5
$\begingroup$

Definitions and some motivation:

Thomae’s function, also known as the raindrop function has several curious properties. One of which is the following - both its discontinuity set and continuity set are dense in $[0, 1]$, and also every discontinuity point is a removable discontinuity.

Can this happen in a “measure theoretic“ sense as well? In the sense that all the above conditions hold modulo null sets, in a way that is made precise below.

Question set up:

For a measurable function $f:\mathbb [0, 1] \to \mathbb R$, we say that $x \in \mathbb R$ is a essentially removable discontinuity of $f$ if

i) $f$ is not continuous at $x$,

ii) There exists a measure zero set $N$, and a measurable function $g: \mathbb [0, 1] \to \mathbb R$ such that $g$ is continuous at $x$, and $g$ agrees with $f$ everywhere except possibly on $N$.

Note that here both $N$ and $g$ are allowed to depend on $x$.

Question: Denote by $\mathcal E$ the set of essentially removable discontinuities of $f$, and $\mathcal C$ the set of continuity points respectively. Does there exist a function f such that $[0, 1]= \mathcal E \cup \mathcal C$ and such that both $\mathcal E$ and $\mathcal C$ have nonzero measure in any open interval?

$\endgroup$
8
  • 2
    $\begingroup$ Places to look (I don't have time now to dig into this): Henry Blumberg's papers (especially late 1910s through early 1920s, and measurable boundaries paper, and exceptional sets paper), Ákos Császár's early papers on negligent limits (e.g. 1 and 2 and 3 and 4), and Chapter 3 of Thomson's book. $\endgroup$ Commented Jun 15, 2021 at 8:31
  • 1
    $\begingroup$ Comment on the terminology: typically one defines a discontinuity as essential on the real line if at least one of the one-sided limits fail to exist. The choice to call what you are defining an "essentially removable discontinuity" is a somewhat unfortunate clash. $\endgroup$ Commented Jun 15, 2021 at 14:50
  • $\begingroup$ I managed to look around a bit a few hours after my last comment and found some items and results that are more relevant, but I'm not going to have time to discuss them now and I also do not (yet) know the answer to your specific question. I'll try to write some kind of answer tomorrow, even if it only discusses what follows in a bit more detail. For now I'll mention that your generalized continuity condition is sometimes called "measure continuous", which unfortunately is a useless phrase for internet searching purposes. The notion is weaker than ordinary continuity (clear) (continued) $\endgroup$ Commented Jun 15, 2021 at 16:52
  • $\begingroup$ and stronger than approximate continuity 1 2 (less clear). Moreover, if a function is everywhere measure continuous then it is everywhere continuous, which incidentally is not true for approximate continuity. However, in your case we don't want a function that is everywhere measure continuous. I believe you want both $C$ and $MC - C$ to be metrically dense sets. $\endgroup$ Commented Jun 15, 2021 at 16:52
  • 1
    $\begingroup$ Update: A little over a week ago my contract work went from a nearly nonexistent trickle (past 2 months) to a wide open deluge, and there is also a huge amount of material I have on hand to sort through, so this will take a few more days. However, I think the result will be a useful survey of a lot of very little known literature, some of it I haven't thought about for 25+ years and some of it I've had unresolved questions about (from the 1990s) that I've never gotten around to sorting out (but I'm trying to now). $\endgroup$ Commented Jun 19, 2021 at 19:02

0

You must log in to answer this question.

Browse other questions tagged .