2

Now asked on MO here


The definition of the derivative of a function $f$ at a point $x$ is : $\lim\limits_{h \to 0 }\frac{f(x+h) -f(x)}{h}$ but what if we change that $h$ to be any continuous function $g(h)$ that goes to $0$ as $h$ goes to $0$? Lets focus our attention at powers of $h$ so $g(h)=h^a$ for some $a$. If $f$ is a differentiable function at $x$ any function other than $h$ the absolute value of the derivative will go to either $0$ or $\infty$. I will refer to $\lim\limits_{h \to 0 }\frac{f(x+h) -f(x)}{g(h)}$ as $a-$derivative.

A question that came to my mind Given any continuous function nowhere differentiable on $\mathbb{R}$ is there a $g$ such that the absolute value of the $a-$derivative" exist almost everywhere and how to determine such functions ?

The case where $g(x)$ is big enough to make this $a-$derivative $0$ is annoying because one can say a very big number like $a=1/TREE(3) $ and say, for example, at this number the $a-$derivative exist and it is $0$ for the Weierstrass function,

so let me clear that : Given any continuous function nowhere differentiable what is the $\sup g(x)$ such that $\lim\limits_{h \to 0 }\frac{f(x+h) -f(x)}{g(h)}= 0$ for all $x$ where the $\sup $ is taken of all real $a$ does the absolute value of the $a-$derivative exist at that value ?

Lets take L. van der Waerden's function for example

Define $g(x)= |x|$ for $|x|\in [-1,1]$ , $g(x+2)=g(x)$ $$f(x)= \sum_{n \ge 1} \frac{3^n g\left(4^n x\right) }{4^n}$$

Is an example of continuous function nowhere differentiable, There is a proof that the absolute value secant line's slope goes to infinity as it approaches the point so this made me wonder will the absolute value of the $a-$derivative exist if we decrease the power of $h$?

pie
  • 8,483
  • 1
    "If $f$ is a differentiable function at $x$ any function other than $h$ the absolute value of the derivative will goes to either $0$ or $\infty$". I don't think so. If $f$ is differentiable at $x$ and $g$ is a function such that $g(0) = 0, g'(0) \neq 0$ then shouldn't the limit be $f'(x)/g'(0)$ ? – Thành Nguyễn Apr 17 '24 at 02:53
  • 1
    Why not change the given limit to $$\lim_{h \to 0} \frac{f(x+h)-f(x)}{h} \cdot \left(\frac {g(h)-g(0)} {h}\right)^{-1}$$ then think about the existence of $\displaystyle \lim_{h \to 0}\left(\frac {g(h)-g(0)}{h}\right)^{-1} = \left(g’(0)\right)^{-1}$? – RDK Apr 17 '24 at 02:56
  • 1
    @ThànhNguyễn I was talking about powers of $h$ I mentioned that afterwords but all of the question is meant for powers of $h$. That is a mistake from me apologies! – pie Apr 17 '24 at 02:56
  • @RDK Since the function is only $g(h)=h^a$ for some $a\neq 1$, $g'(0)=a h^{a-1}$, so either $a < 1$ and $g'(0)$ is undefined or $a > 1$ and $g'(0)=0$, so $(g'(0))^{-1}$ is undefined. – Noble Mushtak Apr 17 '24 at 03:07
  • You may be interested in Hölder continuity. – Sangchul Lee Apr 17 '24 at 04:38
  • 1

0 Answers0