3

I have a continuous function $f$ that is defined on a compact set. And $f(x_0)$ is its minimum. If I have a sequence $x_n$ such that $f(x_n)\to f(x_0)$, how can I show that $x_n\to x_0$? I tried proving by contradiction but I got lost. Anyone could help me? Thank you very much.

Update:$f$ only achieves minimum at $x_0$.

1 Answers1

1

With the additional assumption the answer is yes. Assume the contrary, that $x_n\not\to x_0$. Since the space is compact, $x_n$ must have a convergent subsequence, say $x_{n_k}\to a$ as $k\to\infty$, and as we assume $x_n$ is not convergent, it must have another convergent subsequence, say $x_{n_j}\to b\not=a$ as $j\to\infty$. By continuity we will have $f(a)=f(x_0)=f(b)$ that is the minimum is achieved in both points $a$ and $b$, contradicting that $f$ only achieves its minimum at a single point.

Edit. I assumed that the compact domain is a subset of Euclidean space, and hence the Bolzano Weierstrass theorem applies (the link does not properly format, you may need to google Bolzano Weierstrass), namely every bounded sequence in $\mathbb R^n$ has a convergent subsequence. In other words I assume the domain is sequentially compact. There are compact spaces that are not sequentially compact. At any rate, a compact space is countably compact and limit point compact, that is, every sequence has a limit point. (In general this is not the same as the limit of a sequence: A sequence may have more than one limit point.) Hence the sequence $x_n$ (as above) must have a limit point $a$. Then $f(a)$ is the minimum, and since $f$ only achieves minimum at $x_0$ we have $a=x_0$. So the sequence $x_n$ has a unique limit point $x_0$ and it follows that $x_n\to x_0$. This may look like the same argument I wrote first, but in this version I need not pick convergent subsequences. At any rate, I felt the original version using convergent subsequences was more appropriate (even formally incomplete), and in addition you may follow the above links to find out about different versions of compactness.

Mirko
  • 13,445
  • I think I get it. Is it true that if a sequence doesn't converge, then there are at least two convergent subsequences that have different limits? – Yuki Kawabata Apr 27 '15 at 02:40
  • depends on the space (the domain). The sequence $1,2,3,...$ does not converge, and has no convergent subsequences. But if the domain is compact, any sequence must have a convergent subsequence. (Strictly speaking I am cheating, assuming that you also mean the domain is metrizable or first countable.) Say the subsequence converges to $a$, but the original sequence does not. So there is a neighborhood of $a$ missing infinitely many members of the original sequence, and from those you could pick a converging subsequence converging to some $b\not=a$. – Mirko Apr 27 '15 at 02:51
  • I see. Thank you very much! – Yuki Kawabata Apr 27 '15 at 03:40
  • You are welcome! I edited my answer to indicate different possible interpretations, since I originally used so-called sequential compactness which does not necessarily follow from compactness. The answer remains the same, though the details of the proof may be slightly different, at any rate I added some links to relevant definitions in case you might want to take a look. – Mirko Apr 27 '15 at 04:01