3
$\begingroup$

Here is the question:

Let M be a compact smooth manifold, $f:M\rightarrow\mathbb{R}$ be a smooth non-constant function. Show that $f$ has at least two critical points.

I am trying to show by contradiction. Suppose $f$ has less than two critical points. First assume it has none. By Inverse Function Theorem, for $x\in M$, there is an open neighbourhood $U $ containing $x$ such that $f|_U$ is a diffeomorphism. This means $f$ is an open map. Thus, $f(M)$ is open. But this is contradiction to the fact that $f(M)$ is compact (since $f$ is cont's and $M$ is compact).

Is this fine? Now I don't know how to show that there cannot be only one critical point. Only after proving this, I will be done. Some-where I should be using that $f$ is non-constant. I don't know how. I understand that if $f$ is a constant function, then every point in $M$ is trivially critical.

Intuitively I don't understand why compactness is used. I am trying to find an example of non-compact smooth manifold $M$ and a smooth map $f:M\rightarrow \mathbb{R}$ that has no critical points. I will appreciate any hint for this too.

$\endgroup$
  • $\begingroup$ I have absolutely no idea about how you apply the inverse function theorem given that $\dim M$ might be strictly larger that $1$. What you could have done was to note that if $f$ has no critical points than it must be a submersion, therefore an open map. $\endgroup$ – Alex M. Jun 9 '15 at 8:28
6
$\begingroup$

Since $M$ is compact and $f:M \to \Bbb R$ is continuous, there exists points $p_\sigma, p_\gamma \in M$ such that $f(p_\sigma) = \sigma$ and $f(p_\gamma) = \gamma$, where $\sigma$ and $\gamma$ are the global minimum and maximum values of $f$ on $M$. This is a basic result of elementary topology. Since $p_\sigma$ is a global minimum for $f$, we have $df_{p_\sigma} = 0$ and likewise $df_{p_\gamma} = 0$, since $p_\gamma$ is a global maximum. Since $f$ is not constant, $\sigma < \gamma$ whence $p_\sigma \ne p_\gamma$. Thus $p_\sigma$ and $p_\gamma$ are distinct critical points of $f$; we see there are at least two critical points of $f:M \to \Bbb R$. QED.

Hope this helps. Cheers,

and as always,

Fiat Lux!!!

$\endgroup$
  • $\begingroup$ quite awesome proof. I like this very much as you used very elementary ideas. $\endgroup$ – ugstudent1243 Mar 9 '14 at 17:15
  • 1
    $\begingroup$ @ ugstudent: thank you very much! It is indeed pretty "elementary"! And thanks for the "acceptance"! $\endgroup$ – Robert Lewis Mar 9 '14 at 17:17
  • $\begingroup$ you used $M$ for maximum of $f$. Perhaps a different letter would be better as it is used for manifolds already. $\endgroup$ – ugstudent1243 Mar 9 '14 at 17:20
  • $\begingroup$ @ ugstudent: you are correct! Will edit shortly. Thanks! $\endgroup$ – Robert Lewis Mar 9 '14 at 17:22
  • 1
    $\begingroup$ How to show that the property: "'p is a minimum(or max) of a function in a manifold then Dfp=0" holds to manifolds? $\endgroup$ – Luiz Feb 7 '17 at 18:49
1
$\begingroup$

Let's suppose that $M$ is connected. The general case then follows by considering the restriction of $f$ to any connected component of $M$. Then $f(M)$ is a connected compact subset of $\mathbb R$, and since we're assuming $f$ is nonconstant, $f(M)$ is not a single point. What must $f(M)$ look like?

For a counterexample in the case that $M$ is not compact, you could take $M = \mathbb R$ and $f(x)=x$.

$\endgroup$
  • $\begingroup$ Are you trying to give argument for why $f$ cannot have one critical point? $\endgroup$ – ugstudent1243 Mar 9 '14 at 16:56
  • $\begingroup$ I was only giving a hint. Since a complete solution has now been posted, I'll elaborate on this hint. The only compact connected subsets of $\mathbb R$ are singletons and closed intervals, so $f(M)$ must be a closed interval $[a,b]$. Choosing $p,q\in M$ with $f(p)=a$ and $f(q)=b$, you can then show that $p$ and $q$ are critical points (e.g. the inverse function theorem fails at these points). $\endgroup$ – bradhd Mar 9 '14 at 18:07

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.