Let $X$ and $Y$ be Tychonoff (i.e. completely regular Hausdorff) topological spaces and let $\varphi:X\to Y$ be a continuous surjection that also has a property that $\operatorname{int}\overline{\varphi(U)}\ne\varnothing$, for any open nonempty $U\subset X$ (this property is called skeletal, weakly open, almost open etc in various sources; equivalently, the preimage of any nowhere set is nowhere dense).

Let us consider the topology $\tau$ on $Y$ which is the strongest Tychonoff topology with respect to which $\varphi$ is continuous (this property for $\varphi$ as a map into $(Y,\tau)$ is called $\mathbb{R}$-quotient, which may be viewed as a quotient map in the category of Tychonoff spaces).

Is $\varphi$ still skeletal as a map from $X$ into $(Y,\tau)$?

This is true is $X$ is locally compact, so the counterexamples should be looked for outside of this class.


1 Answer 1


The answer to this question is negative. A suitable counterexample can be constructed as follows.

Let $Y=\mathbb R$ be the real line with the standard Euclidean topology. Let $\mathbb Q$ be the subspace of rational numbers in $\mathbb R$. Write $\mathbb R\setminus \mathbb Q$ as the union $\bigcup_{q\in \mathbb Q}X_q$ of pairwise disjoint dense sets in $\mathbb R$. Let $$X=\mathbb Q\oplus\bigoplus_{q\in \mathbb Q}(\{q\}\cup X_q)$$be the topological sum of the spaces $\mathbb Q$ and $\{q\}\cup X_q$ for $q\in\mathbb Q$.

Let $\varphi:X\to Y$ be the natural projection. It is clear that the spaces $X,Y$ are separable, metrizable (and hence Tychonoff) and the function $\varphi:X\to Y$ is skeletal.

On the other hand, in the $\mathbb R$-quotient topology $\tau$ on $Y=\mathbb R$, the set $\mathbb Q$ is closed and nowhere dense in $(Y,\tau)$, witnessing that the function $\varphi:X\to(Y,\tau)$ is not skeletal.

To show that $\mathbb Q$ is closed in $(Y,\tau)$, choose any point $y\in\mathbb R\setminus \mathbb Q$. Find $q\in\mathbb Q$ such that $y\in X_q$. Consider the function $f_q:Y\to\mathbb R$ defined by $f_q(x)=|x-q|$ if $x\in X_q$ and $f_q(x)=0$, otherwise.

Observe that the composition $f_q\circ \varphi:X\to\mathbb R$ is continuous, which implies that the set $X_q=\{x\in Y:f_q(x)>0\}$ is $\tau$-open, contain $y$, and does not intersect $\mathbb Q$. This completes the proof of the closedness of $\mathbb Q$.

Assuming that the closed set $\mathbb Q$ is not nowhere dense in $(Y,\tau)$, we can find a nonempty open set $U\subseteq\mathbb Q$. Choose any point $q\in U$ and observe that the preimage $\varphi^{-1}(U)$ is an open set in $X$ containing the point $q\in \{q\}\cup X_q$. Since $\{q\}$ is nowhere dense in $X_q$, the set $\varphi^{-1}(U)$ has nonempty intersection with the set $X_q$ and then $\emptyset \ne \varphi[X_q\cap \varphi^{-1}(U)]\subseteq U\cap X_q\subseteq U\setminus\mathbb Q$, which contradicts the choice of $U\subseteq\mathbb Q$.

  • $\begingroup$ Thank you! After the formula the second $\mathbb{Q}$ should be multiplied with $0$, right? $\endgroup$
    – erz
    Jan 29, 2022 at 21:44
  • $\begingroup$ Actually, why is $X$ Tychonoff? $\endgroup$
    – erz
    Jan 29, 2022 at 21:44
  • $\begingroup$ @erz Ups! Indeed, $X$ is not even regular. So, you accepted my answer too fast. I should think a bit more. Sorry for this gap. $\endgroup$ Jan 29, 2022 at 22:39
  • $\begingroup$ @erz Now I have corrected the example, so it is became Tychonoff. $\endgroup$ Jan 30, 2022 at 13:09
  • $\begingroup$ I don't think I understand your construction of $X$ and $\varphi$. If $X$ is a disjoint union of subsets of $\mathbb{R}$ which add up to $\mathbb{R}$, then $\varphi$ is the identity map, and so $(Y,\tau)=X$. So it is probably something else but i cannot surmise it from the solution $\endgroup$
    – erz
    Jan 31, 2022 at 5:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.