2
$\begingroup$

Consider the function

$$ f(x)=\lim_{k \to \infty}\bigg(\int_0^x \sum_{n=1}^k e^{\frac{\log n}{\log r}}~dr \bigg)\bigg( \int_0^1 \sum_{n=1}^k e^{\frac{\log n}{\log r}}~dr \bigg)^{-1} $$

I want to find a closed form for

$$\sum_{n=1}^\infty f\bigg( e^{-\sqrt{\log n}} \bigg). $$

I simplified this to

$$\sum_{n=1}^\infty f\bigg( e^{-\sqrt{\log n}} \bigg)=\sum_{n=1}^\infty \bigg( \lim_{k \to \infty}\bigg(\int_0^{e^{-\sqrt{\log n}}} \sum_{n=1}^k e^{\frac{\log n}{\log r}}~dr \bigg)\bigg( \int_0^1 \sum_{n=1}^k e^{\frac{\log n}{\log r}}~dr \bigg)^{-1}\bigg)$$

After simplyfying more I arrived at:


$$ = 1/2+1/4\bigg(\lim_{k\to\infty}\bigg( \sum_{n=1}^k e^{-2\sqrt{\log n}} \bigg)\bigg(\sum_{n=1}^k \sqrt{\log n}~K_1\big(2\sqrt{{\log n}} \big) \bigg)^{-1}\bigg). $$

Is there a closed form for the limit in parentheses?

$K_1$ is the modified bessel function.

But I'm unable to make progress on simplifying the last limit.

I did think to look at the asymptotic expansion for large $x$ here:

$$K_{\nu}(x)\sim \sqrt{\frac{\pi}{2x}}e^{-x}\Big(1+\frac{4\nu^2-1}{8x}+\frac{(4\nu^2-1)(4\nu^2-9)}{2!(8x)^2}+\ldots\Big).$$

Then some calculations based on the first 2 terms of the expansion seem to suggest $$\sum_{n=1}^\infty f\bigg( e^{-\sqrt{\log n}} \bigg)=1/2$$

because we basically have

$$ 1/2+C/4\bigg( \lim_{k\to \infty} \frac{\sum_{n=1}^k e^{-2\sqrt{\log n}}}{\sum_{n=1}^k(\log n)^{1/4}e^{-2\sqrt{\log n}}} \bigg) $$

$\endgroup$
6
  • 1
    $\begingroup$ Why do you believe a closed form exists ? $\endgroup$
    – Sam
    Commented Jan 26 at 23:47
  • $\begingroup$ @Sam The limit might go to zero obtaining a closed form of 1/2 $\endgroup$
    – zeta space
    Commented Jan 30 at 8:32
  • $\begingroup$ what it the meaning of $n$ ? is it variable of the summation or variable for a function $f(x)$ $\endgroup$
    – Faoler
    Commented Feb 10 at 14:38
  • $\begingroup$ @Faoler $n$ is the variable of the summation $\endgroup$
    – zeta space
    Commented Feb 18 at 23:21
  • 1
    $\begingroup$ What's the meaning of the $n$ in $f(\exp(-\sqrt{\log n}))$? That can't be the variable in the summation. Is it just an arbitrary positive integer? If so, then surely finding a closed form for $f(\exp(-\sqrt{\log n}))$ is just the same as finding a closed form for $f(x)$ and setting $x=\exp(-\sqrt{\log n})$. (Unless some sort of magic coincidence happens when $x$ has that particular form, but I don't see why it would.) $\endgroup$ Commented Feb 19 at 1:17

0

You must log in to answer this question.