Originally asked on MSE.

Let $X$ be a normal (or even metrizable) topological space and let $Y$ be a closed subset of $X$. Let $C(X)$ be the linear space of all continuous scalar functions on $X$ endowed with the compact-open topology. Consider a map $R:C(X)\to C(Y)$ defined by $Rf=f|_Y$. This map is obviously linear, and it is easy to see that it is continuous. By Tietze theorem it is also surjective. It seems that this is in fact a quotient map. However, what interests me is the following question:

Let $B$ be a closed convex balanced subset of $C(X)$. Is $RB$ closed in $C(Y)$?

Under additional assumption that $B$ is weakly compact, this is true, since then $B$ is compact in the pointwise topology, and so $RB$ is pointwise compact, and so pointwise closed, and so closed in $C(Y)$. I can also show that this is wrong if we don't require $Y$ to be closed in $X$.

  • $\begingroup$ many years ago I read a theorem in a paper that there is linear isometric embedding from $C(Y)$ to $C(X)$ which act by extension. that is: there is an isometric section for the map you consider. I really do not not remember that paper because I read it more than 10 years ago. but this obviously implies that the restriction map is closed. $\endgroup$ – Ali Taghavi Jul 1 '18 at 8:44
  • $\begingroup$ may be Michael selection lemma is helpful to construct an appropriate right inverse(section)/ $\endgroup$ – Ali Taghavi Jul 1 '18 at 8:57
  • 1
    $\begingroup$ @AliTaghavi do I understand you correctly that there is an isometry $S$ from $C(Y)$ into $C(X)$ such that $RS=Id$? I guess that $X$ is assumed to be compact in this statement. I don't see immediately why this implies that $R$ is a closed map.. $\endgroup$ – erz Jul 1 '18 at 8:59
  • $\begingroup$ I am sorry I was mistaken. my comment was inappropriate. $\endgroup$ – Ali Taghavi Jul 1 '18 at 9:38
  • 1
    $\begingroup$ @AliTaghavi: selection theorems do give a right inverse to any bounded linear surjective operator between Banach spaces, which is however in general not linear (thus also nowhere differentiable). In the present case, the extension is a bounded linear operator by construction. $\endgroup$ – Pietro Majer Jul 1 '18 at 13:10

The answer to the main question is negative:

Consider the compact subset $X=[0,1]\cup \{2\}$ of the real line and let $Y=\{2\}$ be a singleton in $X$. In the function space $C(X)$ consider the closed convex balanced subset $$B:=\{f\in C(X):\sup_{x\in [0,1]}|f(x)|\le 1,\;f(0)=0,\;f(2)=\int_0^1 f(t)dt\}.$$ It is easy to see that for the restriction operator $R:C(X)\to C(Y)$, the image $R(B)=\{f\in C(Y):|f(2)|<1\}$ is not closed in $C(Y)$.


Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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