6

I know this question has been asked before but I feel like my approach to solving the problem is "different" (EDIT: turned out to be different because it's wrong!)

Since $[a,b]$ is closed and bounded we may conclude that, by the Heini-Borel theorem, $[a, b]$ must be compact. By definition of compactness, every sequence in $[a,b]$ must contain a subsequence that converges to a limit that is also in $[a,b]$. Consider any arbitrary increasing subsequence $\{x_n\}$ inside $[a,b]$. By the definition of compactness, we know that there exist some $c \in [a,b]$ such that $\{x_n\} \rightarrow c$.

Since $f$ satisfies the intermediate value property, we know that $f(c)$ exists and is within the range of $f$. One of the "characterizations of continuity" states that:

"For all $\{x_n\} \rightarrow c$, it follows that $f(x_n) → f(c).$".

How do I show that $f(x_n)$ converges to $f(c)$?

Thing is, I know that $f(x_n)$ has to converge to something since the range of $f$ is a compact set ( $[f(a), f(b)]$)… I'm just trying to show that $f(x_n) \rightarrow f(c)$! Any idea of how I can go about this?

talrefae
  • 254
  • 2
  • 7
  • Welcome to stackexchange. My first thoughts. You don't yet know that $f(x_n)$ converges to something, only that some subsequence converges to something. But you haven't yet used the fact that $f$ is increasing. – Ethan Bolker Oct 24 '17 at 23:21
  • You are missing not only the last piece.... You haven't started the proof. I can't find if this is asked before, if you can find it, try to read the answer and see if you can follow it. If you can't, point out where you are stuck. –  Oct 24 '17 at 23:25
  • Ethan: Just a shot in the dark here: couldn't we argue that $f(x_n)$ converges by the monotone convergence theorem? Since we assumed ${x_n}$ to be increasing, $f(x_n)$ must also be increasing. $f$ is surely bounded above by $f(b)$ -- thus, $f(x_n)$ must converge to something? – talrefae Oct 24 '17 at 23:39
  • @talrefae why could you assume the sequence is increasing? – adfriedman Oct 24 '17 at 23:43
  • @adfriedman because $x_n$ and $f(x)$ are both increasing. – Marcelo Fornet Oct 24 '17 at 23:48
  • 1
    You have to show this for all possible sequences ${x_n}$, not just increasing ones. – adfriedman Oct 24 '17 at 23:49
  • @adfriedman good point… Your comment makes me believe that this 'proof' is going nowhere… – talrefae Oct 25 '17 at 00:08

2 Answers2

1

Often it is better just to go back to the $\epsilon$-$\delta$ definitions. I will show that $f$ is right continuous first. Left continuity should be easy enough for you to show yourself. Let $\epsilon > 0$. For a fixed $x \in [a,b)$, let $\epsilon' = \min\{\epsilon, f(b)-f(x)\}$, then

$$f(x) < f(x)+\epsilon' \leq f(b)$$

By the intermediate value theorem there exists a $y'\in (x, b]$ so that $f(y')=f(x)+\epsilon'$, hence because $f$ is increasing, $f(x)<f(y)<f(y')=f(x)+\epsilon'$ for all $y\in (x,y')$.

In other words, take $\delta=y'-x$, then for all $y$ such that $y-x < \delta$, $$f(y)-f(x) < f(y')-f(x) =\epsilon' \leq \epsilon$$ and we are done.

Alternative Proof Another useful (and more general) characterization of continuity is that a function is continuous if and only if the preimage of any open interval is open. We may use this as follows:

Take any open interval $(y_1,y_2) \subsetneq [f(a),f(b)]$. Given that $f$ is monotone we know that it is one-to-one and therefore permits a unique inverse $f^{-1}$ on its range. Moreover the preimage must be a subset of the interval $\left(f^{-1}(y_1), f^{-1}(y_2)\right)$. By the intermediate value theorem every value in that interval is attained, so the preimage is the interval we just defined and is therefore open. Hence $f$ is continuous.

adfriedman
  • 3,641
0

Let $x\in[a,b]$ and $\epsilon\gt0$, if $f(x)+\epsilon\le f(b)$, then there is an $x_+$ so that $x\lt x_+\le b$ and $$ f(x)+\epsilon=f(x_+) $$ If $f(x)+\epsilon\gt f(b)$, then let $x_+=b$.

If $f(x)-\epsilon\ge f(a)$, then there is an $x_-$ so that $a\le x_-\lt x$ and $$ f(x)-\epsilon=f(x_-) $$ If $f(x)-\epsilon\lt f(a)$, then let $x_-=a$.

Let $$ \delta=\left\{\begin{array}{} x-x_-&\text{if }x_+=b\\ x_+-x&\text{if }x_-=a\\ \min(x_+-x,x-x_-)&\text{otherwise} \end{array}\right. $$ then for all $t\in[a,b]$ so that $|x-t|\le\delta$, we have $|f(t)-f(x)|\le\epsilon$.

robjohn
  • 345,667