7

Let $(X,T)$ be a topological Hausdorff space. By $C_b(X)$ denote the continuous bounded function $f\colon X\to\mathbb{R}$, by $C_c(X)$ the continuous functions $f\colon X\to\mathbb{R}$ which have compact support $\text{supp}f=\overline{\left\{x\in X: f(x)\neq 0\right\}}$ and, finally, by $C_0(X)$ denote the continuous functions $f\colon X\to\mathbb{R}$ that vanish at infinity.

Show that the closure of $C_c(X)$ with respect to the Banachspace $(C_b(X),\lVert\cdot\rVert_{\infty})$ is $C_0(X)$, i.e. $$ \overline{C_c(X)}=C_0(X). $$

I think the inclusion "$\subseteq$" is the easier one.

Let $\varepsilon >0$ be arbitrary and let $(f_n)$ be a sequence in $C_c(X)$ that converges to $f$ w.r.t. to $\lVert\cdot\rVert_{\infty}$. This implies uniform convergence of $f_n$ to $f$. Hence, $f$ is continuous. Moreover, there exists some $N(\varepsilon)\in\mathbb{N}$ such that $$ \lvert f_n(x)-f(x)\rvert < \varepsilon~\forall n\geq N(\varepsilon) $$ for all $x\in X$. For all $x\notin\text{supp}f_{N(\varepsilon)}$, $f_{N(\varepsilon)}(x)=0$, hence $$ \lvert f(x)\rvert=\lvert f(x)-f_{N(\varepsilon)}\rvert <\varepsilon $$ for all $x\notin\text{supp}f_{N(\varepsilon)}$.

This means that $f\in C_0(X)$.


Now, I do not know exactly how to prove "$\supseteq$". I guess the general idea is the following. Let $f\in C_0(X)$. Now, construct a sequence $(\varphi_n)$ in $C_c(X)$ with $\varphi_n\to f$ w.r.t. the supremum norm $\lVert\cdot\rVert_{\infty}$.

Rhjg
  • 2,029
  • Let $\varphi_n = f\chi_{[-n,n]}$, then $$|f-\varphi_n|\infty = |f\chi{(-\infty,n)\ \cup\ (n,\infty)}|_\infty\stackrel{n\to\infty}\longrightarrow0.$$ – Math1000 Apr 05 '16 at 19:04
  • @Math1000: Why should $\varphi_n$ be continuous? – carmichael561 Apr 05 '16 at 19:06
  • @carmichael561 You mentioned before that it might be necessary to have local compactness. Looking here http://planetmath.org/sites/default/files/texpdf/41281.pdf (Corollary 2 which uses the Urysohn-Lemma that needs local compactness) seems to confirm your comment. I'll ask the author of the task if this might be assumed. – Rhjg Apr 05 '16 at 19:10
  • Sorry, $[-n,n]$ would be ${x\in X: |x|\leqslant n}$, but you are correct, that function would be continuous on the compact set, not on all of $X$. – Math1000 Apr 05 '16 at 19:11
  • 2
    @ Rhjg: That was my initial impression, but now I'm not sure. I would definitely ask. – carmichael561 Apr 05 '16 at 19:15
  • @Math1000 ${ x \in X : | x | \leqslant n }$ still makes no sense because $X$ is an arbitrary Hausdorff topological space, not a normed vector space. – Adayah May 31 '18 at 20:23

1 Answers1

8

I don't think you need to assume that $X$ is locally compact:

Let $f\in C_{0}\left(X\right)$ and $\varepsilon>0$. Note that $U:=\left\{ x\in X\,\mid\,\left|f\left(x\right)\right|>\frac{\varepsilon}{2}\right\} $ is open with closure $\overline{U}\subset\left\{ x\in X\,\mid\,\left|f\left(x\right)\right|\geq\frac{\varepsilon}{2}\right\} $, which is compact as a closed subset of a compact set. Furthermore, $K:=\left\{ x\in X\,|\,\left|f\left(x\right)\right|\geq\varepsilon\right\} $ is a compact subset of $U$, so that the Urysohn-Lemma (applied to the compact Hausdorff space $\overline{U}$, with $U\subset\overline{U}$ open and $K\subset U$ compact) yields $\varphi\in C(\overline{U})$ with $\varphi|_{K}\equiv1$, $0\leq\varphi\leq1$ and $\operatorname{supp}\varphi\subset U$, where the closure for the support is taken in $\overline{U}$.

Now, define $$ g:X\to\left[0,1\right],x\mapsto\begin{cases} 0, & x\in X\setminus U,\\ \varphi\left(x\right), & x\in\overline{U}. \end{cases} $$ Note that $X\setminus U$ and $\overline{U}$ are both closed, so that $g$ is continuous by the gluing lemma (note that $g$ is well-defined, since for $x\in\left(X\setminus U\right)\cap\overline{U}$, we have $\varphi\left(x\right)=0$ since $\operatorname{supp}\varphi\subset U$.

Thus, $g$ has compact support (in fact $\operatorname{supp}g=\operatorname{supp}\varphi$), so that the same is true of $\tilde{f}:=g\cdot f\in C_{c}\left(X\right)$. It is not hard to see $\left\Vert \tilde{f}-f\right\Vert _{\sup}\leq\varepsilon$.

PhoemueX
  • 35,087
  • I guess with 'gluing lemma' you mean this one: http://topospaces.subwiki.org/wiki/Gluing_lemma_for_closed_subsets ? I've never heard of it but it seems to be well-known and very important. – Rhjg Apr 06 '16 at 09:12
  • @Rhjg Munkres calls it the "pasting lemma." – Math1000 Apr 06 '16 at 15:00
  • 1
    @Math1000 Thanks, I found it here: https://en.wikipedia.org/wiki/Pasting_lemma -- it's a nice statement I did not know before.. – Rhjg Apr 06 '16 at 15:12
  • Why is $\left{ x\in X,\mid,\left|f\left(x\right)\right|\geq\frac{\varepsilon}{2}\right}$ compact? – user193319 Nov 29 '19 at 16:38
  • @PhoemueX I thought that $C_c$ was the collection of functions with compact support, whereas $C_0$ was the collection of functions which vanish at infinity. Vanishing at infinity doesn't imply compact support, right? – user193319 Nov 29 '19 at 21:30
  • @user193319: Yes, you are right, sorry about that. Still, vanishing at infinity means that the set you are asking about is compact, by definition. See here: https://en.m.wikipedia.org/wiki/Vanish_at_infinity – PhoemueX Nov 29 '19 at 23:04