1
$\begingroup$

I am given a function $f(t) \in \mathbb{R}$ which is continuous; bounded above by $M$ and below by $0$. $f$ is differentiable everywhere except at $f=0$. Also, $\lim_{t \to \infty} f = 0$ and $t \in [0, \infty)$.

How do I show that a global maximum/supremum exists and where it exists?

I thought of dividing this into 3 cases: when function is non-decreasing, non-increasing and when there exists a local extremum.

$(i) $ $f$ can not be non-decreasing since it converges to $0$, except when $f(0) = 0$ which will be a trivial case.

$(ii)$ If $f$ is non-increasing then, $f(0)$ is the global maxima as $f(0) \ge f(t) \forall \ t$.

$(iii)$ That leaves us with the case when there exists at least $1$ local extremum. Now, if the domain were bounded, I could have used Extreme value theorem and say that global maximum exists either at $t=0$ or at a local maximum, but the domain is unbounded in my case, so that can't be used. However I do have information about the limit, which could be helpful.

Intuitively, I still think the result will be same as that of Extreme value theorem, but I want to prove it using proper results. How should I modify my process?

$\endgroup$
  • 2
    $\begingroup$ When you write $\lim_{t\to\infty}$, do you mean $\lim_{t\to\pm\infty}$? $\endgroup$ – José Carlos Santos Jul 6 '18 at 9:54
  • $\begingroup$ I would claim that the supremum exists (be careful, but it works since you are bounded from above and you converge to $0$). Then take a sequence converging to this supremum and by continuity this makes sure that the supremum and the maximum coincide $\endgroup$ – Stan Tendijck Jul 6 '18 at 9:57
  • $\begingroup$ @JoséCarlosSantos No just positive one. I think I should mention it, $t \in [0, \infty)$ $\endgroup$ – Manish Jul 6 '18 at 9:57
  • $\begingroup$ @StanTendijck Can you please elaborate a little, especially the argument regarding supremum and maximum coinciding? Also, it should be same as local maximum, right? $\endgroup$ – Manish Jul 6 '18 at 9:59
  • 1
    $\begingroup$ You can attain the supremum as a limit of a sequence, i.e., there exist $a_n$ such that $\lim f(a_n) = S$ (with $S$ the supremum). You can argue by convergence to $0$ that the supremum is 'attained' at a bounded number, i.e. $a_n$ does not converge to $\infty$. Hence, there exist an $a$ which occurs at the limit of $a_n$. Moreover, by continuity you can interchange the limit and $f$ such that $\lim f(a_n) = f(a)$. Is this better? $\endgroup$ – Stan Tendijck Jul 6 '18 at 10:11
1
$\begingroup$

If $f:[0,\infty)\to[0,\infty)$ is continuous and $\lim_{t\to\infty}f(t)=0$, then you can prove the existence of a global maximum. If $f\equiv 0$, the statement is trivial. Otherwise, there exists a point $t_0\in\mathbb R$ such that $f(t_0)>0$. Further, from the limit you can conclude the existence of $R>0$ such that $f(t)<f(t_0)$ for all $t>R$. Since $I:=[0,R]$ is compact and $f$ continuous there exists a global maximum of $f\mid_I$ which is a global maximum of $f$.

To find the extreme point, you have to do it in two steps.

1) Consider the extreme points of $f\mid_{(0,\infty)}$ using $f'$.

2) Compare the maxima on $(0,\infty)$ with the value at $0$.

You can have 3 cases:

a) If there are no maxima on $(0,\infty)$, then the global maximum has to be at $0$.

b) If there exist maxima on $(0,\infty)$ but their heights are below $f(0)$, then the global maximum is at $0$.

c) If there exist maxima on $(0,\infty)$ and the highest value is above $f(0)$, then the maxima with the highest value are the global maxima.

$\endgroup$
1
$\begingroup$

If $f$ is the null function, then it attains its maximum everywhere. Otherwise, tak $t_0>$ such that $f(t_0)>0$. Then there is a $R>0$ such that $t>R\implies f(t)<f(t_0)$. By the extreme value theorem, the restriction of $f$ to $[0,R]$ has a maximum in some $t_1$ in that interval and since outside the interva $f$ always takes values smaller that $f(t_0)$, $f$ attains its global maximum at $t_1$. Therefore, either $f'(t_1)=0$ or the maximum is attained at $0$.

$\endgroup$
  • $\begingroup$ I can understand the logic properly now. I think both the answers convey the same explanation. It's sad that both can't be tagged accepted :( $\endgroup$ – Manish Jul 6 '18 at 10:17
  • 1
    $\begingroup$ I accept the one which has more appeal to you. It's as simple as that. $\endgroup$ – José Carlos Santos Jul 6 '18 at 10:21

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.