Suppose that $f:(0,1)\to\mathbb{R}$ is differentiable, and that $f(x_1)=f(x_2)=0$ and $f’(x_1)>0$ and $f’(x_2)>0$ for some $0 <x_1<x_2<1$. Then there must exist an $x_0\in(x_1,x_2)$ such that $f(x_0)=0$ and $f’(x_0)\leq0$, as follows: Let $A=\{x\in(x_1,x_2):f(x)\geq0\}$, and note that $A$ is nonempty, since the condition $f’(x_1)>0$ guarantees that there exist points that exceed $x_1$ by arbitrarily small amounts, at which $f$ is strictly positive. Also note that $A$ is bounded above, by $x_2$. Therefore, let $x_0=\sup A$. Note that $x_0 > x_1$. Since $f’(x_2)>0$, there exist points that $x_2$ exceeds by arbitrarily small positive amounts, at which $f$ is strictly negative. It follows that $x_1 < x_0 < x_2$. By continuity of $f$, we have that $f(x_0)=0$. Since $f(x)<0$ for all $x\in(x_0,x_2)$, it follows also that $f’(x_0)\leq0$, as required.

Is this proof correct, and is there a simpler proof, perhaps using a ready-made theorem such as the Intermediate Value Theorem ?

| cite | improve this question | | | | |
  • 2
    $\begingroup$ Quick thought: by the direction of the derivative there must exist points $f(x_3)>0$ and $f(x_4)<0$. Then apply intermediate value theorem? The derivative is such for at least one point as the function crosses over. $\endgroup$ – Dole Dec 27 '18 at 21:15
  • $\begingroup$ @RRL, Please check my argument again - in fact I do not appeal to the IVT. Also the reason why $f(x)<0$ for all $x\in(x_0,x_2)$ is that $x_0$ is an upper bound for the set where $f(x)\geq0$. $\endgroup$ – Simon Dec 28 '18 at 1:31
  • $\begingroup$ @Dole, thank you. This was my first thought, too. Trying to prove the last sentence in your comment is what lead me to the proof I gave in my original question. $\endgroup$ – Simon Dec 28 '18 at 1:39
  • $\begingroup$ @RRL, in fact I am not arguing that there are points in a right neighbourhood of $x_0$ where $f$ is positive. I am arguing that in the interval $(x_0,x_2)$, $f$ must be negative. I am not using the IVT. I believe that I have shown rigorously, that $f(\sup A)=0$ (I am calling $\sup A$ $x_0$), and that $f'(x_0)\leq0$. $\endgroup$ – Simon Dec 28 '18 at 2:00
  • 1
    $\begingroup$ OK I finished my proof. It gets to the same ultimate argument -- finding a smallest or largest zero $x_0$ bounded away from either $x_1$ or $x_2$ so that, consequently, $f'(x_0) \leqslant 0$. $\endgroup$ – RRL Dec 28 '18 at 2:32

By your argument there are points $x_1 < x' < x'' < x_2$ where $f(x') > 0$ and $f(x'') < 0$. By the IVT there is at least one point $y_1$ (and possibly more) where $x' < y_1 < x''$ and $f(y_1) = 0$.

If $f'(y_1) \leqslant 0$, then we are done. On the other hand, if $f'(y_1) > 0$, then we have the same problem with $y_1$ replacing $x_2$ and there exists a point $y_2$ between $x_1$ and $y_1$ such that $f(y_2) = 0$.

Continuing in this way we either find a zero where the derivative is less than or equal to $0$ or generate a sequence $y_n \in (x',x'')$ such that $f(y_n) = 0$ and $f'(y_n) > 0$.

However, it can be shown that if there are no zeros of a function that is differentiable on a closed interval where the derivative is also $0$, then the set of zeros is finite. Since $f$ is differentiable on the closed interval $[x_1,x_2]$ there exists only a finite set of zeros $\{y_1,y_2, \ldots, y_n\}$ between $x_1$ and $x_2$.

Armed with this, you can now show that $f'(y_n) \leqslant 0$ since $y_n$ must be the smallest number between $x'$ and $x''$ with $f(y_n) = 0$. If $f'(y_n) > 0$ then there would be another zero between $x'$ and $y_n$, a contradiction.


Suppose $f$ is differentiable on $[a,b]$ and at no point $x \in [a,b]$ do we have $f(x) = f'(x) = 0$. Then the set of points in $[a,b]$ where $f(x) = 0$ is finite.

To prove this, assume otherwise. Then there is an infinite sequence of zeros and by compactness and continuity a subsequence $(x_n)$ converging to some point $c \in [a,b]$ such that $f(x_n) = f(c) = 0$. Since $f$ is differentiable

$$f'(c) = \lim_{n \to \infty} \frac{f(x_n) - f(c)}{x_n - c} = 0,$$

a contradiction.

| cite | improve this answer | | | | |
  • $\begingroup$ Thank you for your answer, RRL. Can you elaborate on how to prove the statement preceded by "it can be shown that" ? $\endgroup$ – Simon Dec 28 '18 at 1:37
  • $\begingroup$ After the second paragraph why don't you just say: the sequence $y_1\gt y_2\gt y_3\gt\cdots$ is decreasing and bounded, so it converges to a limit $y=\lim_{n\to\infty}y_n$. Then $x_1\le y\lt x_2$, and $f(y)=0$ by continuity since $f(y_n)=0$, and $f'(y)=0$ since $$f'(y)=\lim_{n\to\infty}\frac{f(y_n)-f(y)}{y_n-y}=\lim_{n\to\infty}0=0,$$ so $x_1\lt y$ since $f'(x_1)\gt0=f'(y)$? $\endgroup$ – bof Dec 28 '18 at 3:04
  • $\begingroup$ @bof: That is a great suggestion. On the other hand what I like about this site is the opportunity to pursue less than conventional lines of thought unlike the many more constrained aspects in my life. :-) $\endgroup$ – RRL Dec 28 '18 at 3:45

Your proof is correct (+1) and the key idea is that if $f$ is continuous on $[a, b] $ then the set $A=\{x\mid x\in[a, b], f(x) =k\} $ is closed (inverse image of a closed set under continuous map is closed, similar result holds for open sets also).

Here you choose $a$ near and to the right of $x_1$ so that $f(x) >0\,\forall x\in(x_1,a]$ and $b$ near and to the left of $x_2$ such that $f(x) <0\,\forall x\in[b, x_2)$. By IVT the set $$A=\{x\mid x\in[a, b], f(x) =0\} $$ is non empty and as noted above is closed. Since $A$ is obviously bounded and closed it has a minimum and a maximum. Both $\min A$ and $\max A$ (which can be same) work as the desired point $x_0$.

| cite | improve this answer | | | | |
  • $\begingroup$ Thank you Paramanand. $\endgroup$ – Simon Dec 28 '18 at 10:33

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.