In this paper we shall prove a “parallel” metrization theorem answering a question [1] of the Mathoverflow user116515. The question concerns parallel sets in metric spaces.

Two non-empty sets AB in a metric space (Xd) are called parallel if

$$\begin{aligned} d(a,B)=d(A,B)=d(A,b)\quad \hbox {for any } a\in A \hbox { and } b\in B. \end{aligned}$$

Here \(d(A,B)=\inf \{ d(a,b):a\in A,\, b\in B\}\) and for \(x\in X\). Observe that two closed parallel sets AB in a metric space are either disjoint or coincide.

Let \({\mathscr {C}}\) be a family of non-empty closed subsets of a topological space X. A metric d on X is defined to be \({\mathscr {C}}\)-parallel if any two sets \(A,B\in {\mathscr {C}}\) are parallel with respect to the metric d.

A family \({\mathscr {C}}\) of subsets of X is called a compact cover of X if \(X=\bigcup {\mathscr {C}}\) and each set \(C\in {\mathscr {C}}\) is compact.

FormalPara Problem

([1]) For which compact covers \({\mathscr {C}}\) of a topological space X the topology of X is generated by a \({\mathscr {C}}\)-parallel metric?

A metric generating the topology of a given topological space will be called admissible. A necessary condition for the existence of an admissible \({\mathscr {C}}\)-parallel metric is the upper and lower semicontinuity of the cover \({\mathscr {C}}\).

A family \({\mathscr {C}}\) of subsets of a topological space X is called

  • lower semicontinuous if for any open set \(U\subset X\) its \({\mathscr {C}}\)-star is open in X;

  • upper semicontinuous if for any closed set \(F\subset X\) its \({\mathscr {C}}\)-star is closed in X;

  • continuous if \({\mathscr {C}}\) is both lower and upper semicontinuous;

  • disjoint if any distinct sets \(A,B\in {\mathscr {C}}\) are disjoint.

The following theorem is the main result of the paper, answering Problem.

FormalPara Theorem

For a compact cover \({\mathscr {C}}\) of a metrizable topological space X the following conditions are equivalent:

  1. (i)

    the topology of X is generated by a \({\mathscr {C}}\)-parallel metric;

  2. (ii)

    the family \({\mathscr {C}}\) is disjoint and continuous.

FormalPara Proof

(i) \(\Rightarrow \) (ii). Assume that d is an admissible \({\mathscr {C}}\)-parallel metric on X. The disjointness of the cover \({\mathscr {C}}\) follows from the obvious observation that two closed parallel sets in a metric space are either disjoint or coincide.

To see that \({\mathscr {C}}\) is lower semicontinuous, fix any open set \(U\subset X\) and consider its \({\mathscr {C}}\)-star . To see that is open, take any point and find a set \(C\in {\mathscr {C}}\) such that \(s\in C\) and \(C\cap U\ne \varnothing \). Fix a point \(u\in U\cap C\) and find \(\varepsilon >0\) such that the \(\varepsilon \)-ball \(B(u;\varepsilon )=\{ x\in X:d(x,u)<\varepsilon \}\) is contained in U. We claim that . Indeed, for any \(x\in B(s;\varepsilon )\) we can find a set \(C_x\in {\mathscr {C}}\) containing x and conclude that \(d(C_x,u)=d(C_x,C)\leqslant d(x,s)<\varepsilon \) and hence \(C_x\cap U\ne \varnothing \) and .

To see that \({\mathscr {C}}\) is upper semicontinuous, fix any closed set \(F\subset X\) and consider its \({\mathscr {C}}\)-star . To see that is closed, take any point and find a set \(C\in {\mathscr {C}}\) such that \(s\in C\). It follows from that \(C\cap F=\varnothing \) and hence by the compactness of C. We claim that . Assuming the opposite, we can find a point and a set \(C_x\in {\mathscr {C}}\) such that \(x\in C_x\) and \(C_x\cap F\ne \varnothing \). Fix a point \(z\in C_x\cap F\) and observe that \(d(C,F)\leqslant d(C,z)=d(C,C_x)\leqslant d(s,x)<\varepsilon =d(C,F)\), which is a desired contradiction.

(ii) \(\Rightarrow \) (i). The proof of this implication is more difficult. Assume that \({\mathscr {C}}\) is disjoint and continuous. Fix any admissible metric \(\rho \leqslant 1\) on X.

Let \({\mathscr {U}}_0(C)=\{X\}\) for every \(C\in {\mathscr {C}}\).

FormalPara Claim

For every \(n\in {\mathbb {N}}\) and every \(C\in {\mathscr {C}}\) there exists a finite cover \({\mathscr {U}}_n(C)\) of C by open subsets of X such that

  1. (a)

    each set \(U\in {\mathscr {U}}_n(C)\) has \(\rho \)-diameter \(\leqslant 1/{2^n}\);

  2. (b)

    if a set \(A\in {\mathscr {C}}\) meets some set \(U\in {\mathscr {U}}_n(C)\), then \(A\subset \bigcup {\mathscr {U}}_n(C)\) and A meets each set \(U'\in {\mathscr {U}}_n(C)\).

FormalPara Proof

Using the paracompactness [2, 5.1.3] of the metrizable space X, choose a locally finite open cover \({\mathscr {V}}\) of X consisting of sets of \(\rho \)-diameter \(<1/{2^n}\).

For every compact set \(C\in {\mathscr {C}}\) consider the finite subfamily of the locally finite cover \({\mathscr {V}}\). Since the cover \({\mathscr {C}}\) is upper semicontinuous, the set is closed and disjoint with the set C. Since \({\mathscr {C}}\) is lower semicontinuous, for any open set \(V\in {\mathscr {V}}(C)\) the set is open and hence is an open neighborhood of C.

Put and observe that \({\mathscr {U}}_n\) satisfies condition (a).

Let us show that the cover \({\mathscr {U}}_n(C)\) satisfies condition (b). Assume that a set \(A\in {\mathscr {C}}\) meets some set \(U\in {\mathscr {U}}_n(C)\). First we show that \(A\subset \bigcup {\mathscr {U}}_n(C)\). Find a set \(V\in {\mathscr {V}}(C)\) such that \(U=W(C)\cap V\). It follows from \(\varnothing \ne A\cap U\subset A\cap W(C)\) that the set A meets W(C) and hence is contained in W(C) and is disjoint with \(F_C\). Hence

Next, take any set \(U'\in {\mathscr {U}}_n(C)\) and find a set \(V'\in {\mathscr {V}}(C)\) with \(U'=W(C)\cap V'\). The relation \(A\cap W(C)\cap V=A\cap U\ne \varnothing \) and the definition of the set \(W(C)\supset A\) imply that A intersects the set \(V'\in {\mathscr {V}}(C)\) and hence intersects the set \(U'=W(C)\cap V'\). \(\blacksquare \)

Given two points \(x,y\in X\) let

Adjust the function \(\delta \) to a pseudometric d letting

where the infimum is taken over all sequences \(x=x_0\), \(\dots \), \(x_m=y\). Condition (a) of Claim implies that \(\rho (x,y)\leqslant \delta (x,y)\) and hence \(\rho (x,y)\leqslant d(x,y)\) for any \(x,y\in X\). So, the pseudometric d is a metric on X such that the identity map \((X,d)\rightarrow (X,\rho )\) is continuous. To see that this map is a homeomorphism, take any point \(x\in X\) and \(\varepsilon >0\). Find \(n\in {\mathbb {N}}\) such that \(1/{2^n}<\varepsilon \) and choose a set \(C\in {\mathscr {C}}\) with \(x\in C\) and a set \(U\in {\mathscr {U}}_n(C)\) with \(x\in U\). Then for any \(y\in U\) we get \(d(y,x)\leqslant \delta (x,y)\leqslant 1/{2^n}<\varepsilon \), which means that the map \(X\rightarrow (X,d)\) is continuous.

Finally, let us prove that the metric d is \({\mathscr {C}}\)-parallel. Pick any two distinct compact sets \(A,B\in {\mathscr {C}}\). We need to show that \(d(a,B)=d(A,B)=d(A,b)\) for any \(a\in A\), \(b\in B\). Assuming that this inequality is not true, we conclude that either \(d(a,B)>d(A,B)>0\) or \(d(A,b)>d(A,B)>0\) for some \(a\in A\) and \(b\in B\).

First assume that \(d(a,B)>d(A,B)\) for some \(a\in A\). Choose points \(a'\in A\), \(b'\in B\) such that \(d(a',b')=d(A,B)<d(a,B)\). By the definition of the distance \(d(a',b')<d(a,B)\), there exists a chain \(a'=x'_0, x'_1,\dots ,x'_m=b'\) such that \(\sum _{ i=1}^{ m}\delta (x'_{i-1},x'_i)<d(a,B)\). We can assume that the points \(x'_0,\dots ,x'_m\) are pairwise distinct, so for every \(i\leqslant m\) there exist \(n_i\geqslant 0\) such that \(\delta (x'_{i-1},x'_i)= 1/{2^{n_i}}\) and hence \(x'_{i-1},x'_i\in U_i'\) for some \(C_i\in {\mathscr {C}}\) and \(U_i'\in {\mathscr {U}}_{n_i}(C_i)\). For every \(i\leqslant m\) let \(A_i\in {\mathscr {C}}\) be the unique set with \(x_i'\in A_i\). Then \(A_0=A\) and \(A_m=B\).

Using condition (b), we can inductively construct a sequence of points \(a=x_0,x_1,\dots ,x_m\in B\) such that for every positive \(i\leqslant m\) the point \(x_i\) belongs to \(A_i\) and the points \(x_{i-1},x_i\) belong to some set \(U_i\in {\mathscr {U}}_{n_i}(C_i)\). The chain \(a=x_0,x_1,\dots ,x_m\in A_m=B\) witnesses that

which is a desired contradiction. By analogy we can prove that the case \(d(A,B)<d(A,b)\) leads to a contradiction. \(\square \)