I am working through an old analysis qualifier question, and I was hoping someone could help me verify or correct my proof. I would also appreciate any feedback for how I could strengthen my proof, and for alternative ways of approaching the problem. Thank you in advance for any help!

Theorem: Let $X$ be a sequentially compact metric space, and let $Y \subset X$. Let $f : X \to Y$ be an isometry between $X$ and $Y$. Recall that this means that $f$ is a distance preserving homeomorphism from $X$ onto $Y$. Thus, for each $a, b \in X$, $d(a, b) = d(f(a), f(b))$. Show that $Y = X$.

Proof: Suppose to the contrary that $X \neq Y$. As $f$ is an isometry, it is continuous. So as $f$ is continuous and $X$ is compact, it follows that $f(X)$ is connected and compact. In particular, $f(X)$ is closed, so $X \setminus f(X)$ is open. Let $a \in X \setminus f(X)$ s.t. there exists $y \in f(X)$ s.t. $d(a, y) > \text{diam}(f(X))$, and let $r > 0$ s.t. $B_{r}(a) \subset (X \setminus f(X))$ (such an $r$ exists as $X \setminus f(X)$ is open).

In particular, we have $\overline{B_{r/2}(a)} \subset B_{r}(a)$. As $\overline{B_{r/2}(a)}$ is a closed subset of a compact space, it is compact. So by the Weierstrass Extreme Value Theorem, let:

$$m := \max_{x \in \overline{B_{r/2}(a)}, y \in f(X)} d(x, y)$$

By construction, $m > \text{diam}(f(X))$. Let $x \in \overline{B_{r/2}(a)}, y \in f(X)$ achieving $d(x, y) = m$. By isometry, $d(f(x), f(y)) = m$. So $m \leq \text{diam}(f(X))$, a contradiction. QED.

In my proof, am I correct in selecting:

Let $a \in X \setminus f(X)$ s.t. there exists $y \in f(X)$ s.t. $d(a, y) > \text{diam}(f(X))$

If so, how can I justify such an $a$ exists? If not, is there a way I can correct this hole (as well as the others) in my proof?

Thank you in advance for any help!

  • 1
    $\begingroup$ Why is $f(X)$ connected? Is $X$ connected? $\endgroup$ – ajotatxe Dec 28 '16 at 21:28
  • 1
    $\begingroup$ The existence of $a$ is just wishful thinking on your part, there is no argument leading in that direction. What you probably want is this kind of argument. Also, sequentially compact = compact for metric spaces, so it's unnecessary to say "sequential". $\endgroup$ – user357151 Dec 28 '16 at 21:30
  • $\begingroup$ @ajotatxe- That's a good point. There was no assumption that $X$ was connected. The finite union of disjoint intervals in $\mathbb{R}$ would be a good counterexample here. zaq- I realize. The problem statement included sequential compactness though. Thanks for the link! I'll check it out. $\endgroup$ – ml0105 Dec 28 '16 at 21:30
  • 1
    $\begingroup$ Very short argument at math.stackexchange.com/questions/170989/… $\endgroup$ – user399601 Dec 28 '16 at 21:39
  • $\begingroup$ I'm not trying to show that $f$ is onto- this is given. I'm trying to show $X = Y$. Edit: That is a nice argument! I see now how it applies here. Thank you for sharing it! $\endgroup$ – ml0105 Dec 28 '16 at 21:40
  1. Since you have not assumed $X$ is connected, it does not follow from continuity that $f(X)$ is connected.

  2. The way you choose your point $a\in X\setminus f(X)$ needs clarification (and probably no such $a$ can be proved to exist). It is not obvious that you can choose $a$ such that there is $y\in f(X)$ with $d(a,y) > \text{diam}f(X)$. Imagine, for example, that somehow you had $X = [0,3]$ and $f(X) = [0,1]\cup [2,3]$. Then $\text{diam} f(X) = 3$. But for any $a\notin [0,1] \cup [2,3]$, the largest distance to any point in $[0,1]\cup[2,3]$ is $2$. Since at this point you have not provided any further properties of $f(X)$, it is unclear how to conclude this magical $a$ exists.

  3. You really are being asked to prove that $f$ is surjective as a map $X\to X$. For if $f$ is surjective as a map $X\to X$, then $Y = f(X) = X$, and conversely if $Y = X$, then $f(X) = X$ so $f:X\to X$ is surjective. user399601's link provides a clean proof.

  • $\begingroup$ Thank you for your help! Your explanation along with user399601's link really helped to clear things up. $\endgroup$ – ml0105 Dec 28 '16 at 21:54

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.