The product of locally compact paracompact spaces

It is well known that when X and Y are paracompact spaces, the product space X \times Y is not necessarily normal. Classic examples include the product of the Sorgenfrey line with itself (discussed here) and the product of the Michael line and the space of irrational numbers (discussed here). However, if one of the paracompact factors is “compact”, the product can be normal or even paracompact. This post discusses several classic results along this line. All spaces are Hausdorff and regular.

Suppose that X and Y are paracompact spaces. We have the following results:

  1. If Y is a compact space, then X and Y is paracompact.
  2. If Y is a \sigma-compact space, then X and Y is paracompact.
  3. If Y is a locally compact space, then X and Y is paracompact.
  4. If Y is a \sigma-locally compact space, then X and Y is paracompact.

The proof of the first result makes uses the tube lemma. The second result is a corollary of the first. The proofs of both results are given here. The third result is a corollary of the fourth result. We give a proof of the fourth result.

____________________________________________________________________

Proof of the Fourth Result

The fourth result indicated above is restated as Theorem 2 below. It is a theorem of K. Morita [1]. This is one classic result on product of paracompact spaces. After proving the theorem, comments are made about interesting facts and properties that follow from this result. Theorem 2 is also Theorem 3.22 in chapter 18 in the Handbook of Set-Theoretic Topology [2].

A space W is a locally compact space if for each w \in W, there is an open subset O of W such that w \in O and \overline{O} is compact. When we say Y is a \sigma-locally compact space, we mean that Y=\bigcup_{j=1}^\infty Y_j where each Y_j is a locally compact space. In proving the result discussed here, we also assume that each Y_j is a closed subspace of Y. The following lemma will be helpful.

Lemma 1
Let Y be a paracompact space. Suppose that Y is a \sigma-locally compact. Then there exists a cover \mathcal{C}=\bigcup_{j=1}^\infty \mathcal{C}_j of Y such that each \mathcal{C}_j is a locally finite family consisting of compact sets.

Proof of Lemma 1
Let Y=\bigcup_{n=1}^\infty Y_n such that each Y_n is closed and is locally compact. Fix an integer n. For each y \in Y_n, let O_{n,y} be an open subset of Y_n such that y \in O_{n,y} and \overline{O_{n,y}} is compact (the closure is taken in Y_n). Consider the open cover \mathcal{O}=\left\{ O_{n,y}: y \in Y_j \right\} of Y_n. Since Y_n is a closed subspace of Y, Y_n is also paracompact. Let \mathcal{V}=\left\{ V_{n,y}: y \in Y_j \right\} be a locally finite open cover of Y_n such that \overline{V_{n,y}} \subset O_{n,y} for each y \in Y_n (again the closure is taken in Y_n). Each \overline{V_{n,y}} is compact since \overline{V_{n,y}} \subset O_{n,y} \subset \overline{O_{n,y}}. Let \mathcal{C}_n=\left\{ \overline{V_{n,y}}: y \in Y_n \right\}.

We claim that \mathcal{C}_n is a locally finite family with respect to the space Y. For each y \in Y-Y_n, Y-Y_n is an open set containing y that intersects no set in \mathcal{C}_n. For each y \in Y_n, there is an open set O \subset Y_n that meets only finitely many sets in \mathcal{C}_n. Extend O to an open subset O_1 of Y. That is, O_1 is an open subset of Y such that O=O_1 \cap Y_n. It is clear that O_1 can only meets finitely many sets in \mathcal{C}_n.

Then \mathcal{C}=\bigcup_{j=1}^\infty \mathcal{C}_j is the desired \sigma-locally finite cover of Y. \square

Theorem 2
Let X be any paracompact space and let Y be any \sigma-locally compact paracompact space. Then X \times Y is paracompact.

Proof of Theorem 2
By Lemma 1, let \mathcal{C}=\bigcup_{n=1}^\infty \mathcal{C}_n be a \sigma-locally finite cover of Y such that each \mathcal{C}_n consists of compact sets. To show that X \times Y is paracompact, let \mathcal{U} be an open cover of X \times Y. For each C \in \mathcal{C} and for each x \in X, the set \left\{ x \right\} \times C is obviously compact.

Fix C \in \mathcal{C} and fix x \in X. For each y \in C, the point (x,y) \in U_{y} for some U_{y} \in \mathcal{U}. Choose open H_y \subset X and open K_y \subset Y such that (x,y) \in H_y \times K_y \subset U_{x,y}. Letting y vary, the open sets H_y \times K_y cover the compact set \left\{ x \right\} \times C. Choose finitely many open sets H_y \times K_y that also cover \left\{ x \right\} \times C. Let H(C,x) be the intersection of these finitely many H_y. Let \mathcal{K}(C,x) be the set of these finitely many K_y.

To summarize what we have obtained in the previous paragraph, for each C \in \mathcal{C} and for each x \in X, there exists an open subset H(C,x) containing x, and there exists a finite set \mathcal{K}(C,x) of open subsets of Y such that

  • C \subset \bigcup \mathcal{K}(C,x),
  • for each K \in \mathcal{K}(C,x), H(C,x) \times K \subset U for some U \in \mathcal{U}.

For each C \in \mathcal{C}, the set of all H(C,x) is an open cover of X. Since X is paracompact, for each C \in \mathcal{C}, there exists a locally finite open cover \mathcal{L}_C=\left\{L(C,x): x \in X \right\} such that L(C,x) \subset H(C,x) for all x. Consider the following families of open sets.

    \mathcal{E}_n=\left\{L(C,x) \times K: C \in \mathcal{C}_n \text{ and } x \in X \text{ and } K \in \mathcal{K}(C,x) \right\}

    \mathcal{E}=\bigcup_{n=1}^\infty \mathcal{E}_n

We claim that \mathcal{E} is a \sigma-locally finite open refinement of \mathcal{U}. First, show that \mathcal{E} is an open cover of X \times Y. Let (a,b) \in X \times Y. Then for some n, b \in C for some C \in \mathcal{C}_n. Furthermore, a \in L(C,x) for some x \in X. The information about C and x are detailed above. For example, C \subset \bigcup \mathcal{K}(C,x). Thus there exists some K \in \mathcal{K}(C,x) such that b \in K. We now have (a,b) \in L(C,x) \times K \in \mathcal{E}_n.

Next we show that \mathcal{E} is a refinement of \mathcal{U}. Fix L(C,x) \times K \in \mathcal{E}_n. Immediately we see that L(C,x) \subset H(C,x). Since K \in \mathcal{K}(C,x), H(C,x) \times K \subset U for some U \in \mathcal{U}. Then L(C,x) \times K \subset U.

The remaining point to make is that each \mathcal{E}_n is a locally finite family of open subsets of X \times Y. Let (a,b) \in X \times Y. Since \mathcal{C}_n is locally finite in Y, there exists some open Q \subset Y such that b \in Q and Q meets only finitely many sets in \mathcal{C}_n, say C_1,C_2,\cdots,C_m. Recall that \mathcal{L}_{C_j} is the set of all L(C_j,x) and is locally finite. Thus there exists an open O \subset X such that a \in O and O meets only finitely many sets in each \mathcal{L}_{C_j} where j=1,2,\cdots,m. Thus the open set O meets only finitely many sets L(C,x) for finitely many C \in \mathcal{C}_n and finitely many x \in X. These finitely many C and x lead to finitely many K. Thus it follows that O \times Q meets only finitely many sets L(C,x) \times K in \mathcal{E}_n. Thus \mathcal{E}_n is locally finite.

What has been established is that every open cover of X \times Y has a \sigma-locally finite open refinement. This fact is equivalent to paracompactness (according to Theorem 1 in this previous post). This concludes the proof of the theorem. \square

____________________________________________________________________

Productively Paracompact Spaces

Consider this property for a space X.

    (*) The space X satisfies the property that X \times Y is a paracompact space for every paracompact space Y.

Such a space can be called a productively paracompact space (for some reason, this term is not used in the literature).

According to the four results stated at the beginning, any space in any one of the following four classes

  1. Compact spaces.
  2. \sigma-compact spaces.
  3. Locally compact paracompact spaces.
  4. \sigma-locally compact paracompact spaces.

satisfies this property. Both the Michael line and the space of the irrational numbers are examples of paracompact spaces that do not have this productively paracompact property. According to comments made on page 799 [2], the theorem of Morita (Theorem 2 here) triggered extensive research to investigate this class of spaces. The class of spaces is broader than the four classes listed here. For example, the productively paracompact spaces also include the closed images of locally compact paracompact spaces. The handbook [2] has more references.

____________________________________________________________________

Normal P-Spaces

Consider this property.

    (**) The space X satisfies the property that X \times Y is a normal space for every metric space Y.

These spaces can be called productively normal spaces with respect to metric spaces. They go by another name. Morita defined the notion of P-spaces and proved that a space X is a normal P-space if and only if the product of X with any metric space is normal.

Since the class of metric spaces contain the paracompact spaces, any space has property (*) would have property (**), i.e. a normal P-space.Thus any locally compact paracompact space is a normal P-space. Any \sigma-locally compact paracompact space is a normal P-space. If a paracompact space has any one of the four “compact” properties discussed here, it is a normal P-space.

Other examples of normal P-spaces are countably compact normal spaces (see here) and perfectly normal spaces (see here).

____________________________________________________________________

Looking at Diagrams

Let’s compare these classes of spaces: productively paracompact spaces (the spaces satisfying property (*)), normal P-spaces and paracompact spaces. We have the following diagram.

    Diagram 1

    \displaystyle \begin{array}{ccccc} \text{ } &\text{ } & \text{Productively Paracompact} & \text{ } & \text{ } \\  \text{ } & \swarrow & \text{ } & \searrow & \text{ } \\  \text{Paracompact} &\text{ } & \text{ } & \text{ } & \text{Normal P-space} \\     \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\    \end{array}

Clearly productively paracompact implies paracompact. As discussed in the previous section, productively paracompact implies normal P. If a space X is such that the product of X with every paracompact space is paracompact, then the product of X with every metric space is paracompact and hence normal.

However, the arrows in Diagram 1 are not reversible. The Michael line mentioned at the beginning will shed some light on this point. Here’s the previous post on Michael line. Let \mathbb{M} be the Michael line. Let \mathbb{P} be the space of the irrational numbers. The space \mathbb{M} would be a paracompact space that is not productively paracompact since its product with \mathbb{P} is not normal, hence not paracompact.

On the other hand, the space of irrational numbers \mathbb{P} is a normal P-space since it is a metric space. But it is not productively paracompact since its product with the Michael line \mathbb{M} is not normal, hence not paracompact.

The two classes of spaces at the bottom of Diagram 1 do not relate. The Michael line \mathbb{M} is a paracompact space that is not a normal P-space since its product with \mathbb{P} is not normal. Normal P-space does not imply paracompact. Any space that is normal and countably compact is a normal P-space. For example, the space \omega_1, the first uncountable ordinal, with the ordered topology is normal and countably compact and is not paracompact.

There are other normal P-spaces that are not paracompact. For example, Bing’s Example H is perfectly normal and not paracompact. As mentioned in the previous section, any perfectly normal space is a normal P-space.

The class of spaces whose product with every paracompact space is paracompact is stronger than both classes of paracompact spaces and normal P-spaces. It is a strong property and an interesting class of spaces. It is also an excellent topics for any student who wants to dig deeper into paracompact spaces.

Let’s add one more property to Diagram 1.

    Diagram 2

    \displaystyle \begin{array}{ccccc} \text{ } &\text{ } & \text{Productively Paracompact} & \text{ } & \text{ } \\  \text{ } & \swarrow & \text{ } & \searrow & \text{ } \\  \text{Paracompact} &\text{ } & \text{ } & \text{ } & \text{Normal P-space} \\   \text{ } & \searrow & \text{ } & \swarrow & \text{ } \\  \text{ } &\text{ } & \text{Normal Countably Paracompact} & \text{ } & \text{ } \\     \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\    \end{array}

All properties in Diagram 2 except for paracompact are productive. Normal countably paracompact spaces are productive. According to Dowker’s theorem, the product of any normal countably paracompact space with any compact metric space is normal (see Theorem 1 in this previous post). The last two arrows in Diagram 2 are also not reversible.

____________________________________________________________________

Reference

  1. Morita K., On the Product of Paracompact Spaces, Proc. Japan Acad., Vol. 39, 559-563, 1963.
  2. Przymusinski T. C., Products of Normal Spaces, Handbook of Set-Theoretic Topology (K. Kunen and J. E. Vaughan, eds), Elsevier Science Publishers B. V., Amsterdam, 781-826, 1984.

____________________________________________________________________
\copyright 2017 – Dan Ma

The product of a perfectly normal space and a metric space is perfectly normal

The previous post gives a positive result for normality in product space. It shows that the product of a normal countably compact space and a metric space is always normal. In this post, we discuss another positive result, which is the following theorem.

Main Theorem
If X is a perfectly normal space and Y is a metric space, then X \times Y is a perfectly normal space.

As a result of this theorem, perfectly normal spaces belong to a special class of spaces called P-spaces. K. Morita defined the notion of P-space and he proved that a space Y is a Normal P-space if and only if X \times Y is normal for every metric space X (see the section below on P-spaces). Thus any perfectly normal space is a Normal P-space.

All spaces under consideration are Hausdorff. A subset A of the space X is a G_\delta-subset of the space X if A is the intersection of countably many open subsets of X. A subset B of the space X is an F_\sigma-subset of the space X if B is the union of countably many closed subsets of X. Clearly, a set A is a G_\delta-subset of the space X if and only if X-A is an F_\sigma-subset of the space X.

A space X is said to be a perfectly normal space if X is normal with the additional property that every closed subset of X is a G_\delta-subset of X (or equivalently every open subset of X is an F_\sigma-subset of X).

The perfect normality has a characterization in terms of zero-sets and cozero-sets. A subset A of the space X is said to be a zero-set if there exists a continuous function f: X \rightarrow [0,1] such that A=f^{-1}(0), where f^{-1}(0)=\left\{x \in X: f(x)=0 \right\}. A subset B of the space X is a cozero-set if X-B is a zero-set, or more explicitly if there is a continuous function f: X \rightarrow [0,1] such that B=\left\{x \in X: f(x)>0 \right\}.

It is well known that the space X is perfectly normal if and only if every closed subset of X is a zero-set, equivalently every open subset of X is a cozero-set. See here for a proof of this result. We use this result to show that X \times Y is perfectly normal.

____________________________________________________________________

The Proof

Let X be a perfectly normal space and Y be a metric space. Since Y is a metric space, let \mathcal{B}=\bigcup_{j=1}^\infty \mathcal{B}_j be a base for Y such that each \mathcal{B}_j is locally finite. We show that X \times Y is perfectly normal. To that end, we show that every open subset of X \times Y is a cozero-set. Let U be an open subset of X \times Y.

For each (x,y) \in X \times Y, there exists open O_{x,y} \subset X and there exists B_{x,y} \in \mathcal{B} such that (x,y) \in O_{x,y} \times B_{x,y} \subset U. Then U is the union of all sets O_{x,y} \times B_{x,y}. Observe that B_{x,y} \in \mathcal{B}_{j} for some integer j. For each B \in \mathcal{B} such that B=B_{x,y} for some (x,y) \in X \times Y, let O(B) be the union of all corresponding open sets O_{x,y} for all applicable (x,y).

For each positive integer j, let \mathcal{W}_j be the collection of all open sets O(B) \times B such that B \in \mathcal{B}_j and B=B_{x,y} for some (x,y) \in X \times Y. Let \mathcal{V}_j=\cup \mathcal{W}_j. As a result, U=\bigcup_{j=1}^\infty \mathcal{V}_j.

Since both X and Y are perfectly normal, for each O(B) \times B \in \mathcal{W}_j, there exist continuous functions

    F_{O(B),j}: X \rightarrow [0,1]

    G_{B,j}: Y \rightarrow [0,1]

such that

    O(B)=\left\{x \in X: F_{O(B),j}(x) >0 \right\}

    B=\left\{y \in Y: G_{B,j}(y) >0 \right\}

Now define H_j: X \times Y \rightarrow [0,1] by the following:

    \displaystyle H_j(x,y)=\sum \limits_{O(B) \times B \in \mathcal{W}_j} F_{O(B),j}(x) \ G_{B,j}(y)

for all (x,y) \in X \times Y. Note that the function H_j is well defined. Since \mathcal{B}_j is locally finite in Y, \mathcal{W}_j is locally finite in X \times Y. Thus H_j(x,y) is obtained by summing a finite number of values of F_{O(B),j}(x) \ G_{B,j}(y). On the other hand, it can be shown that H_j is continuous for each j. Based on the definition of H_j, it can be readily verified that H_j(x,y)>0 for all (x,y) \in \cup \mathcal{W}_j and H_j(x,y)=0 for all (x,y) \notin \cup \mathcal{W}_j.

Define H: X \times Y \rightarrow [0,1] by the following:

    \displaystyle H(x,y)=\sum \limits_{j=1}^\infty \biggl[ \frac{1}{2^j} \ \frac{H_j(x,y)}{1+H_j(x,y)} \biggr]

It is clear that H is continuous. We claim that U=\left\{(x,y) \in X \times Y: H(x,y) >0 \right\}. Recall that the open set U is the union of all O(B) \times B \in \mathcal{W}_j for all j. Thus if (x,y) \in \cup \mathcal{W}_j for some j, then H(x,y)>0 since H_j(x,y)>0. If (x,y) \notin \cup \mathcal{W}_j for all j, H(x,y)=0 since H_j(x,y)=0 for all j. Thus the open set U is an F_\sigma-subset of X \times Y. This concludes the proof that X \times Y is perfectly normal. \square

____________________________________________________________________

Remarks

The main theorem here is a classic result in general topology. An alternative proof is to show that any perfectly normal space is a P-space (definition given below). Then by Morita’s theorem, the product of any perfectly normal space and any metric space is normal (Theorem 1 below). For another proof that is elementary, see Lemma 7 in this previous post.

The notions of perfectly normal spaces and paracompact spaces are quite different. By the theorem discussed here, perfectly normal spaces are normally productive with metric spaces. It is possible for a paracompact space to have a non-normal product with a metric space. The classic example is the Michael line (discussed here).

On the other hand, there are perfectly normal spaces that are not paracompact. One example is Bing’s Example H, which is perfectly normal and not paracompact (see here).

Even though a perfectly normal space is normally productive with metric spaces, it cannot be normally productive in general. For each non-discrete perfectly normal space X, there exists a normal space Y such that X \times Y is not normal. This follows from Morita’s first conjecture (now a true statement). Morita’s first conjecture is discussed here.

____________________________________________________________________

P-Space in the Sense of Morita

Morita defined the notion of P-spaces [1] and [2]. Let \kappa be a cardinal number such that \kappa \ge 1. Let \Gamma be the set of all finite ordered sequences (\alpha_1,\alpha_2,\cdots,\alpha_n) where n=1,2,\cdots and all \alpha_i < \kappa. Let X be a space. The collection \left\{F_\sigma \subset X: \sigma \in \Gamma \right\} is said to be decreasing if this condition holds: \sigma =(\alpha_1,\alpha_2,\cdots,\alpha_n) and \delta =(\alpha_1,\alpha_2,\cdots,\alpha_n, \cdots, \alpha_m) with n<m imply that F_{\delta} \subset F_{\sigma}. The space X is a P-space if for any cardinal \kappa \ge 1 and for any decreasing collection \left\{F_\sigma \subset X: \sigma \in \Gamma \right\} of closed subsets of X, there exists open set U_\sigma for each \sigma \in \Gamma such that the following conditions hold:

  • for all \sigma \in \Gamma, F_\sigma \subset U_\sigma,
  • for any infinite sequence (\alpha_1,\alpha_2,\cdots,\alpha_n,\cdots) where each each finite subsequence \sigma_n=(\alpha_1,\alpha_2,\cdots,\alpha_n) is an element of \Gamma, if \bigcap_{n=1}^\infty F_{\sigma_n}=\varnothing, then \bigcap_{n=1}^\infty U_{\sigma_n}=\varnothing.

If \kappa=1 where 1=\left\{0 \right\}. Then the index set \Gamma defined above can be viewed as the set of all positive integers. As a result, the definition of P-space with \kappa=1 implies the a condition in Dowker’s theorem (see condition 6 in Theorem 1 here). Thus any space X that is normal and a P-space is countably paracompact (or countably shrinking or that X \times Y is normal for every compact metric space or any other equivalent condition in Dowker’s theorem). The following is a theorem of Morita.

Theorem 1 (Morita)
Let X be a space. Then X is a normal P-space if and only if X \times Y is normal for every metric space Y.

In light of Theorem 1, both perfectly normal spaces and normal countably compact spaces are P-spaces (see here). According to Theorem 1 and Dowker’s theorem, it follows that any normal P-space is countably paracompact.

____________________________________________________________________

Reference

  1. Morita K., On the Product of a Normal Space with a Metric Space, Proc. Japan Acad., Vol. 39, 148-150, 1963. (article information; paper)
  2. Morita K., Products of Normal Spaces with Metric Spaces, Math. Ann., Vol. 154, 365-382, 1964.

____________________________________________________________________
\copyright \ 2017 \text{ by Dan Ma}

The product of a normal countably compact space and a metric space is normal

It is well known that normality is not preserved by taking products. When nothing is known about the spaces X and Y other than the facts that they are normal spaces, there is not enough to go on for determining whether X \times Y is normal. In fact even when one factor is a metric space and the other factor is a hereditarily paracompact space, the product can be non-normal (discussed here). This post discusses a productive scenario – the first factor is a normal space and second factor is a metric space with the first factor having the additional property that it is countably compact. In this scenario the product is always normal. This is a well known result in general topology. The goal here is to nail down a proof for use as future reference.

Main Theorem
Let X be a normal and countably compact space. Then X \times Y is a normal space for every metric space Y.

The proof of the main theorem uses the notion of shrinkable open covers.

Remarks
The main theorem is a classic result and is often used as motivation for more advanced results for products of normal spaces. Thus we would like to present a clear and complete proof of this classic result for anyone who would like to study the topics of normality (or the lack of) in product spaces. We found that some proofs of this result in the literature are hard to follow. In A. H. Stone’s paper [2], the result is stated in a footnote, stating that “it can be shown that the topological product of a metric space and a normal countably compact space is normal, though not necessarily paracompact”. We had seen several other papers citing [2] as a reference for the result. The Handbook [1] also has a proof (Corollary 4.10 in page 805), which we feel may not be the best proof to learn from. We found a good proof in [3] using the idea of shrinking of open covers.

____________________________________________________________________

The Notion of Shrinking

The key to the proof is the notion of shrinkable open covers and shrinking spaces. Let X be a space. Let \mathcal{U} be an open cover of X. The open cover of \mathcal{U} is said to be shrinkable if there is an open cover \mathcal{V}=\left\{V(U): U \in \mathcal{U} \right\} of X such that \overline{V(U)} \subset U for each U \in \mathcal{U}. When this is the case, the open cover \mathcal{V} is said to be a shrinking of \mathcal{U}. If an open cover is shrinkable, we also say that the open cover can be shrunk (or has a shrinking). Whenever an open cover has a shrinking, the shrinking is indexed by the open cover that is being shrunk. Thus if the original cover is indexed, e.g. \left\{U_\alpha: \alpha<\kappa \right\}, then a shrinking has the same indexing, e.g. \left\{V_\alpha: \alpha<\kappa \right\}.

A space X is a shrinking space if every open cover of X is shrinkable. Every open cover of a paracompact space has a locally finite open refinement. With a little bit of rearranging, the locally finite open refinement can be made to be a shrinking (see Theorem 2 here). Thus every paracompact space is a shrinking space. For other spaces, the shrinking phenomenon is limited to certain types of open covers. In a normal space, every finite open cover has a shrinking, as stated in the following theorem.

Theorem 1
The following conditions are equivalent.

  1. The space X is normal.
  2. Every point-finite open cover of X is shrinkable.
  3. Every locally finite open cover of X is shrinkable.
  4. Every finite open cover of X is shrinkable.
  5. Every two-element open cover of X is shrinkable.

The hardest direction in the proof is 1 \Longrightarrow 2, which is established in this previous post. The directions 2 \Longrightarrow 3 \Longrightarrow 4 \Longrightarrow 5 are immediate. To see 5 \Longrightarrow 1, let H and K be two disjoint closed subsets of X. By condition 5, the two-element open cover \left\{X-H,X-K \right\} has a shrinking \left\{U,V \right\}. Then \overline{U} \subset X-H and \overline{V} \subset X-K. As a result, H \subset X-\overline{U} and K \subset X-\overline{V}. Since the open sets U and V cover the whole space, X-\overline{U} and X-\overline{V} are disjoint open sets. Thus X is normal.

In a normal space, all finite open covers are shrinkable. In general, an infinite open cover of a normal space may or may not be shrinkable. It turns out that finding a normal space with an infinite open cover that is not shrinkable is no trivial matter (see Dowker’s theorem in this previous post). However, if an open cover in a normal space point-finite or locally finite, then it is shrinkable.

____________________________________________________________________

Key Idea

We now discuss the key idea to the proof of the main theorem. Consider the produce space X \times Y. Let \mathcal{U} be an open cover of X \times Y. Let M \subset X \times Y. The set M is stable with respect to the open cover \mathcal{U} if for each x \in X, there is an open set O_x containing x such that O_x \times M \subset U for some U \in \mathcal{U}.

Let \kappa be a cardinal number (either finite or infinite). A space X is a \kappa-shrinking space if for each open cover \mathcal{W} of X such that the cardinality of \mathcal{W} is \le \kappa, then \mathcal{W} is shrinkable. According to Theorem 1, any normal space is 2-shrinkable.

Theorem 2
Let \kappa be a cardinal number (either finite or infinite). Let X be a \kappa-shrinking space. Let Y be a paracompact space. Suppose that \mathcal{U} is an open cover of X \times Y such that the following two conditions are satisfied:

  • Each point y \in Y has an open set V_y containing y such that V_y is stable with respect to \mathcal{U}.
  • \lvert \mathcal{U} \lvert = \kappa.

Then \mathcal{U} is shrinkable.

Proof of Theorem 2
Let \mathcal{U} be any open cover of X \times Y satisfying the hypothesis. We show that \mathcal{U} has a shrinking.

For each y \in Y, obtain the open covers \left\{G(U,y): U \in \mathcal{U} \right\} and \left\{H(U,y): U \in \mathcal{U} \right\} of X as follows. For each U \in \mathcal{U}, define the following:

    G(U,y)=\cup \left\{O: O \text{ is open in } X \text{ such that } O \times V_y \subset U \right\}

Then \left\{G(U,y): U \in \mathcal{U} \right\} is an open cover of X. Since X is \kappa-shrinkable, there is an open cover \left\{H(U,y): U \in \mathcal{U} \right\} of X such that \overline{H(U,y)} \subset G(U,y) for each U \in \mathcal{U}.

Now \left\{V_y: y \in Y \right\} is an open cover of Y. By the paracompactness of Y, let \left\{W_y: y \in Y \right\} be a locally finite open cover of Y such that \overline{W_y} \subset V_y for each y \in Y. For each U \in \mathcal{U}, define the following:

    W_U=\cup \left\{H(U,y) \times W_y: y \in Y \text{ such that } \overline{H(U,y) \times W_y} \subset U \right\}

We claim that \mathcal{W}=\left\{ W_U: U \in \mathcal{U} \right\} is a shrinking of \mathcal{U}. First it is a cover of X \times Y. Let (x,t) \in X \times Y. Then t \in W_y for some y \in Y. There exists U \in \mathcal{U} such that x \in H(U,y). Note the following.

    \overline{H(U,y) \times W_y} \subset \overline{H(U,y)} \times \overline{W_y} \subset G(U,y) \times V_y \subset U

This means that H(U,y) \times W_y \subset W_U. Since (x,t) \in H(U,y) \times W_y, (x,t) \in W_U. Thus \mathcal{W} is an open cover of X \times Y.

Now we show that \mathcal{W} is a shrinking of \mathcal{U}. Let U \in \mathcal{U}. To show that \overline{W_U} \subset U, let (x,t) \in \overline{W_U}. Let L be open in Y such that t \in L and that L meets only finitely many W_y, say for y=y_1,y_2,\cdots,y_n. Immediately we have the following relations.

    \forall \ i=1,\cdots,n, \ \overline{W_{y_i}} \subset V_{y_i}

    \forall \ i=1,\cdots,n, \ \overline{H(U,y_i)} \subset G(U,y_i)

    \forall \ i=1,\cdots,n, \ \overline{H(U,y_i) \times W_{y_i}} \subset \overline{H(U,y_i)} \times \overline{W_{y_i}} \subset G(U,y_i) \times V_{y_i} \subset U

Then it follows that

    \displaystyle (x,t) \in \overline{\bigcup \limits_{j=1}^n H(U,y_j) \times W_{y_j}}=\bigcup \limits_{j=1}^n \overline{H(U,y_j) \times W_{y_j}} \subset U

Thus U \in \mathcal{U}. This shows that \mathcal{W} is a shrinking of \mathcal{U}. \square

Remark
Theorem 2 is the Theorem 3.2 in [3]. Theorem 2 is a formulation of Theorem 3.2 [3] for the purpose of proving Theorem 3 below.

____________________________________________________________________

Main Theorem

Theorem 3 (Main Theorem)
Let X be a normal and countably compact space. Let Y be a metric space. Then X \times Y is a normal space.

Proof of Theorem 3
Let \mathcal{U} be a 2-element open cover of X \times Y. We show that \mathcal{U} is shrinkable. This would mean that X \times Y is normal (according to Theorem 1). To show that \mathcal{U} is shrinkable, we show that the open cover \mathcal{U} satisfies the two bullet points in Theorem 2.

Fix y \in Y. Let \left\{B_n: n=1,2,3,\cdots \right\} be a base at the point y. Define G_n as follows:

    G_n=\cup \left\{O \subset X: O \text{ is open such that } O \times B_n \subset U \text{ for some } U \in \mathcal{U} \right\}

It is clear that \mathcal{G}=\left\{G_n: n=1,2,3,\cdots \right\} is an open cover of X. Since X is countably compact, choose m such that \left\{G_1,G_2,\cdots,G_m \right\} is a cover of X. Let E_y=\bigcap_{j=1}^m B_j. We claim that E_y is stable with respect to \mathcal{U}. To see this, let x \in X. Then x \in G_j for some j \le m. By the definition of G_j, there is some open set O_x \subset X such that x \in O_x and O_x \times B_j \subset U for some U \in \mathcal{U}. Furthermore, O_x \times E_y \subset O_x \times B_j \subset U.

To summarize: for each y \in Y, there is an open set E_y such that y \in E_y and E_y is stable with respect to the open cover \mathcal{U}. Thus the first bullet point of Theorem 2 is satisfied. The open cover \mathcal{U} is a 2-element open cover. Thus the second bullet point of Theorem 2 is satisfied. By Theorem 2, the open cover \mathcal{U} is shrinkable. Thus X \times Y is normal. \square

Corollary 4
Let X be a normal and pseudocompact space. Let Y be a metric space. Then X \times Y is a normal space.

The corollary follows from the fact that any normal and pseudocompact space is countably compact (see here).

Remarks
The proof of Theorem 3 actually gives a more general result. Note that the second factor only needs to be paracompact and that every point has a countable base (i.e. first countable). The first factor X has to be countably compact. The shrinking requirement for X is flexible – if open covers of a certain size for X are shrinkable, then open covers of that size for the product are shrinkable. We have the following corollaries.

Corollary 5
Let X be a \kappa-shrinking and countably compact space and let Y be a paracompact first countable space. Then X \times Y is a \kappa-shrinking space.

Corollary 6
Let X be a shrinking and countably compact space and let Y be a paracompact first countable space. Then X \times Y is a shrinking space.

____________________________________________________________________

Remarks

The main theorem (Theorem 3) says that any normal and countably compact space is productively normal with one class of spaces, namely the metric spaces. Thus if one wishes to find a non-normal product space with one factor being countably compact, the other factor must not be a metric space. For example, if W=\omega_1, the first uncountable ordinal with the ordered topology, then W \times X is always normal for every metric X. For non-normal example, W \times C is not normal for any compact space C with uncountable tightness (see Theorem 1 in this previous post). Another example, W \times L_{\omega_1} is not normal where L_{\omega_1} is the one-point Lindelofication of a discrete space of cardinality \omega_1 (follows from Example 1 and Theorem 7 in this previous post).

Another comment is that normal countably paracompact spaces are examples of Normal P-spaces. K. Morita defined the notion of P-space and he proved that a space Y is a Normal P-space if and only if X \times Y is normal for every metric space X.

____________________________________________________________________

Reference

  1. Przymusinski T. C., Products of Normal Spaces, Handbook of Set-Theoretic Topology (K. Kunen and J. E. Vaughan, eds), Elsevier Science Publishers B. V., Amsterdam, 781-826, 1984.
  2. Stone A. H., Paracompactness and Product Spaces, Bull. Amer. Math. Soc., Vol. 54, 977-982, 1948. (paper)
  3. Yang L., The Normality in Products with a Countably Compact Factor, Canad. Math. Bull., Vol. 41 (2), 245-251, 1998. (abstract, paper)

____________________________________________________________________
\copyright \ 2017 \text{ by Dan Ma}

kappa-Dowker space and the first conjecture of Morita

Recall the product space of the Michael line and the space of the irrational numbers. Even though the first factor is a normal space (in fact a paracompact space) and the second factor is a metric space, their product space is not normal. This is one of the classic examples demonstrating that normality is not well behaved with respect to product space. This post presents an even more striking result, i.e., for any non-discrete normal space Y, there exists another normal space X such that X \times Y is not normal. The example of the non-normal product of the Michael line and the irrationals is not some isolated example. Rather it is part of a wide spread phenomenon. This result guarantees that no matter how nice a space Y is, a counter part X can always be found that the product of the two spaces is not normal. This result is known as Morita’s first conjecture and was proved by Atsuji and Rudin. The solution is based on a generalization of Dowker’s theorem and a construction done by Rudin. This post demonstrates how the solution is put together.

All spaces under consideration are Hausdorff.

____________________________________________________________________

Morita’s First Conjecture

In 1976, K. Morita posed the following conjecture.

    Morita’s First Conjecture
    If Y is a normal space such that X \times Y is a normal space for every normal space X, then Y is a discrete space.

The proof given in this post is for proving the contrapositive of the above statement.

    Morita’s First Conjecture
    If Y is a non-discrete normal space, then there exists some normal space X such that X \times Y is not a normal space.

Though the two forms are logically equivalent, the contrapositive form seems to have a bigger punch. The contrapositive form gives an association. Each non-discrete normal space is paired with a normal space to form a non-normal product. Examples of such pairings are readily available. Michael line is paired with the space of the irrational numbers (as discussed above). The Sogenfrey line is paired with itself. The first uncountable ordinal \omega_1 is paired with \omega_1+1 (see here) or paired with the cube I^I where I=[0,1] with the usual topology (see here). There are plenty of other individual examples that can be cited. In this post, we focus on a constructive proof of finding such a pairing.

Since the conjecture had been affirmed positively, it should no longer be called a conjecture. Calling it Morita’s first theorem is not appropriate since there are other results that are identified with Morita. In this discussion, we continue to call it a conjecture. Just know that it had been proven.

____________________________________________________________________

Dowker’s Theorem

Next, we examine Dowker’s theorem, which characterizes normal countably paracompact spaces. The following is the statement.

Theorem 1 (Dowker’s Theorem)
Let X be a normal space. The following conditions are equivalent.

  1. The space X is countably paracompact.
  2. Every countable open cover of X has a point-finite open refinement.
  3. If \left\{U_n: n=1,2,3,\cdots \right\} is an open cover of X, there exists an open refinement \left\{V_n: n=1,2,3,\cdots \right\} such that \overline{V_n} \subset U_n for each n.
  4. The product space X \times Y is normal for any compact metric space Y.
  5. The product space X \times [0,1] is normal where [0,1] is the closed unit interval with the usual Euclidean topology.
  6. For each sequence \left\{A_n \subset X: n=1,2,3,\cdots \right\} of closed subsets of X such that A_1 \supset A_2 \supset A_3 \supset \cdots and \cap_n A_n=\varnothing, there exist open sets B_1,B_2,B_3,\cdots such that A_n \subset B_n for each n such that \cap_n B_n=\varnothing.

The theorem is discussed here and proved here. Any normal space that violates any one of the conditions in the theorem is said to be a Dowker space. One such space was constructed by Rudin in 1971 [2]. Any Dowker space would be one factor in a non-normal product space with the other factor being a compact metric space. Actually much more can be said.

The Dowker space constructed by Rudin is the solution of Morita’s conjecture for a large number of spaces. At minimum, the product of any infinite compact metric space and the Dowker space is not normal as indicated by Dowker’s theorem. Any nontrivial convergent sequence plus the limit point is a compact metric space since it is homeomorphic to S=\left\{0 \right\} \cup \left\{\frac{1}{n}: n=1,2,3,\cdots \right\} (as a subspace of the real line). Thus Rudin’s Dowker space has non-normal product with S. Furthermore, the product of Rudin’s Dowker space and any space containing a copy of S is not normal.

Spaces that contain a copy of S extend far beyond the compact metric spaces. Spaces that have lots of convergent sequences include first countable spaces, Frechet spaces and many sequential spaces (see here for an introduction for these spaces). Thus any Dowker space is an answer to Morita’s first conjecture for the non-discrete members of these classes of spaces. Actually, the range for the solution is wider than these spaces. It turns out that any space that has a countable non-discrete subspace would have a non-normal product with a Dowker space. These would include all the classes mentioned above (first countable, Frechet, sequential) as well as countably tight spaces and more.

Therefore, any Dowker space, a normal space that is not countably paracompact, is severely lacking in ability in forming normal product with another space. In order to obtain a complete solution to Morita’s first conjecture, we would need a generalized Dowker’s theorem.

____________________________________________________________________

Shrinking Properties

The key is to come up with a generalized Dowker’s theorem, a theorem like Theorem 1 above, except that it is for arbitrary infinite cardinality. Then a \kappa-Dowker space is a space that violates one condition in the theorem. That space would be a candidate for the solution of Morita’s first conjecture. Note that Theorem 1 is for the infinite countable cardinal \omega only. Before stating the theorem, let’s gather all the notions that will go into the theorem.

Let X be a space. Let \mathcal{U} be an open cover of the space X. The open cover \mathcal{U} is said to be shrinkable if there is an open cover \mathcal{V}=\left\{V(U): U \in \mathcal{U} \right\} such that \overline{V(U)} \subset U for each U \in \mathcal{U}. When this is the case, the open cover \mathcal{V} is said to be a shrinking of \mathcal{U}. If an open cover is shrinkable, we also say that the open cover can be shrunk (or has a shrinking).

Let \kappa be a cardinal. The space X is said to be a \kappa-shrinking space if every open cover of cardinality \le \kappa of the space X is shinkable. The space X is a shrinking space if it is a \kappa-shrinking space for every cardinal \kappa.

When a family of sets are indexed by ordinals, the notion of an increasing or decreasing family of sets is possible. For example, the family \left\{A_\alpha \subset X: \alpha<\kappa \right\} of subsets of the space X is said to be increasing if A_\beta \subset A_\gamma whenever \beta<\gamma. In other words, for an increasing family, the sets are getting larger whenever the index becomes larger. A decreasing family of sets is defined in the reverse way. These two notions are important for some shrinking properties discussed here – e.g. using an open cover that is increasing or using a family of closed sets that is decreasing.

In the previous discussion on shrinking spaces, two other shrinking properties are discussed – property \mathcal{D}(\kappa) and property \mathcal{B}(\kappa). A space X is said to have property \mathcal{D}(\kappa) if every increasing open cover of cardinality \le \kappa for the space X is shrinkable. A space X is said to have property \mathcal{B}(\kappa) if every increasing open cover of cardinality \le \kappa for the space X has a shrinking that is increasing. See this previous post for a discussion on property \mathcal{D}(\kappa) and property \mathcal{B}(\kappa).

____________________________________________________________________

An Attempt for a Generalized Dowker’s Theorem

Let \kappa be an infinite cardinal. The space X is said to be a \kappa-paracompact space if every open cover \mathcal{U} of X with \lvert \mathcal{U} \lvert \le \kappa has a locally finite open refinement. Thus a space is paracompact if it is \kappa-paracompact for every infinite cardinal \kappa. Of course, an \omega-paracompact space is a countably paracompact space.

For any infinite \kappa, let D_\kappa be a discrete space of size \kappa. Let p be a point not in D_\kappa. Define the space Y_\kappa=D_\kappa \cup \left\{p \right\} as follows. The subspace D_\kappa is discrete as before. The open neighborhoods at p are of the form \left\{ p \right\} \cup B where B \subset D_\kappa and \lvert D_\kappa-B \lvert<\kappa. In other words, any open set containing p contains all but less than \kappa many discrete points.

Another concept that is needed is the cardinal function called minimal tightness. Let Y be any space. Define the minimal tightness mt(Y) as the least infinite cardinal \kappa such that there is a non-discrete subspace of Y of cardinality \kappa. If Y is a discrete space, then let mt(Y)=0. For any non-discrete space Y, mt(Y)=\kappa for some infinite \kappa. Note that for the space Y_\kappa defined above would have mt(Y_\kappa)=\kappa. For any space Y, mt(Y)=\omega if and only if Y has a countable non-discrete subspace.

The following theorem can be called a \kappa-Dowker’s Theorem.

Theorem 2
Let X be a normal space. Let \kappa be an infinite cardinal. Consider the following conditions.

  1. The space X is a \kappa-paracompact space.
  2. The space X is a \kappa-shrinking space.
    • For each open cover \left\{U_\alpha: \alpha<\kappa \right\} of X, there exists an open cover \left\{V_\alpha: \alpha<\kappa \right\} such that \overline{V_\alpha} \subset U_\alpha for each \alpha<\kappa.
  3. The space X has property \mathcal{D}(\kappa).
    • For each increasing open cover \left\{U_\alpha: \alpha<\kappa \right\} of X, there exists an open cover \left\{V_\alpha: \alpha<\kappa \right\} such that \overline{V_\alpha} \subset U_\alpha for each \alpha<\kappa.
  4. For each decreasing family \left\{F_\alpha: \alpha<\kappa \right\} of closed subsets of X such that \bigcap_{\alpha<\kappa} F_\alpha=\varnothing, there exists a family \left\{G_\alpha: \alpha<\kappa \right\} of open subsets of X such that \bigcap_{\alpha<\kappa} G_\alpha=\varnothing and F_\alpha \subset G_\alpha for each \alpha<\kappa.
  5. The space X has property \mathcal{B}(\kappa).
    • For each increasing open cover \left\{U_\alpha: \alpha<\kappa \right\} of X, there exists an increasing open cover \left\{V_\alpha: \alpha<\kappa \right\} such that \overline{V_\alpha} \subset U_\alpha for each \alpha<\kappa.
  6. The product space X \times Y_\kappa is a normal space.
  7. The product space X \times Y is a normal space for some space Y with mt(Y)=\kappa.

The following diagram shows how these conditions are related.

Diagram 1
\displaystyle \begin{array}{ccccc}  1 &\text{ } & \Longrightarrow & \text{ } & 5 \\   \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\   \Downarrow & \text{ } & \text{ } & \text{ } & \Updownarrow \\   \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\   2 &\text{ } & \text{ } & \text{ } & 6 \\      \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\   \Downarrow & \text{ } & \text{ } & \text{ } & \Downarrow \\   \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\    3 & \text{ } & \Longleftarrow & \text{ } & 7 \\  \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\  \Updownarrow & \text{ } & \text{ } & \text{ } & \text{ } \\  \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\  4 & \text{ } & \text{ } & \text{ } & \text{ }  \end{array}

In addition to Diagram 1, we have the relations 5 \Longrightarrow 3 and 2 \not \Longrightarrow 5.

Remarks
At first glance, Diagram 1 might give the impression that the conditions in the theorem form a loop. It turns out the strongest property is \kappa-paracompactness (condition 1). Since condition 2 does not imply condition 5, condition 2 does not imply condition 1. Thus the conditions do not form a loop.

The implications 1 \Longrightarrow 2 \Longrightarrow 3 \Longleftarrow 5 and 6 \Longrightarrow 7 are immediate. The following implications are established in this previous post.

    3 \Longleftrightarrow 4 (Theorem 4)

    5 \Longleftrightarrow 6 (Theorem 7)

    2 \not \Longrightarrow 5 (Example 1)

The remaining implications to be shown are 1 \Longrightarrow 5 and 7 \Longrightarrow 3.

____________________________________________________________________

Proof of Theorem 2

1 \Longrightarrow 5
Let \mathcal{U}=\left\{U_\alpha: \alpha<\kappa \right\} be an increasing open cover of X. By \kappa-paracompactness, let \mathcal{G} be a locally finite open refinement of \mathcal{U}. For each \alpha<\kappa, define W_\alpha as follows:

    W_\alpha=\cup \left\{G \in \mathcal{G}: G \subset U_\alpha \right\}

Then \mathcal{W}=\left\{W_\alpha: \alpha<\kappa \right\} is still a locally finite refinement of \mathcal{U}. Since the space X is normal, any locally finite open cover is shrinkable. Let \mathcal{E}=\left\{E_\alpha: \alpha<\kappa \right\} be a shrinking of \mathcal{W}. The open cover \mathcal{E} is also locally finite. For each \alpha, let V_\alpha=\bigcup_{\beta<\alpha} E_\beta. Then \mathcal{V}=\left\{V_\alpha: \alpha<\kappa \right\} is an increasing open cover of X. Note that

    \overline{V_\alpha}=\overline{\bigcup_{\beta<\alpha} E_\beta}=\bigcup_{\beta<\alpha} \overline{E_\beta}

since \mathcal{E} is locally finite and thus closure preserving. Since \mathcal{U} is increasing, \overline{E_\beta} \subset W_\beta \subset U_\beta \subset U_\alpha for all \beta<\alpha. This means that \overline{V_\alpha} \subset U_\alpha for all \alpha.

7 \Longrightarrow 3
Since condition 3 is equivalent to condition 4, we show 7 \Longrightarrow 4. Suppose that X \times Y is normal where Y is a space such that mt(Y)=\kappa. Let D=\left\{d_\alpha: \alpha<\kappa \right\} be a non-discrete subset of Y. Let p be a point such that p \ne d_\alpha for all \alpha and such that p is a limit point of D (this means that every open set containing p contains some d_\alpha). Let \mathcal{F}=\left\{F_\alpha: \alpha<\kappa \right\} be a decreasing family of closed subsets of X such that \bigcap_{\alpha<\kappa} F_\alpha=\varnothing. Define H and K as follows:

    H=\cup \left\{F_\alpha \times \left\{d_\alpha \right\}: \alpha<\kappa \right\}

    K=X \times \left\{p \right\}

The sets H and K are clearly disjoint. The set K is clearly a closed subset of X \times Y. To show that H is closed, let (x,y) \in (X \times Y)-H. Two cases to consider: x \in F_0 or x \notin F_0 where F_0 is the first closed set in the family \mathcal{F}.

The first case x \in F_0. Let \beta<\kappa be least such that x \notin F_\beta. Then y \ne d_\gamma for all \gamma<\beta since (x,y) \in (X \times Y)-H. In the space Y, any subset of cardinality <\kappa is a closed set. Let E=Y-\left\{d_\gamma: \gamma<\beta \right\}, which is open containing y. Let O \subset X be open such that x \in O and O \cap F_\beta=\varnothing. Then (x,y) \in O \times E and O \times E misses points of H.

Now consider the second case x \notin F_0. Let O \subset X be open such that x \in O and O misses F_0. Then O \times Y is an open set containing (x,y) such that O \times Y misses H. Thus H is a closed subset of X \times Y.

Since X \times Y is normal, choose open V \subset X \times Y such that H \subset V and \overline{V} \cap K=\varnothing. For each \alpha<\kappa, define G_\alpha as follows:

    G_\alpha=\left\{x \in X: (x,d_\alpha) \in V \right\}

Note that each G_\alpha is open in X and that F_\alpha \subset G_\alpha for each \alpha<\kappa. We claim that \bigcap_{\alpha<\kappa} G_\alpha=\varnothing. Let x \in X. The point (x,p) is in K. Thus (x,p) \notin \overline{V}. Choose an open set L \times M such that (x,p) \in L \times M and (L \times M) \cap \overline{V}=\varnothing. Since p \in M, there is some \gamma<\kappa such that d_\gamma \in M. Since (x,d_\gamma) \notin \overline{V}, (x,d_\gamma) \notin V. Thus x \notin G_\gamma. This establishes the claim that \bigcap_{\alpha<\kappa} G_\alpha=\varnothing.

____________________________________________________________________

\kappa-Dowker Space

Analogous to the Dowker space, a \kappa-space is a normal space that violates one condition in Theorem 2. Since the seven conditions listed in Theorem 7 are not all equivalent, which condition to use? Condition 1 is the strongest condition since it implies all the other condition. At the lower left corner of Diagram 1 is condition 3, which follows from every other condition. Thus condition 3 (or 4) is the weakest property. An appropriate definition of a \kappa-Dowker space is through negating condition 3 or condition 4. Thus, given an infinite cardinal \kappa, a \kappa-Dowker space is a normal space X that satisfies the following condition:

    There exists a decreasing family \left\{F_\alpha: \alpha<\kappa \right\} of closed subsets of X with \bigcap_{\alpha<\kappa} F_\alpha=\varnothing such that for every family \left\{G_\alpha: \alpha<\kappa \right\} of open subsets of X with F_\alpha \subset G_\alpha for each \alpha, \bigcap_{\alpha<\kappa} G_\alpha \ne \varnothing.

The definition of \kappa-Dowker space is through negating condition 4. Of course, negating condition 3 would give an equivalent definition.

When \kappa is the countably infinite cardinal \omega, a \kappa-Dowker space is simply the ordinary Dowker space constructed by M. E. Rudin [2]. Rudin generalized the construction of the ordinary Dowker space to obtain a \kappa-Dowker space for every infinite cardinal \kappa [4]. The space that Rudin constructed in [4] would be a normal space X such that condition 4 of Theorem 2 is violated. This means that the space X would violate condition 7 in Theorem 2. Thus X \times Y is not normal for every space Y with mt(Y)=\kappa.

Here’s the solution of Morita’s first conjecture. Let Y be a normal and non-discrete space. Determine the least cardinality \kappa of a non-discrete subspace of Y. Obtain the \kappa-Dowker space X as in [4]. Then X \times Y is not normal according to the preceding paragraph.

____________________________________________________________________

Remarks

Answering Morita’s first conjecture is a two-step approach. First, figure out what a generalized Dowker’s theorem should be. Then a \kappa-Dowker space is one that violates an appropriate condition in the generalized Dowker’s theorem. By violating the right condition in the theorem, we have a way to obtain non-normal product space needed in the answer. The second step is of course the proof of the existence of a space that violates the condition in the generalized Dowker’s theorem.

Figuring out the form of the generalized Dowker’s theorem took some work. It is more than just changing the countable infinite cardinal in Dowker’s theorem (Theorem 1 above) to an arbitrary infinite cardinal. This is because the conditions in Theorem 1 are unequal when the cardinality is changed to an uncountable one.

We take the cue from Rudin’s chapter on Dowker spaces [3]. In the last page of that chapter, Rudin pointed out the conditions that should go into a generalized Dowker’s theorem. However, the explanation of the relationship among the conditions is not clear. The previous post and this post are an attempt to sort out the conditions and fill in as much details as possible.

Rudin’s chapter did have the right condition for defining \kappa-Dowker space. It seems that prior to the writing of that chapter, there was some confusion on how to define a \kappa-Dowker space, i.e. a condition in the theorem the violation of which would give a \kappa-Dowker space. If the condition used is a stronger property, the violation may not yield enough information to get non-normal products. According to Diagram 1, condition 3 in Theorem 2 is the right one to use since it is the weakest condition and is down streamed from the conditions about normal product space. So the violation of condition 3 would answer Morita’s first conjecture.

We do not discuss the other step in the solution in any details. Any interested reader can review Rudin’s construction in [2] and [4]. The \kappa-Dowker space is an appropriate subspace of a product space with the box topology.

One interesting observation about the ordinary Dowker space (the one that violates a condition in Theorem 1) is that the product of any Dowker space and any space with a countable non-discrete subspace is not normal. This shows that Dowker space is badly non-productive with respect to normality. This fact is actually not obvious in the usual formulation of Dowker’s theorem (Theorem 1 above). What makes this more obvious in the direction 7 \Longrightarrow 3 in Theorem 2. For the countably infinite case, 7 \Longrightarrow 3 is essentially this: If X \times Y is normal where Y has a countable non-discrete subspace, then X is not a Dowker space. Thus if the goal is to find a non-normal product space, a Dowker space should be one space to check.

____________________________________________________________________

Loose Ends

In the course of working on the contents in this post and the previous post, there are some questions that we do not know how to answer and have not spent time to verify one way or the other. Possibly there are some loose ends to tie. They for the most parts are not open questions, but they should be interesting questions to consider.

For the \kappa-Dowker’s theorem (Theorem 2), one natural question is on the relative strengths of the conditions. It will be interesting to find out the implications not shown in Diagram 1. For example, for the three shrinking properties (conditions 2, 3 and 5), it is straightforward from definition that 2 \Longrightarrow 3 and 5 \Longrightarrow 3. The example of X=\omega_1 (the first uncountable ordinal) shows that 2 \not \Longrightarrow 5 and hence 2 \not \Longrightarrow 1. What about 3 \Longrightarrow 2? In [5], Beslagic and Rudin showed that 3 \not \Longrightarrow 2 using \Diamond ^{++}. A natural question would be: can there be ZFC example? Perhaps searching on more recent papers can yield some answers.

Another question is 5 \Longrightarrow 1? The answer is no with the example being a Navy space – Example 7.6 in p. 194 [1]. The other two directions that have not been accounted for are: 7 \Longrightarrow 6 and 3 \Longrightarrow 7? We do not know the answer.

Another small question that we come across is about X=\omega_1 (the first uncountable ordinal). This is an example for showing 2 \not \Longrightarrow 5. Thus condition 6 is false. Thus X \times Y_{\omega_1} is not normal. Here Y_{\omega_1} is simply the one-point Lindelofication of a discrete space of cardinality \omega_1. The question is: is condition 7 true for X=\omega_1? The product of X=\omega_1 and Y_{\omega_1} (a space with minimal tightness \omega_1) is not normal. Is there a normal X \times Y where Y is another space with minimal tightness \omega_1?

Dowker’s theorem and \kappa-Dowker’s theorem show that finding a normal space that is not shrinking is not a simple matter. To find a normal space that is not countably shrinking took 20 years (1951 to 1971). For any uncountable \kappa, the \kappa-Dowker space that is based on the same construction of an ordinary Dowker space is also a space that is not \kappa-shrinking. With an uncountable \kappa, is the \kappa-Dowker space countably shrinking? This is not obvious one way or the other just from the definition of \kappa-Dowker space. Perhaps there is something obvious and we have not connected the dots. Perhaps we need to go into the definition of the \kappa-Dowker space in [4] to show that it is countably shrinking. The motivation is that we tried to find a normal space that is countably shrinking but not \kappa-shrinking for some uncountable \kappa. It seems that the \kappa-Dowker space in [4] is the natural candidate.

____________________________________________________________________

Reference

  1. Morita K., Nagata J.,Topics in General Topology, Elsevier Science Publishers, B. V., The Netherlands, 1989.
  2. Rudin M. E., A Normal Space X for which X \times I is not Normal, Fund. Math., 73, 179-486, 1971. (link)
  3. Rudin M. E., Dowker Spaces, Handbook of Set-Theoretic Topology (K. Kunen and J. E. Vaughan, eds), Elsevier Science Publishers B. V., Amsterdam, (1984) 761-780.
  4. Rudin M. E., \kappa-Dowker Spaces, Czechoslovak Mathematical Journal, 28, No.2, 324-326, 1978. (link)
  5. Rudin M. E., Beslagic A.,Set-Theoretic Constructions of Non-Shrinking Open Covers, Topology Appl., 20, 167-177, 1985. (link)
  6. Yasui Y., On the Characterization of the \mathcal{B}-Property by the Normality of Product Spaces, Topology and its Applications, 15, 323-326, 1983. (abstract and paper)
  7. Yasui Y., Some Characterization of a \mathcal{B}-Property, TSUKUBA J. MATH., 10, No. 2, 243-247, 1986.

____________________________________________________________________
\copyright \ 2017 \text{ by Dan Ma}

Spaces with shrinking properties

Certain covering properties and separation properties allow open covers to shrink, e.g. paracompact spaces, normal spaces, and countably paracompact spaces. The shrinking property is also interesting on its own. This post gives a more in-depth discussion than the one in the previous post on countably paracompact spaces. After discussing shrinking spaces, we introduce three shrinking related properties. These properties show that there is a deep and delicate connection among shrinking properties and normality in products. This post is also a preparation for the next post on \kappa-Dowker space and Morita’s first conjecture.

All spaces under consideration are Hausdorff and normal or Hausdorff and regular (if not normal).

____________________________________________________________________

Shrinking Spaces

Let X be a space. Let \mathcal{U} be an open cover of X. The open cover of \mathcal{U} is said to be shrinkable if there is an open cover \mathcal{V}=\left\{V(U): U \in \mathcal{U} \right\} of X such that \overline{V(U)} \subset U for each U \in \mathcal{U}. When this is the case, the open cover \mathcal{V} is said to be a shrinking of \mathcal{U}. If an open cover is shrinkable, we also say that the open cover can be shrunk (or has a shrinking). Whenever an open cover has a shrinking, the shrinking is indexed by the open cover that is being shrunk. Thus if the original cover is indexed in a certain way, e.g. \left\{U_\alpha: \alpha<\kappa \right\}, then a shrinking has the same indexing, e.g. \left\{V_\alpha: \alpha<\kappa \right\}.

A space X is a shrinking space if every open cover of X is shrinkable. The property can also be broken up according to the cardinality of the open cover. Let \kappa be a cardinal. A space X is \kappa-shrinking if every open cover of cardinality \le \kappa for X is shrinkable. A space X is countably shrinking if it is \omega-shrinking.

____________________________________________________________________

Examples of Shrinking

Let’s look at a few situations where open covers can be shrunk either all the time or on a limited basis. For a normal space, certain covers can be shrunk as indicated by the following theorem.

Theorem 1
The following conditions are equivalent.

  1. The space X is normal.
  2. Every point-finite open cover of X is shrinkable.
  3. Every locally finite open cover of X is shrinkable.
  4. Every finite open cover of X is shrinkable.
  5. Every two-element open cover of X is shrinkable.

The hardest direction in the proof is 1 \Longrightarrow 2, which is established in this previous post. The directions 2 \Longrightarrow 3 \Longrightarrow 4 \Longrightarrow 5 are immediate. To see 5 \Longrightarrow 1, let H and K be two disjoint closed subsets of X. By condition 5, the two-element open cover \left\{X-H,X-K \right\} has a shrinking \left\{U,V \right\}. Then \overline{U} \subset X-H and \overline{V} \subset X-K. As a result, H \subset X-\overline{U} and K \subset X-\overline{V}. Since the open sets U and V cover the whole space, X-\overline{U} and X-\overline{V} are disjoint open sets. Thus X is normal.

In a normal space, all finite open covers are shrinkable. In general, an infinite open cover of a normal space does not have to be shrinkable unless it is a point-finite or locally finite open cover.

The theorem of C. H. Dowker states that a normal space X is countably paracompact if and only every countable open cover of X is shrinkable if and only if the product space X \times Y is normal for every compact metric space Y if and only if the product space X \times [0,1] is normal. The theorem is discussed here. A Dowker space is a normal space that violates the theorem. Thus any Dowker space has a countably infinite open cover that cannot be shrunk, or equivalently a normal space that forms a non-normal product with a compact metric space. Thus the notion of shrinking has a connection with normality in the product spaces. A Dowker space space was constructed by M. E. Rudin in ZFC [2]. So far Rudin’s example is essentially the only ZFC Dowker space. This goes to show that finding a normal space that is not countably shrinking is not a trivial matter.

Several facts can be derived easily from Theorem 1 and Dowker’s theorem. For clarity, they are called out as corollaries.

Corollary 2

  • All shrinking spaces are normal.
  • All shrinking spaces are normal and countably paracompact.
  • Any normal and metacompact space is a shrinking space.

For the first corollary, if every open cover of a space can be shrunk, then all finite open covers can be shrunk and thus the space must be normal. As indicated above, Dowker’s theorem states that in a normal space, countably paracompactness is equivalent to countably shrinking. Thus any shrinking space is normal and countably paracompact.

Though an infinite open cover of a normal space may not be shrinkable, adding an appropriate covering property to any normal space will make it into a shrinking space. An easy way is through point-finite open covers. If every open cover has a point-finite open refinement (i.e. a metacompact space), then the point-finite open refinement can be shrunk (if the space is also normal). Thus the third corollary is established. Note that the metacompact is not the best possible result. For example, it is known that any normal and submetacompact space is a shrinking space – see Theorem 6.2 of [1].

In paracompact spaces, all open covers can be shrunk. One way to see this is through Corollary 2. Any paracompact space is normal and metacompact. It is also informative to look at the following characterization of paracompact spaces.

Theorem 3
A space X is paracompact if and only if every open cover \left\{U_\alpha: \alpha<\kappa \right\} of X has a locally finite open refinement \left\{V_\alpha: \alpha<\kappa \right\} such that \overline{V_\alpha} \subset U_\alpha for each \alpha.

A proof can be found here. Thus every open cover of a paracompact space can be shrunk by a locally finite shrinking. To summarize, we have discussed the following implications.

    Diagram 1

    \displaystyle \begin{aligned} \text{Paracompact} \Longrightarrow & \text{ Normal + Metacompact}  \\&\ \ \ \ \ \ \Big \Downarrow \\&\text{ Shrinking} \\&\ \ \ \ \ \ \Big \Downarrow  \\& \text{ Normal + Countably Paracompact} \\&\ \ \ \ \ \ \Big \Downarrow  \\& \text{ Normal} \end{aligned}

____________________________________________________________________

Three Shrinking Related Properties

None of the implications in Diagram 1 can be reversed. The last implication in the diagram cannot be reversed due to Rudin’es Dowker space. One natural example to look for would be spaces that are normal and countably paracompact but fail in shrinking at some uncountable cardinal. As indicated by the the theorem of C. H, Dowker, the notion of shrinking is intimately connected to normality in product spaces X \times Y. To further investigate, consider the following three properties.

Let X be a space. Let \kappa be an infinite cardinal. Consider the following three properties.

The space X is \kappa-shrinking if and only if any open cover of cardinality \le \kappa for the space X is shrinkable, i.e. the following condition holds.

    For each open cover \left\{U_\alpha: \alpha<\kappa \right\} of X, there exists an open cover \left\{V_\alpha: \alpha<\kappa \right\} such that \overline{V_\alpha} \subset U_\alpha for each \alpha<\kappa.

The space X has Property \mathcal{D}(\kappa) if and only if every increasing open cover of cardinality \le \kappa for the space X is shrinkable, i.e. the following holds.

    For each increasing open cover \left\{U_\alpha: \alpha<\kappa \right\} of X, there exists an open cover \left\{V_\alpha: \alpha<\kappa \right\} such that \overline{V_\alpha} \subset U_\alpha for each \alpha<\kappa.

The space X has Property \mathcal{B}(\kappa) if and only if the following holds.

    For each increasing open cover \left\{U_\alpha: \alpha<\kappa \right\} of X, there exists an increasing open cover \left\{V_\alpha: \alpha<\kappa \right\} such that \overline{V_\alpha} \subset U_\alpha for each \alpha<\kappa.

A family \left\{A_\alpha: \alpha<\kappa \right\} is increasing if A_\alpha \subset A_\beta for any \alpha<\beta<\kappa. It is decreasing if A_\beta \subset A_\alpha for any \alpha<\beta<\kappa.

In general, any space that is \kappa-shrinking for all cardinals \kappa is a shrinking space as defined earlier. Any space that has property \mathcal{D}(\kappa) for all cardinals \kappa is said to have property \mathcal{D}. Any space that has property \mathcal{B}(\kappa) for all cardinals \kappa is said to have property \mathcal{B}.

The first property \kappa-shrinking is simply the shrinking property for open covers of cardinality \le \kappa. The property \mathcal{D}(\kappa) is \kappa-shrinking with the additional requirement that the open covers to be shrunk must be increasing. It is clear that \kappa-shrinking implies property \mathcal{D}(\kappa). The property \mathcal{B}(\kappa) appears to be similar to \mathcal{D}(\kappa) except that \mathcal{B}(\kappa) has the additional requirement that the shrinking is also increasing. As a result \mathcal{B}(\kappa) implies \mathcal{D}(\kappa). The following diagram shows the implications.

    Diagram 2

    \displaystyle \begin{array}{ccccc} \kappa \text{-Shrinking} &\text{ } & \not \longrightarrow & \text{ } & \text{Property } \mathcal{B}(\kappa) \\  \text{ } & \searrow & \text{ } & \swarrow & \text{ } \\  \text{ } &\text{ } & \text{Property } \mathcal{D}(\kappa) & \text{ } & \text{ } \\     \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\    \end{array}

The implications in Diagram 2 are immediate. An example is given below showing that \omega_1-shrinking does not imply property \mathcal{B}(\omega_1). If \kappa=\omega, then all three properties are equivalent in normal spaces, as displayed in the following diagram. The proof is in Theorem 5.

    Diagram 3

    \displaystyle \begin{array}{ccccc} \omega \text{-Shrinking} &\text{ } & \longrightarrow & \text{ } & \text{Property } \mathcal{B}(\omega) \\  \text{ } & \nwarrow & \text{ } & \swarrow & \text{ } \\  \text{ } &\text{ } & \text{Property } \mathcal{D}(\omega) & \text{ } & \text{ } \\     \text{ } & \text{ } & \text{ } & \text{ } & \text{ } \\    \end{array}

The property \mathcal{D}(\kappa) has a dual statement in terms of decreasing closed sets. The following theorem gives the dual statement.

Theorem 4
Let X be a normal space. Let \kappa be an infinite cardinal. The following two properties are equivalent.

  • The space X has property \mathcal{D}(\kappa).
  • For each decreasing family \left\{F_\alpha: \alpha<\kappa \right\} of closed subsets of X such that \bigcap_{\alpha<\kappa} F_\alpha=\varnothing, there exists a family \left\{G_\alpha: \alpha<\kappa \right\} of open subsets of X such that \bigcap_{\alpha<\kappa} G_\alpha=\varnothing and F_\alpha \subset G_\alpha for each \alpha<\kappa.

First bullet implies second bullet
Let \left\{F_\alpha: \alpha<\kappa \right\} be a decreasing family of closed subsets of X with empty intersection. Then \left\{U_\alpha: \alpha<\kappa \right\} is an increasing family of open subsets of X where U_\alpha=X-F_\alpha. Let \left\{V_\alpha: \alpha<\kappa \right\} be an open cover of X such that \overline{V_\alpha} \subset U_\alpha for each \alpha. Then \left\{G_\alpha: \alpha<\kappa \right\} where G_\alpha=X-\overline{V_\alpha} is the needed open expansion.

Second bullet implies first bullet
Let \left\{U_\alpha: \alpha<\kappa \right\} be an increasing open cover of X. Then \left\{F_\alpha: \alpha<\kappa \right\} is a decreasing family of closed subsets of X where F_\alpha=X-U_\alpha. Note that \bigcap_{\alpha<\kappa} F_\alpha=\varnothing. Let \left\{G_\alpha: \alpha<\kappa \right\} be a family of open subsets of X such that \bigcap_{\alpha<\kappa} G_\alpha=\varnothing and F_\alpha \subset G_\alpha for each \alpha. For each \alpha, there is open set W_\alpha such that F_\alpha \subset W_\alpha \subset \overline{W_\alpha} \subset G_\alpha since X is normal. For each \alpha, let V_\alpha=X-\overline{W_\alpha}. Then \left\{V_\alpha: \alpha<\kappa \right\} is a family of open subsets of X required by the first bullet. It is a cover because \bigcap_{\alpha<\kappa} \overline{W_\alpha}=\varnothing. To show \overline{V_\alpha} \subset U_\alpha, let x \in \overline{V_\alpha} such that x \notin U_\alpha. Then x \in W_\alpha. Since x \in \overline{V_\alpha} and W_\alpha is open, W_\alpha \cap V_\alpha \ne \varnothing. Let y \in W_\alpha \cap V_\alpha. Since y \in V_\alpha, y \notin \overline{W_\alpha}, which means y \notin W_\alpha, a contradiction. Thus \overline{V_\alpha} \subset U_\alpha.

Now we show that the three properties in Diagram 3 are equivalent.

Theorem 5
Let X be a normal space. Then the following implications hold.
\omega-shrinking \Longrightarrow Property \mathcal{B}(\omega) \Longrightarrow Property \mathcal{D}(\omega) \Longrightarrow \omega-shrinking

Proof of Theorem 5
\omega-shrinking \Longrightarrow Property \mathcal{B}(\omega)
Suppose that X is \omega-shrinking. By Dowker’s theorem, X \times (\omega+1) is a normal space. We can think of \omega+1 as a convergent sequence with \omega as the limit point. Let \left\{U_n:n=0,1,2,\cdots \right\} be an increasing open cover of X. Define H and K as follows:

    H=\cup \left\{(X-U_n) \times \left\{n \right\}: n=0,1,2,\cdots \right\}

    K=X \times \left\{\omega \right\}

It is straightforward to verify that H and K are disjoint closed subsets of X \times (\omega+1). By normality, let V and W be disjoint open subsets of X \times (\omega+1) such that H \subset W and K \subset V. For each integer n=0,1,2,\cdots, define V_n as follows:

    V_n=\left\{x \in X: \exists \ \text{open } O \subset X \text{ such that } x \in O \text{ and } O \times [n, \omega] \subset V \right\}

The set [n, \omega] consists of all integers \ge n and the limit point \omega. From the way the sets V_n are defined, \left\{V_n:n=0,1,2,\cdots \right\} is an increasing open cover of X. The remaining thing to show is that \overline{V_n} \subset U_n for each n. Suppose that x \in \overline{V_n} and x \notin U_n. Then (x,n) \in H by definition of H. There exists an open set E \times \left\{n \right\} such that (x,n) \in E \times \left\{n \right\} and (E \times \left\{n \right\}) \cap V=\varnothing. Since E is an open set containing x, E \cap V_n \ne \varnothing. Let y \in E \cap V_n. By definition of V_n, there is some open set O such that y \in O and O \times [n, \omega] \subset V, a contradiction since (E \cap O) \times \left\{n \right\} is supposed to miss V. Thus \overline{V_n} \subset U_n for all integers n.

The direction Property \mathcal{B}(\omega) \Longrightarrow Property \mathcal{D}(\omega) is immediate.

Property \mathcal{D}(\omega) \Longrightarrow \omega-shrinking
Consider the dual condition of \mathcal{D}(\omega) in Theorem 4, which is equivalent to \omega-shrinking according to Dowker’s theorem. \square

Remarks
The direction \omega-shrinking \Longrightarrow Property \mathcal{B}(\omega) is true because \omega-shrinking is equivalent to the normality in the product X \times (\omega+1). The same is not true when \kappa becomes an uncountable cardinal. We now show that \kappa-shrinking does not imply \mathcal{B}(\kappa) in general.

Example 1
The space X=\omega_1 is the set of all ordinals less than \omega_1 with the ordered topology. Since it is a linearly ordered space, it is a shrinking space. Thus in particular it is \omega_1-shrinking. To show that X does not have property \mathcal{B}(\omega_1), consider the increasing open cover \left\{U_\alpha: \alpha<\omega_1 \right\} where U_\alpha=[0,\alpha) for each \alpha<\omega_1. Here [0,\alpha) consists of all ordinals less than \alpha. Suppose X has property \mathcal{B}(\omega_1). Then let \left\{V_\alpha: \alpha<\omega_1 \right\} be an increasing open cover of X such that \overline{V_\alpha} \subset U_\alpha for each \alpha.

Let L be the set of all limit ordinals in X. For each \alpha \in L, \alpha \notin U_\alpha and thus \alpha \notin \overline{V_\alpha}. Thus there exists a countable ordinal f(\alpha)<\alpha such that (f(\alpha),\alpha] misses points in \overline{V_\alpha}. Thus the map f: L \rightarrow \omega_1 is a pressing down map. By the pressing down lemma, there exists some \alpha<\omega_1 such that S=f^{-1}(\alpha) is a stationary set in \omega_1, which means that S intersects with every closed and unbounded subset of X=\omega_1. This means that for each \gamma>\alpha, (\alpha, \gamma] would miss \overline{V_\gamma}. This means that for each \gamma>\alpha, \overline{V_\gamma} \subset [0,\alpha]. As a result \left\{V_\alpha: \alpha<\omega_1 \right\} would not be a cover of X, a contradiction. So X does not have property \mathcal{B}(\omega_1). \square

____________________________________________________________________

Property \mathcal{B}(\kappa)

Of the three properties discussed in the above section, we would like to single out property \mathcal{B}(\kappa). This property has a connection with normality in the product X \times Y (see Theorem 7). First, we prove a lemma that is used in proving Theorem 7.

Lemma 6
Show that the property \mathcal{B}(\kappa) is hereditary with respect to closed subsets.

Proof of Lemma 6
Let X be a space with property \mathcal{B}(\kappa). Let A be a closed subspace of X. Let \left\{U_\alpha \subset A: \alpha<\kappa \right\} be an increasing open cover of A. For each \alpha, let W_\alpha be an open subset of X such that U_\alpha=W_\alpha \cap A. Since the open sets U_\alpha are increasing, the open sets W_\alpha can be chosen inductively such that W_\alpha \supset W_\gamma for all \gamma<\alpha. This will ensure that W_\alpha will form an increasing cover.

Then \left\{W_\alpha^* \subset X: \alpha<\kappa \right\} is an increasing open cover of X where W_\alpha^*=W_\alpha \cup (X-A). By property \mathcal{B}(\kappa), let \left\{E_\alpha \subset X: \alpha<\kappa \right\} be an increasing open cover of X such that \overline{E_\alpha} \subset W_\alpha^*. For each \alpha, let V_\alpha=E_\alpha \cap A. It can be readily verified that \left\{V_\alpha \subset A: \alpha<\kappa \right\} is an increasing open cover of A. Furthermore, \overline{V_\alpha} \subset U_\alpha for each \alpha (closure taken in A). \square

Let \kappa be an infinite cardinal. Let D_\kappa=\left\{d_\alpha: \alpha<\kappa \right\} be a discrete space of cardinality \kappa. Let p be a point not in D_\kappa. Let Y_\kappa=D_\kappa \cup \left\{p \right\}. Define a topology on Y_\kappa by letting D_\kappa be discrete and by letting open neighborhood of p be of the form \left\{p \right\} \cup E where E \subset D_\kappa and D_\kappa-E has cardinality less than \kappa. Note the similarity between Y_\kappa and the convergent sequence \omega+1 in the proof of Theorem 5.

Theorem 7
Let X be a normal space. Then the product space X \times Y_\kappa is normal if and only if X has property \mathcal{B}(\kappa).

Remarks
The property \mathcal{B}(\kappa) involves the shrinking of any increasing open cover with the added property that the shrinking is also increasing. The increasing shrinking is just what is needed to show that disjoint closed subsets of the product space can be separated.

Notations
Let’s set some notations that are useful in proving Theorem 7.

  • The set [d_\alpha,p] is an open set in Y_\kappa containing the point p and is defined as follows.
    • [d_\alpha,p]=\left\{d_\beta: \alpha \le \beta<\kappa \right\} \cup \left\{p \right\}.
  • For any two disjoint closed subsets H and K of the product space X \times Y_\kappa, define the following sets.
    • For each \alpha<\kappa, let H_\alpha=H \cap (X \times \left\{d_\alpha \right\}) and K_\alpha=K \cap (X \times \left\{d_\alpha \right\}).
    • Let H_p=H \cap (X \times \left\{p \right\}) and K_p=K \cap (X \times \left\{p \right\}).
    • For each \alpha<\kappa, choose open O_\alpha \subset X such that G_\alpha=O_\alpha \times \left\{d_\alpha \right\}, H_\alpha \subset G_\alpha and \overline{G_\alpha} \cap K_\alpha=\varnothing (due to normality of X).
    • Choose open O_p \subset X such that G_p=O_p \times \left\{p \right\}, H_p \subset G_p and \overline{G_p} \cap K_p=\varnothing (due to normality of X).

Proof of Theorem 7
Suppose that X has property \mathcal{B}(\kappa). Let H and K be two disjoint closed sets of X \times Y_\kappa. Consider the following cases based on the locations of the closed sets H and K.

    Case 1. H \subset X \times D_\kappa and K \subset X \times D_\kappa.
    Case 2a. H=X \times \left\{p\right\}
    Case 2b. Exactly one of H and K intersect the set X \times \left\{p\right\}.
    Case 3. Both H and K intersect the set X \times \left\{p\right\}.

Remarks
Case 1 is easy. Case 2a is the pivotal case. Case 2b and Case 3 use a similar idea. The result in Theorem 7 is found in [1] (Theorem 6.9 in p. 189) and [4]. The authors in these two sources claimed that Case 2a is the only case that matters, citing a lemma in another source. The lemma was not stated in these two sources and the source for the lemma is a PhD dissertation that is not readily available. Case 3 essentially uses the same idea but it has enough differences. For the sake of completeness, we work out all the cases. Case 3 applies property \mathcal{B}(\kappa) twice. Despite the complicated notations, the essential idea is quite simple. If any reader finds the proof too long, just understand Case 2a and then get the gist of how the idea is applied in Case 2b and Case 3.

Case 1.
H \subset X \times D_\kappa and K \subset X \times D_\kappa.

Let M =\bigcup_{\alpha<\kappa} G_\alpha. It is clear that H \subset M and \overline{M} \cap K=\varnothing.

Case 2a.
Assume that H=X \times \left\{p\right\}. We now proceed to separate H and K with disjoint open sets. For each \alpha<\kappa, define U_\alpha as follows:

    U_\alpha=\cup \left\{O \subset X: O \text{ is open such that } (O \times [d_\alpha,p]) \cap K =\varnothing \right\}

Then \left\{U_\alpha: \alpha<\kappa \right\} is an increasing open cover of X. By property \mathcal{B}(\kappa), there is an increasing open cover \mathcal{V}=\left\{V_\alpha: \alpha<\kappa \right\} of X such that \overline{V_\alpha} \subset U_\alpha for each \alpha. The shrinking \mathcal{V} allows us to define an open set G such that H \subset G and \overline{G} \cap K=\varnothing.

Let G=\cup \left\{V_\alpha \times [d_\alpha,p]: \alpha<\kappa \right\}. It is clear that H \subset G. Next, we show that \overline{G} \cap K=\varnothing. Suppose that (x,d_\alpha) \in K. Then (x,d_\alpha) \notin U_\alpha \times [d_\alpha,p]. As a result, (x,d_\alpha) \notin \overline{V_\alpha} \times [d_\alpha,p]. Let O \subset X be open such that x \in O and (O \times \left\{d_\alpha \right\}) \cap (\overline{V_\alpha} \times [d_\alpha,p])=\varnothing. Since V_\beta \subset V_\alpha for all \beta<\alpha, it follows that (O \times \left\{d_\alpha \right\}) \cap (V_\beta \times [d_\beta,p])=\varnothing for all \beta < \alpha. It is clear that (O \times \left\{d_\alpha \right\}) \cap (V_\gamma \times [d_\gamma,p])=\varnothing for all \gamma>\alpha. What has been shown is that there is an open set containing the point (x,d_\alpha) that contains no point of G. This means that (x,d_\alpha) \notin \overline{G}. We have established that \overline{G} \cap K=\varnothing.

Case 2b.
Exactly one of H and K intersect the set X \times \left\{p\right\}. We assume that H is the set that intersects the set X \times \left\{p\right\}. The only difference between Case 2b and Case 2a is that there can be points of H outside of X \times \left\{p\right\} in Case 2b.

Now proceed as in Case 2a. Obtain the open cover \left\{U_\alpha: \alpha<\kappa \right\}, the open cover \left\{V_\alpha: \alpha<\kappa \right\} and the open set G as in Case 2a. Let M=G \cup (\bigcup_{\alpha<\kappa} G_\alpha). It is clear that H \subset M. We claim that \overline{M} \cap K=\varnothing. Suppose that (x,d_\gamma) \in K. Since \overline{G} \cap K=\varnothing (as in Case 2a), there exists open set W=O \times \left\{ d_\gamma \right\} such that (x,d_\gamma) \in W and W \cap \overline{G}=\varnothing. There also exists open W_1 \subset W such that (x,d_\gamma) \in W_1 and W_1 \cap \overline{G_\gamma}=\varnothing. It is clear that W_1 \cap G_\beta=\varnothing for all \beta \ne \gamma. This means that W_1 is an open set containing the point (x,d_\gamma) such that W_1 misses the open set M. Thus \overline{M} \cap K=\varnothing.

Case 3.
Both H and K intersect the set X \times \left\{p\right\}.

Now project H_p and K_p onto the space X.

    H_p^*=\left\{x \in X: (x,p) \in H_p \right\}

    K_p^*=\left\{x \in X: (x,p) \in K_p \right\}

Note that H_p^* is simply the copy of H_p and K_p^* is the copy of K_p in X. Since X is normal, choose disjoint open sets E_1 and E_1 such that H_p^* \subset E_1 and K_p^* \subset E_2.

Let A_1=\overline{E_1} and B_1=X-K_p^*. Let A_2=\overline{E_2} and B_2=X-H_p^*. Note that A_1 is closed in X, B_1 is open in X and A_1 \subset B_1. Similarly A_2 is closed in X, B_2 is open in X and A_2 \subset B_2.

We now define two increasing open covers using property \mathcal{B}(\kappa). Define U_{\alpha,1} and T_{\alpha,1} and U_{\alpha,2} and T_{\alpha,2} as follows:

    U_{\alpha,1}=\cup \left\{O \subset B_1: O \text{ is open such that } (O \times [d_\alpha,p]) \cap K =\varnothing \right\}

    T_{\alpha,1}=U_{\alpha,1} \cap A_1

    U_{\alpha,2}=\cup \left\{O \subset B_2: O \text{ is open such that } (O \times [d_\alpha,p]) \cap H =\varnothing \right\}

    T_{\alpha,2}=U_{\alpha,2} \cap A_2

The open cover \mathcal{T}_1=\left\{T_{\alpha,1}: \alpha<\kappa \right\} is an increasing open cover of A_1. The open cover \mathcal{T}_2=\left\{T_{\alpha,2}: \alpha<\kappa \right\} is an increasing open cover of A_2.By property \mathcal{B}(\kappa) of A_1 and A_2, both covers have the following as shrinking (by Lemma 6). The two shrinkings are:

    \mathcal{V}_1=\left\{V_{\alpha,1} \subset A_1: \alpha<\kappa \right\}

    \mathcal{V}_2=\left\{V_{\alpha,2} \subset A_2: \alpha<\kappa \right\}

such that

    \overline{V_{\alpha,1}} \subset T_{\alpha,1}

    \overline{V_{\alpha,2}} \subset T_{\alpha,2}

for each \alpha<\kappa and such that both \mathcal{V}_1 and \mathcal{V}_2 are increasing open covers. Note that the closure \overline{V_{\alpha,1}} is taken in A_1 and the closure \overline{V_{\alpha,2}} is taken in A_2.

For each \alpha, let W_{\alpha,1} be the interior of V_{\alpha,1} and W_{\alpha,2} be the interior of V_{\alpha,2} (with respect to X). Note that W_{\alpha,1} is meaningful since V_{\alpha,1} is a subset of the closure of the open set E_1. Similar observation for W_{\alpha,2}. To make the rest of the argument easier to see, note the following fact about W_{\alpha,1} and W_{\alpha,2}.

    \overline{W_{\alpha,1}} \subset \overline{V_{\alpha,1}} \subset T_{\alpha,1} \subset U_{\alpha,1} (closure with respect to X)

    \overline{W_{\alpha,2}} \subset \overline{V_{\alpha,2}} \subset T_{\alpha,2} \subset U_{\alpha,2} (closure with respect to X)

For each \alpha<\kappa, choose open set O_\alpha \subset X such that

    L_\alpha=O_\alpha \times \left\{d_\alpha \right\}

    H_\alpha \subset L_\alpha

    \overline{L_\alpha} \cap K_\alpha=\varnothing

    L_\alpha \cap (\overline{W_{\alpha,2}} \times [d_\alpha,p])=\varnothing

The last point is possible because U_{\alpha,2} \times [d_\alpha,p] misses H and \overline{W_{\alpha,2}}  \subset U_{\alpha,2}. Define the open sets G and M as follows:

    G=\cup \left\{W_{\alpha,1} \times [d_\alpha,p]: \alpha<\kappa \right\}

    M=G \cup (\bigcup_{\alpha<\kappa} L_\alpha)

It is clear that H \subset M. We claim that \overline{M} \cap K=\varnothing. To this end, we show that if (x,y) \in K, then (x,y) \notin \overline{M}. If (x,y) \in K, then either (x,y)=(x,d_\gamma) for some \gamma or (x,y)=(x,p).

Let (x,d_\gamma) \in K. Note that (x,d_\gamma) \notin U_{\gamma,1} \times [d_\gamma,p]. Since \overline{W_{\gamma,1}} \subset \overline{V_{\gamma,1}} \subset T_{\gamma,1} \subset U_{\gamma,1}, (x,d_\gamma) \notin \overline{W_{\gamma,1}} \times [d_\gamma,p]. Choose an open set O \subset X such that x \in O and C=O \times \left\{d_\gamma \right\} misses \overline{W_{\gamma,1}} \times [d_\gamma,p]. Note that C misses W_{\beta,1} \times [d_\beta,p] for all \beta<\gamma since W_{\beta,1} \subset W_{\gamma,1} for all \beta<\gamma. It is clear that C misses W_{\beta,1} \times [d_\beta,p] for all \beta>\gamma.

We can also choose open C_1 \subset C such that (x,d_\gamma) \in C_1 and C_1 misses \overline{L_\gamma}. It is clear that C_1 misses L_\beta for all \beta \ne \gamma. Thus there is an open set C_1 containing the point (x,d_\gamma) such that C_1 contains no point of M.

Let (x,p) \in K. First we find an open set Q containing (x,p) such that Q misses G. From the way the open sets U_{\alpha,1} are defined, it follows that (x,p) \notin \overline{W_{\alpha,1}} \times [d_\alpha,p] for all \alpha. Furthermore W_{\alpha,1} \subset \overline{A_1}. Thus Q=(X-\overline{A_1}) \times Y_\kappa is the desired open set. On the other hand, there exists \alpha<\kappa such that x \in W_{\alpha,2}. Note that L_\gamma are chosen so that (W_{\gamma,2} \times [d_\gamma,p]) \cap L_\gamma=\varnothing for all \gamma. Since W_{\alpha,2} \subset W_{\beta,2} for all \beta \ge \alpha, (W_{\alpha,2} \times [d_\alpha,p]) \cap L_\beta=\varnothing for all \beta \ge \alpha. Thus the open set W_{\alpha,2} \times [d_\alpha,p] contains no points of L_\gamma for any \gamma. Then the open set Q \cap (W_{\alpha,2} \times [d_\alpha,p]) contains no point of M. This means that (x,p) \notin \overline{M}. Thus \overline{M} \cap K=\varnothing.

In each of the four cases (1, 2a, 2b and 3), there exists an open set M \subset X \times Y_\kappa such that H \subset M and \overline{M} \cap K=\varnothing. This completes the proof that X \times Y_\kappa is normal assuming that X has property \mathcal{B}(\kappa).

Now the other direction. Suppose that X \times Y_\kappa is normal. Then it can be shown that X has property \mathcal{B}(\kappa). The proof is similar to the proof for \omega-shrinking \Longrightarrow Property \mathcal{B}(\omega) in Theorem 5. \square

____________________________________________________________________

Reference

  1. Morita K., Nagata J.,Topics in General Topology, Elsevier Science Publishers, B. V., The Netherlands, 1989.
  2. Rudin M. E., A Normal Space X for which X \times I is not Normal, Fund. Math., 73, 179-486, 1971. (link)
  3. Rudin M. E., Dowker Spaces, Handbook of Set-Theoretic Topology (K. Kunen and J. E. Vaughan, eds), Elsevier Science Publishers B. V., Amsterdam, (1984) 761-780.
  4. Yasui Y., On the Characterization of the \mathcal{B}-Property by the Normality of Product Spaces, Topology and its Applications, 15, 323-326, 1983. (abstract and paper)
  5. Yasui Y., Some Characterization of a \mathcal{B}-Property, TSUKUBA J. MATH., 10, No. 2, 243-247, 1986.

____________________________________________________________________
\copyright \ 2017 \text{ by Dan Ma}

Countably paracompact spaces

This post is a basic discussion on countably paracompact space. A space is a paracompact space if every open cover has a locally finite open refinement. The definition can be tweaked by saying that only open covers of size not more than a certain cardinal number \tau can have a locally finite open refinement (any space with this property is called a \tau-paracompact space). The focus here is that the open covers of interest are countable in size. Specifically, a space is a countably paracompact space if every countable open cover has a locally finite open refinement. Even though the property appears to be weaker than paracompact spaces, the notion of countably paracompactness is important in general topology. This post discusses basic properties of such spaces. All spaces under consideration are Hausdorff.

Basic discussion of paracompact spaces and their Cartesian products are discussed in these two posts (here and here).

A related notion is that of metacompactness. A space is a metacompact space if every open cover has a point-finite open refinement. For a given open cover, any locally finite refinement is a point-finite refinement. Thus paracompactness implies metacompactness. The countable version of metacompactness is also interesting. A space is countably metacompact if every countable open cover has a point-finite open refinement. In fact, for any normal space, the space is countably paracompact if and only of it is countably metacompact (see Corollary 2 below).

____________________________________________________________________

Normal Countably Paracompact Spaces

A good place to begin is to look at countably paracompactness along with normality. In 1951, C. H. Dowker characterized countably paracompactness in the class of normal spaces.

Theorem 1 (Dowker’s Theorem)
Let X be a normal space. The following conditions are equivalent.

  1. The space X is countably paracompact.
  2. Every countable open cover of X has a point-finite open refinement.
  3. If \left\{U_n: n=1,2,3,\cdots \right\} is an open cover of X, there exists an open refinement \left\{V_n: n=1,2,3,\cdots \right\} such that \overline{V_n} \subset U_n for each n.
  4. The product space X \times Y is normal for any compact metric space Y.
  5. The product space X \times [0,1] is normal where [0,1] is the closed unit interval with the usual Euclidean topology.
  6. For each sequence \left\{A_n \subset X: n=1,2,3,\cdots \right\} of closed subsets of X such that A_1 \supset A_2 \supset A_3 \supset \cdots and \cap_n A_n=\varnothing, there exist open sets B_1,B_2,B_3,\cdots such that A_n \subset B_n for each n such that \cap_n B_n=\varnothing.

Dowker’s Theorem is proved in this previous post. Condition 2 in the above formulation of the Dowker’s theorem is not in the Dowker’s theorem in the previous post. In the proof for 1 \rightarrow 2 in the previous post is essentially 1 \rightarrow 2 \rightarrow 3 for Theorem 1 above. As a result, we have the following.

Corollary 2
Let X be a normal space. Then X is countably paracompact if and only of X is countably metacompact.

Theorem 1 indicates that normal countably paracompact spaces are important for the discussion of normality in product spaces. As a result of this theorem, we know that normal countably paracompact spaces are productively normal with compact metric spaces. The Cartesian product of normal spaces with compact spaces can be non-normal (an example is found here). When the normal factor is countably paracompact and the compact factor is upgraded to a metric space, the product is always normal. The connection with normality in products is further demonstrated by the following corollary of Theorem 1.

Corollary 3
Let X be a normal space. Let Y be a non-discrete metric space. If X \times Y is normal, then X is countably paracompact.

Since Y is non-discrete, there is a non-trivial convergent sequence (i.e. the sequence represents infinitely many points). Then the sequence along with the limit point is a compact metric subspace of Y. Let’s call this subspace S. Then X \times S is a closed subspace of the normal X \times Y. As a result, X \times S is normal. By Theorem 1, X is countably paracompact.

C. H. Dowker in 1951 raised the question: is every normal space countably paracompact? Put it in another way, is the product of a normal space and the unit interval always a normal space? As a result of Theorem 1, any normal space that is not countably paracompact is called a Dowker space. The search for a Dowker space took about 20 years. In 1955, M. E. Rudin showed that a Dowker space can be constructed from assuming a Souslin line. In the mid 1960s, the existence of a Souslin line was shown to be independent of the usual axioms of set theorey (ZFC). Thus the existence of a Dowker space was known to be consistent with ZFC. In 1971, Rudin constructed a Dowker space in ZFC. Rudin’s Dowker space has large cardinality and is pathological in many ways. Zoltan Balogh constructed a small Dowker space (cardinality continuum) in 1996. Various Dowker space with nicer properties have also been constructed using extra set theory axioms. The first ZFC Dowker space constructed by Rudin is found in [2]. An in-depth discussion of Dowker spaces is found in [3]. Other references on Dowker spaces is found in [4].

Since Dowker spaces are rare and are difficult to come by, we can employ a “probabilistic” argument. For example, any “concrete” normal space (i.e. normality can be shown without using extra set theory axioms) is likely to be countably paracompact. Thus any space that is normal and not paracompact is likely countably paracompact (if the fact of being normal and not paracompact is established in ZFC). Indeed, any well known ZFC example of normal and not paracompact must be countably paracompact. In the long search for Dowker spaces, researchers must have checked all the well known examples! This probability thinking is not meant to be a proof that a given normal space is countably paracompact. It is just a way to suggest a possible answer. In fact, a good exercise is to pick a normal and non-paracompact space and show that it is countably paracompact.

____________________________________________________________________

Some Examples

The following lists out a few classes of spaces that are always countably paracompact.

  • Metric spaces are countably paracompact.
  • Paracompact spaces are countably paracompact.
  • Compact spaces are countably paracompact.
  • Countably compact spaces are countably paracompact.
  • Perfectly normal spaces are countably paracompact.
  • Normal Moore spaces are countably paracompact.
  • Linearly ordered spaces are countably paracompact.
  • Shrinking spaces are countably paracompact.

The first four bullet points are clear. Metric spaces are paracompact. It is clear from definition that paracompact spaces, compact and countably compact spaces are countably paracompact. One way to show perfect normal spaces are countably paracompact is to show that they satisfy condition 6 in Theorem 1 (shown here). Any Moore space is perfect (closed sets are G_\delta). Thus normal Moore space are perfectly normal and hence countably paracompact. The proof of the countably paracompactness of linearly ordered spaces can be found in [1]. See Theorem 5 and Corollary 6 below for the proof of the last bullet point.

As suggested by the probability thinking in the last section, we now look at examples of countably paracompact spaces among spaces that are “normal and not paracompact”. The first uncountable ordinal \omega_1 is normal and not paracompact. But it is countably compact and is thus countably paracompact.

Example 1
Any \Sigma-product of uncountably many metric spaces is normal and countably paracompact.

For each \alpha<\omega_1, let X_\alpha be a metric space that has at least two points. Assume that each X_\alpha has a point that is labeled 0. Consider the following subspace of the product space \prod_{\alpha<\omega_1} X_\alpha.

    \displaystyle \Sigma_{\alpha<\omega_1} X_\alpha =\left\{f \in \prod_{\alpha<\omega_1} X_\alpha: \ f(\alpha) \ne 0 \text{ for at most countably many } \alpha \right\}

The space \Sigma_{\alpha<\omega_1} X_\alpha is said to be the \Sigma-product of the spaces X_\alpha. It is well known that the \Sigma-product of metric spaces is normal, in fact collectionwise normal (this previous post has a proof that \Sigma-product of separable metric spaces is collectionwise normal). On the other hand, any \Sigma-product always contains \omega_1 as a closed subset as long as there are uncountably many factors and each factor has at least two points (see the lemma in this previous post). Thus any such \Sigma-product, including the one being discussed, cannot be paracompact.

Next we show that T=(\Sigma_{\alpha<\omega_1} X_\alpha) \times [0,1] is normal. The space T can be reformulated as a \Sigma-product of metric spaces and is thus normal. Note that T=\Sigma_{\alpha<\omega_1} Y_\alpha where Y_0=[0,1], for any n with 1 \le n<\omega, Y_n=X_{n-1} and for any \alpha with \alpha>\omega, Y_\alpha=X_\alpha. Thus T is normal since it is the \Sigma-product of metric spaces. By Theorem 1, the space \Sigma_{\alpha<\omega_1} X_\alpha is countably paracompact. \square

Example 2
Let \tau be any uncountable cardinal number. Let D_\tau be the discrete space of cardinality \tau. Let L_\tau be the one-point Lindelofication of D_\tau. This means that L_\tau=D_\tau \cup \left\{\infty \right\} where \infty is a point not in D_\tau. In the topology for L_\tau, points in D_\tau are isolated as before and open neighborhoods at \infty are of the form L_\tau - C where C is any countable subset of D_\tau. Now consider C_p(L_\tau), the space of real-valued continuous functions defined on L_\tau endowed with the pointwise convergence topology. The space C_p(L_\tau) is normal and not Lindelof, hence not paracompact (discussed here). The space C_p(L_\tau) is also homeomorphic to a \Sigma-product of \tau many copies of the real lines. By the same discussion in Example 1, C_p(L_\tau) is countably paracompact. For the purpose at hand, Example 2 is similar to Example 1. \square

Example 3
Consider R. H. Bing’s example G, which is a classic example of a normal and not collectionwise normal space. It is also countably paracompact. This previous post shows that Bing’s Example G is countably metacompact. By Corollary 2, it is countably paracompact. \square

Based on the “probabilistic” reasoning discussed at the end of the last section (based on the idea that Dowker spaces are rare), “normal countably paracompact and not paracompact” should be in plentiful supply. The above three examples are a small demonstration of this phenomenon.

Existence of Dowker spaces shows that normality by itself does not imply countably paracompactness. On the other hand, paracompact implies countably paracompact. Is there some intermediate property that always implies countably paracompactness? We point that even though collectionwise normality is intermediate between paracompactness and normality, it is not sufficiently strong to imply countably paracompactness. In fact, the Dowker space constructed by Rudin in 1971 is collectionwise normal.

____________________________________________________________________

More on Countably Paracompactness

Without assuming normality, the following is a characterization of countably paracompact spaces.

Theorem 4
Let X be a topological space. Then the space X is countably paracompact if and only of the following condition holds.

  • For any decreasing sequence \left\{A_n: n=1,2,3,\cdots \right\} of closed subsets of X such that \cap_n A_n=\varnothing, there exists a decreasing sequence \left\{B_n: n=1,2,3,\cdots \right\} of open subsets of X such that A_n \subset B_n for each n and \cap_n \overline{B_n}=\varnothing.

Proof of Theorem 4
Suppose that X is countably paracompact. Suppose that \left\{A_n: n=1,2,3,\cdots \right\} is a decreasing sequence of closed subsets of X as in the condition in the theorem. Then \mathcal{U}=\left\{X-A_n: n=1,2,3,\cdots \right\} is an open cover of X. Let \mathcal{V} be a locally finite open refinement of \mathcal{U}. For each n=1,2,3,\cdots, define the following:

    B_n=\cup \left\{V \in \mathcal{V}: V \cap A_n \ne \varnothing  \right\}

It is clear that A_n \subset B_n for each n. The open sets B_n are decreasing, i.e. B_1 \supset B_2 \supset \cdots since the closed sets A_n are decreasing. To show that \cap_n \overline{B_n}=\varnothing, let x \in X. The goal is to find B_j such that x \notin \overline{B_j}. Once B_j is found, we will obtain an open set V such that x \in V and V contains no points of B_j.

Since \mathcal{V} is locally finite, there exists an open set V such that x \in V and V meets only finitely many sets in \mathcal{V}. Suppose that these finitely many open sets in \mathcal{V} are V_1,V_2,\cdots,V_m. Observe that for each i=1,2,\cdots,m, there is some j(i) such that V_i \cap A_{j(i)}=\varnothing (i.e. V_i \subset X-A_{j(i)}). This follows from the fact that \mathcal{V} is a refinement \mathcal{U}. Let j be the maximum of all j(i) where i=1,2,\cdots,m. Then V_i \cap A_{j}=\varnothing for all i=1,2,\cdots,m. It follows that the open set V contains no points of B_j. Thus x \notin \overline{B_j}.

For the other direction, suppose that the space X satisfies the condition given in the theorem. Let \mathcal{U}=\left\{U_n: n=1,2,3,\cdots \right\} be an open cover of X. For each n, define A_n as follows:

    A_n=X-U_1 \cup U_2 \cup \cdots \cup U_n

Then the closed sets A_n form a decreasing sequence of closed sets with empty intersection. Let B_n be decreasing open sets such that \bigcap_{i=1}^\infty \overline{B_i}=\varnothing and A_n \subset B_n for each n. Let C_n=X-B_n for each n. Then C_n \subset \cup_{j=1}^n U_j. Define V_1=U_1. For each n \ge 2, define V_n=U_n-\bigcup_{j=1}^{n-1}C_{j}. Clearly each V_n is open and V_n \subset U_n. It is straightforward to verify that \mathcal{V}=\left\{V_n: n=1,2,3,\cdots \right\} is a cover of X.

We claim that \mathcal{V} is locally finite in X. Let x \in X. Choose the least n such that x \notin \overline{B_n}. Choose an open set O such that x \in O and O \cap \overline{B_n}=\varnothing. Then O \cap B_n=\varnothing and O \subset C_n. This means that O \cap V_k=\varnothing for all k \ge n+1. Thus the open cover \mathcal{V} is a locally finite refinement of \mathcal{U}. \square

___________________________________

We present another characterization of countably paracompact spaces that involves the notion of shrinkable open covers. An open cover \mathcal{U} of a space X is said to be shrinkable if there exists an open cover \mathcal{V}=\left\{V(U): U \in \mathcal{U} \right\} of the space X such that for each U \in \mathcal{U}, \overline{V(U)} \subset U. If \mathcal{U} is shrinkable by \mathcal{V}, then we also say that \mathcal{V} is a shrinking of \mathcal{U}. Note that Theorem 1 involves a shrinking. Condition 3 in Theorem 1 (Dowker’s Theorem) can rephrased as: every countable open cover of X has a shrinking. This for any normal countably paracompact space, every countable open cover has a shrinking (or is shrinkable).

A space X is a shrinking space if every open cover of X is shrinkable. Every shrinking space is a normal space. This follows from this lemma: A space X is normal if and only if every point-finite open cover of X is shrinkable (see here for a proof). With this lemma, it follows that every shrinking space is normal. The converse is not true. To see this we first show that any shrinking space is countably paracompact. Since any Dowker space is a normal space that is not countably paracompact, any Dowker space is an example of a normal space that is not a shrinking space. To show that any shrinking space is countably paracompact, we first prove the following characterization of countably paracompactness.

Theorem 5
Let X be a space. Then X is countably paracompact if and only of every countable increasing open cover of X is shrinkable.

Proof of Theorem 5
Suppose that X is countably paracompact. Let \mathcal{U}=\left\{U_1,U_2,U_3,\cdots \right\} be an increasing open cover of X. Then there exists a locally open refinement \mathcal{V}_0 of \mathcal{U}. For each n, define V_n=\cup \left\{O \in \mathcal{V}_0: O \subset U_n \right\}. Then \mathcal{V}=\left\{V_1,V_2,V_3,\cdots \right\} is also a locally finite refinement of \mathcal{U}. For each n, define

    G_n=\cup \left\{O \subset X: O \text{ is open and } \forall \ m > n, O \cap V_m=\varnothing \right\}

Let \mathcal{G}=\left\{G_n: n=1,2,3,\cdots \right\}. It follows that G_n \subset G_m if n<m. Then \mathcal{G} is an increasing open cover of X. Observe that for each n, \overline{G_n} \cap V_m=\varnothing for all m > n. Then we have the following:

    \displaystyle \begin{aligned} \overline{G_n}&\subset X-\cup \left\{V_m: m > n \right\} \\&\subset \cup \left\{V_k: k=1,2,\cdots,n \right\} \\&\subset \cup \left\{U_k: k=1,2,\cdots,n \right\}=U_n  \end{aligned}

We have just established that \mathcal{G} is a shrinking of \mathcal{U}, or that \mathcal{U} is shrinkable.

For the other direction, to show that X is countably paracompact, we show that the condition in Theorem 4 is satisfied. Let \left\{A_1,A_2,A_3,\cdots \right\} be a decreasing sequence of closed subsets of X with empty intersection. Then \mathcal{U}=\left\{U_1,U_2,U_3,\cdots \right\} be an open cover of X where U_n=X-A_n for each n. By assumption, \mathcal{U} is shrinkable. Let \mathcal{V}=\left\{V_1,V_2,V_3,\cdots \right\} be a shrinking. We can assume that \mathcal{V} is an increasing sequence of open sets.

For each n, let B_n=X-\overline{V_n}. We claim that \left\{B_1,B_2,B_3,\cdots \right\} is a decreasing sequence of open sets that expand the closed sets A_n and that \bigcap_{n=1}^\infty \overline{B_n}=\varnothing. The expansion part follows from the following:

    A_n=X-U_n \subset X-\overline{V_n}=B_n

The part about decreasing follows from:

    B_{n+1}=X-\overline{V_{n+1}} \subset X-\overline{V_n}=B_n

We show that \bigcap_{n=1}^\infty \overline{B_n}=\varnothing. To this end, let x \in X. Then x \in V_n for some n. We claim that x \notin \overline{B_n}. Suppose x \in \overline{B_n}. Since V_n is an open set containing x, V_n must contain a point of B_n, say y. Since y \in B_n, y \notin \overline{V_n}. This in turns means that y \notin V_n, a contradiction. Thus we have x \notin \overline{B_n} as claimed. We have established that every point of X is not in \overline{B_n} for some n. Thus the intersection of all the \overline{B_n} must be empty. We have established the condition in Theorem 4 is satisfied. Thus X is countably paracompact. \square

Corollary 6
If X is a shrinking space, then X is countably paracompact.

____________________________________________________________________

Reference

  1. Ball, B. J., Countable Paracompactness in Linearly Ordered Spaces, Proc. Amer. Math. Soc., 5, 190-192, 1954. (link)
  2. Rudin, M. E., A Normal Space X for which X \times I is not Normal, Fund. Math., 73, 179-486, 1971. (link)
  3. Rudin, M. E., Dowker Spaces, Handbook of Set-Theoretic Topology (K. Kunen and J. E. Vaughan, eds), Elsevier Science Publishers B. V., Amsterdam, (1984) 761-780.
  4. Wikipedia Entry on Dowker Spaces (link)

____________________________________________________________________
\copyright \ 2016 \text{ by Dan Ma}

A stroll in Bing’s Example G

In this post we take a leisurely walk in Bing’s Example G, which is a classic example of a normal but not collectionwise normal space. Hopefully anyone who is new to this topological space can come away with an intuitive feel and further learn about it. Indeed this is a famous space that had been extensively studied. This example has been written about in several posts in this topology blog. In this post, we explain how Example G is defined, focusing on intuitive idea as much as possible. Of course, the intuitive idea is solely the perspective of the author. Any reader who is interested in building his/her own intuition on this example can skip this post and go straight to the previous introduction. Other blog posts on various subspaces of Example G are here, here and here. Bing’s Example H is discussed here.

At the end of the post, we will demonstrate that the product of Bing’s Example G with the closed unit interval, F \times [0,1], is a normal space.

____________________________________________________________________

The Product Space Angle

The topology in Example G is tweaked from the product space topology. It is thus a good idea to first examine the relevant product space. Let P be any uncountable set. Let Q be the set of all subsets of P. In other words, Q is the power set of P. Consider the product of \lvert Q \lvert many copies of the two element set \left\{0,1 \right\}. The usual notation of this product space is 2^Q. The elements of 2^Q are simply the functions from Q into \left\{0,1 \right\}. An arbitrary element of 2^Q is a function f that maps every subset of P to either 0 or 1.

Though the base set P can be any uncountable set, it is a good idea to visualize clearly what P is. In the remainder of this section, think of P as the real line \mathbb{R}. Then Q is simply the collection of all subsets of the real line. The elements of the product space are simply functions that map each set of real numbers to either 0 or 1. Or think of each function as a 2-color labeling of the subsets of the real line, where each subset is either red or green for example. There are 2^c many subsets of the real line where c is the cardinality of the continuum.

To further visualize the product space, let’s look at a particular subspace of 2^Q. For each real number p, define the function f_p such that f_p always maps any set of real numbers that contains p to 1 and maps any set of real numbers that does not contain p to 0. For example, the following are several values of the function f_0.

    f_0([0,1])=1

    f_0([1,2])=0

    f_0(\left\{0 \right\})=1

    f_0(\mathbb{R}-\left\{0 \right\})=0

    f_0(\mathbb{R})=1

    f_0(\varnothing)=0

    f_0(\mathbb{P})=0

where \mathbb{P} is the set of all irrational numbers. Consider the subspace F_P=\left\{f_p: p \in P \right\}. Members of F_P are easy to describe. Each function in F_P maps a subset of the real line to 0 or 1 depending on whether the subscript belongs to the given subset. Another reason that F_P is important is that Bing’s Example is defined by declaring all points not in F_P isolated points and by allowing all points in F_P retaining the open sets in the product topology.

Any point f in F_P determines f(q)=0 \text{ or } 1 based on membership (whether the reference point belongs to the set q). Points not in F_P have no easy characterization. It seems that any set can be mapped to 0 or 1. Note that any f in F_P maps equally to 0 or 1. So the constant functions f(q)=0 and f(q)=1 are not in F_P. Furthermore, any f such that f(q)=1 for at most countably many q would not be in F_P.

Let’s continue focusing on the product space for the time being. When F_P is considered as a subspace of the product space 2^Q, F_P is a discrete space. For each p \in P, there is an open set W_p containing f_p such that W_p contains no other points of F_P. So F_P is relatively discrete in the product space 2^Q. Of course F_P cannot be closed in 2^Q since 2^Q is a compact space. The open set W_p is defined as follows:

    W_p=\left\{f \in 2^Q: f(\left\{p \right\})=1 \text{ and } f(P-\left\{p \right\})=0 \right\}

It is clear that f_p \in W_p and that f_t \notin W_p for any real number t \ne p.

Two properties of the product space 2^Q would be very relevant for the discussion. By the well known Tychonoff theorem, the product space 2^Q is compact. Since P is uncountable, 2^Q always has the countable chain condition (CCC) since it is the product of separable spaces. A space having CCC means that there can only be at most countably many pairwise disjoint open sets. As a result, the uncountably many open sets W_p cannot be all pairwise disjoint. So there exist at least a pair of W_p, say W_{a} and W_{b}, with nonempty intersection.

The last observation can be generalized. For each p \in P, let V_p be any open set containing f_p (open in the product topology). We observe that there are at least two a and b from P such that V_a \cap V_b \ne \varnothing. If there are only countably many distinct sets V_p, then there are uncountably many V_p that are identical and the observation is valid. So assume that there are uncountably many distinct V_p. By the CCC in the product space, there are at least two a and b with V_a \cap V_b \ne \varnothing. This observation shows that the discrete points in F_P cannot be separated by disjoint open sets. This means that Bing’s Example G is not collectionwise Hausdorff and hence not collectionwise normal.

Another observation is that any disjoint A_1, A_2 \subset F_P can be separated by disjoint open sets. To see this, define the following two open sets E_1 and E_2 in the product topology.

    q_1=\left\{p \in P: f_p \in A_1 \right\}

    q_2=\left\{p \in P: f_p \in A_2 \right\}

    E_1=\left\{f \in 2^Q: f(q_1)=1 \text{ and } f(q_2)=0 \right\}

    E_2=\left\{f \in 2^Q: f(q_1)=0 \text{ and } f(q_2)=1 \right\}

It is clear that A_1 \subset E_1 and A_2 \subset E_2. Furthermore, E_1 \cap E_2=\varnothing. This observation will be the basis for showing that Bing’s Example G is normal.

____________________________________________________________________

The Topology of Bing’s Example G

The topology for Bing’s Example G is obtained by tweaking the product topology on 2^Q. Let P be any uncountable set. Let Q be the set of all subsets of P. The set F_P is defined as above. Bing’s Example G is F=2^Q with points in F_P retaining the open sets in the product topology and with points not in F_P declared isolated. For some reason, in Bing’s original paper, the notation F is used even though the example is identified by G. We will follow Bing’s notation.

The subspace F_P is discrete but not closed in the product topology. However, F_P is both discrete and closed in Bing’s Example G. Based on the discussion in the previous section, one immediate conclusion we can made is that the space F is not collectionwise Hausdorff. This follows from the fact that points in the uncountable closed and discrete set F_P cannot be separated by disjoint open sets. By declaring points not in F_P isolated, the countable chain condition in the original product space 2^Q is destroyed. However, there is still a strong trace of CCC around the points in the set F_P, which is sufficient to prevent collectionwise Hausdorffness, and consequently collectionwise normality.

To show that F is normal, let H and K be disjoint closed subsets of F. To make it easy to follow, let H=A_1 \cup B_1 and K=A_2 \cup B_2 where

    A_1=H \cap F_P \ \ \ \ B_1=H \cap (F-F_P)

    A_2=K \cap F_P \ \ \ \ B_2=K \cap (F-F_P)

In other words, A is the non-isolated part and B is the isolated part of the respective closed set. Based on the observation made in the previous section, obtain the disjoint open sets E_1 and E_2 where A_1 \subset E_1 and A_2 \subset E_2. Set the following open sets.

    O_1=(E_1 \cup B_1) - K

    O_2=(E_2 \cup B_2) - H

It follows that O_1 and O_2 are disjoint open sets and that A_1 \subset O_1 and A_2 \subset O_2. Thus Bing’s Example G is a normal space.

____________________________________________________________________

Bing’s Example G is Countably Paracompact

We discuss one more property of Bing’s Example G. A space X is countably paracompact if every countable countable open cover of X has a locally finite open refinement. In other words, such a space satisfies the property of being a paracompact space but just for countable open covers. A space is countably metacompact if every countable open cover has a point-finite open refinement (i.e. replacing locally finite in the paracompact definition with point-finite). It is well known that in the class of normal spaces, the two notions are equivalent (see Corollary 2 here). Since Bing’s Example G is normal, we only need to show that it is countably metacompact. Note that Bing’s Example G is not metacompact (see here).

Let \mathcal{U} be a countable open cover of F. Let \mathcal{U}^*=\left\{U_1,U_2,U_3,\cdots \right\} be the set of all open sets in \mathcal{U} that contain points in F_P. For each i, let A_i=U_i \cap F_P. From the perspective of Bing’s Example G, the sets A_i are discrete closed sets. In any normal space, countably many discrete closed sets can be separated by disjoint open sets (see Lemma 1 here). Let O_1,O_2,O_3,\cdots be disjoint open sets such that A_i \subset O_i for each i.

We now build a point-finite open refinement of \mathcal{U}. For each i, let V_i=U_i \cap O_i. Let V=\cup_{i=1}^\infty V_i. Consider the following.

    \mathcal{V}=\left\{V_i: i=1,2,3,\cdots \right\} \cup \left\{\left\{ x \right\}: x \in F-V \right\}

It follows that \mathcal{V} is an open cover of F. All points of F_P belong to the open sets V_i. Any point that is not in one of the V_i belongs to a singleton open set. It is also clear that \mathcal{V} is a refinement of \mathcal{U}. For each i, V_i \subset U_i and each singleton set is contained in some member of \mathcal{U}. It follows that each point in F belongs to at most finitely many sets in \mathcal{V}. In fact, each point belongs to exactly one set in \mathcal{V}. Each point in F_P belongs to exactly one V_i since the open sets O_i are disjoint. Any point in V belongs to exactly one singleton open set. What we just show is slightly stronger than countably metacompact. The technical term would be countably 1-bounded metacompact.

Since among normal spaces, countably paracompactness is equivalent to countably metacompact, we can now say that Bing’s Example G is a topological space that is normal and countably paracompact. By Dowker’s Theorem, we can conclude that the product of Bing’s Example G with the closed unit interval, F \times [0,1], is a normal space.

____________________________________________________________________

Previous Posts

____________________________________________________________________
\copyright \ 2016 \text{ by Dan Ma}