7
$\begingroup$

Let $S\subset\mathbb R$ be a $G_\delta$ set. A variation on the construction of the Thomae function (which is discontinuous on the rationals and continuous elsewhere) shows that there is a function $\mathbb R\to\mathbb R$ that is continuous exactly on $S$.

I'm trying to find a published reference for this result. Note: the wikipedia page on Thomae's function mentions an equivalent result, phrased in terms of $F_\sigma$'s, without giving a reference. So it's clear the result is well-known; but I'd like a reference in an article or book (even as an exercise) rather than just the wikipedia page.

$\endgroup$
7
  • 1
    $\begingroup$ Have you tried Kechris' book? Specifically Theorem 3.8 seems relevant here. $\endgroup$
    – Asaf Karagila
    Oct 28, 2014 at 18:55
  • 1
    $\begingroup$ Are you referring to Classical Descriptive Set Theory? Theorem 3.8 is Kuratowski's theorem, and I don't see how it easily implies the result I quoted. (I'm not looking for a proof of the result, but for a source where it appears, preferably in a form that's not too distant from what I quoted.) $\endgroup$ Oct 28, 2014 at 19:17
  • $\begingroup$ Well, take $A$ to be the $G_\delta$ set and take its characteristic function. $\endgroup$
    – Asaf Karagila
    Oct 28, 2014 at 19:26
  • $\begingroup$ Sorry if I'm being dense but the theorem then returns a set $G$ possibly bigger than the starting set, and a function $G\to \mathbb R$ that is continuous on all of $G$. $\endgroup$ Oct 28, 2014 at 19:34
  • 3
    $\begingroup$ A nice argument can be found in A second course on real functions by van Rooij and Schikhof: First, prove that if $O$ is open, there is a function $f_O$ that is zero outside of $O$, and discontinuous precisely at points in $O$. Second, prove that if $F$ is an $F_\sigma$ that contains no intervals, say $F=\bigcup_n C_n$, each $C_n$ closed nowhere dense, then $f_F=\sum_n \chi_{C_n}/2^n$ is continuous precisely at points not in $F$. Third, any $F_\sigma$ set is the disjoint union of an $F_\sigma$ set $F$ that contains no intervals, and an open set $O$, and take $f=f_O+f_F$. $\endgroup$ Oct 29, 2014 at 5:53

1 Answer 1

10
$\begingroup$

The result you are looking for can be found in most advanced references in topology or analysis. Since I am away from my office at the time of this writing, I will provide you with a source readily available on the internet: Theorem 7.2 (p.30) in Oxtoby's Measure and Category.

Note that Theorem 7.2 shows that any $F_{\sigma}$ set is precisely the set of discontinuities of some real valued function on $\Bbb{R}$; the result you want the reference for is equivalent to Theorem 7.2 thanks to De Morgan Laws. This theorem, I suspect, is due to Lebesgue, but at the moment I do not have a reference at hand to corroborate my suspicion.

Addendum: According to this source (see history of Theorem 2') the above result was first proved by William H. Young in 1903, and perhaps independently by Lebesgue in 1904.

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.