2

If $f$ is continuous and derivable, prove that if $\lim_{x\to 0}f'(x)=L$, then $f'(0) = L$

I need help proving this, thank you.

Sahiba Arora
  • 10,847
TheNicouU
  • 361

4 Answers4

4

The function must be defined and continuous at $0$ for this to hold.

  • If $f$ is not defined at $0$, it cannot have a derivative at $0$.
  • If $f$ is defined at $0$, but not continuous, it cannot have a derivative at $0$.

The result is now a simple application of l’Hôpital: $$ \lim_{x\to0}\frac{f(x)-f(0)}{x-0}=\lim_{x\to0}\frac{f'(x)}{1} $$ which exists by assumption

egreg
  • 238,574
2

Thanks for asking this. It's used as a Lemma in Spivak's book, and I've never been quite certain how to prove it until now.

This is a consequence of Darboux's theorem that derivatives have the intermediate value property (i.e., if $f'(a) = A$ and $f'(b)= B$ and $A < C< B$, then there's a $c$ between $a$ and $b$ with $f'(c) = C$.)

For a discussion of that theorem, see: How to prove that derivatives have the Intermediate Value Property

To prove the claim from Darboux:

First, let

$$ g(x) = f(x) - Lx. $$ Then we have that $$ \lim_{x \to 0} g'(x) = 0 $$ and want to prove that $g'(0) = 0$. What's given is that $g$ is everywhere differentiable on $\Bbb R$ (I assume!). In particular, $g'(0)$ exists. $$ \newcommand{\ep} {\epsilon} $$ Suppose that $A = g'(0) \ne 0$. Without loss of generality, suppose then that $A > 0$ (the other case is very similar).

Consider $\ep = A/3$. For some $\delta$, we know that $0 < |x| < \delta \implies |f'(x)| < \ep$ by the limit assumption. In particular, picking $x_0 = \delta/2$, we have $$ |f'(x_0)| < \frac{A}{3}. $$

Darboux's theorem then tells us that for some $x_1$ between $0$ and $x_0$, we have $$ f'(x_1) = \frac{2A}{3} $$ because $\frac{2A}{3}$ is between $A = f'(0)$ and $\pm \frac{A}{3}$, which is $f'(x_0)$. But this is a contradiction, for $|x_1| < |x_0| = \frac{\delta}{2} < \delta$, and the limit condition tells us that $$ |f'(x_1)| < \epsilon = \frac{A}{3}. $$ But $\frac{2A}{3}$ is NOT less than $\frac{A}{3}$, because $A$ is positive.

John Hughes
  • 93,729
  • The question says it is differentiable on $\mathbb{R} \setminus {0}$. So $g'(0)$ may not exist, right? – Sahiba Arora Jun 07 '17 at 12:53
  • The question does not say that -- it says it's "derivable", which I take to mean "differentiable", and since it doesn't say where, I took that to mean "everywhere differentiable." The Spivak Lemma actually claims (I think) the thing you're suggesting: if it's differentiable away from $0$, and continuous at $0$, and the derivative limit at $0$ exists, then it's differentiable at $0$ and the derivative at $0$ is the limit of nearby derivatives. To prove that would take a bit more work. – John Hughes Jun 07 '17 at 13:12
  • If you see the title.. – Sahiba Arora Jun 07 '17 at 13:15
  • Oh, jeez. I hate it when the question and title don't match. I'm leaving this as an answer to the question asked (rather than the one asked in the title), because as egreg points out, the one in the title is false without the additional assumption of continuity. Overall, though, I prefer egreg's answer. :) – John Hughes Jun 07 '17 at 13:18
1

This is not true, x=0 is not necessarily on the domain of $f'(x)$.

For example:

$$f:(0,1) \to R, \\ f(x)=\sin(x), f'(x)=\cos(x), \\ \lim_{x \to 0} f'(x)=1$$

but nor $f(0)$ neither $f'(0)$ exist, because they are outside of the domain, which exclude the point $x=0$.

if $f$ is differenciable, then by definition, the derivative exist on its domain, the value $f'(0)$ exist, and in particular, the limit $\lim_{x\to o}f'(x)$ exist.

Brethlosze
  • 3,010
1

You want to prove that $\lim_{x\to0}\frac{f(x)-f(0)}x=L$. Take $\varepsilon>0$. Since $\lim_{x\to0}f'(x)=L$, there is a $\delta>0$ such that $|x|<\delta\Longrightarrow|f'(x)-L|<\varepsilon$. If $|x|<\delta$, take $c\in(0,x)$ such that $\frac{f(x)-f(0)}x=f'(c)$; such a $c$ exists by the mean value theorem. So,$$\left|\frac{f(x)-f(0)}x-L\right|=\bigl|f'(c)-L\bigr|<\varepsilon.$$

  • That shows that for some $c$ in $(0, x)$, you have $|f'(c) - L| < \epsilon$ (in your answer, you lost the $L$ somewhere!). But it doesn't show that $|f(0) - L| < \epsilon$ -- that would require continuity of the derivative, which isn't always the case. – John Hughes Jun 07 '17 at 12:24
  • I have added the $L$ (twice). What I proved was that $|x|<\delta\Longrightarrow\left|\frac{f(x)-f(0)}x-L\right|<\varepsilon$. – José Carlos Santos Jun 07 '17 at 12:27
  • Agreed -- my criticism on that aspect was an overstep -- apologies. – John Hughes Jun 07 '17 at 13:15