Let $f$ be a countinously differentiable function on $[a,b]$. Suppose that there exists $c \in(a,b)$ such that $f'(c)=0$. Prove that there exist $\xi\in (a,b)$ such that $$f(\xi)=f(a)+f'(\xi)(b-a) $$

I tried looking at function $H(x)=f(x)-f(a)-f'(x)(b-a)$ no good came from it. How to solve this ?

As suggesed in an answer lets solve for $f'(\xi)$: $$f'(\xi)=\frac{f(\xi)-f(a)}{b-a} $$ This is not the statement of mean value theorem, if it were $f(b)$ it would have been

  • $\begingroup$ I missed that it was an $f(\xi)$ on the left instead of an $f(b)$. If we had an $f(b)$, this would be the mean value theorem. $\endgroup$ Jan 16 '20 at 21:37
  • $\begingroup$ You seem to have copied something incorrectly. Note that for the function $f(x) = x$ over the interval $[0,1]$, there is no such $\xi$. $\endgroup$ Jan 16 '20 at 22:34
  • $\begingroup$ @Omnomnomnom $f(x)=x$ doesn't satisfy the condition that there exist a $c\in(a,b)$ such that $f'(c)=0$ $\endgroup$
    – Milan
    Jan 16 '20 at 22:38
  • $\begingroup$ @Omnomnomnom maybe i didn't phrase that part of the question well. I am not familiar with how it is usually said in English $\endgroup$
    – Milan
    Jan 16 '20 at 22:41
  • $\begingroup$ I think I misread that part as well. Anyway, now that I understand the question I'm not sure where to start either, good luck $\endgroup$ Jan 16 '20 at 23:02

For simplicity, one may prove a special case (see Proposition below), and then use it to derive OP's result (see Corollary below).

Proposition. Let $g(x)$ be a continuously differentiable function on $[0,1]$ such that $g(0)=0$ and there exists $c\in (0,1)$ with $g'(c)=0$. Then there exists $\xi\in (0,1)$ such that $g(\xi)=g'(\xi).$

Proof. One needs to prove the existence of $\xi\in (0,1)$ such the $g(\xi)=g'(\xi)$ or equivalently $H(\xi)=0$ for $H(x)=g(x)-g'(x)$. Up to multiplication by $-1$, there are two cases to consider:

Case 1: $g'(0)>0$.

Let $c_1=\inf\{c|g'(c)=0\}$. By continuity of $g'$, one has $g'(c_1)=0$. Furthermore $g'(x)\geq 0$ on $[0,c_1]$ (otherwise by intermediate value theorem for $g'$, there would exist $c_2\in (0,c_1)$ with $g'(c_2)=0$ contradicting the assumption of $c_1$), so $g$ is increasing on $[0,c_1]$ and $g(c_1)>0$. It follows that $$H(0)=g(0)-g'(0)=-g'(0)<0,~H(c_1)=g(c_1)-g'(c_1)=g(c_1)>0.$$ So by intermediate value theorem, there exists $\xi\in (0,c_1)\subseteq (0,1)$ such that $H(\xi)=0$.

Case 2: $g(0)=g'(0)=0.$

One has $H(0)=0$. If $H(x)=0$ or changes sign in $(0,1)$, then the result is either trivial or follows from the intermediate value theorem. It remains to rule out the case when $H(x)\neq 0$ in $(0,1).$ Up to multiplication by $-1$, one proves that the following case is impossible: $$g(0)=g'(0)=0,g(x)-g'(x)>0,\forall x\in(0,1),g'(c)=0~ ({\rm so~}g(c)>0){\rm~for~some~}c\in (0,1)$$

By mean value theorem, there exists $\eta\in(0,c)$ such that $$g'(\eta)=\frac{g(c)-g(0)}{c-0}=\frac{g(c)}c>g(c)~(\because 0<c<1)$$ $$\Rightarrow g(\eta)>g'(\eta)>g(c),$$ which shows that the absolute maximum of $g$ on $[0,c]$ occurs at some $\eta_1\in (0,c),$ so $g'(\eta_1)=0$. By mean value theorem as above, there exists $\eta_2\in (0,\eta_1)$ such that $$g'(\eta_2)=\frac{g(\eta_1)-g(0)}{\eta_1-0}=\frac{g(\eta_1)}{\eta_1}>g(\eta_1)$$ $$\Rightarrow g(\eta_2)>g'(\eta_2)>g(\eta_1),$$ contradicting the assumption that $g(\eta_1)$ is the absolute maximum of $g$ on $[0,c]$. This completes the proof of Case 2.

Combining Case 1 and Case 2, the Proposition is proven.

Corollary. Let $f$ be continuously differentiable on $[a,b]$ such that there exists $c\in (a,b)$ with $f'(c)=0$. Then there exists $\xi\in (a,b)$ such that $$f(\xi)=f(a)+f'(\xi)(b-a).$$

Proof. Let $f$ be as given. Let $c_1$ be such that $c=a+(b-a)c_1$ (so $0<c_1<1$). Define $g$ on $[0,1]$ by $$g(x)=f(a+(b-a)x)-f(a).$$ Then one has $g(0)=0$ and $g'(c_1)=f'(a+(b-a)c_1)(b-a)=f'(c)(b-a)=0.$ So $g(x)$ satisfies the assumptions of the Proposition. It follows that there exists $\xi_1\in (0,1)$ such that $$g(\xi_1)=g'(\xi_1)$$ $$\Leftrightarrow f(a+(b-a)\xi_1)-f(a)=f'(a+(b-a)\xi_1)(b-a)$$ $$\Leftrightarrow f(\xi)=f(a)+f'(\xi)(b-a),$$ where $\xi:=a+(b-a)\xi_1\in (a,b)$, as required. QED

  • $\begingroup$ What function do you mean when writing $f$ $\endgroup$
    – Milan
    Jan 17 '20 at 22:13
  • $\begingroup$ What is the relationship of your f to the original f. Use of g confuses me $\endgroup$
    – Milan
    Jan 17 '20 at 22:52
  • $\begingroup$ @Milan I have rewritten the proof. Hope this is no longer confusing. $\endgroup$
    – Pythagoras
    Jan 18 '20 at 2:20

Taking $\phi(x) = f(x) - f(a) - f'(x)(b-a)$ we have $\phi(a) = -f'(a)(b-a)$ and $\phi(c) = f(c) - f(a)$ (where $f'(c) = 0$). Also since $f$ is continuously differentiable it follows that $\phi$ is continuous.

If $f'(a) > 0$ and $f(c) > f(a)$, then $\phi(a) < 0$ and $\phi(c) > 0$ and there exists a point $\xi \in (a,c)$ such that $\phi(\xi) = 0$ by the intermediate value theorem. On the other hand if $f(c) \leqslant f(a)$ then by continuity with $f'(a) > 0$ there exists a relative maximum at some point $c_1 < c$ where $f(c_1) > f(a)$ and $f'(c_1) = 0$. Again we find a point $\xi$ such that $\phi(\xi) = 0$.

Note that $\phi(\xi) = 0$ implies

$$f(\xi) = f(a) + f'(\xi)(b-a)$$

A similar argument applies if $f'(a) < 0$. See if you can take care of the case where $f'(a) = 0$.


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.