5
$\begingroup$

This might be a duplicate but I tried googling the MSE site and could not find a satisfactory answer.

Let $(X, d)$ be a metric space and $f$ be a real valued continuous function on $X$. Suppose $f$ has a compact support. Does this imply the uniform continuity of $f$?

I tried proving this statement, but only for locally connected spaces have I succeeded in doing so. Is thus true for general metric spaces? I just couldn't provide a proof (or a counterexample) by myself. Please enlighten me.

$\endgroup$
  • 1
    $\begingroup$ Isn't it a direct application of Heine theorem? (which is easier than locally connected spaces btw) $\endgroup$ – md5 Jan 6 '17 at 13:52
  • $\begingroup$ @md5 : Could briefly explain how to choose $\epsilon$ for the whole space? Not just for the points in $\operatorname{supp}(f)$? $\endgroup$ – Dilemian Jan 6 '17 at 14:02
  • $\begingroup$ I added it as an answer. You may need an other argument of compacity to get information at the frontier of the support. $\endgroup$ – md5 Jan 6 '17 at 14:41
3
$\begingroup$

Let $f:X\to\mathbf{R}$ satisfying your hypotheses. There exists $K$ compact s.t. $f_{|X \setminus K}=0$.

Let's show $f$ is uniformly continuous. Let $\epsilon>0$. By Heine theorem applied to $f_{|K}$, $f$ is uniformly continuous on $K$:

$$\exists \delta_1>0 \; \forall (x, y) \in K^2 \; d(x,y) < \delta_1 \implies |f(x)-f(y)| < \epsilon$$

If $(x, y) \in (X\setminus K)^2$, $\delta_1$ works.

Now suppose by contradiction that there exists $\epsilon>0$ s.t. for all $n \in \mathbf{N}$, there exists $x_n \in K$, $y_n \in X\setminus K$ s.t. $d(x_n, y_n) < 2^{-n}$ and $|f(x_n) - f(y_n)| > \epsilon$. $(x_n)$ lies in a compact space, so there exists $\sigma$ strictly increasing and $x \in K$ s.t. $x_{\sigma(n)} \to x$. Then $y_{\sigma(n)} \to x$, but $|f(x_{\sigma(n)}) -f(y_{\sigma(n)})| > \epsilon$, thus $f$ is not continuous in $x$, contradiction.

Hence:

$$\exists \delta_2>0 \; \forall (x, y) \in K \times (X \setminus K) \; d(x,y) < \delta_2 \implies |f(x)-f(y)| < \epsilon$$

Finally $\delta = \min(\delta_1, \delta_2)$ works.

$\endgroup$
0
$\begingroup$

Let $X_1=X$ \ $f^{-1}\{0\}.$

(1). For $r>0$ there exists $s>0$ such that $\forall x\in \overline {X_1}\;(|f(x)|\geq r\implies B_d(x,s)\subset \overline {X_1}).$

If not, then for $n\in \mathbb N,$ take $x_n\in \overline {X_1}$ and $y_n\in X$ \ $\overline {X_1}$ with $|f(x_n)|\geq r$ and $d(x_n,y_n)<1/n.$ By compactness of $\overline {X_1}$, take a convergent subsequence $(x_{n_i})_i$ converging to $p\in \overline {X_1}.$ Then $(f(x_{n_i}))_i$ converges to $f(p),$ so $|f(p)|\geq r.$ But $(y_{n_i})_i$ also converges to $p,$ and $f(y_{n_i})=0,$ contradicting the continuity of $f.$

(2). For $r>0,$ take $s>0$ such that $$\cup \{B_d(x,s): |f(x)|\geq r \} \subset \overline {X_1}.$$ Now $f|_{\overline {X_1}}$ is uniformly continuous because $\overline {X_1}$ is compact. So Take $s'\in (0,s)$ such that $$\forall x,y\in \overline {X_1}\;(d(x,y)<s'\implies |f(x)-f(y)|<r).$$ Now if $x\in \overline {X_1}$ and $y\in X$ \ $\overline {X_1}$ with $d(x,y)<s'<s$ then $|f(x)|<r$, so $|f(x)-f(y)|=|f(y)-0|<r.$ And of course if $x,y \in X$ \ $X_1$ then $|f(x)-f(y)|=0<r.$

$\endgroup$
  • $\begingroup$ @rt6. Sorry for my first comment, which only makes sense if $X=\mathbb R$. I meant my answer to be the finishing part of yours, as it was too long for a comment. I saw your A is deleted, so I re-wrote my A. $\endgroup$ – DanielWainfleet Jan 6 '17 at 16:33

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.