This question concerns the asymptotic maximum of $L(1,\chi_d)/ \log \log d$ when $d$ is even. I computed it for even $d$ less than 7.5 million. The largest value is $1.230126$, occurring for $d= 2552806$. This value is never exceeded for the next five million $d$. Languasco (arXiv:2005.04664v2 [math.NT] 26 May 2020) made similar numerical computations without assuming $d$ is even and got a much larger maximum of 2.21. Can anyone explain why $L(1,\chi_d)$ should be smaller for even $d$? In particular, for infinitely many $d$ we know (on GRH maybe) that the value $e^{\gamma} = 1.781\ldots $ is exceeded, but apparently not by even $d$.

  • $\begingroup$ Also, if someone can kindly point me to instructions for including an image with a question, I would post a cute graph here. $\endgroup$ Commented Nov 18, 2022 at 16:47
  • 6
    $\begingroup$ One heuristic model for the distribution of $L(1,\chi_d)$ is a product of Euler factors $(1 - \chi(p)/p)^{-1}$ where each $\chi(p)$ is chosen independently from $\{-1,0,+1\}$ with appropriate probabilities. The $p=2$ factor has the largest effect, and can be $2$, $1$, or $2/3$. Requiring $2|d$ makes that factor $1$, which we can expect to roughly halve the maximal $L(1,\chi_d)$. $\endgroup$ Commented Nov 18, 2022 at 17:00
  • $\begingroup$ For images, see the last paragraphs at mathoverflow.net/help/formatting $\endgroup$ Commented Nov 18, 2022 at 22:07


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.