Since $A$ is compact and $f$ is continuous, then $f(A)$ is compact. Suppose $f$ is not uniformly continuous. Then $\exists \epsilon >0.\forall \delta >0. \exists x,y\in A$ such that $d(x,y)<\delta$ and $d(f(x),f(y))>\epsilon$. Therefore we can choose some $x$ such that, for some specific choice of $\epsilon$, we have $d(x,y)<\delta$ and $d(f(x),f(y))>\epsilon$ for all $\delta >0$ and $y\in A$. Denote that $x$ as $x_0$

Let ${x_k}$ be a sequence with a subsequence converging to $x_0$ (that at least one such sequence exists is clear). Call this subsequence ${x_n}$. Since $f$ is continuous, then as $x_n \rightarrow x_0$ we have $f(x_n)\rightarrow f(x_0)$. Therefore, for every $\epsilon >0$ we can find some $\delta >0$ such that there exists some element $x_i$ of the sequence ${x_n}$ such that $d(x_i,d_0)<\delta$ and $df((x_i),f(x_0))$, in contradiction to our assumption that $f$ is not uniformly continuous.

Therefore, $f$ is uniformly continuous.

I'm not sure if this is a good proof, or if there are holes in it. I struggle quite a bit with uniform continuity; I don't really have a good handle on it. Please let me know if you think the proof is good, or how I can improve it.


  • 1
    $\begingroup$ It's strange to take an $\varepsilon$ s.t. $d(x,y)<\delta\implies d(f(x),f(y))>\varepsilon$. It's $\varepsilon>0$ such that $d(x,y)<\delta$ and $d(f(x),f(y))>\varepsilon$ (for certain $x,y\in A$). $\endgroup$ – Surb Nov 11 '15 at 16:11
  • $\begingroup$ I was trying to take the logical negation of the statement of uniform continuity but it looks like I messed it up a little. $\endgroup$ – Zelzy Nov 11 '15 at 16:27
  • 1
    $\begingroup$ Yes, there is a sequence converging to $x_0$. But for all you know, this sequence may be constant. $\endgroup$ – egreg Nov 11 '15 at 17:46

Suppose has given $\epsilon > 0$ has given. for $x\in X$ since $f$ is continous at $x$, Let $\delta_x$ $$ \exists \delta_x > 0 ~~\forall y \in \mathbb R: |x-y|< \delta_x \rightarrow |f(x)-f(y)| < \frac{\epsilon}{2} $$ we have $ X\subseteq \bigcup_{x\in X} B(x,\frac{\delta_{x}}2) $ by compactness of $X$ exists $N$ such that $ X\subseteq \bigcup_{i}^N$ $B(x_i,\frac{\delta_{x_i}}2) $. Let $\delta = \min_{i=1,\cdots,N} \frac {\delta_{x_i}} 2$

consider $x, y\in X$ such that $|x-y|<\delta$, there are $1\leq i, j\leq N$ such that $x\in,B(x_i,\frac{\delta_{x_i}}2) $ and $y\in B(x_j,\frac{\delta_{x_j}}2)$ thus $$d(x_i, y)\leq d(x_i, x)+d(x,y)\leq \frac {\delta_{x_i}} 2+\delta\leq \frac {\delta_{x_i}} 2+\frac {\delta_{x_i}} 2=\delta_{x_i}$$ thus $$|f(x)-f(y)|\leq |f(x)-f(x_i)|+|f(x_i)-f(y)|<\frac{\epsilon}{2}+\frac{\epsilon}{2}=\epsilon$$

  • 1
    $\begingroup$ Could you détails your proof ? I don't think that the conclusion of uniform continuous is that obvious... $\endgroup$ – Surb Nov 11 '15 at 16:49
  • 1
    $\begingroup$ Ok, I see now. You proved in fact that if $x,y$ are such that $|x-y|<\delta$ then they are in the same ball. So now it's correct. $\endgroup$ – Surb Nov 11 '15 at 17:54
  • 1
    $\begingroup$ So, the proof ultimately relies on the continuity of $f$ to choose small enough $\delta_x$ and the compactness of $A$ to choose a finite number of these sets and use the triangle inequality to show that the distance between any two points, if less than $\delta$, must be contained within one of the $\delta_x$ balls we have defined above, and therefore we have the desired $f(x)-f(y)<\epsilon$. I think I've got it. Thanks. $\endgroup$ – Zelzy Nov 11 '15 at 19:30
  • $\begingroup$ you are welcome :) $\endgroup$ – R.N Nov 11 '15 at 19:41

let $\varepsilon>0$ s.t. $$\forall n\in\mathbb N, \exists x_n,y_n\in A: d(x_n,y_n)<\frac{1}{n}\quad\text{and}\quad d(f(x_n),f(y_n))>\varepsilon.$$

By Heine-Borel, $A$ is closed and bounded (since $\mathbb R^n$ is metrizable). Then, by Bolzano-weierstrass $(x_n)$ has a subsequence $(x_{n_k})$ that converge, and since $A$ is closed, it's limit is in $A$. Let $\ell$ it's limit. Since $$d(y_{n_k},\ell)\leq d(x_{n_k},y_{n_k})+d(x_{n_k},\ell)\underset{k\to\infty }{\longrightarrow }0$$ and thus $(y_{n_k})$ converge also to $\ell$. By continuity, you have that $$\lim_{k\to\infty }d(f(x_{n_k}),f(y_{n_k}))=0$$ and this is the contradiction with the fact that $d(f(x_{n_k}),f(x_{n_k}))>\varepsilon$ for all $k$.

To be a more explicit

By continuity, there is a $N$ such that $$d(f(x_{n_K}),\ell)<\frac{\varepsilon}{2}\quad\text{and}\quad d(f(y_{n_N}),\ell)<\frac{\varepsilon}{2}$$
and thus $$d(f(x_{n_N}),f(y_{n_N}))<d(f(x_{n_N}),f(\ell))+d(f(y_{n_N}),f(\ell))$$

and thus, you have that $$d(f(x_{n_N}),f(y_{n_N}))<\varepsilon$$ and here is the contradiction.

  • $\begingroup$ Sorry, i wasn't saying that if $x_n$ is a sequence then it is clear it has a subsequence that converges to $x_0$. I was saying that it is clear that there exists some sequence $x_n$ with a subsequence converging to $x_0$ so we could look generally at sequences with that property. But I guess that's not good enough. $\endgroup$ – Zelzy Nov 11 '15 at 16:26
  • 1
    $\begingroup$ If $x_n$ is unspecified, you can't do a lot of things... But can you continue with my hint or not really ? $\endgroup$ – Surb Nov 11 '15 at 16:29
  • $\begingroup$ But isn't it enough to show that, if at least one sequence with a subsequence converging to $x_0$ gives you a contradiction, then we have a contradiction in general? $\endgroup$ – Zelzy Nov 11 '15 at 16:31
  • $\begingroup$ First, why is there a sequence with a subsequence converging ? For example, if you are in $\mathbb Q$, $x_n=(1+\frac{1}{n})^n$ has no subsequence that converge (and $x_n$ is bounded, increasing...). In $\mathbb R$, $x_n=n$ has no subsequence that converge (but this one is not bounded). But let suppose it. You took $x,y\in A$ such that $d(x,y)<\delta$ and $d(f(x),f(y))> \varepsilon$. Notice that $x$ and $y$ are specified. By convergence ox $x_n$, there is a $N$ s.t. $d(x_n,x_0)<\delta$ if $n>N$. But now, why $|f(x_n)-f(x_0)|>\varepsilon$ ? $x_n$ and $x_0$ are neither $x$ nor $y$, $\endgroup$ – Surb Nov 11 '15 at 16:37
  • 1
    $\begingroup$ Still not working. I have $d(f(x_n),f(y_n))>\varepsilon$ not $d(f(x_n),f(x_0))>\varepsilon$. So it doesn't work ! $\endgroup$ – Surb Nov 11 '15 at 16:45

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.