5
$\begingroup$

While analysing the average runtime of an algorithm, I came across the following identity, and would like to know if anybody knows of any references for it?

For $i \in \mathbb{N}$, let $\bar{s}(i)$ denote the square-free part of $i$, eg., $\bar{s}(12) = 3$ (and $\bar{s}(1)=1$). Then $$ \lim_{n \rightarrow \infty} \frac{\sum_{i=1}^{n} \bar{s}(i)}{n^2} = \frac{\pi^2}{30}. $$

Many thanks.

$\endgroup$
2
  • 2
    $\begingroup$ You have $\prod_p (1+p/p^s+1/p^{2s}+p/p^{3s}+\cdots)$, alternately $1$ and $p$ in coefficients, and likely can write this as a $\zeta$ expression. $\endgroup$
    – Junkie
    Jun 8 '11 at 21:36
  • 1
    $\begingroup$ I guess it is $\zeta(2s)\zeta(s-1)/\zeta(2s-2)$. $\endgroup$
    – Junkie
    Jun 8 '11 at 21:46
1
$\begingroup$

The question asked for a reference, not a proof. A reference is Karl Greger, Square divisors and square-free numbers, Mathematics Magazine 51, No. 4 (Sept. 1978), 211-219. I make no claim that the result was unknown before Greger's paper.

$\endgroup$
0
15
$\begingroup$

If my understanding is correct, for "squarefree part" can be "squarefree kernel" in other cases, the generating Dirichlet series is $${\zeta(2s)\zeta(s-1)\over\zeta(2s-2)}=\prod_p\biggl(1+{p\over p^s}+{1\over p^{2s}}+{p\over p^{3s}}+\cdots\biggr)=\sum_n{\bar s(n)\over n^s}$$ alternating $1$ and $p$ as the coefficients, which is a $\zeta$ quotient as indicated. The residue at $s=2$ is $\zeta(4)/\zeta(2)={\pi^4/90\over\pi^2/6}={\pi^2\over 15}$, so that by Perron's formula $$\sum_{n\le X} \bar s(n)={1\over 2\pi i}\int_{(\sigma)}{\zeta(2s)\zeta(s-1)\over\zeta(2s-2)}{X^s ds\over s} \sim {\zeta(4)\over 2\zeta(2)}X^2={\pi^4/90\over2\pi^2/6}X^2={\pi^2\over 30}X^2,$$ with usual conditions about convergence in vertical strips, which are OK here. Dividing by $X^2$ gives the desired limit.

$\endgroup$
7
  • 3
    $\begingroup$ Junkie, by Perron's formula we have $$\sum_{n\le X} \bar s(n)=\int_{(3)}\frac{\zeta(2s)\zeta(s-1)}{\zeta(2s-2)}\frac{X^s}{s}\,ds\sim \frac{\pi^2}{30}X^2,$$ which is the limit conjectured by the OP. So there is no need to do any partial summation: the extra factor $2$ comes from the $s$ in the denominator of the integrand. $\endgroup$
    – GH from MO
    Jun 8 '11 at 22:41
  • $\begingroup$ Ah yes, you are right, and really there should be more fretting about convergence. $\endgroup$
    – Junkie
    Jun 8 '11 at 22:56
  • 1
    $\begingroup$ Thanks guys, but I was only asking if to anyones knowledge this had been considered before (hence reference request). It's sufficiently natural that I thought it must be classical? $\endgroup$
    – Granger
    Jun 9 '11 at 9:35
  • 1
    $\begingroup$ This is a standard exercise in analytic number theory, regardless if it had been considered before. You got a solution, use it and enjoy it. It is possible that there is a more elementary proof, using convolutions and the hyperbola method. $\endgroup$
    – GH from MO
    Jun 11 '11 at 4:20
  • 3
    $\begingroup$ @GH - I think it is standard etiquette (and saves space) to determine whether or not a result is already in print, before contributing a proof. Therefore whether a result has been considered before is important. I think you and Junkie misunderstood what I was asking. As it is, writing `this is a standard exercise' would have been a sufficient answer, so thank you. $\endgroup$
    – Granger
    Jun 13 '11 at 16:50

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.