@ Applied General Topology c© Universidad Politécnica de Valencia Volume 4, No. 1, 2003 pp. 71–77 Dense Sδ-diagonals and linearly ordered extensions Masami Hosobuchi Abstract. The notion of the Sδ-diagonal was introduced by H. R. Bennett to study the quasi-developability of linearly ordered spaces. In an earlier paper, we obtained a characterization of topological spaces with an Sδ-diagonal and we showed that the Sδ-diagonal property is stronger than the quasi-Gδ-diagonal property. In this paper, we define a dense Sδ-diagonal of a space and show that two linearly ordered extensions of a generalized ordered space X have dense Sδ-diagonals if the sets of right and left looking points are countable. Keywords: Sδ-diagonal, dense Sδ-diagonal, linearly ordered space (LOTS), generalized ordered space (GO-space), linearly ordered extension. 2000 AMS Classification: 54F05. 1. Sδ-diagonals We review in this section the definitions of Sδ-set and Sδ-diagonal, and state our results obtained in [5]. The following definition is a generalization of a Gδ-set and was introduced by H. R. Bennett [2] to study the quasi-developability of linearly ordered (topo- logical) spaces. Definition 1.1. Let X be a topological space. A subset A of X is called an Sδ-set if there exists a countable collection {U(1),U(2), . . .} of open subsets of X such that, for two points p ∈ A and q ∈ X \A, there exists an n such that p ∈ U(n) and q /∈ U(n). It is easy to see that a Gδ-set is an Sδ-set. Hence the notion of Sδ-set is a generalization of Gδ-set. See [3] for a description of S-normal spaces whose closed subsets are Sδ-sets. Definition 1.2. Let X be a topological space. X has an Sδ-diagonal if the diagonal subset ∆X of X × X is an Sδ-set of X × X, where ∆X denotes the 72 M. Hosobuchi diagonal set {(x,x) : x ∈ X} in the Cartesian product X×X. The symbol ( , ) is used to stand for a point of X ×X. It is useful to show the following lemma that relates to the property (∗) given in [2]. N denotes the set of natural numbers. Lemma 1.3. [5] Let X be a topological space. Let {G(n) : n ∈ N} be a family of countable collections of open subsets of X. Suppose that, for any three points x,y and z with y 6= z, there exists an m ∈ N such that x ∈ ⋃ G(m) and that no element of G(m) contains the set {y,z}, where ⋃ G(m) denotes⋃ {U : U ∈ G(m)}. Then, there exists a family {F(n) : n ∈ N} of countable collections of open subsets of X such that, for such three points above, there exists an m ∈ N such that x ∈ ⋃ F(m) and any two distinct points of {x,y,z} do not belong to the same member of F(m). Theorem 1.4. [5] Let X be a topological space. X has an Sδ-diagonal if and only if there exists a family {G(n) : n ∈ N} of countable collections of open subsets of X such that, for three points x,y and z with y 6= z, there exists an m ∈ N such that x ∈ ⋃ G(m) and any two distinct points of {x,y,z} do not belong to the same member of G(m). 2. Two linearly ordered extensions and notation Recall that a generalized ordered space (GO-space) is a triple (X,τ,<), where < is a linear ordering of the set X and τ a Hausdorff topology on X having a base of order-convex sets. We will denote by λ the order topology on (X,<). It is known that λ ⊂ τ. A space of the form (X,λ,<) is called a linearly ordered topological space (LOTS). Every LOTS is a GO-space, but not conversely. In fact it is known that the class of GO-spaces coincides with the class of subspaces of LOTS. Given a GO-space X there are two well-known linearly ordered extensions of X. One of these is X∗ and was defined by D. J. Lutzer [7]. The other one is L(X) and was studied in [8]. We review here the definitions of those linearly ordered extensions. The intervals in a GO-space or a LOTS are written in the form [a,b], [a,b[, ]a,b] and ]a,b[. For example, [a,b] = {x : a ≤ x ≤ b}, [a,b[ = {x : a ≤ x < b} and so on. For a GO-space X, we set R = {x ∈ X : [x,→ [ ∈ τ −λ} and L = {x ∈ X : ]← ,x] ∈ τ −λ}, where λ denotes the order topology as mentioned above. R (resp. L) is called the set of right (resp. left ) looking points. Then X∗ is defined as follows: X∗ = (X ×{0}) ∪{(x,k) : x ∈ R,k < 0,k ∈ Z}∪{(x,k) : x ∈ L,k > 0,k ∈ Z} ⊂ X ×Z, where Z denotes the set of integers. On the other hand, L(X) is defined as follows: L(X) = (X ×{0}) ∪{(x,−1) : x ∈ R}∪{(x, 1) : x ∈ L}⊂ X ×{−1, 0, 1}. X∗ and L(X) are linearly ordered topological spaces equipped with the lex- icographic order topologies. We, furthermore, need some technical notation Dense Sδ-diagonals 73 for the proof of the theorems in Section 5. For a convex open subset U of a GO-space X, we define a convex open subset Ũ of E(X), where E(X) denotes either X∗ or L(X). Then eight cases can occur. In the following, the intervals must be considered in E(X). (1) If a is the minimum point of U, then we define Ũ1 = [(a, 0),→ [ ⊂ E(X). (2) Let a = inf U and a ∈ X \ U. If E(X) = X∗, then Ũ1 = {(x,k) ∈ X∗ : a < x} = ](a, +∞),→ [ ⊂ X∗, where (a, +∞) ∈ X × (Z∪{+∞}) and the interval is taken in X∗. Likewise, if E(X) = L(X), then Ũ1 = {(x,k) ∈ L(X) : a < x} = ](a, 1),→ [ ⊂ L(X). Note that (a, 1) may not belong to L(X). (3) If there is a gap u = (A,B) such that u is the left end-point of U, then we define Ũ1 =](u, 0),→ [ ⊂ E(X). (4) If none of Cases 1 − 3 occurs, then we define Ũ1 = E(X). (5) If b is the maximum point of U, then we define Ũ2 = ] ← , (b, 0)] ⊂ E(X). (6) Let b = sup U and b ∈ X \ U. If E(X) = X∗, then Ũ2 = {(x,k) ∈ X∗ : x < b} = ] ← , (b,−∞)[ ⊂ X∗ (cf. (2)). If E(X) = L(X), then Ũ2 = {(x,k) ∈ L(X) : x < b} = ] ← , (b,−1)[ ⊂ L(X). Note that (b,−1) may not belong to L(X). (7) If there is a gap v = (A,B) such that v is the right end-point of U, then we define Ũ2 = ]← , (v, 0)[ ⊂ E(X). (8) If none of Cases 5 − 7 occurs, then we define Ũ2 = E(X). We set Ũ = Ũ1 ∩ Ũ2. Ũ is called the convex open set associated with U. Let U be an open set of a GO-space X. Then U is decomposed into a union of open convex subsets {Uα : α ∈ A}. In this case, we define Ũ = ⋃ {Ũα : α ∈ A}, where Ũα is the open set associated with Uα. Then Ũ is an open subset of E(X), and called the open set associated with U. 3. Sδ-diagonals in linearly ordered extensions The following theorems are proved in our paper [5]. Let X be a GO-space. It is easily seen that X∗ contains X as a closed subset and L(X) contains X as a dense subset. See [7, 8] for further information about X∗ and L(X). In both cases, X and X ×{0} are identified by the correspondence of x to (x, 0). Theorem 3.1. [5] Let X be a generalized ordered space with an Sδ-diagonal. If R∪L is countable, then X∗ has an Sδ-diagonal. To prove a similar theorem concerning L(X), it is necessary to assume the existence of sequences in X that witnesses first-countability for points of R∪L: Theorem 3.2. [5] Let X be a GO-space with an Sδ-diagonal. Assume that, for every point s ∈ L, there exists a decreasing sequence {x(s,n) : n ∈ N} in X such that inf{x(s,n)} = s and, for every point s ∈ R, there exists an increasing sequence {y(s,n) : n ∈ N} in X such that sup{y(s,n)} = s. If R∪L is countable, then L(X) has an Sδ-diagonal. 74 M. Hosobuchi 4. Dense Sδ-diagonals The following definition gives an analogy to the dense Gδ-diagonal in [1]. Definition 4.1. A Hausdorff space X has a dense Sδ-diagonal if there exists a dense subset D of ∆X such that D is an Sδ-subset of X × X, where ∆X denotes the diagonal subset of the Cartesian product X ×X. We show the following theorem that is analogous to a result concerning spaces that have a dense Gδ-diagonal [1]. Theorem 4.2. Let X be a Hausdorff space. Then X has a dense Sδ-diagonal if and only if X has a dense subset Y such that Y is an Sδ-subset of X and Y has an Sδ-diagonal. Proof. If D ⊂ ∆X is a dense Sδ-set in X×X, then D∩∆X is a dense Sδ-set in ∆X. Now the map h : ∆X → X defined by h(x,x) = x is a homeomorphism, and the homeomorphic image of a dense Sδ-set is a dense Sδ-set. Conversely, suppose Y is a dense Sδ-subset of X. Then h−1(Y ) is a dense Sδ-subset of ∆X. The rest is easily verified. � 5. Theorems concerning dense Sδ-diagonals of linearly ordered extensions Theorem 5.1. Let X = (X,τ) be a GO-space with a dense Sδ-diagonal. If R∪L is countable, then X∗ has a dense Sδ-diagonal. We first show the following lemma. Lemma 5.2. Let X be a GO-space and X∗ the linearly ordered extension of X. For a subspace Y of X, set Z = Y ∪ (X∗ \X). If Y is dense in X and an Sδ-subset of X, then Z is a dense subspace of X∗ and an Sδ-subset of X∗. Proof. To see that Z is dense in X∗, let x ∈ X∗ \ Z = X \ Y and V be a neighborhood of x in X∗, where X is identified with X ×{0} as usual. Since V ∩ X is a neighborhood of x in X, it follows that V ∩ X ∩ Y 6= ∅. Since V ∩X ∩Y ⊂ V ∩Z, it follows that V ∩Z 6= ∅. Hence Z is a dense subspace of X∗. To show the last part, let {U(n) : n ∈ N} be a countable collection of open subsets of X such that, for y ∈ Y and x ∈ X \ Y, there exists an m ∈ N such that y ∈ U(m) and x 6∈ U(m). For every n ∈ N, let Ũ(n) be the open subset associated with U(n) as in Section 3. Set Ũ(0) = X∗ \ X. Then it is obvious that Ũ(0) is an open subset of X∗. We show that the countable collection {Ũ(n) : n ≥ 0} of open subsets of X∗ assures that Z is an Sδ-subset of X∗. Let z ∈ Z and x ∈ X∗ \Z = X \Y. Case 1. Let z ∈ X∗\X. Then it is easy to see that z ∈ Ũ(0) and x 6∈ Ũ(0). Case 2. Let z ∈ Y. Since x ∈ X \ Y, there exists an m ∈ N such that z ∈ U(m) and x 6∈ U(m). By the definition of Ũ(m), it follows that z ∈ Ũ(m) and x 6∈ Ũ(m). This completes the proof. � Dense Sδ-diagonals 75 Now we shall prove Theorem 5.1. Proof of Theorem 5.1. By Theorem 4.2, there exists a dense subspace Y of X such that Y is an Sδ-subset of X and that Y has an Sδ-diagonal. Let {G(n) : n ∈ N} be a family of countable collections of open subsets of Y such that, for three points x,y and z of Y with y 6= z, there exists an m ∈ N such that x ∈ ⋃ G(m) and no element of G(m) contains {y,z}. The existence of the above family is guaranteed by Theorem 1.4. For an open subset V of Y, there exists an open set VX of X such that VX ∩Y = V. Let ṼX be the open subset of X∗ associated with VX as explained in Section 2. Set VZ = ṼX ∩Z, where Z is as mentioned in Lemma 5.2. It is clear that VZ is open in Z and that VZ ∩Y = V . For every n ∈ N, set G̃(n) = {VZ : V ∈ G(n)} and G̃(0) = {{x} : x ∈ X∗ \ X}. Let S = R ∪ L = {si : i ∈ N} be an enumeration of the countable set S. Let (si,k) ∈ X∗\X. Set G̃+(si,k) = {](si,k),→ [ ∩Z} and G̃−(si,k) = {]← , (si,k)[∩Z}, where these intervals are considered in X∗. By virtue of Lemma 5.2 and Theorem 4.2, it is sufficient to show that a family of those countable collections of open subsets of Z witnesses the Sδ-diagonal of Z. To see this, let x,y and z be three points of Z with y 6= z. We may assume without loss of generality that y < z. Case 1. If {x,y,z} ⊂ Y, then there exists an m ∈ N such that x ∈ ⋃ G(m) and {y,z} 6⊂ V for any V ∈G(m). Hence it follows that x ∈ ⋃ G̃(m) and that {y,z} 6⊂ VZ for any VZ ∈ G̃(m). Case 2. Let x ∈ Y and y or z belong to Z \Y. (i) We assume that y ∈ Z \ Y . Then we can write y = (si,k), where k 6= 0. If x < y, then x ∈ ]← ,y[∩Z and {y,z} 6⊂ ]← ,y[. Hence, by the definition, it follows that x ∈ ⋃ G̃−(si,k) and that {y,z} 6⊂ V for V ∈ G̃−(si,k). If y < x, then x ∈ ]y,→ [∩Z and {y,z} 6⊂ ]y,→ [. Hence it follows that x ∈ ⋃ G̃+(si,k) and that {y,z} 6⊂ V for V ∈ G̃+(si,k). (ii) Let z ∈ Z \Y. Then the proof is analogous to (i). Case 3. Let x ∈ Z \ Y. Then it follows that x ∈ ⋃ G̃(0) and {y,z} 6⊂ V for any V ∈ G̃(0). Therefore, by virtue of Lemma 1.3 and Theorem 1.4, X∗ has a dense Sδ-diagonal. This completes the proof. � Theorem 5.3. Let X = (X,τ) be a GO-space with a dense Sδ-diagonal. If R∪L is countable, then L(X) has a dense Sδ-diagonal. Proof. By Theorem 4.2, there exists a dense subspace Y of X such that Y is an Sδ-subset of X and Y has an Sδ-diagonal. Since X is dense in L(X), it follows that Y is a dense subspace of L(X). To prove that L(X) has a dense Sδ-diagonal, it is sufficient to show, by Theorem 4.2, that Y is an Sδ-subset of L(X). Let {U(n) : n ∈ N} be a countable collection of open subsets of X such that, for y ∈ Y and x ∈ X \Y, there exists an m ∈ N such that y ∈ U(m) and x 6∈ U(m). For every n ∈ N, let Ũ(n) be the open subset of L(X) associated with U(n). For si ∈ S = R ∪ L and ε ∈ {−1, 1}, set Ũ+(si,ε) = ](si,ε),→ [ and Ũ−(si,ε) = ]← , (si,ε)[, where the intervals are considered in L(X). The 76 M. Hosobuchi countable collection {Ũ(n), Ũ+(si,ε), Ũ−(si,ε) : n ∈ N, i ∈ N, ε = ±1} of open subsets of L(X) guarantees that Y is an Sδ-subset of L(X). To see this, let y ∈ Y and z ∈ L(X) \Y. Case 1. Let z ∈ X \Y. Then there exists an m ∈ N such that y ∈ U(m) and z 6∈ U(m). Hence it follows that y ∈ Ũ(m) and z 6∈ Ũ(m). Case 2. Let z ∈ L(X) \ X. We can write z = (si,ε), where ε ∈ {−1, 1}. If y < z, then it follows that y ∈ Ũ−(si,ε) and z 6∈ Ũ−(si,ε). If z < y, then it follows that y ∈ Ũ+(si,ε) and z 6∈ Ũ+(si,ε). Hence Y is an Sδ-subset of L(X). This completes the proof of Theorem 5.3. � 6. Examples Example 6.1. Theorems 3.1 and 3.2 do not hold without the assumption of the countability of the set R∪L. Let us consider the Sorgenfrey line X = (R,S). In this case, the right looking points R = R is uncountable. Since X has a Gδ- diagonal, X has an Sδ-diagonal. However, X∗ does not have an Sδ-diagonal. To prove this, it is sufficient to see that X∗ does not have a quasi-Gδ-diagonal [5]. We easily see that there does not exist a family of countable collections of open subsets of X∗ that separates two points of the form (x, 0) and (x, 1), where x ∈ X. Example 6.2. Theorem 3.2 does not hold without the existence of the se- quences for points of R ∪ L. To show a counterexample, let Y be the set of countable ordinals [0,ω1[ with the discrete topology. The right looking points of Y comprise the set of limit ordinals. Let Y ∗ be the linear extension of Y defined in Section 2. Let X = Y ∗∪{(ω1, 0)}, where X is ordered as (ω1, 0) > α for all α ∈ Y ∗, and given the discrete topology. Then R = {(ω1, 0)} is a single- ton and L(X) = Y ∗ ∪{(ω1,−1)}∪{(ω1, 0)}, where α < (ω1,−1) < (ω1, 0) for all α ∈ Y ∗. There does not exist an increasing sequence in Y ∗ that converges to (ω1,−1). Furthermore, L(X) does not have a quasi-Gδ-diagonal, because the points of X and the point {(ω1,−1)} are not separated by a family of countable collections of open subsets of L(X). Hence L(X) does not have an Sδ-diagonal. Example 6.3. A generalized ordered space does not necessarily have a dense Sδ-diagonal. To show this, consider the linearly ordered space Z that was constructed by H. R. Bennett and D. J. Lutzer [4]. They proved that Z is not first-countable at any point. Z is defined as follows: Z = {(α1,α2, . . . ,αn,ω1,ω1, . . .) : αi < ω1, 1 ≤ i ≤ n, αi = ω1, i > n, n ≥ 1}, with the lexicographic order. Since Z is densely-ordered, a dense subset Y of Z is a LOTS. If Y has a quasi-Gδ-diagonal, Y is quasi-developable. Since a quasi-developable space is first-countable, Y does not have a quasi-Gδ-diagonal. Therefore, Z does not have a dense Sδ-diagonal. Dense Sδ-diagonals 77 Acknowledgements. The author is grateful to the referees for their helpful comments. References [1] A. V. Arhangel’skĭı and Lj. D. Kočinac, On a dense Gδ-diagonal, Publ. L’Institut Math. 47 (61) (1990), 121–126. [2] H. R. Bennett, LOTS with Sδ-diagonals, Topology Proc. 12 (1987), 211–216. [3] H. R. Bennett, M. Hosobuchi and D. J. Lutzer, A note on perfect generalized ordered spaces, Rocky Mountain J. Math. 29 (4) (1999), 1195–1207. [4] H. R. Bennett and D. J. Lutzer, Point countability in generalized ordered spaces, Topology Appl. 71 (1996), 149–165. [5] M. Hosobuchi, Sδ-diagonals and generalized ordered spaces, J. Tokyo Kasei Gakuin Univ. (Nat. Sci. Tech.) 41 (2001), 1–7. [6] D. J. Lutzer, A metrization theorem for linearly orderable spaces, Proc. Amer. Math. Soc. 22 (1969), 557–558. [7] D. J. Lutzer, On generalized ordered spaces, Dissertationes Math. 89 (1971), 1–32. [8] T. Miwa and N. Kemoto, Linearly ordered extensions of GO spaces, Topology Appl. 54 (1993), 133–140. Received January 2002 Revised December 2002 Masami Hosobuchi Tokyo Kasei Gakuin University, Department of Housing and Planning, 2600 Aihara, Machida, Tokyo 194-0292, Japan. E-mail address : mhsbc@kasei-gakuin.ac.jp Dense S-diagonals and linearly ordered extensions. By M. Hosobuchi