2

This question is continuation of this question. (And last of this type)

Let $f: [a,b]\rightarrow \mathbb{R}$, where $f(x)$ is monotonically non-decreasing positive continuous function and differentiable on $(a,b)$ or simply $f'(x)\geq 0$ for $x\in (a,b)$.

can we find such function for which $f'(x)$ is not continuous?

I could not find any,obviously. Is there such function even exists!

user467365
  • 1,223
  • 1
  • 11
  • 16

3 Answers3

2

Take $x\in(-1,1)$ $$ f(x) =100000+ 100x+ x^2\sin\frac{1}{x} \qquad\text{with $f(0)=100000$}$$ which is differentiable, and $$ f'(x) = 100+ 2x\sin\frac{1}{x}-\cos\frac{1}{x} \qquad\text{with $f'(0)=100$}$$ $f'$ is not continuous a $x=0$ and you might choose the interval $[a,b]$ around $x=0$ as it suite to you. Indeed

$$ f'(x)=100+2x\sin\frac{1}{x}-\cos\frac{1}{x} \ge 100+ 2x\sin\frac{1}{x}-1$$

Since $|\sin a| \le |a|$ and $-1\le-\cos a\le 1$ then $|2x\sin\frac{1}{x}|\le 2$ i.e $$2x\sin\frac{1}{x}\ge -2 $$

therefore

$$ f'(x)=100+2x\sin\frac{1}{x}-\cos\frac{1}{x} \ge 100-2-1= 97>0$$ for every $x\in \mathbb R$. so $f $ is increasing and $f'$ is not continuous at $x=0$.

Also $|x^2\sin\frac{1}{x}|\le 1$ for all $x\in (-1,1)$ so that $$f(x) =100000+ 100x+ x^2\sin\frac{1}{x} > 100000-100 -1 >0$$ for every $x\in (-1,1).$

Guy Fsone
  • 25,237
  • thanks again,... – user467365 Aug 17 '17 at 14:30
  • 1
    I don't want to be overly critical, but what is the point of reproducing pretty much exactly the answer to the previous question with all the analysis, while it suffices to explain "take the same function and add a constant, since all that was missing was $f>0$"? – Clement C. Aug 17 '17 at 14:30
0

Define $f(x)=1+3x+x^2\sin\left(\frac1x\right)$ if $x\in(0,1]$ and $f(0)=1$. It satisfies your conditions, but $f'$ is discountinuous at $0$.

0

Take the function $f$ from the question you link. It is already a monotonically increasing continuous function and differentiable on $[a,b]$, yet $f'$ is not continuous. It is only missing positivity.

Make it positive by adding $\lVert f\rVert_\infty+1$ (possible since $f$ is continuous on $[a,b]$, so bounded). This does not change the derivative.

You then have an answer to your new question.

Clement C.
  • 68,437