16
$\begingroup$

After this question : Does every real function have this weak continuity property?

Natrualy there are an other (more difficult) :

Is it true that for every real function $f:\mathbb{R}\to\mathbb{R}$, there exists a real sequence $(x_n)_n$, taking values different from $c$, converging to some real number $c$, such that the sequence $\left(\dfrac{f(c)-f(x_n)}{c-x_n}\right)_n$ converges ?

$\endgroup$
6
  • 6
    $\begingroup$ See auburn.edu/~brownj4/tatras.pdf and for a higher dimensional generalization arxiv.org/pdf/1108.1756.pdf $\endgroup$
    – Ashutosh
    Aug 26 '18 at 14:09
  • 4
    $\begingroup$ @Ashutosh: If I'm reading correctly, Theorem 6 in the Brown paper (which is attributed to Ceder) gives a positive answer to this question. Perhaps you'd like to post as an answer? $\endgroup$ Aug 26 '18 at 16:00
  • $\begingroup$ What you're asking is whether an arbitrary function has, at some point in its domain, a finite derived number. I initially thought the answer could be no for some functions, but the example I had in mind clearly (in retrospect) doesn't work. Perhaps the theorem @Ashutosh and Nate Eldredge mentioned answers the question. However, I don't have time to look now (in a rush to get out the door) --- for related ideas, google the phrases "derived number" and "sequential derivative". $\endgroup$ Aug 26 '18 at 22:04
  • 5
    $\begingroup$ Here's another way to see this. By Denjoy-Young-Saks theorem, for any real function $f$, the set of points $x$ for which $\lim_{h \to 0^+} |f(x + h) - f(x)|/h = \infty$ has measure zero. It follows that for for a.e. $x$, there exists an injective sequence $x_n \downarrow x$ such that $(f(x_n) - f(x))/(x_n - x)$ converges. $\endgroup$
    – Ashutosh
    Aug 27 '18 at 1:07
  • 1
    $\begingroup$ @Ashutosh: I think you have enough for a decent answer. To give a bit more detail, the DYS theorem implication you mention follows from the fact that at any a.e. such point $x$ and for each $\delta>0,$ there exists a positive real number $M$ such that infinitely many of the difference quotients with $0<h<\delta$ lie within the compact set $[-M,M].$ Also, "almost all" cannot be strengthened in the sense of size, because given any measure zero set there exists a function (even a continuous and strictly increasing function) that has an infinite derivative at each point of that measure zero set. $\endgroup$ Aug 27 '18 at 10:55
1
$\begingroup$

the answer is yes : http://webhome.auburn.edu/~brownj4/tatras.pdf

PS : the answer was in the comments, but no one gave an answer

$\endgroup$

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.