TkachenkoAGT.dvi @ Applied General Topology c© Universidad Politécnica de Valencia Volume 10, No. 2, 2009 pp. 269-276 Compact self T1-complementary spaces without isolated points Mikhail Tkachenko Abstract. We present an example of a compact Hausdorff self T1-complementary space without isolated points. This answers Ques- tion 3.11 from [A compact Hausdorff topology that is a T1-complement of itself, Fund. Math. 175 (2002), 163–173] affirmatively. 2000 AMS Classification: 54A10, 54A25, 54D30 Keywords: Alexandroff duplicate; Čech–Stone compactification; Compact; Isolated point; T1-complementary; Transversal topology 1. Introduction We deal with the concept of complementarity in the lattice of T1-topologies on a given infinite set. Two elements a, b of an abstract lattice {L, ∨, ∧, 0, 1} with the smallest and greatest elements 0 and 1, respectively, are called com- plementary if a ∨ b = 1 and a ∧ b = 0. Birkhoff noted in [1] that the family L(X) of all topologies on a nonempty set X becomes a lattice when the infi- mum τ1 ∧ τ2 of τ1, τ2 ∈ L(X) is defined to be the intersection τ1 ∩ τ2 and the supremum τ1 ∨ τ2 is the topology on X with the subbase τ1 ∪ τ2. Clearly, the smallest element 0 of L(X) is the coarsest topology {∅, X}, while the greatest element 1 of L(X) is the discrete topology of X. In the case of the lattice L1(X) of all T1-topologies on X, the smallest element 0 of L1(X) is the cofinite topology cf in(X) = {∅} ∪ {X \ F : F ⊆ X, F is finite}. Therefore, two topologies τ1, τ2 ∈ L1(X) are complementary in L1(X) if τ1 ∩ τ2 = cf in(X) and τ1 ∪ τ2 is a subbase for the discrete topology on X. It is said that τ1 and τ2 are T1-complementary in this case. The study of complementarity in L1(X) was initiated by A. Steiner and E. Steiner in [6, 8, 7]. Later on, S. Watson used an elaborated combinatorics in 270 M. Tkachenko [10] to prove that a set X of cardinality c+, where c = 2ω, admits a Tychonoff self T1-complementary topology τ . Self T1-complementarity of τ means that there exists a bijection f of X onto itself such that the topologies τ and σ = {f −1(U ) : U ∈ τ} are T1-complementary. In [4], D. Shakhmatov and the author applied a recursive construction to show that the Alexandroff duplicate A(βω \ ω) of βω \ ω is a T1-complement of itself. A(βω \ ω) was the first example of an infinite compact Hausdorff space with this property. It is clear that |A(βω \ ω)| = 2c > c, which looks quite similar to the cardinality of Watson’s self T1-complementary space in [10]. The necessity of working with topologies on big sets was explained in [4, Corollary 3.6]—the existence of a compact Hausdorff self T1-complementary space of cardinality less than or equal to c is independent of ZF C. The concept of T1-complementarity of topologies is naturally split into trans- versality and T1-independence. Following [5, 9], we say that topologies τ1, τ2 ∈ L1(X) are transversal if τ1 ∨ τ2 is the discrete topology, and T1-independent if τ1 ∧τ2 is the cofinite topology on X. In addition, if the topologies τ1 and τ2 are homeomorphic (i.e., τ2 is obtained from τ1 by means of a bijection of X), we come to the notions of self-transversality and self T1-independence, respectively. A usual way to produce self-transversal topologies is to work with a space that has many isolated points. Indeed, suppose that X is a space with topology τ , Y ⊆ X, |Y | = |X| = |X \ Y |, and each point of Y is isolated in X. Take any bijection f : X → X such that f (X \ Y ) = Y and put σ = {f −1(U ) : U ∈ τ}. It is easy to see that every point of X is isolated either in τ or in σ, so τ ∨ σ is the discrete topology on X. In other words, the space (X, τ ) is self-transversal. This approach was also adopted in [4, Corollary 3.8] to show that the compact space A(βω \ ω) is self-transversal (as a part of the proof that the space is self T1-complementary). This explains Question 3.11 from [4]: Does there exist a self T1-complementary compact Hausdorff space without isolated points? Theorem 2.1 answers this question in the affirmative. Our space (or, better to say, a series of spaces) is A(βω \ ω) × Y , where Y is any dense-in-itself compact Hausdorff space of cardinality c. It is worth mentioning that the idea of the proof of Theorem 2.1 is a natural refinement of the arguments in [4] and [2]. Taking Y to be the closed unit interval or the Cantor set, we obtain in ZF C an example of a compact Hausdorff space without isolated points which is a T1-complement of itself (see Corollary 2.2). Further, assuming that 2 ℵ1 = c and taking Y = {0, 1}ω1, we get an example of a compact Hausdorff space without points of countable character which is again a T1-complement of itself (see Corollary 2.3). We finish the article with three open problems about possible cardinalities of compact Hausdorff self T1-complementary spaces. 2. The Alexandroff duplicate of βω \ ω and products In what follows K denotes βω \ ω, the remainder of the Čech–Stone com- pactification of the countable discrete space ω. It is clear that every nonempty Compact T1-complementary spaces without isolated points 271 open subset of K has cardinality 2c. We will also use the fact that K contains a pairwise disjoint family λ of open sets such that |λ| = c. The Alexandroff duplicate of K is A(K). It is easy to verify that every infinite closed subset of A(K) has cardinality 2c. The reader can find a detailed discussion of the properties of A(X), for an arbitrary space X, in [3]. Theorem 2.1. For every compact Hausdorff space Y with |Y | ≤ c, the product space A(K) × Y is self T1-complementary. Proof. Let Z = A(K) × Y . Let also τ be the product topology of Z. By recursion of length κ = 2c we will construct a bijection f : Z → Z such that (1) f ◦ f = idZ ; (2) the topology σ = {f (U ) : U ∈ τ} is T1-complementary to τ . Let K∗ = A(K) \ K. One of the main ideas of our construction is to use open fibers {x} × Y ⊆ Z, with x ∈ K∗, to guarantee that each point z ∈ Z will be isolated in (Z, τ ∨ σ). More precisely, we will construct the bijection f to satisfy the following additional conditions: (3) f (K × Y ) = K∗ × Y ; (4) for every x ∈ K∗, the image f ({x} × Y ) is a discrete subset of K × Y . Let us show first that every bijection f satisfying conditions (1), (3), and (4) produces the topology σ = f (τ ) transversal to τ . Indeed, let π : A(K) × Y → A(K) be the projection. Take a point z ∈ Z such that x = π(z) ∈ K∗. Clearly, z ∈ {x} × Y and, by (4), f ({x} × Y ) is a discrete subset of K × Y . Hence there exists an open set U in Z such that (∗) {f (z)} = U ∩ f ({x} × Y ). Since the point x is isolated in A(K), the set {x} × Y is τ -open in A(K) × Y . Hence (∗) implies that f (z) is an isolated point of the space (Z, τ ∨σ). Further, it follows from (1) and (3) that K × Y = f (K∗ × Y ), and we conclude that every point of K × Y is isolated in (Z, τ ∨ σ). Applying f to both parts of (∗) and taking into account (1), we obtain the equality {z} = f (U ) ∩ ({x} × Y ). This means that every point of K∗ × Y is isolated in (Z, τ ∨ σ). We have thus proved that the topology τ ∨ σ is discrete, i.e., τ and σ are transversal. To guarantee the T1-independence of τ and σ is a more difficult task. We can reformulate the latter relation between τ and σ by saying that f (F ) is not τ -closed in Z, for every proper infinite τ -closed set F ⊆ Z. Let us describe the recursive construction of the bijection f in detail. In what follows the space Z always carries the topology τ unless the otherwise is specified. We start with three observations that will be used in our construction of f . The first and the third of them are evident. Fact 1. If B is an infinite subset of A(K), then the set B ∩ K has cardinality κ = 2c, where B is the closure of B in A(K). Fact 2. If C ⊆ Z and the set π(C) is infinite, then the projection π(C∩(K×Y )) has cardinality κ, where C is the closure of C in Z. 272 M. Tkachenko Indeed, since the projection π is a closed mapping, we have the equality π(C) = π(C). It follows from |π(C)| ≥ ω and Fact 1 that the set π(C) ∩ K has cardinality κ. Again, since the mapping π is closed, we see that π−1(x)∩C 6= ∅ for each x ∈ π(C) ∩ K. Hence |π(C ∩ (K × Y ))| = κ. Fact 3. If U is open in Z and U ∩ (K × Y ) 6= ∅, then |U \ (K × Y )| = κ. It is clear that χ(K) ≤ w(K) = c, χ(A(K)) = χ(K) ≤ c, and w(Y ) ≤ |Y | ≤ c. Therefore, χ(z, Z) ≤ c for every z ∈ Z. Since |K × Y | = |K| = κ, there exists a base B for K × Y in Z with |B| ≤ κ. In other words, B is a family of open sets in Z with the property that for every z ∈ K × Y and every open neighbourhood O of z in Z, there exists U ∈ B such that z ∈ U ⊆ O. Clearly, we can assume that U ∩ (K × Y ) 6= ∅ for each U ∈ B. Since κ = κω, we see that |[Z]ω × B| = κ, where [Z]ω denotes the family of all countably infinite subsets of Z. Let {(Cα, Uα) : α < κ} be an enumeration of the set [Z] ω × B such that for every pair (C, U ) ∈ [Z]ω × B, the set {α < κ : (C, U ) = (Cα, Uα)} is cofinal in κ. Let {zα : α < κ} be a faithful enumeration of Z. By recursion on α < κ we will construct sets Zα ⊆ Z and mappings fα : Zα → Zα satisfying the following conditions: (iα) |Zα| ≤ |α| · c; (iiα) if γ < α, then Zγ ⊆ Zα; (iiiα) zα ∈ Zα+1; (ivα) fα is a bijection of Zα onto itself and fα ◦ fα = idZα ; (vα) if γ < α, then fα↾Zγ = fγ ; (viα) if z ′, z′′ ∈ Zα, π(z ′) = π(z′′), and z′ 6= z′′, then π(fα(z ′)) 6= π(fα(z ′′)); (viiα) fα+1(Uα∩Zα+1)∩fα+1(Cα ∩ Zα+1) 6= ∅ provided that the set πfα(Cα∩ Zα) is infinite; (viiiα) π −1(x) ⊆ Zα for each x ∈ π(Zα) ∩ K ∗; (ixα) if x ∈ π(Zα) ∩ K ∗, then fα({x} × Y ) is a discrete subset of K × Y ; (xα) fα(Zα ∩ (K × Y )) ⊆ K ∗ × Y . Put Z0 = ∅ and f0 = ∅. Clearly, Z0 and f0 satisfy (i0)–(x0). Let α < κ, and suppose that a set Zβ ⊆ Z and a mapping fβ of Zβ to itself satisfying conditions (iβ )–(xβ ) have already been defined for all β < α. If α > 0 is limit, we put Zα = ⋃ {Zβ : β < α} and fα = ⋃ {fβ : β < α}. Then the subset Zα of Z and the mapping fα : Zα → Zα satisfy (iα)–(xα), except for (iiiα) and (viiα) which are valid for all β < α. Suppose now that α = γ+1. Let Z′γ = Zγ ∪{zγ}. Since Uγ ∩(K×Y ) 6= ∅, the cardinality of the set Uγ\(K×Y ) is κ by Fact 3. It follows from |Z ′ γ| ≤ |Zγ|+1 ≤ |γ+1|·c < κ and |π−1π(Z′γ )| ≤ |Z ′ γ|·|Y | < κ that |(Uγ \(K×Y ))\π −1π(Z′γ )| = κ. Therefore, we can pick a point sα ∈ Uγ \ π −1(K ∪ π(Z′γ )). If πfγ (Cγ ∩ Zγ ) is infinite, then fγ (Cγ ∩ Zγ ) ∩ (K × Y ) is a closed subset of Z whose projection to A(K) has cardinality κ by Fact 2. We then use the inequalities |Z′γ| < κ and |Y | ≤ c to pick a point tα ∈ (K × Y ) ∩ fγ (Cγ ∩ Zγ ) \ Compact T1-complementary spaces without isolated points 273 π−1π(Z′γ ). Otherwise pick an arbitrary point tα ∈ π −1(K \ π(Z′γ )); again, such a point exists because |π(Z′γ )| ≤ |Z ′ γ| < κ = |K|. In either case, tα ∈ K × Y . Suppose that zγ = (xγ , yγ ), sα = (x ′ α, y ′ α), and tα = (x ′′ α, y ′′ α). Notice that x′α ∈ K ∗ \ π(Z′γ ) and x ′′ α ∈ K \ π(Z ′ γ ). To define Zα, we consider the following possible cases. Case 1. zγ ∈ Zγ . Then Z ′ γ = Zγ and we choose a discrete set Dα ⊆ K × {y ′′ α} such that tα ∈ Dα, π(Dα) ∩ π(Zγ ) = ∅, and |Dα| = |Y |. This is possible since x′′α = π(tα) /∈ π(Zγ ) and K contains c pairwise disjoint nonempty open sets, each of cardinality κ. Put Zα = Zγ ∪ Dα ∪ ({x ′ α} × Y ). It follows from the definition that {zγ, sα, tα} ⊆ Zα. Since the sets Dα, {x ′ α} × Y , and Zγ are pairwise disjoint, there exists an idempotent bijection fα of Zα onto itself such that fα extends fγ , fα({x ′ α} × Y ) = Dα, and fα(sα) = tα. Case 2. zγ /∈ Zγ . Again, we split this case into two subcases. Case 2.1. zγ ∈ K × Y , i.e., xγ ∈ K. Then we choose a discrete subset Dα of K × Y such that {zγ , tα} ⊆ Dα, Dα ∩ Zγ = ∅, the restriction of π to Dα is one-to-one, and |Dα| = |Y |. Again, this is possible since neither zγ nor tα is in Zγ and, by the choice of tα, xγ = π(zγ ) 6= π(tα) = x ′′ α. As in Case 1, we put Zα = Zγ ∪ Dα ∪ ({x ′ α} × Y ). Then {zγ , sα, tα} ⊆ Zα. Since the sets Dα, {x ′ α} × Y , and Zγ are pairwise disjoint, there exists an idempotent bijection fα : Zα → Zα such that fα extends fγ , fα(sα) = tα, and fα({x ′ α} × Y ) = Dα. Case 2.2. xγ ∈ K ∗. We choose a discrete set Dα ⊆ K × {y ′′ α} such that tα ∈ Dα, π(Dα) ∩ π(Zγ ) = ∅, and |Dα| = |Y |. Then we put Zα = Zγ ∪ Dα ∪ ({xγ , x ′ α} × Y ). Clearly, {zγ , sα, tα} ⊆ Zα. Since {xγ , x ′ α} ⊆ K ∗ and {zγ, sα} ∩ Zγ = ∅, it follows from (viiiγ ) that ({xγ , x ′ α} × Y ) ∩ Zγ = ∅. In addition, the set Dα is disjoint from both Zγ and {xγ , x ′ α} × Y , so there exists an idempotent bijection fα of Zα onto itself such that fα extends fγ , fα({xγ , x ′ α} × Y ) = Dα, and fα(sα) = tα. Clearly, conditions (iα), (iiα), (iiiγ ), (ivα), (vα), and (viiiα)–(xα) hold true. Let us verify conditions (viα) and (viiγ ). We verify (viα) only in Case 2.1—the argument in the rest of cases is anal- ogous or even simpler. Suppose that z′ and z′′ are distinct elements of Zα such that π(z′) = π(z′′). If {z′, z′′} ⊆ Zγ , then (vα) and (viγ ) imply that π(fα(z ′)) = π(fγ (z ′)) 6= π(fγ (z ′′)) = π(fα(z ′′)). If {z′, z′′} ⊆ {x′α} × Y , then π(fα(z ′)) 6= π(fα(z ′′)) since fα({x ′ α} × Y ) = Dα and the restriction of π to Dα is one-to-one. The case {z′, z′′} ⊆ Dα is clearly impossible. Finally, suppose that z′ ∈ Zγ and z ′′ ∈ Zα \Zγ (or vice versa). Since x ′ α /∈ π(Zγ ), if follows from π(z′) = π(z′′) and the definition of Zα that z ′′ ∈ Dα. Our choice of fα implies that fα(Dα) = {x ′ α} × Y because fα is an idempotent bijection of Zα onto itself. Hence π(fα(z ′′)) = x′α /∈ π(Zγ ) and, therefore, π(fα(z ′′)) 6= π(fα(z ′)). 274 M. Tkachenko To check (viiγ ), suppose that πfγ (Cγ ∩ Zγ ) is infinite. It follows from our construction that sα ∈ Uγ ∩ Zα and fα(sα) = tα ∈ fγ (Cγ ∩ Zγ ) which yields tα ∈ fα(Uγ ∩ Zα) ∩ fα(Cγ ∩ Zα) 6= ∅. The recursive step is completed. We can now define the bijection f : Z → Z. From (iiiα) for all α < κ it follows that Z = ⋃ {Zα : α < κ}. Let f = ⋃ {fα : α < κ}. Since (iiα), (ivα) and (vα) hold for all α < κ, f is an idempotent bijection of Z onto itself. This means that (1) holds. It also follows from (viiiα) and (ixα) for all α < κ that f (K∗ × Y ) ⊆ K × Y , while (xα) implies that f (K × Y ) ⊆ K ∗ × Y . Since f is a bijection, we conclude that f (K∗ × Y ) = K × Y and f (K × Y ) = K∗ × Y , i.e., (3) holds. Similarly, conditions (viiiα) and (ixα) for all α < κ together imply the validity of (4). It was shown before the recursive construction that for any bijection f : Z → Z satisfying (1), (3), and (4), the topologies τ and σ = f (τ ) on Z are transver- sal. It only remains to prove that τ and σ = f (τ ) are T1-independent, for this special bijection f . In other words, we have to verify that for every proper infinite closed subset Φ of Z, the image f (Φ) is not closed in Z. Let us consider two cases. Case A. The projection π(Φ) is finite. Since Φ ⊆ π−1π(Φ) and each fiber π−1(x) has cardinality |Y | ≤ c, we see that |Φ| ≤ c. Also, since κc = κ, the cofinality of the cardinal κ is greater than c. Applying the equality Z = ⋃ {Zα : α < κ} and (iiα) for α < κ, we see that Φ ⊆ Zβ for some β < κ. It is also clear that π−1(x) ∩ Φ is infinite for some x ∈ A(K). Then (viβ ) yields that the set π(f (Φ)) = π(fβ (Φ)) is infinite. In its turn, it follows from Fact 2 that the closure of f (Φ) in Z has cardinality κ and, since |Φ| ≤ c, the set f (Φ) cannot be closed in Z. Case B. The set π(Φ) is infinite. Then |Φ| = κ, by Fact 2. Again, we split this case into two subcases. Case B.1. (K × Y ) \ Φ 6= ∅. Since cf (κ) > c > ω, the set πfβ (Φ ∩ Zβ) must be infinite for some β < κ. Indeed, otherwise πf (Φ) is finite and hence |Φ| = |f (Φ)| ≤ c, a contradiction. Choose a countable set C ⊆ Φ ∩ Zβ such that πf (C) is infinite. Take a point z ∈ (K × Y ) \ Φ and an element U ∈ B such that z ∈ U ⊆ Z \ Φ. This is possible because B is a base for K × Y in Z. Note that (C, U ) ∈ [Z]ω × B. Since the set {α < κ : (C, U ) = (Cα, Uα)} is cofinal in κ, (C, U ) = (Cα, Uα) for some α with β ≤ α < κ. From Zα ⊇ Zβ and Cα = C ⊆ Zβ we get Cα ∩ Zα ⊇ Cα ∩ Zβ = C and, since πf (C) is infinite, so is πf (Cα ∩ Zα) = πfα(Cα ∩ Zα). Then (viiα) shows that fα+1(Uα ∩ Zα+1) ∩ fα+1(Cα ∩ Zα+1) 6= ∅. Since f extends fα and Φ ⊇ C = Cα, it follows that f (Uα) ∩ f (Φ) ⊇ f (Uα) ∩ f (Cα) ⊇ fα+1(Uα ∩ Zα+1) ∩ fα+1(Cα ∩ Zα+1) 6= ∅. Therefore, there exists z∗ ∈ Uα such that f (z ∗) ∈ f (Φ). It follows from Uα = U ⊆ Z \ Φ that z∗ /∈ Φ. Since f is a bijection of Z, this yields f (z∗) /∈ f (Φ). Thus f (z∗) ∈ f (Φ) \ f (Φ), that is, the set f (Φ) is not closed in Z. Compact T1-complementary spaces without isolated points 275 Case B.2. K × Y ⊆ Φ. Suppose to the contrary that f (Φ) is closed in Z. Since f (K × Y ) = K∗ × Y and the latter set is dense in Z, we see that K∗ × Y ⊆ f (Φ) = Z. This contradicts our choice of Φ as a proper subset of Z. We have thus proved that f (Φ) fails to be closed in Z, i.e., the topologies τ and σ = f (τ ) are T1-independent. Since we already know that τ and σ are transversal, this finishes the proof of the theorem. � Taking Y in Theorem 2.1 to be the Cantor set or the closed unit interval I = [0, 1], we obtain the following result which answers Question 3.11 from [4] in the affirmative: Corollary 2.2. There exists an infinite compact Hausdorff self T1-complement- ary space without isolated points. Under additional set-theoretic assumptions, one can refine Corollary 2.2 as follows: Corollary 2.3. Let κ be a cardinal with ω ≤ κ < c. It is consistent with ZF C that there exists a compact Hausdorff self T1-complementary space Z such that χ(z, Z) ≥ κ for each z ∈ Z. Proof. One can assume that 2κ = 2ω = c and take Y = Iκ in Theorem 2.1. � The following questions remain open. Problem 2.4. Let K = βω \ ω. Is the product space A(K) × K self T1- complementary? Problem 2.5. Is it true that for every cardinal λ, there exists a compact Hausdorff self T1-complementary space Z with |Z| ≥ λ? Here is a stronger version of the above problem: Problem 2.6. Is it true that for every cardinal λ, there exists a compact Hausdorff self T1-complementary space Z such that χ(z, Z) ≥ λ for all z ∈ Z? References [1] G. Birkhoff, On the combination of topologies, Fund. Math. 26 (1936), 156–166. [2] A. B laszczyk and M. G. Tkachenko, Transversal and T1-independent topologies and the Alexandroff duplicate, submitted. [3] R. Engelking, On the Double Circumference of Alexandroff, Bull. Acad. Polon. Sci. 16 (1968), 629–634. [4] D. Shakhmatov and M. G. Tkachenko, A compact Hausdorff topology that is a T1- complement of itself, Fund. Math. 175 (2002), 163–173. [5] D. Shakhmatov, M. G. Tkachenko, and R. G. Wilson, Transversal and T1-independent topologies, Houston J. Math. 30 (2004), no. 2, 421–433. [6] A. K. Steiner, Complementation in the lattice of T1-topologies, Proc. Amer. Math. Soc. 17 (1966), 884–885. [7] A. K. Steiner and E. F. Steiner, Topologies with T1-complements, Fund. Math. 61 (1967), 23-38. [8] A. K. Steiner and E. F. Steiner, A T1-complement of the reals, Proc. Amer. Math. Soc. 19 (1968), 177-179. 276 M. Tkachenko [9] M. G. Tkachenko and Iv. Yaschenko, Independent group topologies on Abelian groups, Topol. Appl. 122 (2002), no. 1-2, 425–451. [10] W. S. Watson, A completely regular space which is the T1-complement of itself, Proc. Amer. Math. Soc. 124 (1996), no. 4, 1281–1284. Received September 2009 Accepted November 2009 Mikhail Tkachenko (mich@xanum.uam.mx) Departamento de Matemáticas, Universidad Autónoma Metropolitana, Av. San Rafael Atlixco 186, Col. Vicentina, Iztapalapa, C.P. 09340, México D.F., México