I'm currently practicing how to approach problems in my own without mimicking solutions in textbooks. So, the problem I am solving is this: Let $f:[a,b]\to \Bbb{R}$ be continuous. Then $f$ is uniformly continuous

Here is what I have done:

Let $x_0\in[a,b]$ be arbitrary, such that for every $\{x_n\}\subset [a,b],\;\;x_n\to x_0$ as $n\to\infty.$ By the continuity of $f$ at $x_0$, $f(x_n)\to f(x_0)$ as $n\to\infty$ for some $f(x_0)\in \Bbb{R}$. Since $[a,b]$ is compact, i.e., closed and bounded, then for any $\{x_n\},\{y_n\}\subset [a,b]$ with $x_n\to x_0$ and $y_n\to x_0$ as $n\to\infty,$ we must have $f(x_n)\to f(x_0)$ and $f(y_n)\to f(x_0)$ as $n\to\infty$. So, let $\epsilon>0$ be given, then $\exists\;\delta=\delta(x_0,\epsilon)>0$ s.t. $|x_n-x_0|<\delta/2$ and $|y_n-x_0|<\delta/2 \implies|f(x_n)-f(x_0)|<\epsilon/2$ and $|f(y_n)-f(x_0)|<\epsilon/2$.

Now, $|x_n-y_n|=|x_n-x_0+x_0-y_n|\leq |x_n-x_0|+|y_n-x_0|<\delta/2+<\delta/2=\delta$.

Remark: This new $\delta$ only depends on $\epsilon.$

Thus, for every $\{x_n\},\{y_n\}\subset [a,b]$ with $|x_n-y_n|<\delta$, we have $$|f(x_n)-f(y_n)|=|f(x_n)-f(x_0)+f(x_0)-f(y_n)|$$ $$\leq|f(x_n)-f(x_0)|+|f(y_n)-f(x_0)|$$ $$<\epsilon/2+\epsilon/2=\epsilon$$

Hence, for any $\epsilon>0,\;\exists\;\delta=\delta(\epsilon)$ such that, for every $\{x_n\},\{y_n\}\subset [a,b]$ with $|x_n-y_n|<\delta$, we have $|f(x_n)-f(y_n)|<\epsilon$. So, $f$ is uniformly continuous.

So, can anyone help me debug this? What are my errors? Any better way of going about this? Kindly provide what I should have incorporated instead, if I was wrong!


It has several errors:

  • You wrote “such that for every $\{x_n\}\subset [a,b],\;\;x_n\to x_0$ as $n\to\infty.$” What this means is that every sequence of elements of $[a,b]$ converges to $x_0$. This is impossible.
  • You wrote “Since $[a,b]$ is compact, i.e., closed and bounded, then for any $\{x_n\},\{y_n\}\subset [a,b]$ with $x_n\to x_0$ and $y_n\to x_0$ as $n\to\infty,$ we must have $f(x_n)\to f(x_0)$ and $f(y_n)\to f(x_0)$ as $n\to\infty$.” Actually, the fact that $[a,b]$ is compact is irrelevant here. The only thing that matters is that $f$ is continuous.
  • The $\delta$ that you chose depends on $x_0$ too, obviously.
  • $\begingroup$ Hmm... Still learning though! So, what I'm I supposed to incorporate instead of those untrue statements? $\endgroup$ May 12 '18 at 10:15
  • $\begingroup$ @Mike You'll find a proof here. $\endgroup$ May 12 '18 at 10:26
  • $\begingroup$ @ Sir José Carlos Santos: Thanks! $\endgroup$ May 12 '18 at 10:41

The errors of the proposed proof (the choice of sequences $(x_n), (y_n)$ doesn't capture the premise of the sequential criterion for (i.e. two arbitrary sequences which satisfy $\lim\limits_{n\to\infty} x_n - y_n = 0$)) has been analysed by another answer (which I upvoted), so I'll focus on the last prompt.

Any better way of going about this?

The compactness of $[a,b]$, which you mention in your proposed proof, is a good way to condense the proof. This proof is classic and you can find it in many real analysis textbooks (e.g. in Royden's book).

  1. Use the continuity of $f$ to get an open cover $\{I_x\}_{x \in [a,b]}$ for $[a,b]$ so that $I_x = (x-\delta_x/2,x+\delta_x/2)$, where $\delta_x$ responds to the $\epsilon > 0$ challenge. (i.e. $\forall x \in [a,b] \forall \epsilon > 0 \exists \delta_x > 0\forall x' \in [a,b]: (|x'-x|<\delta_x \implies |f(x')-f(x)| < \epsilon$)
  2. Use Heine-Borel Theorem to obtain a finite subcover $\{I_{x_k}\}_{k\in\{1,\dots,n\}}$ of $\{I_x\}_{x \in [a,b]}$.
  3. Claim that $\delta = \frac12\min\limits_{k \in \{1,\dots,n\}} \delta_{x_k}$ responds to the $2\epsilon$ challenge.
    • Given $|x'-x| < \delta$. $x \in I_{x_k}$ for some $k$, so $|x-x_k|<\delta_{x_k}/2$ and $|f(x)-f(x_k)| < \epsilon$
    • It remains to show that $|f(x')-f(x_k)| < \epsilon$. Similarly, a sufficient condition is $|x'-x_k| < \delta_{x_k}$. To prove this, note that $|x'-x_k| \le |x'-x|+|x-x_k|<\delta+\delta_{x_k}/2 \le \delta_{x_k}$.
  • $\begingroup$ Thank you very much for the unique solution. It made me think really deep. Wow, this is great! $\endgroup$ May 12 '18 at 12:44

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.