DMMeNaAGT.dvi @ Applied General Topology c© Universidad Politécnica de Valencia Volume 9, No. 1, 2008 pp. 109-132 Symmetric Bombay topology Giuseppe Di Maio, Enrico Meccariello∗ and Somashekhar Naimpally Abstract. The subject of hyperspace topologies on closed or closed and compact subsets of a topological space X began in the early part of the last century with the discoveries of Hausdorff metric and Vietoris hit-and-miss topology. In course of time, several hyperspace topologies were discovered either for solving some problems in Applied or Pure Mathematics or as natural generalizations of the existing ones. Each hyperspace topology can be split into a lower and an upper part. In the upper part the original set inclusion of Vietoris was generalized to proximal set inclusion. Then the topologization of the Wijsman topo- logy led to the upper Bombay topology which involves two proximities. In all these developments the lower topology, involving intersection of finitely many open sets, was generalized to locally finite families but in- tersection was left unchanged. Recently the authors studied symmetric proximal topology in which proximity was used for the first time in the lower part replacing intersection with its generalization: nearness. In this paper we use two proximities also in the lower part and we obtain the lower Bombay hypertopology. Consequently, a new hypertopology arises in a natural way: the symmetric Bombay topology which is the join of a lower and an upper Bombay topology. 2000 AMS Classification: 54B20, 54E05, 54E15, 54E35. 1. Introduction and preliminaries. Given a topological space, a topological vector space or a Banach space X, frequently it is necessary to study a family of closed or compact (convex) subsets of X, called a hyperset of X, in (a) Optimization (b) Measure Theory (c) Function space topologies ( each function f : X → Y , as a graph, is a subset of X × Y ) (d) Geometric Functional Analysis (e) Image Processing (f ) Convex Analysis etc. So there is a need to put an appropriate topology on the hyperset ∗We regret to announce the sad demise of our friend and collaborator Enrico. 110 G. Di Maio, E. Meccariello and S. Naimpally and so we construct an hyperspace. Two early discoveries were the Hausdorff metric topology (1914) [12] [earlier studied by Pompeiu (1905)] when X is a metric space and the Vietoris hit-and-miss topology (1922) [26] when X is a T1 space. Since then many hyperspace topologies have been studied (see [20]). All hyperspace topologies have a lower and an upper part. A typical member of the lower hyperspace topology consists of members which hit a finite or a locally finite family of open sets. The authors showed that all upper hyperspace topologies known until last year can be expressed with the use of two proximities ([7]) and called the resulting upper hyperspace topology, the upper Bombay topology. In this project the lower part was left unchanged using the hit sets. Recently the authors radically changed the lower part by replacing the hit sets by near sets and thus getting symmetric hyperspaces [8]. In this paper we generalize hyperspace topologies by using two proximities in the lower part obtaining the lower Bombay topology. Combining the lower and the upper Bombay topologies, we have the symmetric Bombay topology. Henceforth, (X, τ ), or X, denotes a T1 space. For any E ⊂ X, clE, intE and Ec stand for the closure, interior and complement of E in X, respectively. A binary relation δ on the power set of X is a basic proximity iff (i) AδB implies BδA; (ii) Aδ(B ∪ C) implies AδB or AδC; (iii) AδB implies A 6= ∅, B 6= ∅; (iv) A ∩ B 6= ∅ implies AδB. A basic proximity δ is a LO-proximity iff it satisfies (LO) AδB and bδC for every b ∈ B together imply AδC. A basic proximity δ is an R-proximity iff it satisfies (R) xδA ( where δ means the negation of δ) implies there exists E ⊂ X such that xδE and EcδA. Moreover, a proximity δ which is both LO and R is called a LR-proximity. A basic proximity δ is an EF-proximity iff it satisfies (EF) AδB implies there exists E ⊂ X such that AδE and EcδB. Note that each EF-proximity is a LR-proximity, but, in general, the converse does not hold. If δ is a LO-proximity, then for each A ⊂ X, we denote Aδ = {x ∈ X : xδA}. Then τ (δ) is the topology on X induced by the Kuratowski closure operator A → Aδ. The proximity δ is compatible with the topology τ iff τ = τ (δ) ( see [10], [21], [25] or [28]). A T1 space X admits a compatible LO-proximity. A space X has a compa- tible LR- ( respectively, EF-) proximity iff it is T3 ( respectively, Tychonoff). Moreover, δ is a compatible LR-proximity iff (LR) for each xδA, there is a closed nbhd. W of x such that W δA. Symmetric Bombay topology 111 If AδB, then we say A is δ-near B; if AδB we say A is δ-far from B. A ≪δ B stands for AδB c and A is said to be strongly δ-contained in B whereas A≪δB stands for its negation, i.e. AδB c. In the sequel η1, η2, η or γ1, γ2, γ denote (compatible) proximities on X. We recall that η2 is coarser than η1 (or equivalently η1 is finer than η2), written η2 ≤ η1, iff Aη 2 B implies Aη 1 B. The most important and well studied proximity is the Wallman or fine LO- proximity η0 given by Aη0B ⇔ clA ∩ clB 6= ∅. The Wallman proximity η0 is the finest compatible LO-proximity on a T1 space X. We note that η0 is a LR-proximity iff X is regular (see [13] Lemma 2). Moreover, η0 is an EF-proximity iff X is normal (Urysohn’s Lemma). If X is a metric space with metric d, the metric proximity η is given by AηB iff D(A, B) = inf{d(a, b), a ∈ A, b ∈ B} = 0. Another useful proximity is the discrete proximity η⋆ given by Aη⋆B ⇔ A ∩ B 6= ∅. We note that η⋆ is the finest proximity on X, but is not a compatible one, unless (X, τ ) is discrete. We point out that in this paper η⋆ is the only proximity that might be non compatible. U , V denote open sets. CL(X) is the family of all nonempty closed subsets of X. For any set E in X we use the notation: E++γ = {F ∈ CL(X) : F ≪γ E or equivalently F γE c}. E++γ0 = E + = {F ∈ CL(X) : F ≪γ0 E or equivalently F ⊂ intE}. E ++ η∗ = {F ∈ CL(X) : F ≪η∗ E or equivalently F ⊂ E}. E−η = {F ∈ CL(X) : F ηE}. E − η⋆ = {F ∈ CL(X) : F η ⋆E or equivalently F ∩ E 6= ∅}. Now, we define the lower Bombay topology. Let η1, η2 be two proximities on a T1 space X with η2 ≤ η1. A typical nbhd. of A ∈ CL(X) in the lower Bombay topology σ(η2, η1) − consists of the sets of the form U −η2 with Aη1U , i.e. {E ∈ CL(X) : Eη2U, where U is open and Aη1U}. We stress that in the description of the lower Bombay topology σ(η2, η1) − the order in which the coordinate proximities are written (η2, η1) emphasizes the fact that the first coordinate proximity η2 is coarser than the second one η1, 112 G. Di Maio, E. Meccariello and S. Naimpally i.e. η2 ≤ η1. Furthermore, the proximity η1 selects the open subsets U which delineate the nbhds of A (U fulfills the property Aη1U ), whereas η2 describes the nbhd. labelled by U , namely U −η2 . If η2 = η1 = η, then we have the lower η-proximal topology (cf. [8]) denoted by σ(η−) = σ(η, η)−. As for the upper part we have two compatible LO-proximities γ1, γ2 with γ1 ≤ γ2, and ∆ ⊂ CL(X) a cobase, i.e. ∆ is closed under finite unions and contains singletons. A typical nbhd. of A ∈ CL(X) in the upper Bombay topology σ(γ1, γ2; ∆) + consists of the sets of the form U ++γ2 , where U c ∈ ∆ and A ≪γ1 U , i.e. {E ∈ CL(X) : E ≪γ2 U, where U c ∈ ∆ and A ≪γ1 U}. Similarly, in the description of the upper Bombay topology σ(γ1, γ2; ∆) +, the order in which the coordinate proximities are written (γ1, γ2) stresses the fact that the first coordinate proximity γ1 is coarser than the second one γ2, i.e. γ1 ≤ γ2. Furthermore, the proximity γ1 and the cobase ∆ ⊂ CL(X) together select the open subsets U which delineate the nbhds of A (U fulfills the property A ≪γ1 U and U c ∈ ∆), whereas γ2 describes the nbhd. indexed by U , namely U ++γ2 . If γ2 = γ1 = γ, then we have the upper γ-∆-proximal topology (cf. [8]) σ(γ+; ∆) = σ(γ, γ; ∆)+. For futher details on proximities and hyperspaces see [7], [8], [25]. 2. Basic results on lower Bombay topology. Since we have already studied the upper Bombay topology in our previous paper, we investigate here the salient properties of the lower Bombay topology. First, we recall that if (X, τ ) is a T1 space, then a topology τ ′ on CL(X) is declared admissible if the map i : (X, τ ) → (CL(X), τ ′), defined by i(x) = {x}, is an embedding. We point out that all the classical hypertopologies, namely, lower Vietoris topology, upper Vietoris topology, Wijsman topology, Fell topology, Hausdorff topology etc. are admissible. On the contrary, if the involved proximities η1, η2 are different from the discrete proximity η⋆, then the map i : (X, τ ) → (CL(X), σ(η2, η1) −) is, in general, not even continuous as the following example shows. Example 2.1. Let X = [−1, 1] with the metric proximity η, η0 the Wal- lman proximity. Let A = {0}, An = { 1 n } for all n ∈ N. Then 1 n converges to 0 in X, but An does not converge to A in the topology σ(η, η0) −, because if U = (−1, 0), then 0η0U , but AnηU for all n. Hence the map i : (X, τ ) → (CL(X), σ(η, η0) −), where i(x) = {x}, is not an embedding. Symmetric Bombay topology 113 Lemma 2.2. Let X be a T1 space, η1, η2, LO-proximities on X with η2 ≤ η1 and η⋆ the discrete proximity. The following results hold: (a) σ(η⋆−) = τ (V −), the lower Vietoris topology. (b) σ(η2, η1) − ⊂ σ(η−1 ) ∩ σ(η − 2 ). Example 2.3. We give examples to show that, in general, σ(η2, η1) − is not comparable with the lower Vietoris topology τ (V −). (a) Let X = R with the usual metric and η the metric proximity, A = [0, 1], An = [0, 1− 1 n ], n ∈ N. Then An → A in τ (V −), but does not converge in the topology σ(η, η0) −, since for U = (1, 2) Aη0U , but AnηU for all n. Hence σ(η, η0) − 6⊂ τ (V −) ( see also Example 2.1). (b) Let l2 denote the space of all square summable sequences. X = B(θ, 1) ∪ {(1 + 1 k )ek : k ∈ N} ⊂ l2, endowed with the Alexandroff proximity ηa (i.e. EηaF iff clE ∩ clF 6= ∅ or both clE, clF are non- compact) and the Wallman proximity η0. Then An = {(1 + 1 k )ek : k ≥ n} converges to A = {θ} in the topology σ(ηa, η0) −, but not in τ (V −). Hence τ (V −) 6⊂ σ(ηa, η0) −. Theorem 2.4. Let (X, τ ) be a T1 regular space and η1, η2 LO-proximities on X. If η2 is also a compatible LR-proximity on X, then τ (V −) ⊂ σ(η2, η1) −. Proof. Suppose that the net (Aλ) of closed sets converges to a closed set A in the topology σ(η2, η1) −. If A ∈ U −, where U ∈ τ , then there is an a ∈ A ∩ U . Since η2 is a compatible LR-proximity, there is an open set V such that a ∈ V and clV η 2 U c by (LR) axiom. Therefore, a ∈ V ⊂ clV ⊂ U and clV η 2 U c. We claim that eventually Aλ intersects U . For if not, then frequently Aλ ⊂ U c and so frequently Aλη 2 V ; a contradiction. � Corollary 2.5. If η is the metric proximity on a metric space (X, d), then (a) τ (V −) ⊂ σ(η, η0) − ⊂ σ(η−). (b) τ (V −) ⊂ σ(η, η0) − ⊂ σ(η−0 ). Remark 2.6. Example 2.3 (b) points out that the assumption η2 is a compa- tible LR-proximity on X cannot be dropped in Theorem 2.4. Example 2.3 (a) shows that the inclusion in Theorem 2.4 might be strict even in nice spaces. We note that the base space X is one of the best possible spaces and the sets involved are also compact. Theorem 2.7. If η is the metric proximity on a metric space (X, d), then the following are equivalent: (a) X is a UC space; (b) σ(η−) ⊂ σ(η, η0) −; (c) σ(η−) = σ(η, η0) −. 114 G. Di Maio, E. Meccariello and S. Naimpally Proof. We need prove only (b) ⇒ (a). Suppose X is not a UC space. Then, there are two disjoint closed sets of distinct points A = {an : n ∈ N}, B = {bn : n ∈ N} such that d(an+1, bn+1) < d(an, bn) → 0. Then, An = {ak : k ≤ n} converges to A in the topology σ(η, η0) −, but not in σ(η−). In fact, for each natural number n choose 0 < εn < ( 1 4 )n d(an, bn) and for n 6= m, Sd(bn, εn) ∩ Sd(bm, εm) = ∅ (where Sd(x, r) is the open sphere centered at x with radius r). Let U = ⋃ n∈N Sd(bn, εn). Then, U is open and AηU (in fact B ⊂ U and d(an, bn) → 0). But, AnηU for each n ∈ N (in fact, it is easy to check that for each n ∈ N, 0 < εn < inf{d(ak, u) : ak ∈ An and u ∈ U}). � Let (X, U) be a T2 uniform space. We recall that for U ∈ U U (x) = {y ∈ X : (x, y) ∈ U} and for any subset E of X, U (E) = ⋃ e∈E U (e). Moreover, E is declared U-discrete if U (e) ∩ E = {e} for each e ∈ E. Definition 2.8 (cf. [12] or [1]). Let (X, U) be a T2 uniform space. (a) A typical nbhd. of A ∈ CL(X) in the lower Hausdorff-Bourbaki or lower H-B topology τ (H−) consists of the sets of the form {B ∈ CL(X) : A ⊂ U (B)}, where U ∈ U. (b) A typical nbhd. of A ∈ CL(X) in the upper Hausdorff-Bourbaki or upper H-B topology τ (H+) consists of the sets of the form {B ∈ CL(X) : B ⊂ U (A)}, where U ∈ U. Note that the lower H-B uniform topology τ (H−) is a topology of locally finite type as Naimpally first proved in [19]. More precisely, Naimpally showed that the topology τ (H−) is generated by hit sets LU = the collection of families of the form {U (x) : x ∈ Q ⊂ A}, A ∈ CL(X), U ∈ U and Q is U -discrete. Note also that the upper H-B uniform topology τ (H+) is a topology of upper proximal type, i.e. τ (H+) = σ(δ+) (see [5]), where δ is the EF-proximity associated to U (see [21] or [10]). Now, we give examples to show that the lower Hausdorff-Bourbaki or H-B uniform topology τ (H−) is not comparable with σ(η2, η1) −. Example 2.9. (a) Again, let X = R endowed with the Euclidean metric and η the metric proximity, A = [0, 1], An = [0, 1 − 1 n ], n ∈ N. Then, An → A in τ (H−), but it does not converge in the topology σ(η, η0) −. In fact, if U = (1, 2), then Aη0U , but AnηU for all n. Hence σ(η, η0) − 6⊂ τ (H−). (b) Let X = N = A with the usual metric, An = {1, 2, . . . , n}. Here η = η0 = η ⋆. Moreover, the sequence (An) converges to A with respect the topology σ(η, η0) − = σ(η⋆−) = τ (V −), but (An) does not converge to A with respect to the lower H-B uniform topology τ (H−). Therefore, τ (H−) 6⊂ σ(η, η0) −. Symmetric Bombay topology 115 Let η1, η2 be LO-proximities on X with η2 ≤ η1, A ∈ CL(X) and E a locally finite family of open sets such that Aη1U for all U ∈ E. Then E − η2 is the set {B ∈ CL(X) : Bη2U for all U ∈ E}. Definition 2.10. Let η1, η2 be LO-proximities on X with η2 ≤ η1. Given a collection L of locally finite families of open sets, LA denotes the subcollection of L such that if E is a family of L verifying Aη1U for all U ∈ E, then E ∈ LA. Suppose L is a collection of locally finite families of open sets satisfying the filter condition: (♯) for each A ∈ CL(X) whenever E, F ∈ LA implies there exists a G ∈ LA such that G−η2 ⊂ E − η2 ∩ F−η2 . Under the above condition, the topology σ(η2, η1; L) − on CL(X) is the topo- logy which has as a basic nbhds system of A ∈ CL(X) all sets of the form {B ∈ CL(X) : Bη2U for each U ∈ E}, where E ∈ LA. It is obvious that if L is a collection of locally finite families of open sets satisfying the above filter condition (♯) and η⋆ is the discrete proximity, then τ (L−) = σ(η⋆−; L) = σ(η⋆, η⋆; L)−. We say that a collection L of locally finite families of open sets is stable under locally finite families if E ∈ L and F is a locally finite family of open sets such that for all V ∈ F there exists U ∈ E with V ⊆ U , then F ∈ L. Theorem 2.11. Let X be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1 and L a collection of locally finite families of open sets which satisfies the filter condition (♯) and is stable under locally finite families. If η2 is a compatible LR-proximity on X, then τ (L−) ⊂ σ(η2, η1; L) −. Proof. Suppose η2 is a LR-proximity and A ∈ E − where E ∈ L. Then for each U ∈ E there exists aU ∈ U ∩ A. So, for each U ∈ E there is an open set VU such that aU ∈ VU ≪η2 U . Let F = {VU : aU ∈ VU ≪η2 U and U ∈ E}. Hence, F ∈ LA and W = {B ∈ CL(X) : Bη2VU for each VU ∈ F} is a σ(η2, η1; L) − nbhd. of A which is contained in E−. � Corollary 2.12. Let (X, d) be a metric space and η the associated metric pro- ximity. Let L be the collection of families of open sets of the form {Sd(x, 1 n ) : x ∈ Q ⊂ A, where A ∈ CL(X), n ∈ N, Q is 1 n -discrete}. Then we have the following: σ(η, η0) − τ (V −) ⊂ ⊂ σ(η, η0; L) − τ (H−) = σ(η⋆; L)− Furthermore, in general, σ(η, η0) − and τ (H−) are not comparable. 116 G. Di Maio, E. Meccariello and S. Naimpally Theorem 2.13. Let (X, d) be a locally compact metric space and η the asso- ciated metric proximity. Then τ (V −) ⊂ σ(η, η0) − = σ(η−0 ) ⊂ σ(η −). Proof. We only prove σ(η, η0) − = σ(η−0 ). It suffices to show σ(η − 0 ) ⊂ σ(η, η0) −, since by Corollary 2.5 (b) σ(η, η0) − ⊂ σ(η−0 ). So, suppose that the net (Aλ) of closed sets converges to a closed set A in the topology σ(η, η0) −. Let V be an open set and let A ∈ V −η0 . Then, there is an a ∈ A ∩ clV . Let U be a compact nbhd. of a and set W = U ∩V . Note that clW is compact and that a closed set is η-near a compact set iff it is η0-near. As a result, eventually Aλ ∈ W − η ⊂ V − η0 and the claim holds. � Remark 2.14. More generally, the equality σ(η, η0) − = σ(η−0 ) in the above Theorem 2.13 holds if one of the following conditions is satisfied: (i) the base space X is compact; (ii) the involved proximities η, η0 are LR, the net of closed sets (Aλ) is even- tually locally finite and converges to A in the topology σ(η, η0) −. Now, we compare two different lower Bombay topologies σ(γ2, γ1) −, σ(η2, η1) −. If η is a LO-proximity on X, then for A ⊂ X, η(A) = {E ⊂ X : EηA}. Theorem 2.15 (Main Theorem). Let (X, τ ) be a T1 space with LO-proxi- mities γ1, γ2, η1, η2 with γ2 ≤ γ1 and η2 ≤ η1. If γ2 and η2 are compatible, then the following are equivalent: (a) σ(γ2, γ1) − ⊂ σ(η2, η1) −; (b) whenever A ∈ CL(X) and U ∈ τ with Aγ1U , there exists a V ∈ τ such that: (i) Aη1V , and (ii) η2(V ) ⊂ γ2(U ). Proof. (a) ⇒ (b). Let A ∈ CL(X) and U ∈ τ with Aγ1U . Then U − γ2 is a σ(γ2, γ1) −-nbhd. of A. By assumption there exists a σ(η2, η1) −-nbhd. L of A such that A ∈ L ⊂ U −γ2 . L has the form: n⋂ k=1 {(Vk) − η2 with Vk ∈ τ and Aη1Vk for each k ∈ {1, . . . , n}}. We claim that there exists k0 ∈ {1, . . . , n} such that η2(Vk0 ) ⊂ γ2(U ). As- sume not. Then, for each k ∈ {1, . . . , n} there is a closed set Fk with Fk ∈ η2(Vk) \ γ2(U ). Set F = ⋃n k=1 Fk. Then F ∈ L, but F 6∈ U − γ2 ; a contra- diction. (b) ⇒ (a). Let A ∈ CL(X) and U −γ2 (U ∈ τ and Aγ1U ) be a subbasic σ(γ2, γ1) −-nbhd. of A. By assumption, there is an open subset V with Aη1V and η2(V ) ⊂ γ2(U ). It follows that V − η2 is a σ(η2, η1) −-nbhd. of A with V −η2 ⊂ U −γ2 . � Corollary 2.16. Let (X, τ ) be a T1 space with a compatible LO-proximity η. If η⋆ is the discrete proximity, then the following are equivalent: (a) τ (V −) = σ(η⋆−) ⊂ σ(η, η⋆)−; (b) whenever A ∈ CL(X) and U ∈ τ with A ∩ U 6= ∅, there exists a V ∈ τ such that A ∩ V 6= ∅ and η(V ) ⊂ η⋆(U ). Symmetric Bombay topology 117 3. First and second countability of lower and upper Bombay topologies. Lemma 3.1. Let (X, τ ) be a T1 space with a compatible LO-proximity η and W , V open subsets of X. The following are equivalent: (a) clW ⊂ clV ; (b) η(W ) ⊂ η(V ). Proof. (a) ⇒ (b). Let F ∈ η(W ). Then F ηW . Since clW ⊂ clV , we have F ηV and hence F ∈ η(V ). (b) ⇒ (a). Assume not. Then, there exists an x ∈ clW \ clV . Set F = {x}. We have F ∈ η(W ), but F 6∈ η(V ); a contradiction. � Definition 3.2 (see [8]). Let (X, τ ) be a T1 space, η a compatible LO-proxi- mity on X and A ∈ CL(X). A family NA of open subsets of X is an external proximal local base of A with respect to η (or, briefly a η-external proximal local base of A) if for any U open subset of X with AηU , there exists V ∈ NA satisfying AηV and clV ⊂ clU . The external proximal character of A with respect to η ( or, briefly the η- external proximal character of A) is defined as the smallest (infinite) cardinal number of the form |NA|, where NA is a η-external proximal local base of A, and it is denoted by Eχ(A, η). The external proximal character of CL(X) with respect to η (or, briefly the η-external proximal character ) is defined as the supremum of all cardinal numbers Eχ(A, η), where A ∈ CL(X) and is denoted by Eχ(CL(X), η). Note that if η = η⋆, then a family NA of open subsets of X is an external local base of A if for any U , an open subset of X with Aη⋆U ( i.e. A ∩ U 6= ∅), there exists a V ∈ NA satisying Aη ⋆V (i.e. A ∩ V 6= ∅) and V ⊂ U (cf. [2]). In a similar way, we can define the external character Eχ(A) of A and the external character Eχ(CL(X)) of CL(X) (cf. [2]). Remark 3.3. Obviously, if the external character Eχ(CL(X)) is coun- table, then the base space X is first countable and each closed subset of X is separable. On the other hand, if we consider the proximal case, X might not be separable even if the η-external character Eχ(CL(X), η) is countable. However, we have the following: If X is a T3 space, η is a compatible LR-proximity and the η-external pro- ximal character Eχ(CL(X), η) is countable, then X is separable. We now consider the first countability of the lower Bombay topology σ(η2, η1) −. If η2 = η1 = η ⋆, then σ(η2, η1) − is the lower Vietoris topology τ (V −). Its first countability has been studied since 1971 ( [2], see also [6] and [15]) and holds if and only if X is first countable and each closed subset of X is separable. Hence, we investigate the case σ(η2, η1) − 6= τ (V −), i.e. η2 6= η ⋆. First, we study the case η2 ≤ η1 6= η ⋆. 118 G. Di Maio, E. Meccariello and S. Naimpally Theorem 3.4. Let (X, τ ) be a T1 space with compatible LO-proximities η1, η2, η2 ≤ η1 and η1 6= η ⋆. The following are equivalent: (a) (CL(X), σ(η2, η1) −) is first countable; (b) the η1-external proximal character Eχ(CL(X), η1) is countable; (c) (CL(X), σ(η−1 )) is first countable. Proof. (a) ⇒ (b). It suffices to show that for each A ∈ CL(X) there exists a countable family NA of open subsets of X which is a η1-external proximal local base of A. So, let A ∈ CL(X) and Z be a countable σ(η2, η1) −-nbhd. system of A. Then, Z = {Ln : n ∈ N}, where each Ln has the form ⋂ k∈In {(Vk) − η2 with Vk ∈ τ, Aη1Vk for every k ∈ In and In finite subset of N}. Set NA = {Vk : k ∈ In, n ∈ N}. NA is a countable family and if Vk ∈ NA, then Aη1Vk (by construction). We claim that for each open set U with Aη1U , there is a Vk ∈ NA with clVk ⊂ clU . So, let U be open with Aη1U and consider U − η2 (which is a σ(η2, η1) −-nbhd. of A). By assumption, there is some Ln ∈ Z with Ln ⊂ U − η2 . Since Ln has the form ⋂ k∈In {(Vk) − η2 with Vk ∈ τ , Aη1Vk for every k ∈ In and In finite subset of N}, we have that ⋂ k∈In (Vk) − η2 ⊂ U −η2 . We claim that there exists some k0 ∈ In such that clVk0 ⊂ clU . Assume not and for each k ∈ In let xk ∈ clVk \ clU . Set E = ⋃ k∈In {xk}. Then E ∈ CL(X) as well as E ∈ Ln, but E 6∈ U − η2 ; a contradiction. (b) ⇒ (a). Let A ∈ CL(X) and NA = {Vn : n ∈ N} be a countable η1- external proximal local base of A. Obviously, Z = {(Vn) − η2 : Vn ∈ NA} is a countable σ(η2, η1) −-subbasic nbhds system of A. (b) ⇔ (c). Theorem 2.11 in [8]. � The case η2 ≤ η1 = η ⋆ can be handled similarly. So, we have Theorem 3.5. Let (X, τ ) be a T1 space with a compatible LO-proximity η, and the discrete proximity η⋆ on X. The following are equivalent: (a) (CL(X), σ(η, η⋆)−) is first countable; (b) the external character Eχ(CL(X)) is countable; (c) (CL(X), τ (V −)) is first countable. Remark 3.6. From the above discussion the following unexpected, but natu- ral result holds. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1. The following are equivalent: (a) (CL(X), σ(η2, η1) −) is first countable; (b) (CL(X), σ(η−1 )) is first countable. Now, we study the first countability of the upper Bombay ∆ topology σ(γ1, γ2; ∆) +. First, we need the following definition. Symmetric Bombay topology 119 Definition 3.7. Let (X, τ ) be a T1 space with a compatible LO-proximity γ, A ∈ CL(X) and ∆ ⊂ CL(X) a cobase. A family LA of open nbhds. of A is a local proximal ∆ base with respect to γ ( or, briefly a γ-local proximal ∆ base of A) if for any open subset U of X with U c ∈ ∆ and A ≪γ U , there exists V ∈ LA with V c ∈ ∆ and A ≪γ V ⊂ U . The γ-proximal ∆ character of A is defined as the smallest ( infinite) cardinal number of the form |LA|, where LA is a γ-local proximal ∆ base of A, and it is denoted by χ(A, γ, ∆). The γ-proximal ∆ character of CL(X) is defined as the supremum of all car- dinal numbers χ(A, γ, ∆), where A ∈ CL(X), and is denoted by χ(CL(X), γ, ∆). Theorem 3.8. Let (X, τ ) be a T1 space with compatible LO-proximities γ1, γ2, γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. The following are equivalent: (a) (CL(X), σ(γ1, γ2; ∆) +) is first countable; (b) the γ1-proximal ∆ character χ(CL(X), γ1, ∆) of CL(X) is countable; (c) (CL(X), σ(γ+1 ; ∆)) is first countable. Proof. (a) ⇒ (b). Let A ∈ CL(X) and UA be a countable σ(γ1, γ2; ∆) +-nbhd. system of A. Then, UA = {On : n ∈ N}, where On = V ++ γ2 with V c ∈ ∆ and A ≪γ1 V . Let LA = {V : A ≪γ1 V and V ++ γ2 = On for some On ∈ UA}. By construction LA is countable. We claim that LA is a γ1-proximal local ∆ base of A. So, let U be an open subset of X with A ≪γ1 U and U c ∈ ∆. Then, U ++γ2 is a σ(γ1, γ2; ∆) +-nbhd. of A. By assumption, there is On ∈ UA with A ∈ On ⊂ U ++ γ2 . Note that On = V ++ γ2 , where V c ∈ ∆ and A ≪γ1 V . We claim V ⊂ U . Assume not. Let x ∈ V \ U and set F = {x}. Then, F ∈ On, but F 6∈ U ++γ2 ; a contradiction. (b) ⇒ (a). Let A ∈ CL(X) and LA be a countable γ1-proximal local ∆ base of A. Set UA = {V ++ γ2 : V ∈ LA}. By construction UA is a countable family of open σ(γ1, γ2; ∆) +-nbhd. of A. We claim that UA is a σ(γ1, γ2; ∆) +-nbhd. system of A. So, let B be a σ(γ1, γ2; ∆) +-nbhd. of A. Then, B has the form {U ++γ2 : U c ∈ ∆ and A ≪γ1 U}. Let V be an open subset with V ∈ LA and A ≪γ1 V ⊂ U and consider V ++ γ2 . Then V ++γ2 ∈ UA and since V ⊂ U we have V ++γ2 ⊂ U ++ γ2 . (b) ⇔ (c). It is straightforward. � Remark 3.9. Let (X, τ ) be a T1 space with compatible LO-proximities γ1, γ2, γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. The following are equivalent: (a) (CL(X), σ(γ1, γ2; ∆) +) is first countable; (b) (CL(X), σ(γ+1 ; ∆)) is first countable. Definition 3.10 (cf. [8]). Let (X, τ ) be a T1 space and η a compatible LO- proximity on X. A family N of open subsets of X is an external proximal base with respect to η (or, briefly a η-external proximal base) if for any A ∈ CL(X) and any U ∈ τ with AηU , there exists V ∈ N satisfying AηV and clV ⊂ clU . 120 G. Di Maio, E. Meccariello and S. Naimpally The external proximal weight of CL(X) with respect to η (or, briefly the η-external proximal weight of CL(X)) is the smallest (infinite) cardinality of its η-external proximal bases and it is denoted by EW (CL(X), η). Note that if η = η⋆, then a family N of open subsets of X is an external base if for any A ∈ CL(X) and any U ∈ τ with Aη⋆U (i.e. A ∩ U 6= ∅), there exists a V ∈ N satisfying Aη⋆V (i.e. A ∩ V 6= ∅) and V ⊂ U ( see [2]). The external character Eχ(A) of A and the external weight EW (CL(X)) of CL(X) can be defined similarly (see [2]). Now, we study the second countability of the lower Bombay topology σ(η2, η1) −. If η2 = η1 = η ⋆, then σ(η2, η1) − is the lower Vietoris topology τ (V −). Its second countability has been studied by [6] and [15] and holds if and only if X is second countable. Hence we investigate the case σ(η2, η1) − 6= τ (V −), i.e. η2 6= η ⋆. First, the case η2 ≤ η1 6= η ⋆. Theorem 3.11. Let (X, τ ) be a T1 space with compatible LO-proximities η1, η2, η2 ≤ η1 and η1 6= η ⋆. The following are equivalent: (a) (CL(X), σ(η2, η1) −) is second countable; (b) the η1-external proximal weight EW (CL(X), η1) of CL(X) is count- able; (c) (CL(X), σ(η−1 )) is second countable. Now, we discuss the case η2 ≤ η1 = η ⋆. Theorem 3.12. Let (X, τ ) be a T1 space with a compatible LO-proximity η and η⋆ the discrete proximity on X. The following are equivalent: (a) (CL(X), σ(η, η⋆)−) is second countable; (b) the external proximal weight EW (CL(X)) of CL(X) is countable; (c) (CL(X), τ (V −)) is second countable. Remark 3.13. Let (X, τ ) be a T1 space and η1, η2 LO-proximities on X with η2 ≤ η1. The following are equivalent: (a) (CL(X), σ(η2, η1) −) is second countable; (b) (CL(X), σ(η−1 )) is second countable. Definition 3.14. Let (X, τ ) be a T1 space with a compatible LO-proximity γ and ∆ ⊂ CL(X) a cobase. A family B of open subsets of X is a γ-proximal base with respect to ∆ if whenever A ≪γ U with U c ∈ ∆, there exists V ∈ B with V c ∈ ∆ and A ≪γ V ⊂ U . The γ-proximal weight of CL(X) with respect to ∆ (or, briefly the γ-proximal weight with respect to ∆) is the smallest (infinite) cardinality of its γ-proximal bases with respect to ∆ and it is denoted by W (CL(X), γ, ∆). Symmetric Bombay topology 121 Theorem 3.15. Let (X, τ ) be a T1 space with compatible LO-proximities γ1, γ2, γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. The following are equivalent: (a) (CL(X), σ(γ1, γ2; ∆) +) is second countable; (b) the γ1-proximal weight W (CL(X), γ1, ∆) of CL(X) with respect to ∆ is countable; (c) (CL(X), σ(γ+1 ; ∆)) is second countable. Remark 3.16. Let (X, τ ) be a T1 space with compatible LO-proximities γ1, γ2, γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. The following are equivalent: (a) (CL(X), σ(γ1, γ2; ∆) +) is second countable; (b) (CL(X), σ(γ+1 ; ∆)) is second countable. 4. Symmetric Bombay topology and some of its properties. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. The lower Bombay topology σ(η2, η1) − combined with the upper one σ(γ1, γ2; ∆) + yields a new hypertopology, namely the ∆-symmetric Bombay topology with respect to η2, η1, γ1, γ2 denoted by π(η2, η1, γ1, γ2; ∆) = σ(η2, η1) − ∨ σ(γ1, γ2; ∆) +. If η2 = η1 = η ⋆, then we have the standard γ1-γ2-∆-Bombay topology σ(γ1, γ2; ∆) = π(η ⋆, η⋆, γ1, γ2; ∆) = τ (V −) ∨ σ(γ1, γ2; ∆) +, investigated in [7]. If η2 = η1 = η and γ1 = γ2 = γ, then we have the η-γ-∆-symmetric proximal topology π(η, γ; ∆) = σ(η, η)− ∨σ(γ, γ; ∆)+ = σ(η−)∨σ(γ+; ∆), studied in [8]. We now consider some basic properties of π(η2, η1, γ1, γ2; ∆). In general, the space X is not embedded in (CL(X), π(η2, η1, γ1, γ2; ∆)) (cf. Example 2.1) and so π(η2, η1, γ1, γ2; ∆) is not an admissible topology. Lemma 4.1. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2, ∆ ⊂ CL(X) a cobase and A ∈ CL(X). If γ1 ≤ η1, then a base for the nbhd. system of A with respect to the π(η2, η1, γ1, γ2; ∆) topology consists of all sets of the form: V ++γ2 ∩ ⋂ j∈J (Sj ) − η2 , with A ≪γ1 V , V c ∈ ∆, Aη1Sj , Sj ∈ τ for each j ∈ J, J finite and ⋃ j∈J Sj ⊂ V . Proof. Let A ∈ V ++γ2 ∩ ⋂ j∈J (Sj ) − η2 , where A ≪γ1 V , V c ∈ ∆, Aη1Sj , Sj ∈ τ for each j ∈ J and J finite. We may replace each Sj with Sj ∩ V . In fact, from γ1 ≤ η1 we have Aη 1 V c and thus Aη1Sj iff Aη1(Sj ∩ V ). � Remark 4.2. The condition γ1 ≤ η1 in the above Lemma 4.1 is indeed a nat- ural one. In fact, in the presentation V ++γ2 ∩ ⋂ j∈J (Sj ) − η2 we may assume that⋃ j∈J Sj ⊂ V as in the classic Vietoris topology. When γ1 ≤ η1, they associated symmetric Bombay topology π(η2, η1, γ1, γ2; ∆) is called standard, otherwise abstract. We will see that the most significant result hold for standard sym- metric Bombay topologies. Often, we will omit the term standard. 122 G. Di Maio, E. Meccariello and S. Naimpally We point out that all the symmetric Bombay topologies investigated in this section are standard. Remark 4.3. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2, ∆ ⊂ CL(X) a cobase, γ1 ≤ η1, i.e. π(η2, η1, γ1, γ2; ∆) is standard. If D is a dense subset of X, then the family of all finite subsets of D is dense in (CL(X), π(η2, η1, γ1, γ2; ∆)). Theorem 4.4. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2, ∆ ⊂ CL(X) a cobase and γ1 ≤ η1. The following are equivalent: (a) (CL(X), π(η2, η1, γ1, γ2; ∆)) is first countable; (b) (CL(X), σ(η2, η1) −) and (CL(X), σ(γ1, γ2; ∆) +) are both first count- able. Proof. (b) ⇒ (a) is clear, since π(η2, η1, γ1, γ2; ∆) = σ(η2, η1) − ∨ σ(γ1, γ2; ∆) +. (a) ⇒ (b). Let A ∈ CL(X). Assume Z = {L = V ++γ2 ∩ ⋂ j∈J (Sj ) − η2 , with A ≪γ1 V , V c ∈ ∆, Sj open, Aη1Sj and J finite } is a countable local base of A with respect to the topology π(η2, η1, γ1, γ2; ∆). We claim that the family Z+ = {V ++γ2 : V ++ γ2 occurs in some L ∈ Z} ⋃ {CL(X)} forms a local base of A with respect to the topology σ(γ1, γ2; ∆) +. Indeed, if there is no open subset U with A ≪γ1 U , U c ∈ ∆, then CL(X) is the only open set in σ(γ1, γ2; ∆) + containing A. If there is U with A ≪γ1 U , U c ∈ ∆, then U ++γ2 is a π(η2, η1, γ1, γ2; ∆)-nbhd. of A. Therefore, there exists L ∈ Z with A ∈ L ⊂ U ++γ2 . Note that L cannot be of the form ⋂ j∈J (Sj ) − η2 , otherwise by setting F = A ∪ U c, we have F ∈ L, but F 6∈ U ++γ2 ; a contradiction. Thus, L has the form V ++γ2 ∩ ⋂ j∈J (Sj ) − η2 . We claim that V ++γ2 ⊂ U ++ γ2 . Assume not and let E ∈ {V ++γ2 , with A ≪γ1 V , V c ∈ ∆} \ U ++γ2 . Set F = E ∪ A, we have F ∈ L \ U ++γ2 ; a contradiction. Now, we show that there is a countable local base of A with respect to the topology σ(η2, η1) −. Without any loss of generality, we may assume that in the expression of every element from Z the family of index set J is non-empty. In fact {V ++γ2 : A ≪γ1 V, V c ∈ ∆} = {V ++γ2 : A ≪γ1 V, V c ∈ ∆} ∩ V −η2 . Moreover, by Lemma 4.1, if L ∈ Z then L = V ++γ2 ∩ ⋂ j∈J (Sj ) − η2 , where A ≪γ1 V , V c ∈ ∆, Aη1Sj , Sj ∈ τ for each j ∈ J, ⋃ j∈J Sj ⊂ V and J finite. Set Z− = {(Sj) − η2 : (Sj ) − η2 occurs in some L ∈ Z}. We claim that the family Z− is a local subbase of A with respect to the topology σ(η2, η1) −. Take S open with Aη1S. Then, S − η2 is a π(η2, η1, γ1, γ2; ∆)-nbhd. of A. Hence, there exists L ∈ Z with A ∈ L = V ++γ2 ∩ ⋂ j∈J (Sj ) − η2 ⊂ S−η2 . We claim that there exists a j0 ∈ J such that (Sj0 ) − η2 ⊂ S−η2 . It suffices to show that there exists a j0 ∈ J such that Sj0 ⊂ S η2 (where Sη2 = {x ∈ X : xη2S}). Assume not and for each j ∈ J let xj ∈ Sj \ S η2 . The set F = ⋃ j∈J {xj} ∈ L \ S − η2 ; a contradiction. � Symmetric Bombay topology 123 By Theorems 3.4, 3.8 and 4.3 we have Corollary 4.5. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2, ∆ ⊂ CL(X) a cobase and γ1 ≤ η1. The following are equivalent: (a) (CL(X), π(η2, η1, γ1, γ2; ∆)) is first countable; (b) the η1-external proximal character Eχ(CL(X), η1) of CL(X) and the γ1-proximal ∆ character χ(CL(X), γ1, ∆) of CL(X) are both countable; (c) the η1-γ1-∆-symmetric proximal topology π(η1, γ1; ∆) on CL(X) is first countable. Proof. Note that (b) ⇔ (c) follows from Theorem 4.9 in [8]. � Theorem 4.6. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2, ∆ ⊂ CL(X) a cobase and γ1 ≤ η1. The following are equivalent: (a) (CL(X), π(η2, η1, γ1, γ2; ∆)) is second countable; (b) (CL(X), σ(η2, η1) −) and (CL(X), σ(γ1, γ2; ∆) +) are both second coun- table. The next Corollary follows from Theorems 3.11, 3.15 and Corollary 4.5. Corollary 4.7. Let (X, τ ) be a T1 space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2, ∆ ⊂ CL(X) a cobase and γ1 ≤ η1. The following are equivalent: (a) (CL(X), π(η2, η1, γ1, γ2; ∆)) is second countable; (b) the η1-external proximal weight EW (CL(X), η1) of CL(X) and the γ1- proximal weight W (CL(X), γ1, ∆) of CL(X) with respect to ∆ are both countable; (c) the η1-γ1-∆-symmetric proximal topology π(η1, γ1; ∆) on CL(X) is se- cond countable. Proof. Note that (b) ⇔ (c) follows from Theorem 4.12 in [8]. � Theorem 4.8. Let (X, τ ) be a Tychonoff space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2, ∆ ⊂ CL(X) a cobase, γ1 ≤ η1. If η1 is a compatible LR-proximity, γ1 a compatible EF- proximity, then the following are equivalent: (a) (CL(X), π(η2, η1, γ1, γ2; ∆)) is metrizable; (b) (CL(X), π(η2, η1, γ1, γ2; ∆)) is second countable and uniformizable; (c) (CL(X), π(η1, γ1; ∆)) is metrizable. Proof. (b) ⇒ (a). It follows from the Urysohn’s Metrization Theorem. (a) ⇒ (b). Since (CL(X), π(η2, η1, γ1, γ2; ∆)) is first countable, X is separa- ble ( use Theorem 4.4 and Remark 3.3). Thus, (CL(X), π(η2, η1, γ1, γ2; ∆)) is second countable ( see Remark 4.3). (b) ⇔ (c). Use Corollary 4.7. � 124 G. Di Maio, E. Meccariello and S. Naimpally Now, we compare two symmetric standard Bombay topologies. Theorem 4.9. Let (X, τ ) be a T1 space; η1, η2, α1, α2 LO-proximities on X with η2 ≤ η1, α2 ≤ α1; γ1, γ2, δ1, δ2 compatible LO-proximities on X with γ1 ≤ γ2, δ1 ≤ δ2 and ∆ and Λ ⊂ CL(X) cobases. If η2, α2 are compatible and γ1 ≤ η1 as well as δ1 ≤ α1, then the following are equivalent: (a) π(η2, η1, γ1, γ2; ∆) ⊂ π(α2, α1, δ1, δ2; Λ); (b) (1) for each F ∈ CL(X) and U ∈ τ with F η1U there are W ∈ τ and L ∈ Λ such that (1i) F ∈ [α1(W ) \ δ1(L)], and (1ii) [α2(W ) \ δ2(L)] ⊂ η2(U ); (2) for each B ∈ ∆ and W ∈ τ , W 6= X with B ≪γ1 W there exists M ∈ Λ such that (2i) M ≪δ1 W , and (2ii) γ2(B) ⊂ δ2(M ). Proof. (a) ⇒ (b). We start by showing (1). So, let F ∈ CL(X) and U ∈ τ with F η1U . Then U − η2 is a π(η2, η1, γ1, γ2; ∆)-nbhd. of F . By assumption there is a π(α2, α1, δ1, δ2; Λ)-nbhd. W of F such that W ⊂ U − η2 . W = W ++ δ2 ∩ ⋂n i=1 (Wi) − α2 , F α1Wi, Wi ∈ τ , i ∈ {1, . . . , n}, ⋃n i=1 Wi ⊂ W , W c ∈ Λ and F ≪δ1 W . Set L = W c. By construction F δ1L as well as F α1Wi, i.e. F ∈ [α1(Wi) \ δ1(L)], for i ∈ {1, . . . , n}. We claim that there exists i0 ∈ {1, . . . , n} such that [α2(Wi0 ) \ δ2(L)] ⊂ η2(U ). Assume not. Then, for each i ∈ {1, . . . , n} there exists Ti ∈ CL(X) with Ti ∈ [α2(Wi) \ δ2(L)] and Ti 6∈ η2(U ), i.e. Tiα2Wi, Ti ≪δ2 W = L c and Tiη 2 U . Set T = ⋃n i=1 Ti. T ∈ CL(X), T ∈ W = W ++ δ2 ∩ ⋂n i=1 (Wi) − α2 and T 6∈ U −η2 which contradicts W ⊂ U −η2 . Now, we show (2). So, let B ∈ ∆ and W ∈ τ , W 6= X with B ≪γ1 W . Set A = W c. Then, A ∈ CL(X) and A ∈ (Bc)++γ2 ∈ π(η2, η1, γ1, γ2; ∆). Thus, there exists a π(α2, α1, δ1, δ2; Λ)-nbhd. O of A such that O ⊂ (B c)++γ2 . O = O++ δ2 ∩ ⋂n i=1(Oi) − α2 , Aα1Oi, Oi ∈ τ for each i ∈ {1, . . . , n}, ⋃n i=1 Oi ⊂ O, Oc ∈ Λ and A ≪δ1 O. Set M = O c. By construction Aδ1M . Therefore, M ≪δ1 W = A c. We claim that γ2(B) ⊂ δ2(M ). Assume not and let E ∈ [γ2(B) \ δ2(M )] with E ∈ CL(X). Set F = A ∪ E. Then F ∈ CL(X), F ∈ O = O++ δ2 ∩ ⋂n i=1 (Oi) − α2 and F 6∈ (Bc)++γ2 ; a contradiction. (b) ⇒ (a). Let F ∈ CL(X), U = U ++γ2 ∩ ⋂n i=1 (Ui) − η2 be a π(η2, η1, γ1, γ2; ∆)- nbhd. of F . Then, F η1Ui, Ui ∈ τ , i ∈ {1, . . . , n}, ⋃n i=1 Ui ⊂ U , F ≪γ1 U and B = U c ∈ ∆. By (1) for each i ∈ {1, . . . , n} there exist Wi ∈ τ and Li ∈ Λ such that F ∈ [α1(Wi) \ δ1(Li)] and [α2(Wi) \ δ2(Li)] ⊂ η2(Ui). By (2), there exists M ∈ Λ such that M ≪δ1 F c ( i.e. M δ1F ) and γ2(B) ⊂ δ2(M ). Set N = ⋃n i=1 Li ∪ M ∈ Λ, O = N c and for each i ∈ {1, . . . , n} Oi = Wi \ N . Note that F δ1N and by construction Oi ⊂ O for each i ∈ {1, . . . , n}. Thus⋃n i=1 Oi ⊂ O. We claim that F α1Oi for each i ∈ {1, . . . , n}. Assume not. Since, F α1N and F α1Wi0 ∩ N c, then F α1N ∪ (Wi0 ∩ N c), and hence F ≪α1 N c ∩ W ci0 ⊆ W c i0 and hence F α1Wi0 . But F α1Wi0 , a contradiction. It follows that O = O++ δ2 ∩ ⋂n i=1 (Oi) − α2 is a π(α2, α1, δ1, δ2; Λ)-nbhd. of F . We claim that O ⊂ U. Assume not. Then there exists E ∈ O, but E 6∈ U. Hence either (♦⋆) Eη2Ui for some i or (♦ ⋆♦⋆) Eγ2U c. Symmetric Bombay topology 125 If (♦⋆) occurs, then since Eα2Oi, Oi ⊂ Wi, E ≪δ2 O = N c and Li ⊂ N we have E ∈ [α2(Wi) \ δ2(Li)] \ η2(Ui), and hence E ∈ [α2(Wi) \ δ2(Li)] 6⊂ η2(Ui) which contradicts (1ii). If (♦⋆♦⋆) occurs, then since Eγ2B = U c, E ≪δ2 O = N c and M ⊂ N we have E ∈ γ2(B) \ δ2(M ), i.e. γ2(B) 6⊂ δ2(M ); which contradicts (2ii). � 5. Uniformizable symmetric abstract Bombay topologies. This section is devoted to find conditions which guarantee the uniformi- zability of a ∆-symmetric abstract Bombay topology π(η2, η1, γ1, γ2; ∆). First we need the following definitions. Definition 5.1 (cf. [8]). Let (X, τ ) be a T1 space, δ a compatible LO-proximity on X and ∆ ⊂ CL(X). (a) ∆ is δ-Urysohn iff whenever D ∈ ∆ and A ∈ CL(X) are δ-far, there exists an E ∈ ∆ with D ≪δ E ≪δ A c (see also [9], [3]). (b) ∆ is Urysohn iff (a) above is true w.r.t. the LO-proximity δ0, i.e. when- ever D ∈ ∆ and A ∈ CL(X) are disjoint, there exists E ∈ ∆ with D ⊂ intE ⊂ E ⊂ Ac. Lemma 5.2 (cf. Theorem 1.6 in [9]). Let (X, τ ) be a Tychonoff space, γ a compatible LO-proximity on X and ∆ ⊂ CL(X) a cobase. If ∆ is γ-Urysohn, then the relation δ defined on the power set of X by (∗) AδB iff clAγclB and either clA ∈ ∆ or clB ∈ ∆ is a compatible EF-proximity on X. Moreover, δ ≤ γ and ∆ is γ-Urysohn iff ∆ is δ-Urysohn. We recall that if (X, τ ) is a Tychonoff space with a compatible EF-proximity δ, then a uniformity U on X is compatible w.r.t. δ if the proximity relation δ(U) defined by Aδ(U)B iff A ∩ U (B) 6= ∅ for each U ∈ U equals δ (see [21] or [10]). We point out that δ admits a unique compatible totally bounded uniformity Uw(δ) ([21], [10]). We will omit reference to δ if this is clear from the context. Let U be a compatible uniformity on X and ∆ a cobase. For each D ∈ ∆ and U ∈ U set [D, U ] = {(A1, A2) ∈ CL(X)×CL(X) : A1∩D ⊂ U (A2) and A2 ∩D ⊂ U (A1)}. The family {[D, U ] : D ∈ ∆ and U ∈ U} is a base for a filter U∆ on CL(X) called the ∆-Attouch-Wets filter . U∆ induces the topology τ (U∆) called the ∆-Attouch-Wets topology (cf. [1] and [9]). We recall that a cobase ∆ is a cover on X iff it is closed hereditary (cf. [9]). The following Theorem is given in [9]. 126 G. Di Maio, E. Meccariello and S. Naimpally Theorem 5.3 (cf. Theorem 2.1 in [9]). Let (X, τ ) be a Tychonoff space with a compatible EF-proximity δ, Uw the unique totally bounded uniformity which induces δ and ∆ ⊆ CL(X) a cover of X. Then the following are equivalent: (a) ∆ is δ-Urysohn; (b) 1) the ∆-Attouch-Wets filter Uw∆ is a Hausdorff uniformity, and 2) the proximal ∆-topology σ(δ, ∆) equals τ (Uw∆). Lemma 5.4 (cf. Theorem 2.2 in [9]). Let (X, τ ) be a Tychonoff space, γ1, γ2, compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cover of X. If ∆ is γ1-Urysohn, δ the compatible EF-proximity on X defined by (∗) AδB iff clAγ 1 clB and either clA ∈ ∆ or clB ∈ ∆ and Uw the unique totally bounded uniformity on X compatible w.r.t. δ, then the Bombay topology σ(γ1, γ2; ∆), the proximal ∆-topology σ(δ; ∆) and the topology τ (Uw∆) induced by the ∆-Attouch-Wets uniformity Uw∆ all coincide. Thus the Bombay topology σ(γ1, γ2; ∆) is Tychonoff. Proof. We omit the proof that is similar to that of Theorem 2.2 in [9]. � By Theorem 2.4 and Lemma 2.2 (b) we know that if η2 is a compatible LR-proximity on X, then τ (V −) ⊂ σ(η2, η1) − ⊂ σ(η−2 ) ∩ σ(η − 1 ). So, in order to get σ(η2, η1) − we have to augment a typical entourage [D, U ] ∈ Uw∆ by adding sets of the type P{Vk} = {(A, B) ∈ CL(X) × CL(X) : Aη1Vk and Bη1Vk} and Q{Vk} = {(A, B) ∈ CL(X) × CL(X) : Aη2Vk and Bη2Vk} for a finite family of open sets {Vk}. Then, we have Theorem 5.5. Let (X, τ ) be a Tychonoff space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cover of X. If η2 is a compatible LR-proximity, ∆ γ1-Urysohn, δ the compatible EF-proximity defined by (∗) AδB iff clAγ 1 clB and either clA ∈ ∆ or clB ∈ ∆ and Uw the unique totally bounded uniformity on X compatible w.r.t. δ, then the family S = Uw∆ ∪ {[D, U ] ∩ P{Vk} : [D, U ] ∈ Uw∆, {Vk} finite family of open sets } ∪{[D, U ] ∩ Q{Vk} : [D, U ] ∈ Uw∆, {Vk} finite family of open sets} defines a compatible uniformity on (CL(X), π(η2, η1, γ1, γ2; ∆)). Proof. It is easy to show that the above family S is a base for a uniformity on CL(X). Nevertherless, it is indeed tricky to prove the compatibility, i.e. that τ (S) equals π(η2, η1, γ1, γ2; ∆) on CL(X). So, let (Aλ) be a net converging to A w.r.t. τ (S). If A ∈ V −η1 , where V ∈ τ , then V − η1 ∈ τ (S) and eventually Aλ ∈ V − η1 . But V −η1 ⊂ V − η2 ( because η2 ≤ η1). Thus, eventually Aλ ∈ V − η2 . If A ∈ V ++γ1 , where V c ∈ ∆, then V c ≪δ A c, where D = V c and δ is the Symmetric Bombay topology 127 compatible EF-proximity defined in (∗). By Lemma 5.2 there is S ∈ ∆ such that D ≪δ S ≪δ A c. Hence, there is W ∈ Uw such that W (A) ∩ S = ∅. Eventually Aλ ∈ [S, W ](A) ⊂ V ++ γ1 . So, eventually Aλ ∈ V ++ γ2 (because γ1 ≤ γ2). Thus, π(η2, η1, γ1, γ2; ∆) ⊂ τ (S). On the other hand, let (Aλ) be a net converging to A w.r.t. π(η2, η1, γ1, γ2; ∆), D ∈ ∆ and U ∈ Uw. Let W ∈ Uw, W symmetric, be such that W ◦ W ⊂ U . By Lemma 5.4 two cases arise: i) A ∈ (Dc)++ δ . Then eventually Aλ ∈ (D c)++ δ and obviously, Aλ ∩ D = ∅ ⊂ W (A) and A ∩ D = ∅ ⊂ W (Aλ). ii) A 6∈ (Dc)++ δ . Then W (A) ∩ D 6= ∅. Since W is totally bounded, there are xk ∈ A, k ∈ {1, . . . , n}, such that A ⊂ ⋃n k=1 W (xk) ⊂ W 2(A). Note that since η2 is a compatible LR-proximity we have τ (V −) ⊂ σ(η2, η1) − ( cf. Theorem 2.4). Now, A ∩ W (xk) 6= ∅ for k ∈ {1, . . . , n}. Hence, for each k there is a Vk with xk ∈ Vk and clVkη 2 [W (xk)] c. But, eventually Aλη2Vk and since clVkη 2 [W (xk)] c we have that eventually Aλ ∩ W (xk) 6= ∅. Therefore, eventually xk ∈ W (Aλ). Hence, eventually A ∩ D ⊂⋃n k=1 W (xk) ⊂ W 2(Aλ) ⊂ U (Aλ). Furthermore, note that [D ∩ (W (A)) c] ∈ ∆ and A ∈ [Dc∪W (A)]++ δ . So, eventually Aλ ∈ [D c∪W (A)]++ δ . Thus, eventually Aλ ∩ D = [Aλ ∩ D ∩ W (A)] ⊂ U (A), i.e. eventually Aλ ∈ [D, U ](A). Therefore, τ (S) ⊂ π(η2, η1, γ1, γ2; ∆). Combining the earlier part we get τ (S) = π(η2, η1, γ1, γ2; ∆). � 6. Appendix (Admissibility). It is a well known fact that if (X, τ ) is a T1 space, then the lower Vietoris topology τ (V −) is an admissible topology, i.e. the map i : (X, τ ) → (CL(X), τ (V −)), defined by i(x) = {x}, is an embedding. On the other hand ( as observed in Example 2.1), if the involved proximities η1, η2 are different from the discrete proximity η ⋆, then the map i : (X, τ ) → (CL(X), σ(η2, η1) −) is, in general, not even continuous. So, we study the behaviour of i : (X, τ ) → (CL(X), σ(η2, η1) −), when η2 ≤ η1 and η1 6= η ⋆. First, we state the following Lemma. Lemma 6.1. Let (X, τ ) be a T1 space, U ∈ τ with clU 6= X and V = (clU ) c. If z ∈ clU ∩ clV , then there exists a net (zλ) τ -converging to z such that for all λ either i) zλ ∈ U and zλ 6= z, or ii) zλ ∈ V and zλ 6= z. Proof. Let N (z) be the filter of open neighbouhoods of z. For each I ∈ N (z), select wI ∈ I ∩ V and yI ∈ I ∩ U . Then, the net (wI ) τ -converges to z and (wI ) ⊂ V as well as the net (yI ) τ -converges to z and (yI ) ⊂ U . We claim that for all I ∈ N (z) either wI 6= z or yI 6= z. Assume not. Then there exist I and J ∈ N (z) such that yI = z and wJ = z. As a result, z ∈ U ∩ V ⊂ clU ∩ V = ∅; a contradiction. � 128 G. Di Maio, E. Meccariello and S. Naimpally Recall that a Hausdorff space X is extremally disconnected if for every open set U ⊂ X, clU is open in X (see [10] Page 368). Proposition 6.2. Let (X, τ ) be a Hausdorff space and η1, η2 two compatible LO-proximities on X with η2 ≤ η1 and η1 6= η ⋆. Then the following are equivalent: (a) X is extremally disconnected; (b) the map i : (X, τ ) → (CL(X), σ(η2, η1) −), defined by i(x) = {x}, is continuous. Proof. (a) ⇒ (b). Let x ∈ X and (xλ) a net τ -converging to x. Let V ⊂ X with V open and {x}η1V . Since {x}η1V and η2 ≤ η1, then {x}η2V and so x ∈ clV . By assumption clV is an open subset of X and the net (xλ) τ -converges to x. Thus, eventually xλ ∈ clV . (b) ⇒ (a). By contradiction, suppose (a) fails. Then, there exists an open set U ⊂ X such that clU is not open in X. Then, clU 6= X. Set V = (clU )c. V is non-empty and open in X. We claim that clU ∩ clV 6= ∅. Assume not, i.e. clU ∩ clV = ∅. Then, clU ⊂ (clV )c ⊂ V c = clU . Thus, clU = (clV )c, i.e. clU is open; a contradiction. Let z ∈ clU ∩ clV . By Lemma 6.1, there exists a net (zλ) τ -converging to z such that for all λ either (1) zλ ∈ U and zλ 6= z or (2) zλ ∈ V and zλ 6= z. In both cases, there exists an open subset W such that z ∈ clW and zλ 6∈ clW for all λ. In fact if (1) holds, then set W = V , otherwise set W = U . Thus, the net (zλ) τ -converges to z, but there exists an open subset W such that {z}η1W ( because z ∈ clW and η1 is a compatible LO-proximity on X) as well as {zλ}η 2 W ( again because zλ 6∈ clW and η2 is a compatible LO-proximity on X) for all λ. Hence, the map i : (X, τ ) → (CL(X), σ(η2, η1) −) fails to be continuous. � Remark 6.3. If η1 = η ⋆, then the map i : (X, τ ) → (CL(X), σ(η2, η1) −) is always continuous. Definition 6.4 (cf. Definition 6.3 in [8]). A T1 space (X, τ ) is nearly regular iff whenever x ∈ U with U ∈ τ there exists V ∈ τ with x ∈ clV ⊂ U . Proposition 6.5. Let (X, τ ) be T1 space and η1, η2 LO-proximities on X with η2 ≤ η1. If η2 is a compatible LO-proximity, then the following are equivalent: (a) (X, τ ) is nearly regular; (b) the map i : (X, τ ) → (CL(X), σ(η2, η1) −) is open. Proof. Left to the reader. � Note that if (X, τ ) is a T1 space, γ1, γ2 are compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase, then the map i : (X, τ ) → (CL(X), σ(γ1, γ2; ∆) +) is always continuous. So, we have: Symmetric Bombay topology 129 Proposition 6.6. Let (X, τ ) be a T1 space, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. The following are equivalent: (a) the map i : (X, τ ) → (CL(X), σ(γ1, γ2; ∆) +), defined by i(x) = {x}, is an embedding; (b) the map i : (X, τ ) → (CL(X), σ(γ1, γ2; ∆) +), defined by i(x) = {x}, is an open map; (c) whenever U ∈ τ and x ∈ U , there exists a B ∈ ∆ such that x ∈ Bc ⊂ U . Finally, we have the following results dealing with admissibility of the sym- metric standard Bombay ∆ topology π(η2, η1, γ1, γ2; ∆). Obviously, we in- vestigate just the significant case η2 6= η ⋆ ( the standard Bombay ∆ topology σ(γ1, γ2; ∆) is always admissible). We have to distinguish the two subcases (1) η1 6= η ⋆, (2) η1 = η ⋆. Proposition 6.7. Let (X, τ ) be a regular Hausdorff space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. Suppose that η1 6= η ⋆ and γ1 ≤ η1. Then the map i : (X, τ ) → (CL(X), π(η2, η1, γ1, γ2; ∆)) is an embedding if and only if the following three conditions are fulfilled: (a) i : (X, τ ) → (CL(X), σ(η2, η1) −) is continuous; (b) i : (X, τ ) → (CL(X), σ(η2, η1) −) is open; (c) i : (X, τ ) → (CL(X), σ(γ1, γ2; ∆) +) is open. Proof. Only necessity requires a proof. Namely just (b) and (c). Now, we show that i : (X, τ ) → (CL(X), σ(η2, η1) −) is open. Assume not. Then there exist x ∈ X and U ∈ N (x), where N (x) is the filter of open nhoods at x such that i(U ) 6∈ σ(η2, η1) − ∩ i(X). So, there is y ∈ U such that for each W −η2 with W ⊂ X open and yη1W , we have W −η2 ∩ i(X) 6⊂ i(U ). But i : (X, τ ) → (CL(X), π(η2, η1, γ1, γ2; ∆)) is an embedding. Thus, for each U ∈ N (x), i(U ) ∈ π(η2, η1, γ1, γ2; ∆) ∩ i(X). Hence there exists O ∈ π(η2, η1, γ1, γ2; ∆) such that {y} ∈ O ∩ i(X) ⊂ i(U ). Note that O has the form O++γ2 ∩ ⋂n j=1 (Oj ) − η2 with yη1Oj , Oj ∈ τ , O c ∈ ∆ and y ≪γ1 O. Furthermore, since π(η2, η1, γ1, γ2; ∆) is standard we may assume that ∪nj=1Oj ⊂ O. Now, since i : (X, τ ) → (CL(X), σ(η2, η1) −) is continuous, there exists a Vj ∈ N (y) such that i(Vj ) ⊂ (Oj ) − η2 for each j ∈ {1, . . . , n}. Because X is regular, there exists Lj, j ∈ {1, . . . , n}, Lj ∈ N (y) such that Lj ⊂ clLj ⊂ Vj . Since X is regular select W ∈ N (y) with W = ∩nj=1Lj and clW ⊂ O. It follows i(W ) ⊂ i(clW ) = W −η2 ∩ i(X) ⊂ i(U ). Thus i(U ) ∈ σ(η2, η1) − ∩ i(X), a contradiction. Now we prove that i : (X, τ ) → (CL(X), σ(γ1, γ2; ∆) +) is open. Assume not. So, there exists x′ ∈ X and V ∈ N (x′), such that i(V ) 6∈ σ(γ1, γ2; ∆) + ∩ i(X). So there is y′ ∈ V such that for each W ++γ2 with W c ∈ ∆ and y′ ≪γ1 W , we have W ++γ2 ∩ i(X) 6⊂ i(V ). But i : (X, τ ) → (CL(X), π(η2, η1, γ1, γ2; ∆)) is an embedding. Thus, for each V ∈ N (x′), i(V ) ∈ π(η2, η1, γ1, γ2; ∆) ∩ i(X). 130 G. Di Maio, E. Meccariello and S. Naimpally Hence there exists T ∈ π(η2, η1, γ1, γ2; ∆) such that {y ′} ∈ T ∩ i(X) ⊂ i(V ). Note that T has the form T ++γ2 ∩ ⋂n j=1 (Tj ) − η2 with y′η1Tj, Tj ∈ τ , T c ∈ ∆ and y′ ≪γ1 T . Moreover, since π(η2, η1, γ1, γ2; ∆) is standard, we may assume that Tj ⊂ T for each j ∈ {1, . . . , n}. Again, since i : (X, τ ) → (CL(X), σ(η2, η1) −) is continuous, there exists a Sj ∈ N (y ′) such that i(Sj ) ⊂ (Tj) − η2 for each j ∈ {1, . . . , n}. Select Mj ∈ N (y ′), j ∈ {1, . . . , n} such that Mj ⊂ clMj ⊂ Sj . Set W = T ∩ ⋂n j=1 Mj . It follows that W ∈ N (y ′). Again, since i : (X, τ ) → (CL(X), π(η2, η1, γ1, γ2; ∆)) is an embedding, there exists O ∈ π(η2, η1, γ1, γ2; ∆) such that {y′} ∈ O∩i(X) ⊂ i(W ). Note that O has the form O++γ2 ∩ ⋂m k=1 (Ok) − η2 with y′η1Ok, Ok ∈ τ , O c ∈ ∆ and y′ ≪γ1 O. We may assume ∪ n k=1Ok ⊂ O. Moreover, from O∩i(X) ⊂ i(W ) we have O ⊂ W (otherwise, select z ∈ O\W and zk ∈ Ok; the set F = {z} ∪ ⋃n k=1 zk ∈ O \ T, a contradiction). As a result, i(O) = i(X) ∩ O++γ2 ⊂ T ⊂ i(V ). Thus i(V ) ∈ σ(γ1, γ2; ∆) + ∩ i(X); a contradiction. � Theorem 6.8. Let (X, τ ) be a regular Hausdorff space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. If η1 6= η ⋆ and γ1 ≤ η1, then the map i : (X, τ ) → (CL(X), π(η2, η1, γ1, γ2; ∆)) is an embedding if and only if the following condi- tions are fulfilled: (a) X is extremally disconnected; (b) whenever U ∈ τ and x ∈ U , there exists a B ∈ ∆ such that x ∈ Bc ⊂ U . Theorem 6.9. Let (X, τ ) be a regular Hausdorff space, η1, η2 LO-proximities on X with η2 ≤ η1, γ1, γ2 compatible LO-proximities on X with γ1 ≤ γ2 and ∆ ⊂ CL(X) a cobase. If η1 = η ⋆ and η2 is a compatible LO-proximity, then the following are equivalent: (a) the map i : (X, τ ) → (CL(X), π(η2, η1, γ1, γ2; ∆)) is an embedding; (b) whenever U ∈ τ and x ∈ U , there exists a B ∈ ∆ such that x ∈ Bc ⊂ U . Acknowledgements. The authors thank the referee for a careful reading of the manuscript and a number of valuable comments and suggestions, which led to an improvement of the paper. References [1] G. Beer, Topologies on Closed and Closed Convex Sets, Kluwer Academic Publishers, 1993. [2] M. Coban, Note sur la topologie exponentielle, Fund. Math. LXXI (1971), 27–41. [3] D. Di Caprio and E. Meccariello, Notes on Separation Axioms in Hyperspaces, Q. & A. in General Topology 18 (2000), 65–86. [4] D. Di Caprio and E. Meccariello, G-uniformities, LR-proximities and hypertopologies, Acta Math. Hungarica 88 (1-2) (2000), 73–93. Symmetric Bombay topology 131 [5] A. Di Concilio, S. Naimpally and P.L. Sharma, Proximal hypertopologies, Proceedings of the VI Brasilian Topological Meeting, Campinas, Brazil (1988) [unpublished]. [6] G. Di Maio and Ľ. Holá, On hit-and miss topologies, Rend. Acc. Sc. Fis. Mat. Napoli 57 (1995), 103–124. [7] G. Di Maio, E. Meccariello and S. Naimpally, Bombay hypertopologies, Applied General Topology 4 (2) (2003), 421–424. [8] G. Di Maio, E. Meccariello and S. Naimpally, Symmetric proximal hypertopology, Ros- tock. Math. Kolloq. 58 (2003), 2–25. [9] G. Di Maio, E. Meccariello and S. Naimpally, Uniformizing (proximal) ∆-topologies, Topology and its Applications, 137(2004), 99–113. [10] R. Engelking, General topology, Revised and completed version, Helderman Verlag, Hel- derman, Berlin, 1989. [11] J. Fell, A Hausdorff topology for the closed subsets of a locally compact non-Hausdorff space, Proc. Amer. Math. Soc. 13 (1962), 472–476. [12] F. Hausdorff, Grundzüge der Mengenlehre, Leipzing, 1914. [13] D. Harris, Regular-closed spaces and proximities, Pacif. J. Math. 34 (1970), 675–685. [14] D. Harris, Completely regular proximities and RC-proximities, Fund. Math. LXXXV (1974), 103–111. [15] Ľ. Holá and S. Levi, Decomposition properties of hyperspace topologies, Set-Valued Anal- ysis 5 (1997), 309–321. [16] M. Marjanovic, Topologies on collections of closed subsets, Publ. Inst. Math. (Beograd) 20 (1966), 196–130. [17] E. Michael, Topologies on spaces of subsets, Trans. Amer. Math. Soc. 71 (1951), 152– 182. [18] C. J. Mozzochi, M. Gagrat and S. Naimpally, Symmetric generalized topological struc- tures, Exposition Press, (Hicksville, New York, 1976.) [19] S. Naimpally, All Hypertopologies are hit-and-miss, Applied General Topology 3 (2002), 45–53. [20] S. Naimpally, A short hystory of hyperspace topologies, in preparation. [21] S. Naimpally and B. Warrack, Proximity spaces, Cambridge Tracts in Mathematics 59, Cambty Press, 1970. [22] H. Poppe, Eine Bemerkung über Trennungsaumen abgeschlossenen Teilmengen topolo- gischer Räume, Arch. Math. 16 (1965), 197–199. [23] H. Poppe, Einigee Bemerkungen über den Rs̈chlossenen Mengen, Fund. Math. 59 (1966), 159–169. [24] R. F. Snipes, Functions that preserve Cauchy sequences, Nieuw Arch. Voor Wiskunde (3) 25 (1977), 409–422. [25] W. J. Thron, Topological Structures, Holt, Rinehart and Winston, New York, 1966. [26] L. Vietoris, Bereiche zweiter Ordnung, Monatsh. fur Math. und Phys. 32 (1922), 258– 280. [27] R. Wijsman, Convergence of sequences of convex sets, cones, and functions, II, Trans. Amer. Math. Soc. 123 (1966), 32–45. [28] S. Willard, General Topology, Addison-Wesley, 1970. Received November 2006 Accepted December 2007 132 G. Di Maio, E. Meccariello and S. Naimpally Giuseppe Di Maio (giuseppe.dimaio@unina2.it) Seconda Università degli Studi di Napoli, Facoltà di Scienze, Dipartimento di Matematica, Via Vivaldi 43, 81100 Caserta, Italia Somashekhar Naimpally (somnaimpally@yahoo.ca) 96 Dewson Street, Toronto, Ontario, M6H 1H3, Canada