2

If $f$ is a derivative then, is $|f|$ also a derivative?

If $f$ is Riemann integrable then it's true. But, if it's not the case,then is it true?

SOUL
  • 1,042
  • By saying, $f$ is a derivative, I want to mean that there exists a function $g$ such that $g'(x)=f(x)$ – SOUL Jan 24 '19 at 04:40
  • So, technically, a function is a derivative iff $\int_0^x f(t)dt= g(x)$ exists for all $x$ and is differentiable everywhere. Is that it? – stressed out Jan 24 '19 at 04:41

2 Answers2

4

This is not true even if $f$ is integrable.

Consider $f(x) = \begin{cases}\cos(x^{-1}), & 0 < x \leqslant 1 \\ 0, & x= 0 \end{cases}$ and $g(x) = |f|(x) = \begin{cases}|\cos(x^{-1})|, & 0 < x \leqslant 1 \\ 0, & x= 0 \end{cases}$

We can show that $f = F'$ where $F(x) = \int_0^x f(t) \, dt$. This is easy to show for $0 < x \leqslant 1$ where $f$ is continuous. It is also true but harder to show that $F'(0) = f(0) = 0$.

Now if $g = G'$ on $[0,1]$ and $g$ is integrable then $\int_0^x g(t) \, dt = G(x)$ by the FTC with $G'(0) = g(0)$.

However,

$$G'(0) = \lim_{h \to 0} \frac{1}{h}\int_0^h g(t) \, dt = \lim_{h \to 0} \frac{1}{h}\int_0^h |\cos(t^{-1})| \, dt = \frac{2}{\pi} \neq g(0)$$

See here for a proof that $G'(0) = 2/\pi$.

Addendum

To prove that $F'(0) = 0$, note that $F(0) = 0$ and with a variable change $y = t^{-1}$,

$$F(x) = \int_0^x \cos (t^{-1}) \, dt = \int_{x^{-1}}^\infty \frac{\cos y}{y^2} \, dy$$

Integrating by parts we get,

$$F(x) = \frac{\sin(x^{-1})}{x^{-2}} + 2 \int_{x^{-1}}^\infty \frac{\sin y}{y^3} \, dy$$

Thus,

$$0 \leqslant \left|\frac{F(x)-F(0)}{x} \right| \leqslant \frac{|\sin(x^{-1})|}{x^{-1}} + \frac{2}{x} \int_{x^{-1}}^\infty \frac{|\sin y|}{y^3} \, dy \\ \leqslant x + \frac{2}{x} \int_{x^{-1}}^\infty \frac{1}{y^3} \, dy = 2x$$

By the squeeze theorem it follows that

$$F'(0) = \lim_{x \searrow 0}\frac{F(x)-F(0)}{x} = 0 $$

RRL
  • 90,707
  • Well, I'm replying to your comment here. Yes, I agree with you. All I was saying is that the OP hadn't worded the question as precisely as possible. Anyway, +1 for your answer. – stressed out Jan 24 '19 at 06:31
  • 1
    @stressedout: Thank you. It would be appropriate for OP to respond to your request for clarification. Although think the first comment does it for me. – RRL Jan 24 '19 at 06:33
  • @RRL, can you give a proof of your claim i.e. $F'(0) = f(0) = 0$? – SOUL Jan 24 '19 at 08:19
  • @Tom: I added the proof of that. Note that functions which are derivatives on an interval $[a,b]$ are said to belong to the space $\mathcal{D}^1([a,b])$ and I am sure (aside from the counterexample I came up with) that $f \in \mathcal{D}^1([a,b])$ does not imply $|f| \in \mathcal{D}^1([a,b])$, but I have not found a reference yet. – RRL Jan 24 '19 at 08:51
  • @RRL, you have assumed that $g(x)$ is integrable, which is not necessary for $g$ to being a derivative. Am I correct? – SOUL Jan 24 '19 at 10:09
  • 1
    @Tom: If $f$ is Riemann integrable then $g = |f|$ is always Riemann integrable. In my example $f$ is Riemann integrable. I don’t know what you mean by “I assumed”. In general if $g$ is a derivative it does not have to be integrable. Look up Volterra’s function. – RRL Jan 24 '19 at 15:09
0

Consider $f(x)=x$ which is differentiable everywhere but $|f(x)|$ is not differentiable at zero.

CyclotomicField
  • 11,018
  • 1
  • 12
  • 29
  • 1
    Did you answer the question? – Jacky Chong Jan 24 '19 at 04:48
  • It's not your fault. The question is ambiguous. You haven't actually answered the question. Check the comments. – stressed out Jan 24 '19 at 04:50
  • 1
    @stressedout: The question is not ambiguous. This answer is about differentiability of $|f|$ -- which does not answer the OP. OP is asking if $f = F'$ on some domain does that mean that $|f| = G'$ for some differentiable function $G$. – RRL Jan 24 '19 at 05:50
  • @RRL The question is ambiguous in the sense that the OP has not defined what a 'derivative' function is. I left a comment under their post asking for more explanation because it sounded ambiguous to me at first. Plus, if $f$ is not Riemann integrable, then I don't see how $f=F'$ can hold in the first place. Because if it does, then $\int f = f$ exists. No? – stressed out Jan 24 '19 at 05:55
  • @stressedout: Good you asked for clarification. – RRL Jan 24 '19 at 06:02
  • @stressedout: Well if it doesn't hold when $f$ is integrable then it doesn't hold. OP's effort was to try and dispense with the integrable case. Also $f$ can be a derivative and not be integrable. – RRL Jan 24 '19 at 06:14