7
$\begingroup$

Define the sum of the non-negative numbers $\{r_s \mid s \in S\}$ $S$ uncountable to be

$$\sup _{D \subseteq S} \sum _{d \in D} r_d$$

($D$ being finite), which exists if this supremum is finite.

Define a point function to be a function from $[0, 1]$ to $\mathbb R$ that is $0$ everywhere except for a single point, where it takes a positive value.

Suppose we have an uncountable family of point functions $f_r: [0, 1] \to \mathbb R$ indexed by $r \in [0, 1]$. Define the pointwise sum function $S[a, b]: [0, 1] \to \mathbb R$ as

$$S[a, b] (x) = \sum _{r \in [a, b]} f_r (x) \ .$$

It can be shown that if $S[0, 1]$ is well defined, then so is $S[0, a]$ for any $a$ such that $0 \le a < 1$.

Assume that if $S[0, 1]$ is well defined. Does it follow that for Lebesgue almost every $a \in [0, 1]$ the function $S[0, a]$ is discontinuous at at least one point?

$\endgroup$
10
  • 4
    $\begingroup$ It’s not a homework problem. Why do you feel it belongs on stack exchange? $\endgroup$
    – Ng Ze-An
    Commented Jun 7, 2018 at 12:25
  • 4
    $\begingroup$ It’s hard to give a coherent account of what I’ve tried since I’ve used a lot of non-rigorous reasoning which didn’t end up yielding any rigorous results. Also, why do you feel it belongs on math.stackexchange.com? It’s not a homework problem, and I’ve asked a few professional mathematicians who haven’t been able to solve it. $\endgroup$
    – Ng Ze-An
    Commented Jun 7, 2018 at 12:38
  • 4
    $\begingroup$ Idk, I’ve seen textbook exercises being posted here and the response seemed fine. Most recently I saw one from Hairer’s Rough Paths book which got a very good answer. I checked the faq before posting, and it stated as long as it’s a question you’re genuinely thinking about and trying to answer, it’s okay to post it here. I’ll try to update with what I’ve been able to get so far, but tbh they’ve all been dead ends. $\endgroup$
    – Ng Ze-An
    Commented Jun 7, 2018 at 13:03
  • 3
    $\begingroup$ @GeraldEdgar "A decreasing sequence of continuous functions converging to a continuous limit converges uniformly,", right? That is, indeed, homework but I'm not sure I would give it even as a qualifier problem in such a disguise. $\endgroup$
    – fedja
    Commented Jun 7, 2018 at 15:13
  • 1
    $\begingroup$ @fedja Does that statement help solve this problem? $\endgroup$
    – Ng Ze-An
    Commented Jun 7, 2018 at 15:40

1 Answer 1

7
$\begingroup$

OK, here goes as promised.

Fix $\delta>0$ and consider the set $A_\delta$ of all points $a$ such that $S[0,a]$ is continuous and $\max f_a>\delta$ (this maximum is just that exceptional positive point value in your case but we can do arbitrary non-negative not identically $0$ functions). If we could find $a_n,a\in A_\delta$ such that $a_1>a_2>\dots\to a$, then $S[0,a_n]$ would be a decreasing sequence of continuous functions converging to the continuous limit $S[0,a]$ pointwise but not uniformly (because $\max(S[0,a_n]-S[0,a])\ge\max f_{a_n}>\delta$). Thus, for every $a\in A_\delta$, there exists an open interval $(a,b_a)$ free from points in $A_\delta$. Since we can place at most countably many disjoint open intervals on $\mathbb R$, $A_\delta$ is countable for each $\delta>0$. Hence the set of $a$ for which $S[0,a]$ is continuous (which is, say $\cup_{k\ge 1}A_{1/k}$) is also countable and, thereby, of Lebesgue measure $0$.

Just made it community wiki not to collect reputation from homeworks :-)

$\endgroup$
4
  • $\begingroup$ Edit: never mind I’m an idiot LOL $\endgroup$
    – Ng Ze-An
    Commented Jun 8, 2018 at 13:08
  • $\begingroup$ No, you aren't :lol: $\endgroup$
    – fedja
    Commented Jun 8, 2018 at 13:10
  • $\begingroup$ Cute. Do you have any positive results, showing that several $S[0,a]$ could be continuous? Obviously, $f_r(x)=\delta_r(x)$ gives that $S[0,a]$ is continuous if and only if $a=1$. But I don't see how to get anything else. $\endgroup$ Commented Jun 8, 2018 at 21:18
  • $\begingroup$ @AndrésE.Caicedo Just consider some mapping $\psi$ that maps several intervals bijectively to $[0,1]$ one after another and use $f_r(x)=\delta_{\psi(r)}(x)$. If you want a homework with a star on it, try to figure out if every countable set can appear as the set of parameters of continuity. I'll give no hints on that one, but, perhaps, somebody else will ;-) $\endgroup$
    – fedja
    Commented Jun 8, 2018 at 21:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .