1
$\begingroup$

Let $X$ be a compact metrizable space and $Y$ a metrizable space. Denote by $C(X,Y)$ the space of continuous functions from $X$ to $Y$ with the topology induced by the uniform metric $$d_u(f,g)=sup_{x\in X}d(f(x),g(x)).$$ where $d$ is a compatible metric for $Y$. At pp. $24$ "Classical Descriptive Set Theory", Kechris states:

A simple compactness argument shows that this topology is indipendente of the choice of $d$.

The problem is that it is not clear to me at all how such a simple argument should work, at least looking at definitions.

My attempt (while the below answer was edited): Assume $d_1$, $d_2$ are compatible metrics on $Y$. Let $$B_{d_1}(f,\epsilon)=\{g\in C\mid sup_{x\in X} d_1(f(x),g(x))<\epsilon\} \supseteq\{g\in C\mid d_1(f(x),g(x))<\epsilon'\, \text{for every }x\in X\}=B(f,\epsilon')\supseteq\{g\in C\mid sup_{x\in X} d_1(f(x),g(x))<\frac{\epsilon}{2}\}=B_{d_1}(f,\frac{\epsilon}{2}).$$ if we choose $\frac{\epsilon}{2}<\epsilon'<\epsilon$. By compactness, as the family of open balls with radius $\frac{\epsilon}{2}$ is an open cover of $f(X)$, we can assume the condition holds at points $f(x_1),\dots,f(x_n)$. As $d_1$, $d_2$ are compatible, there is $\delta $ s.t. $B_{d_2}(f(x_i),\delta)\subseteq B_{d_1}(f(x_i),\frac{\epsilon}{2})$ for every $i=1,\dots,n$. Does this suffice to prove that for every $\epsilon$ there is $\delta$ s.t. $B_{d_2}(f,\delta)\subseteq B_{d_1}(f,\epsilon)$? If this is the case, the proof applies switching $d_1$, $d_2$.

$\endgroup$
2
$\begingroup$

Let $d'$ be another metric on $Y$ and $d_u'$ be the associated uniform metric. Fix $f\in C(X,Y)$ and $\epsilon>0$. For each $x\in X$, pick $\epsilon_x>0$ such that $d'(f(x),y)<\epsilon_x$ implies $d(f(x),y)<\epsilon$. Also pick a neighborhood $U_x$ of $x$ such that $d'(f(x),f(y))<\epsilon_x/2$ for all $y\in U_x$. By compactness, $X$ is covered by finitely many of these neighborhoods $U_{x_1},\dots,U_{x_n}$. Let $\epsilon'=\min(\epsilon_{x_1},\dots,\epsilon_{x_n})$.

Now suppose $g\in C(X,Y)$ is such that $d_u'(f,g)<\epsilon'/2$ and let $x\in X$. There is some $i$ such that $x\in U_{x_i}$. Since $x\in U_{x_i}$, $d'(f(x),f(x_i))<\epsilon_{x_i}/2<\epsilon_{x_i}$ and so $d(f(x),f(x_i))<\epsilon$. Also, $$d'(g(x),f(x_i))\leq d'(g(x),f(x))+d'(f(x),f(x_i))<\epsilon'/2+\epsilon_{x_i}/2\leq \epsilon_{x_i}$$ so also $d(g(x),f(x_i))<\epsilon$. Thus $$d(f(x),g(x))\leq d(f(x),f(x_i))+d(g(x),f(x_i))<2\epsilon.$$ That is, $d_u'(f,g)<\epsilon'/2$ implies $d_u(f,g)\leq 2\epsilon$. Since $f$ and $\epsilon>0$ were arbitrary, this means the $d_u'$-topology is finer than the $d_u$ topology, and swapping the roles of $d$ and $d'$ we conclude the two topologies are the same.


Alternatively, you can show that the topology induced by $d_u$ is the compact-open topology, that is the topology generated by sets of the form $\{f:f(K)\subseteq U)$ where $K\subseteq X$ is compact and $U\subseteq Y$ is open. This manifestly does not depend on the choice of a metric on $Y$. The hard part of the proof is to show a $d_u$-ball around $f$ contains a compact-open neighborhood of $f$. The idea is that you can cover $X$ by finitely many closed balls $K_i$ such that $f(K_i)$ has small diameter, and let $U_i$ be a small open neighborhood of $f(K_i)$. Then if $g$ is such that $g(K_i)\subseteq U_i$ for all $i$, $d_u(f,g)$ must be small since each $U_i$ has small diameter.

$\endgroup$
  • $\begingroup$ I read your very plain proof and everything is clear. While you wrote your answer, I was editing an attempt. I have only one request: could you check if the above argument is correct, please? Thank you $\endgroup$ – LBJFS Apr 5 at 8:21
  • 1
    $\begingroup$ Well, it's quite obviously not a complete proof since you haven't explained why $B_{d_2}(f,\delta)\subseteq B_{d_1}(f,\epsilon)$. And I don't think the argument can work: you need to use the continuity of $f$ somewhere. $\endgroup$ – Eric Wofsey Apr 5 at 15:24
  • $\begingroup$ Ok, I thank you for your explanations :) $\endgroup$ – LBJFS Apr 5 at 15:28

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.