Suppose that $f : (a, b) → \mathbb{R}$ is differentiable at $x ∈ (a, b)$. Prove that $\lim_{h→0}\frac{f(x + h) − f(x − h)}{2h}$ exists and equals $f'(x)$ Give an example of a function where the limit exists but the function is not differentiable.

Proof: $\lim_{h→0}\frac{f(x + h) − f(x − h)}{2h}$ =$\lim_{h→0}\frac{f((x-h) + 2h) − f(x − h)}{2h}$ $= f'(x)$ by definition which exists as $f:(a,b) \to \mathbb{R}$ is differentiable $x \in (a,b)$.

Can anyone verify what I have done? I think I'm going wrong somewhere. Can anyone please help? Also, I can't think of a function whose limit exists but it is not differentiable. I was thinking of $f(x) =|x|$ but don't know if its correct.

Thank you.

EDIT (After the comments and hints received):

$\lim_{h→0}\frac{f(x + h) − f(x − h)}{2h}$=$\lim_{h→0}\frac{[f(x + h)-f(x)]+[f(x) − f(x − h)]}{2h}$=$\frac{1}{2}\lim_{h→0}\frac{[f(x + h)-f(x)]+[f(x) − f(x− h)]}{h}$ =$\frac{1}{2}[\lim_{h→0}\frac{f(x + h)-f(x)}{h}$ + $\lim_{h→0}\frac{f(x) − f(x − h)}{h}]=\frac{1}{2}[f'(x)+f'(x)]=f'(x)$

Is this ok? I'm particularly concerned about breaking the limits into two. Is that allowed and can anyone give me a proper theorem or a result/justification that allows me to do that?

Thank you!!

  • $\begingroup$ Your EDIT is fine. However, you should have in mind that it is true only because both $\frac{1}{2}\lim_{h→0}\frac{f(x + h)-f(x)}{h}$ and $\frac{1}{2}\lim_{h→0}\frac{f(x )-f(x-h)}{h}$ exist. I mean: the limits of a sum may exist, while the limit of both term do not. $\endgroup$ – mathcounterexamples.net May 16 '18 at 19:55
  • $\begingroup$ The question says that the function is differentiable on its domain. Can that be used to justify that? $\endgroup$ – A.Asad May 16 '18 at 20:00
  • $\begingroup$ What you wrote is fine even if you suppose that $f$ is differentiable at $x$ only. I'm just warning on the fact that the existence of the limit of a sum do not imply the existence of the limits of both terms. $\endgroup$ – mathcounterexamples.net May 16 '18 at 20:06
  • $\begingroup$ The first line of your proof seems incorrect. That is not the definition of $f'(x).$ $\endgroup$ – zhw. May 16 '18 at 20:46
  • $\begingroup$ @A.Asad We can write simply $$\lim_{h→0}\frac{f(x + h) − f(x − h)}{2h}=\frac12\lim_{h→0}\frac{f(x + h) − f(x))}{h}-\frac12\lim_{h→0}\frac{f(x - h) − f(x)}{-h}$$ $\endgroup$ – user May 17 '18 at 6:44

Recall that for $f(x)$ differentiable at $x ∈ (a, b)$ we have

  • $f(x+h)=f(x)+f'(x)h+o(h)$
  • $f(x-h)=f(x)-f'(x)h+o(h)$


$$\lim_{h→0}\frac{f(x + h) − f(x − h)}{2h}=\lim_{h→0}\frac{2f'(x)h+o(h)}{2h}=\lim_{h→0}\frac{2f'(x)+o(1)}{2}=f'(x)$$

Yes your example for $|x|$ is fine.

  • $\begingroup$ Where did you get the two results in the bullet points from? I'm afraid I haven't seen them (or at least I can't recall). Do they come from a particular theorem or or result? $\endgroup$ – A.Asad May 16 '18 at 19:53
  • $\begingroup$ @A.Asad For functions of one variable the concept is completely equivalent to that of derivative, refer to en.wikipedia.org/wiki/Differential_of_a_function#Definition . For functions of several variables the two concept are different and in some sense differential is the most important one. $\endgroup$ – user May 16 '18 at 20:03
  • $\begingroup$ @A.Asad Note that the little-o notation indicates something which goes to zero with an order $h^n$ with $n>1$, thus $$f(x+h)=f(x)+f'(x)h+o(h)\iff f'(x)=\frac{f(x+h)-f(x)}{h}+\frac{o(h)}{h}$$ and taking the limit since $\frac{o(h)}{h}\to 0$ we obtain the equivalent definition of derivative. $\endgroup$ – user May 16 '18 at 20:06

If you want a clearer proof, here is one:

From the definition $f'(x) = lim_{h\to 0} \frac{f(x+h)-f(x)}{h}$, try to prove $f'(x) = lim_{h\to 0} \frac{f(x)-f(x-h)}{h}$. Then use the trick that $f(x+h)-f(x-h) = [f(x+h)-f(x)] + [f(x)-f(x-h)]$.

The whole point is we should prove things based on the definitions. A priori, we only have $f'(x-a)=\lim_{h\to 0} \frac{f(x-a+2h)-f(x-a)}{2h}$ for any fixed $a$. While it is tempting to substitute $h$ into $a$ and takes $h$ to $0$, I would recommend not to "jump" in your early learning of mathematics.

  • 2
    $\begingroup$ Downvoting. The question asked to "verify what [the questioner has] done"; you have given a completely different proof. $\endgroup$ – BallBoy May 16 '18 at 19:32
  • $\begingroup$ Thank you. Edit: pointed out what the questioner lacks of. $\endgroup$ – Student May 16 '18 at 19:38
  • $\begingroup$ Thank you for the edit. Removing the downvote and upvoting. $\endgroup$ – BallBoy May 16 '18 at 19:40
  • $\begingroup$ I did that. However, the problem is that I should be getting $h$ in the numerator which I'm not. Can I take the $1/2$ outside, compute the two limits, giving me $\frac{1}{2}f'(x)$ which added together would give me $f(x)$? $\endgroup$ – A.Asad May 16 '18 at 19:44
  • $\begingroup$ So the question you should focus on thinking is: can I pull a constant out of a limit? $\endgroup$ – Student May 16 '18 at 19:46

When you say that the expression you have is equal to $f'(x)$ "by definition," you should elaborate. What you have written certainly isn't the usual definition of $f'(x)$. I'm not sure you can make your strategy work, and I think you're going to need a different tack. (Hint: The definition of $f'(x)$ requires there be an $f(x)$ somewhere in the numerator; can you add one in?)

Your $|x|$ example seems like it should work. To convince yourself it's correct, try evaluating $\frac{|h|-|-h|}{2h}$

  • $\begingroup$ I'm not quite sure how to incorporate that into the numerator. I was thinking of adding and subtracting $f(x)$ to the expression in the numerator but that wasn't leading me anywhere. $\endgroup$ – A.Asad May 16 '18 at 19:38
  • $\begingroup$ @A.Asad I think that's a good idea. The next step would be to separate that into two derivative-like expressions. $\endgroup$ – BallBoy May 16 '18 at 19:39
  • $\begingroup$ Can you please check my edit? $\endgroup$ – A.Asad May 16 '18 at 19:51
  • 1
    $\begingroup$ @A.Asad Looks very good! The only thing I'd be a little more clear about is your claim that $\lim_{h\to0}\frac{f(x)-f(x-h)}{h}=f'(x)$. It is true, but it's not quite the definition of the derivative, so perhaps you could justify it. The "limit breaking" is allowed according to the limit law that the limit of a sum is the sum of the limits. $\endgroup$ – BallBoy May 16 '18 at 19:54
  • $\begingroup$ Thank you! And, for the justification, I could write $f(x) = f((x-h)+h)$ so it is in the form as that required by the definition? $\endgroup$ – A.Asad May 16 '18 at 19:58

For the first point, I'm very suspicious of your "by definition proof"...

To be more precise, as $f$ is supposed differentiable at $x$, it exists a map $\epsilon : h \mapsto \epsilon(h)$ with $\lim\limits_{h \to 0} \epsilon(h)=0$ such that $f(x+h) = f(x) + f^\prime(x)h + h\epsilon(h)$. Then:

$$\frac{f(x+h) - f(x-h)}{2h} = \frac{f^\prime(x)h - f^\prime(x)(-h) + \epsilon(h)h - h \epsilon(-h)}{2h}= f^\prime(x) + \frac{\epsilon(h)-\epsilon(-h)}{2},$$

which converges to $f^\prime(x)$ with $h$ as $\lim\limits_{h \to 0}\epsilon(h) = 0$.


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.