I am trying to prove that, given an arbitrary function $f$, the set of discontinuities of $f$, denoted $D_f=[{x\in{\Bbb{R}}:\mbox{f is not continuous at $x$}}$], is a $F_\sigma$ set - that is, a countable union of closed sets. This question may have been asked before, but after searching thoroughly, this is the only one to approach it in a certain manner with specific goals (taken from Stephen Abbott's Understanding Analysis).

First, a crucial piece of terminology (for those who do not know it): "Let $f$ be defined on $\Bbb{R}$, and let $\alpha>0$. The function $f$ is said to be $\alpha$-continuous at $x\in{\Bbb{R}}$ if there exists a $\delta>0$ such that for all $y,z\in{V_\delta{(x)}}$ it follows that $|f(y)-f(z)|<\alpha$. Also, we let $D_\alpha=[x\in{\Bbb{R}}:\mbox{f is not $\alpha$-continuous at $x$}]$

There are 4 steps in this proof, 3 of which I have already proved:

1) Show that, for a fixed $\alpha>0$, $D_\alpha$ is closed (already proved myself)

2) Show that, if $\alpha_1<\alpha_2$, then $D_{\alpha_2}\subset{D_{\alpha_1}}$ (already proved myself)

3) Let $\alpha>0$ be given. Show that if $f$ is continuous at $x$, then it is also $\alpha$-continuous at $x$. Show therefore that $D_\alpha\subset{D_f}$ (already proved myself)

4) Show that if $f$ is not continuous at $x$, then it is not $\alpha$-continuous for some $\alpha>0$. Show why this guarantees that $D_f=\bigcup_{n=1}^{\infty}D_{1/n}$. As, by point 1, each $D_{1/n}$ is closed, it thus follows that $D_f$ is a $F_\sigma$ set.

Now, I have proved the first sentence of point 4 myself, however, the bit where I run into trouble is how to show that this guarantees that $D_f=\bigcup_{n=1}^{\infty}D_{1/n}$. My idea was that this is true because continuity and $\alpha$-continuity are equivalent, so we simply take the union over all $\alpha$, and by point 2, the only $\alpha$ that 'matter' are those infinitesimally close to 0, and so as (1/n) converges to 0, all possible $\alpha$ infinitesimally close to 0 are 'covered'. This is a very loose argument that I would like to make formal, however, this only allows for $\alpha$ of the form $1/n$ even though $\alpha$ can be an irrational number... Please could someone try and explain how this is guaranteed thus (out of this v. long proof only bit I don't get bloody hell).

Thank you.

EDIT: $f$ is a function from the reals to the reals (although this argument can easily be extended to general complete metric spaces)

  • $\begingroup$ Based on your description, $f: \mathbb{R} \rightarrow \mathbb{R} $ ? $\endgroup$ – ertl Aug 1 '18 at 22:56
  • $\begingroup$ @erti Yes, sorry for not making that clear, I'll edit that in now $\endgroup$ – Daniele1234 Aug 1 '18 at 22:58
  • 1
    $\begingroup$ If $x \in D_f$ then $x \in D_\alpha$ for some $\alpha>0$ and hence by choosing ${1 \over n} \le \alpha$ we have $x \in D_{1 \over n}$, hence $x $ is in the union. $\endgroup$ – copper.hat Aug 1 '18 at 23:01
  • $\begingroup$ @copper.hat Oh yes seems rather trivial now, thanks for that. $\endgroup$ – Daniele1234 Aug 1 '18 at 23:10

If $0<b\leq a$ then $D_b\supset D_a.$ For any $a>0$ let $n(a)$ be the least (or any) $n\in \Bbb N$ such that $1/n\leq a.$ Hence $D_a\subset D_{n(a)}.$ Therefore $$D_f=\cup_{a>0}D_a\subset \cup_{a>0}D_{1/n(a)}\subset \cup_{m\in \Bbb N}D_{1/m}.$$ But since $\{1/m:m\in \Bbb N\}\subset \{a:a>0\}$ we also have $$\cup_{m\in \Bbb N}D_{1/m}\subset \cup_{a>0}D_a=D_f.$$

| cite | improve this answer | |
  • $\begingroup$ There is a converse : If $D$ is an $F_{\sigma}$ subset of $\Bbb R$ there exists $f:\Bbb R\to \Bbb R$ such that $D=D_f.$ $\endgroup$ – DanielWainfleet Aug 3 '18 at 19:23

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.