,T_k \in\mathcal{S}. An empty group means that KK is contained in the other open set and is immediate; a whole-space group is equally immediate. We may therefore take m,n≥1m,n \geq1.

For each j,kj,k, use K⊆Sj∪TkK \subseteq S_j \cup T_k to obtain compact saturated Ajk,Bjk⊆KA_{jk},B_{jk} \subseteq K covering KK, with Ajk⊆SjA_{jk} \subseteq S_j and Bjk⊆TkB_{jk} \subseteq T_k. Put

C1=⋂j=1m⋃k=1nAjk,C2=⋃j=1m⋂k=1nBjk.C_1 = \bigcap_{j=1}^{m}\bigcup_{k=1}^{n} A_{jk},\qquad C_2 = \bigcup_{j=1}^{m}\bigcap_{k=1}^{n} B_{jk}.

Finite unions and coherence make both sets compact saturated, and C1⊆G1C_1 \subseteq G_1, C2⊆G2C_2 \subseteq G_2. If x∈K∖C1x \in K \setminus C_1, then some jj satisfies x∉Ajkx \notin A_{jk} for every kk, so x∈Bjkx \in B_{jk} for every kk and hence x∈C2x \in C_2. Thus K⊆C1∪C2K \subseteq C_1 \cup C_2.

For the next correspondence we temporarily include the empty closed set. Let A(X)\mathcal{A}(X) be the space of all closed subsets of XX with the lower Vietoris topology, and use ⋄U\diamond U for {F∈A(X):F∩U≠∅}\{F \in\mathcal{A}(X) : F \cap U \ne\varnothing\} when working in A(X)\mathcal{A}(X). Then PH(X)=⋄XP_H(X) = \diamond X is its open subspace of nonempty closed sets.

Lemma 7.14. The space A(X)\mathcal{A}(X) is Wilker if and only if PH(X)P_H(X) is Wilker. Moreover, A(X)\mathcal{A}(X) is coherent.

Proof. The forward implication follows from open heredity. An open set in A(X)\mathcal{A}(X) containing the point ∅\varnothing is the whole space. If a two-set cover contains such an open set, take the entire compact set as one piece and the empty set as the other. Otherwise the open sets and the compact set being covered lie in PH(X)P_H(X), so Wilker-ness there proves the converse.

For coherence, the proof of Proposition 7.2 applies unchanged with the empty closed set included: union is continuous, and its image on the product of two compact saturated families is their intersection.

Let K\mathcal{K} be a compact saturated subset of A(X)\mathcal{A}(X), possibly empty, and let H∈σ(O(X))\mathcal{H} \in\sigma(\mathcal{O}(X)). Define

h(K)={U∈O(X):K⊆⋄U},h(\mathcal{K}) = \{U \in\mathcal{O}(X) : \mathcal{K} \subseteq\diamond U\},
k(H)=⋂U∈H⋄U.(7)k(\mathcal{H}) = \bigcap_{U \in\mathcal{H}} \diamond U. \tag*{(7)}

We use the following established order correspondence, without any countable-basis or consonance hypothesis.

Proposition 7.15 (de Brecht–Kawai [2 Definition 6.7 and Theorem 6.10]) The maps in (7.1) are mutually inverse, inclusion-reversing bijections between the compact saturated subsets of A(X)\mathcal{A}(X) and the Scott-open subsets of O(X)\mathcal{O}(X). In particular,

U∈h(K)⟺K⊆⋄U(8)U \in h(\mathcal{K}) \Longleftrightarrow\mathcal{K} \subseteq\diamond U \tag*{(8)}
h(K1∪K2)=h(K1)∩h(K2),(9)h(\mathcal{K}_1 \cup\mathcal{K}_2) = h(\mathcal{K}_1) \cap h(\mathcal{K}_2), \tag*{(9)}
k(H1∩H2)=k(H1)∪k(H2).(10)k(\mathcal{H}_1 \cap\mathcal{H}_2) = k(\mathcal{H}_1) \cup k(\mathcal{H}_2). \tag*{(10)}

In the Smyth order on compact families these maps are order-preserving; “inclusion-reversing” refers to ordinary set inclusion on both sides. The empty cases are included: h(∅)=O(X)h(\varnothing) = \mathcal{O}(X) and k(∅)=A(X)k(\varnothing) = \mathcal{A}(X). Only the order correspondence is used below, not the stronger topological assertion in [2 Theorem 6.11], which requires a countable base.

Theorem 7.16 (Hoare–Scott union criterion) For every T0T_0-space XX, the following are equivalent:

(1) PH(X)P_H(X) is Wilker;

(2) the union operation

∪:ΣO(X)×ΣO(X)⟶ΣO(X)\cup: \Sigma\mathcal{O}(X) \times\Sigma\mathcal{O}(X) \longrightarrow\Sigma\mathcal{O}(X)

is jointly continuous, where the domain is the ordinary product of the two Scott spaces.

Proof. Assume (1), and let U∪V∈H∈σ(O(X))U \cup V \in\mathcal{H} \in\sigma(\mathcal{O}(X)). By Proposition 7.15, the compact saturated family k(H)\mathfrak{k}(\mathcal{H}) satisfies

k(H)⊆◊(U∪V)=◊U∪◊V.\mathfrak{k}(\mathcal{H}) \subseteq\Diamond(U \cup V) = \Diamond U \cup\Diamond V.

Use Lemma 7.14 and Wilker-ness to choose compact saturated families K1⊆◊U\mathcal{K}_1 \subseteq\Diamond U and K2⊆◊V\mathcal{K}_2 \subseteq\Diamond V covering k(H)\mathfrak{k}(\mathcal{H}). Put Hi=h(Ki)\mathcal{H}_i = \mathfrak{h}(\mathcal{K}_i). These are Scott-open neighborhoods of U,VU,V, respectively, and (7.3) gives

H1∩H2⊆H.\mathcal{H}_1 \cap\mathcal{H}_2 \subseteq\mathcal{H}.

For U′∈H1U' \in\mathcal{H}_1 and V′∈H2V' \in\mathcal{H}_2, upperness implies U′∪V′∈H1∩H2⊆HU' \cup V' \in\mathcal{H}_1 \cap\mathcal{H}_2 \subseteq\mathcal{H}. This proves joint continuity.

Conversely, assume (2). Suppose that a compact saturated family K⊆A(X)\mathcal{K} \subseteq\mathcal{A}(X) is covered by ◊U\Diamond U and ◊V\Diamond V. Then U∪V∈h(K)U \cup V \in\mathfrak{h}(\mathcal{K}). Continuity gives Scott-open neighborhoods H1,H2\mathcal{H}_1,\mathcal{H}_2 of U,VU,V such that U′∪V′∈h(K)U' \cup V' \in\mathfrak{h}(\mathcal{K}) whenever U′∈H1U' \in\mathcal{H}_1 and V′∈H2V' \in\mathcal{H}_2. Taking U′=V′U'=V' shows

H1∩H2⊆h(K).\mathcal{H}_1 \cap\mathcal{H}_2 \subseteq\mathfrak{h}(\mathcal{K}).

Apply (7.4) to obtain

K⊆k(H1)∪k(H2),k(H1)⊆◊U,k(H2)⊆◊V.\mathcal{K} \subseteq\mathfrak{k}(\mathcal{H}_1) \cup\mathfrak{k}(\mathcal{H}_2), \qquad\mathfrak{k}(\mathcal{H}_1) \subseteq\Diamond U,\quad\mathfrak{k}(\mathcal{H}_2) \subseteq\Diamond V.

These are compact saturated pieces. The diamonds, together with A(X)\mathcal{A}(X), form an open subbasis closed under finite unions; a cover using A(X)\mathcal{A}(X) itself splits trivially. Lemma 7.14 gives coherence, and Lemma 7.13 upgrades this subbasic splitting property to Wilker-ness of A(X)\mathcal{A}(X). Another application of Lemma 7.14 proves (1).

Remark 7.17 Union always preserves directed suprema as a map from the order product O(X)×O(X)\mathcal{O}(X) \times\mathcal{O}(X). This does not imply the joint continuity in Theorem 7.16: its domain has topology σ(O(X))×σ(O(X))\sigma(\mathcal{O}(X)) \times\sigma(\mathcal{O}(X)), not the possibly finer topology σ(O(X)×O(X))\sigma(\mathcal{O}(X) \times\mathcal{O}(X)). Thus the theorem is not a consequence of order-theoretic Scott continuity alone.

Consonant bases and the reflection question

Recall that consonance is the equality ϕ(O(X))=σ(O(X))\phi(\mathcal{O}(X)) = \sigma(\mathcal{O}(X)). The preceding two operation-theoretic characterizations therefore agree on consonant base spaces.

Corollary 7.18 (Hoare preservation and reflection on consonant bases) For every consonant T0T_0-space XX,

X is Wilker⟺PH(X) is Wilker.X\text{ is Wilker} \quad\Longleftrightarrow\quad P_H(X)\text{ is Wilker}.

In particular, Question 7.11 has an affirmative answer for consonant XX, and every counterexample in the general case must be non-consonant.

Proof. Consonance identifies Ok(X)O_k(X) with ΣO(X)\Sigma O(X). Apply Proposition 7.12 and Theorem 7.16.

For consonant spaces, the equivalence between Wilker-ness and joint Scott continuity of union was already obtained by de Brecht and is recorded in [7 Proposition J], where it is attributed to a conversation with Keimel. We use this established connection; the Hoare consequence here comes from combining it with Theorem 7.16.

Remark 7.19 (Compact transversals). There is also a direct proof of the reflection direction that isolates the use of consonance. Let K⊆U1∪U2K \subseteq U_1 \cup U_2 be compact in XX, and put η(x)=cl⁡X{x}\eta(x) = \operatorname{cl}_X\{x\}. If PH(X)P_H(X) is Wilker, split η(K)\eta(K) into compact families Ki⊆◊UiK_i \subseteq\Diamond U_i. Saturate the families, which keeps them compact and inside ◊Ui\Diamond U_i. Consonance gives compact saturated Ci⊆UiC_i \subseteq U_i such that

Φ(Ci)⊆h(Ki).\Phi(C_i) \subseteq h(K_i).

Every F∈KiF \in K_i meets CiC_i: otherwise X∖FX \setminus F belongs to Φ(Ci)\Phi(C_i), but cannot belong to h(Ki)h(K_i). If η(x)∈Ki\eta(x) \in K_i, then cl⁡X{x}∩Ci≠∅\operatorname{cl}_X\{x\} \cap C_i \ne\varnothing, and saturation of CiC_i implies x∈Cix \in C_i. Thus K⊆C1∪C2K \subseteq C_1 \cup C_2. This is the compact transversal mechanism of [2 Lemma 6.12 and Theorem 6.13].

For arbitrary XX, Question 7.11 is now equivalent to whether the following implication always holds:

∪:ΣO(X)×ΣO(X)→ΣO(X) is continuous⇒∪:Ok(X)×Ok(X)→Ok(X) is continuous.(11)\begin{aligned} \cup&: \Sigma\mathcal{O}(X) \times\Sigma\mathcal{O}(X) \to\Sigma\mathcal{O}(X)\text{ is continuous}\\ &\qquad\qquad\qquad\qquad\qquad\qquad\Rightarrow\\ \cup&: O_k(X) \times O_k(X) \to O_k(X)\text{ is continuous}. \tag*{(11)} \end{aligned}

Both domains here have the ordinary product topology. The inclusion ϕ(O(X))⊆σ(O(X))\phi(\mathcal{O}(X)) \subseteq\sigma(\mathcal{O}(X)) alone does not justify this transfer of joint continuity. We have proved it when XX is consonant, but do not claim it for arbitrary XX. In particular, the general Hoare-reflection problem and the core-compactness question 4.2 remain unresolved here.

The Smyth power construction

Let PS(X)P_S(X) be Q(X)\mathcal{Q}(X) equipped with the upper Vietoris topology, whose basic open sets are

□U={K∈Q(X):K⊆U}, U∈O(X).\Box U = \{K \in\mathcal{Q}(X) : K \subseteq U\},\ U \in\mathcal{O}(X).

The specialization order is reverse inclusion.

The local-compactness behavior of the Smyth construction is very different from that of the Hoare construction.

Theorem 8.1 (Lyu–Chen–Jia [12]) For every topological space XX, the following are equivalent:

(1) XX is locally compact;

(2) PS(X)P_S(X) is locally compact;

(3) PS(X)P_S(X) is core-compact.

Hence local compactness gives a first positive answer to the Wilker-preservation problem for Smyth powerspaces.

Corollary 8.2 If XX is locally compact, then PS(X)P_S(X) is Wilker.

We now prove a reflection theorem. We use the following standard compact-union lemma.

Lemma 8.3 (Schalk [15]) If K\mathcal{K} is a compact subset of PS(X)P_S(X), then

⋃K\bigcup\mathcal{K}

is a compact saturated subset of XX.

Proof. The union is saturated because every member of K\mathcal{K} is saturated. Let {Uj:j∈J}\{U_j:j\in J\} be an open cover of ⋃K\bigcup\mathcal{K}. For each Q∈KQ\in\mathcal{K}, compactness of QQ gives a finite JQ⊆JJ_Q\subseteq J such that Q⊆VQ:=⋃j∈JQUjQ\subseteq V_Q:=\bigcup_{j\in J_Q}U_j. The sets □VQ\Box V_Q form an open cover of K\mathcal{K}. Choose finitely many Q1,…,QmQ_1,\ldots,Q_m whose corresponding boxes cover K\mathcal{K}. The finitely many UjU_j appearing in the JQkJ_{Q_k} cover every member of K\mathcal{K}, hence cover ⋃K\bigcup\mathcal{K}.

Theorem 8.4 (Smyth reflection of Wilker-ness) If PS(X)P_S(X) is Wilker, then XX is Wilker.

Proof. Let K⊆XK\subseteq X be compact and K⊆U1∪U2K\subseteq U_1\cup U_2 with UiU_i open. Consider the canonical map

ξX:X→PS(X),ξX(x)=↑x.\xi_X:X\to P_S(X),\qquad\xi_X(x)=\mathord{\uparrow}x.

The set ↑x\mathord{\uparrow}x is compact saturated by Lemma 2.3, applied to the compact singleton {x}\{x\}. The map ξX\xi_X is continuous, since ξX−1(□U)=U\xi_X^{-1}(\Box U)=U. Therefore ξX(K)\xi_X(K) is compact in PS(X)P_S(X), and

ξX(K)⊆□U1∪□U2.\xi_X(K)\subseteq\Box U_1\cup\Box U_2.

Since PS(X)P_S(X) is Wilker, there are compact subsets Ki⊆□UiK_i\subseteq\Box U_i such that

ξX(K)⊆K1∪K2.\xi_X(K)\subseteq K_1\cup K_2.

Set Ki=⋃KiK_i=\bigcup K_i. By Lemma 8.3, KiK_i is compact saturated, and Ki⊆UiK_i\subseteq U_i. If x∈Kx\in K, then ↑x=ξX(x)\mathord{\uparrow}x=\xi_X(x) belongs to one of the KiK_i, hence x∈Kix\in K_i. Thus K⊆K1∪K2K\subseteq K_1\cup K_2.

So the difficult direction is preservation.

Question 8.5 (Smyth preservation) If XX is Wilker, must PS(X)P_S(X) be Wilker?

Again there are large positive subclasses. Besides the locally compact case from Corollary 8.2, the Smyth powerspace of a quasi-Polish space is quasi-Polish, hence Wilker [2].

The Hofmann–Lawson space provides a particularly natural test case: it is core-compact, Wilker and not locally compact. By Theorem 8.1, its Smyth powerspace is not core-compact and not locally compact. Thus Question 8.5 is genuinely outside the automatic locally compact regime already for this classical example.

Two central open problems and further directions

Two central open problems on Wilker spaces

Theorem 3.16 settles the binary product problem negatively, even for countable compact hereditarily Wilker factors. Corollary 3.17 gives a square counterexample, and Corollary 3.19 settles categorical reflectivity negatively. We therefore single out two remaining structural problems.

(W1) Smyth preservation. If XX is Wilker, must PS(X)P_S(X) be Wilker?

This is Question 8.5. The reverse implication is Theorem 8.4, and preservation holds for locally compact and quasi-Polish spaces. The Hofmann–Lawson space remains a natural test case: it is core-compact and Wilker but not locally compact, and its Smyth powerspace lies outside the automatic locally compact regime.

(W2) Core-compactness versus Wilker-ness. Is every core-compact T0T_0-space Wilker?

This is Question 4.2. The complete-lattice Scott-space case is positive by Proposition 4.6. Corollary 4.4 shows that core-compactness together with Wilker-ness does not force local compactness, so an affirmative answer cannot be reduced to a general local-compactness assertion.

The product counterexamples do not answer either question. They give failure of a binary permanence property, whereas (W1) concerns a specific powerspace construction and (W2) concerns the existence of compact-cover refinements in all core-compact spaces.

The Hoare connection and related directions

Question 7.11 is now equivalent to the transfer of joint continuity in (11). It has a positive answer on consonant bases by Corollary 7.18, but its general form remains open. This question is closely tied to (W2): a negative answer to (W2) would give a core-compact non-Wilker space XX whose Hoare powerspace is locally compact and Wilker by Corollary 7.9. Conversely, a positive answer to general Hoare reflection would imply a positive answer to (W2). The new criterion does not by itself supply this transfer outside consonance.

The order-topological variants remain useful test problems rather than additional central questions. Question 4.7 asks for (W2) on Scott spaces of arbitrary posets or dcpos; Question 4.12 concerns countable frames and countable complete meet-continuous lattices. The unrestricted Boolean and upper-topology counterexamples do not decide those countable cases. The quotient behavior of Wilker spaces also remains separate from the product and reflection results proved here.

On the function-space side, it remains natural to ask whether sobriety of Ok(X)\mathcal{O}_k(X), equivalently RR-supercompactness of every compact subset of XX, forces Wilker-ness. A negative answer would isolate a weakening of compact-cover splitting. Other directions concern strictness and possible collapses in the supercompactness hierarchy of Remark 5.2, and detection of Wilker-type properties by property SS or HH-sobriety of Ck(X,Y0)C_k(X,Y_0) for a fixed finite Y0Y_0, such as S2S^2 or a small Alexanderoff space. These connect naturally with consonance, co-consonance and Hofmann–Mislove correspondences; (W1) and (W2) remain the two principal open problems emphasized here.

Conclusions and outlook

Wilker’s compact-cover condition admits descriptions in terms of both topological structure and continuity of union. The KC characterization supplies a sufficient separation hypothesis for the mixed product theorem. The co-finite modifications, by contrast, give hereditarily Wilker factors with a non-Wilker product and square. Thus the class is stable under retracts and sums, but not under products, and it does not form a reflective full subcategory of Top0\mathbf{Top}_0.

The compact-open results use HH-supercompactness to describe the role of the Wilker property. A compact refinement on the domain yields HH-sobriety of Ok(X)\mathcal{O}_k(X), property SS for compact-open function spaces, and preservation of HH-sobriety by Ck(X,−)C_k(X,-). The order-topological examples delimit this mechanism: quasicontinuous Scott spaces are Wilker, whereas complete Boolean algebras need not have Wilker Scott spaces and complete lattices need not have Wilker upper topologies.

The Hoare–Scott union criterion supplies another link between the topology of a space and the topology of its open-set lattice. Wilker-ness of XX is continuity of union on Ok(X)\mathcal{O}_k(X), whereas Wilker-ness of PH(X)P_H(X) is continuity of union on ΣO(X)\Sigma\mathcal{O}(X). Consonance identifies these topologies and gives preservation and reflection on that class of bases. Outside consonance, preservation fails and the general reflection implication is not settled here. Smyth powerspaces always reflect Wilker-ness, but their preservation problem remains open.

The resulting picture leaves (W1), Smyth preservation, and (W2), the core-compactness implication, as the two main remaining questions. The counterexamples remove productivity from that list, while the Hoare criterion gives a precise additional route toward (W2) without claiming an answer to it.

Statements and Declarations

Funding. This research was supported by the National Natural Science Foundation of China (Nos. 12471070 and 12071199).

Competing interests. The author has no relevant financial or non-financial interests to disclose.

Author contributions. Xiaoquan Xu is the sole author and is responsible for the research and the preparation of the manuscript.

Data availability. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.

Use of AI-assisted tools. During the preparation of this manuscript, the author used ChatGPT (OpenAI) to assist with language editing, organization of the exposition, bibliographic checks, and the development and checking of mathematical arguments. The author is responsible for verifying all mathematical statements, proofs and references and takes full responsibility for the final text.

References

  1. [1]de Brecht, M., Goubault-Larrecq, J., Jia, X., Lyu, Z.: Domain-complete and LCS-complete spaces. Electron. Notes Theor. Comput. Sci. 345, 3–35 (2019). https://doi.org/10.1016/j.entcs.2019.07.014
  2. [2]de Brecht, M., Kawai, T.: On the commutativity of the powerspace constructions. Log. Methods Comput. Sci. 15(3), 13:1–13:25 (2019). https://doi.org/10.23638/LMCS-15(3:13)2019
  3. [3]Choquet, G.: Theory of capacities. Ann. Inst. Fourier (Grenoble) 5, 131–295 (1953–1954). https://doi.org/10.5802/aif.53
  4. [4]van Douwen, E.K.: An anti-Hausdorff Fréchet space in which convergent sequences have unique limits. Topology Appl. 51, 147–158 (1993). https://doi.org/10.1016/0166-8641(93)90147-6
  5. [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. [6]Goubault-Larrecq, J.: Non-Hausdorff Topology and Domain Theory: Selected Topics in Point-Set Topology. New Mathematical Monographs, vol. 22. Cambridge University Press, Cambridge (2013). https://doi.org/10.1017/CBO9781139524438
  7. [7]Goubault-Larrecq, J.: Choquet–Wilker spaces. Electronic supplement to Non-Hausdorff Topology and Domain Theory, 20 April 2026, Proposition J. https: //topology.lmf.cnrs.fr/choquet-wilker-spaces/. Accessed 14 October 2026
  8. [8]Hofmann, D., Nora, P.: Dualities for modal algebras from the point of view of triples. Algebra Universalis 73, 297–320 (2015). https://doi.org/10.1007/ s00012-015-0324-5
  9. [9]Hofmann, K.H., Lawson, J.D.: The spectral theory of distributive continuous lattices. Trans. Amer. Math. Soc. 246, 285–310 (1978). https://doi.org/10.1090/ S0002-9947-1978-0515540-7
  10. [10]Keimel, K., Lawson, J.D.: Measure extension theorems for T₀-spaces. Topology Appl. 149, 57–83 (2005). https://doi.org/10.1016/j.topol.2004.02.019
  11. [11]Kovár, M.M.: The Hofmann–Mislove theorem for general topological structures. In: Spatial Representation: Discrete vs. Continuous Computational Models. Dagstuhl Seminar Proceedings 04351, pp. 1–9. Schloss Dagstuhl (2005). https://doi.org/10. 4230/DagSemProc.04351.21
  12. [12]Lyu, Z., Chen, Y., Jia, X.: Core-compactness, consonance and the Smyth power-spaces. Topology Appl. 312, 108066 (2022). https://doi.org/10.1016/j.topol. 2022.108066
  13. [13]Mao, X., Xu, L.: Quasicontinuity of posets via Scott topology and sobrification. Order 23(4), 359–369 (2006). https://doi.org/10.1007/s11083-007-9054-4
  14. [14]Papadopoulos, B.K.: A topological lattice on the set of multifunctions. Int. J. Math. Math. Sci. 12, 665–668 (1989). https://doi.org/10.1155/S0161171289000815
  15. [15]Schalk, A.: Algebras for Generalized Power Constructions. Ph.D. thesis, Technische Hochschule Darmstadt (1993)
  16. [16]Wen, X., Xu, X.: Sober is not always co-sober. Topology Appl. 250, 48–52 (2018). https://doi.org/10.1016/j.topol.2018.10.003
  17. [17]Wilker, P.: Adjoint product and hom functors in general topology. Pacific J. Math. 34, 269–283 (1970). https://doi.org/10.2140/pjm.1970.34.269
  18. [18]Xi, X., Lawson, J.: On well-filtered spaces and ordered sets. Topology Appl. 228, 139–144 (2017). https://doi.org/10.1016/j.topol.2017.06.002
  19. [19]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
  20. [20]Xu, X., Bao, M., Zhang, X.: On function spaces equipped with Isbell topology and Scott topology. Math. Structures Comput. Sci. 32, 1099–1116 (2022). https://doi.org/10.1017/S0960129523000014
  21. [21]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
  22. [22]Xu, X., Yang, Y., Chen, L.: Strong well-filteredness of upper topology on sup-complete posets. Math. Structures Comput. Sci. 36, e19 (2026). https://doi.org/10.1017/S0960129526100589
  23. [23]Zhao, B., He, Z., Wang, K.: On co-sober spaces. Houston J. Math. 49(4), 971–988 (2023)DOI
  24. [24]Zhao, D., Ho, W.: On topologies defined by irreducible sets. J. Log. Algebra Methods Program. 84, 185–195 (2015). https://doi.org/10.1016/j.jlamp.2014.10.003

Paper details

vlnc:W_e2b7bakm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents