7
$\begingroup$

This is a reference request. Below is a prelude to the main questions.

Let $n$ be a positive integer. Characterize all subsets $\Omega$ of $\mathbb{R}^n$ such that any continuous function $f:\Omega\to\mathbb{C}$ can be extended to a continuous function $F:\mathbb{R}^n\to\mathbb{C}$. Here, $\Omega$ inherits the subspace topology from $\mathbb{R}^n$, and of course, $\mathbb{R}^n$ and $\mathbb{C}$ are equipped with their standard topologies.

The answer to the question above is that $\Omega$ must be a closed subset of $\mathbb{R}^n$. A slightly more generalized question is given below.

Question I: What are all subsets $\Omega$ of a topological space $X$ such that any continuous function $f:\Omega\to \mathbb{C}$ can be extended continuously to $F:X\to\mathbb{C}$? As usual, $\Omega$ is equipped with the subspace topology inherited from $X$.

Of course, without any nice properties on $X$, it seems impossible to make a characterization of such subspaces $\Omega$. Hence, with some additional properties on the topology of $X$, the answer may be completely known.

I know that, if $X$ is normal, then all closed subsets $\Omega$ of $X$ have the required property, using the Tietze Extension Theorem. Is there an example of a normal topological space $X$ which contains a non-closed subset $\Omega$ such that any continuous function $f:\Omega\to\mathbb{C}$ has a continuous extension to $X$?

Question II: Even more generally, let $X$ and $Y$ be topological spaces. Find all extendable subsets $\Omega$ of $X$, namely, all subsets $\Omega\subseteq X$ such that any continuous function $f:\Omega\to Y$ can be extended continuously to $F:X\to Y$.

Of course, niceness constraints on $X$ and $Y$ may be needed for a complete characterization. For example, if $X$ is a metric space, $Y$ is a convex subset of a locally convex topological vector space, and $\Omega$ is a closed subspace of $X$, then any continuous function $f:\Omega\to Y$ can be extended to a continuous function $F:X\to Y$. See for example here (although there is no mentioning that closed sets are all such sets).

This paper discusses strongly good pairs, namely, pairs $(X,Y)$ of topological spaces such that every $\Omega\subseteq X$ is extendable. It also discusses good pairs, namely pairs $(X,Y)$ of topological spaces such that, for every $\Omega\subseteq X$ and for any continuous function $f:\Omega\to Y$, there exists a function $F:X\to Y$ extending $f$ such that $F$ is continuous at every point in $\Omega$. This gives me an idea for my final question.

Question III: Let $X$ and $Y$ be topological spaces. What are all weakly extendable subsets $\Omega$ of $X$, namely, subsets $\Omega\subseteq X$ such that, for any continuous function $f:\Omega\to Y$, there exists a function $F:X\to Y$ extending $f$ such that $F$ is continuous at every point in $\Omega$. We shall call such an extension $F$ a weakly continuous extension of $f$ (relative to $\Omega$).

In the case where $X=\mathbb{R}^n$ for some $n\in\mathbb{N}$ and $Y=\mathbb{C}$, it is known that any subset $\Omega$ of $X$ is weakly extendable. In fact, if $Y$ is a locally compact Hausdorff space and $X$ is metrizable, then $(X,Y)$ is a good pair. What are other resources where a full characterization of weakly extendable subsets is known?

$\endgroup$
  • 2
    $\begingroup$ For the first question consider $\Omega = \omega_1 \subseteq X = \omega_1 + 1$, as ordinals in the order topology. $\endgroup$ – Henno Brandsma Aug 21 '16 at 5:28
1
$\begingroup$

I have simple partial answers.

Question I. A subset $\Omega$ of $X$ is extendable iff $\Omega$ is $C$-embedded in $X$. The sufficiency can be proved by the coordinatewise extension, the necessity can be proved using that $\Bbb R$ is a retract of $\Bbb C$.

Henno Brandsma’s example is quite sharp, because if $X$ is a Tychonoff space in which each point is $G_\delta$-set then each $C$-embedded subset of $X$ is closed. Indeed, for each point $x_0\in\overline{\Omega}\setminus\Omega$ it is easy to construct a continuous function $g:X\to\Bbb [0,1]$ such that $g(x)=0$ iff $x=x_0$. Then a function $f(x)=(1/g(x))|_\Omega$ cannot be extended from the set $\Omega$ to the whole space $X$.

Each closed subset of a $T_1$ space $X$ is $C$-embedded iff the space $X$ is normal.

Question II. I have the following trivial remarks. Given a pair $X$, $Y$ of topological spaces I shall call a space $X$ $Y$-collapsing, provided any continuous function from $X$ to $Y$ is constant. For instance, there exists a regular $\Bbb R$-collapsing space. Any $Y$-collapsing subset of the space $X$ is extendable. Conversely, each extendable subset of a $Y$-collapsing space $X$ is $Y$-collapsing too. In particular, if all [linearly] connected subsets of $Y$ are one-point then each [linearly] connected subset of $X$ is extendable. Conversely, if $Y$ is not connected and $X$ is connected then each extendable subset of $X$ is connected too.

If $Y$ is metrizable then any closed subset of any metrizable space $X$ is extendable iff $Y$ is an absolute extensor.

Question III. If $Y$ is an absolute neighborhood extensor then any closed subset of any metrizable space $X$ is weakly extendable.

$\endgroup$

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.