Let $\{a_n\}$ and $\{r_n\}$ be two sequences of real numbers such that $\sum_{n=1}^\infty |a_n|< \infty.$ Prove that

$$ \displaystyle \sum_{n=1}^{\infty} \frac{a_n}{\sqrt{|x-r_n|}} $$ converges absolutely for almost every $x \in \mathbb{R}$.

Can anyone provide a useful hint to solve the problem ? I am unable to figure out how does almost every $x$ come into picture. Should I use some lebesgue integral ?

  • 1
    $\begingroup$ It seems like some property of $r_n$ should be specified. For example, what happens if $r_n \geq x$ for one or more terms $n$? $\endgroup$ – Michael Dec 13 '15 at 2:07
  • $\begingroup$ @Michael: Yeah, I forgot to add the modulus. Edited it. $\endgroup$ – pikachuchameleon Dec 13 '15 at 2:09
  • $\begingroup$ That does not seem to help. For example, what if $r_4=x$? $\endgroup$ – Michael Dec 13 '15 at 2:09
  • 1
    $\begingroup$ @Michael that may be fine, this would automatically exclude only countably many points, the question is about almost every, I assume for all but measure zero set $\endgroup$ – Mirko Dec 13 '15 at 2:11
  • $\begingroup$ @Michael: As Mirko has pointed out, you can as well remove all $\{r_n\}$ from your desired set of points for which the above property holds, since measure won't change. $\endgroup$ – pikachuchameleon Dec 13 '15 at 2:13

For any bounded measurable $A$ $$\int_A\sum_{n=1}^{\infty}\frac{|a_n|}{\sqrt{|x-r_n|}}=\sum_{n\ge 1}\int_A\frac {|a_n|}{\sqrt{|x-r_n|}}<\infty$$ Hence the sum is finite a.s. This can be extended to show a.s. absolute convergence of $$\sum_{n=1}^{\infty}a_nf_n(x)$$ for any locally uniformly integrable $f_n$.

| cite | improve this answer | |
  • $\begingroup$ What does $E$ denote ? $\endgroup$ – Mirko Dec 13 '15 at 4:23
  • 2
    $\begingroup$ I don't quite get it, can't connect the dots. I vaguely believe you need to take limit of the above inequalities as $A$ gets bigger and bigger (so $\sup A\to \mathbb R$) and then the sum may go to $\infty$. Perhaps what you have is right, but is too cryptic for me, or I overlook something related to expectations (or don't see the correct interpretation of what you are doing in terms of suitable probabilities). $\endgroup$ – Mirko Dec 13 '15 at 4:30
  • $\begingroup$ @mirko I've shown that $f<\infty$ a.s. on all bounded $A_l$. Take $A_l=[l,l+1]$ to cover all of $R$. $\endgroup$ – A.S. Dec 13 '15 at 4:40
  • 1
    $\begingroup$ I start to get the idea that perhaps it boils down to using that improper integral $\int_0^1\frac1{\sqrt{x}} dx$ is convergent. I assume your $f_n=\frac{|a_n|}{\sqrt{|x-r_n|}}$. I need to think about the "uniformly" or "locally uniformly" part (first I do not know the definition and need to figure it and make sense of it, may take a while, second once I know what it says I need to convince myself that it works, and that it would work even if the $r_n$ form a dense set). $\endgroup$ – Mirko Dec 13 '15 at 4:52
  • 1
    $\begingroup$ The nub of the answer by A.S. is :For bounded measurable $ A$ let $f_n(x)=\sum_{j\leq n}|a_n|/\sqrt {|x-r_n|}$ for $x\in A\setminus (\{0\}\cup \{r_n\}_n). $ We have $0\leq f_n\leq f_{n+1}$ a.e. on $A$, but $\sup_n\int_Af_n(x)dx<\infty.$ $\endgroup$ – DanielWainfleet Dec 13 '15 at 10:01

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.