EVT: Let $f: [a,b] \to \mathbb{R}$ be continuous, then $f$ achieves maximum and minimum

I think it is very easy to prove using continuous image of compact is compact + Heine Borel but I am stuck on showing that the $\sup$ and $\inf$ are actually $\max$ and $\min$

Proof attempt:

Since $[a,b]$ is compact, $f$ continuous, therefore $f([a,b])$ is compact.

By Heine Borel, $f([a,b])$ is closed and bounded. By boundedness, $f([a,b]) \subseteq [-N,N], N \in \mathbb{R}_{+}$.

Let $u := \sup f([a,b])$, then we wish to show that $u \in f([a,b])$

I'm guessing if I were to continue, it would be something like this. Since $u$ is the $\sup f([a,b])$, then there exists $u_1 \in [u-\epsilon, u]$ for some $\epsilon >0$, otherwise $u$ is not the supremum. Then we can find $u_2 \in [u_1-\epsilon, u]$...this builds a Cauchy sequence. Since any closed interval in $\mathbb{R}$ is complete, the Cauchy sequence converges. Since $f([a,b])$ is closed, and $u_1, u_2,\ldots$ converges to $u$, therefore $u \in f([a,b])$.

Is this correct?

Note: I realized that the above would rely on sequential compactness rather than covering compactness. What is another way to prove this that is more topologically oriented?

  • 1
    $\begingroup$ Not quite. Based on the way you have it written, you can choose the constant sequence $\{u_1,u_1,\ldots\}$. $\endgroup$ – Aweygan Jul 30 '16 at 22:55
  • $\begingroup$ keep in mind the trivial case where the function is constant, I think that's along the lines of what Aweygan was saying. $\endgroup$ – Alan Jul 30 '16 at 22:59
  • $\begingroup$ It is much simpler to argue via contradiction. If $M = \sup f$ and $f$ never takes value $M$ in $[a, b]$ then $g(x) = 1/\{M - f(x)\}$ is continuous on $[a, b]$ and hence bounded. But since $f$ can take values arbitrarily close to $M$ the expression $M - f(x)$ can be made arbitrarily small and thus $g(x)$ takes arbitrarily large values and this contradicts that $g$ is bounded. $\endgroup$ – Paramanand Singh Jul 31 '16 at 9:08

Since the image is compact, the supremum (now seen as the supremum of the set of attained values of $f$) does belong to the image (as a set). (A compact set contains its supremum.) But what does it mean for a value $y$ to be in the image? It means that there is an $x$ in the domain such that $f(x)=y$.

| cite | improve this answer | |

There are a few errors with your proof, which I mentioned in the comments. Here is a proof, following similar lines to yours (constructing a convergent sequence, etc.)

Put $M=\sup\limits_{x\in[a,b]}f(x)$. There is a sequence $\{x_n\}$ in $[a,b]$ with $$M\leq f(x_n)+\frac{1}{n},$$ And thus $f(x_n)\to M$ as $n\to\infty$. Since $[a,b]$ is compact, there is a subsequence $\{x_{n_k}\}$ of $\{x_n\}$ convergent to some $x_0\in[a,b]$, and the continuity of $f$ guarantees that $f(x_{n_k})\to f(x_0)$ as $k\to\infty$. But $f(x_n)$ is convergent, so $f(x_{n_k})$ must converge to the same value, and thus $f(x_0)=M$.

The corresponding proof for $\inf$ follows by applying the proceeding argument to $-f$.

| cite | improve this answer | |
  • $\begingroup$ I like the proof. But I realized half way that this actually relies on a weaker version of compactness which is sequentially compact. Is there a more topological oriented proof that does not involve sequences? $\endgroup$ – Olórin Jul 30 '16 at 23:23
  • $\begingroup$ @MSEisadatingsite Ahh you caught that. Well the other answer does not rely on sequences, but it does rely on the Heine-Borel theorem, which one could argue is less topological than compactness being equivalent to sequential compactness in metric spaces. $\endgroup$ – Aweygan Jul 30 '16 at 23:37
  • $\begingroup$ @MSEisadatingsite After seeing your subsequent question, I feel some explanation is in order. Boundedness is a definition which only makes sense in either a metric space or a topological vector space, while sequential compactness is a property that any subset of any topological space may have. $\endgroup$ – Aweygan Jul 31 '16 at 7:38

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.