I'm trying to find the best asymptotic expansions for $\pi(x; a, q)$ in various states:

  1. Unconditionally we have \begin{equation} \pi(x; a, q) = \frac{\operatorname{li(x)}}{\phi(q)} + O\left(x \operatorname{exp}\left(-c_1 \sqrt{\log x}\right)\right) \end{equation} for some constant $c_1$
  2. Under GRH we have \begin{equation} \pi(x; a, q) = \frac{\operatorname{li}(x)}{\phi(q)} + O\left(x^{1/2 + \epsilon}\right) \end{equation} for all $\epsilon > 0$.

Do we know any further terms in these asymptotic expansions and are there any other conjectures which give better error terms?

  • 7
    $\begingroup$ The first claim (Siegel-Walfisz) is only known under the additional hypothesis that $q \leq \log^A x$ for some fixed $A$ (with $c_1$ depending (ineffectively) on $A$). A more precise estimate is known if one admits the possibility of a term coming from an exceptional zero; see e.g. Davenport's "Multiplicative number theory". $\endgroup$ – Terry Tao Jan 16 '15 at 4:33
  • $\begingroup$ Ah yes, sorry. I sacrificed accuracy for brevity. I'll have a look at Davenport. $\endgroup$ – Stijn Jan 16 '15 at 4:35
  • 4
    $\begingroup$ An error term of roughly $x^\epsilon (x/q)^{1/2}$ for $q \leq x$ is predicted by the Cramer model, and was made explicitly by Montgomery (possibly with some log factor also), see ams.org/mathscinet-getitem?mr=427249 $\endgroup$ – Terry Tao Jan 16 '15 at 4:48
  • 1
    $\begingroup$ Under GRH, the error is just $O(\sqrt x\log x)$ (with explicit dependence of the constant on $q$), see e.g. Iwaniec and Kowalski. $\endgroup$ – Emil Jeřábek Jan 16 '15 at 12:30
  • 1
    $\begingroup$ See also page 426 of Montgomery-Vaughan "Multiplicative Number Theory I", particularly Conjecture 13.9. $\endgroup$ – Terry Tao Jan 16 '15 at 18:09

To the last comment of Terry Tao: For a Vinograodov-Korobov estimate in APs (proved "elementarily" no less), see PRETENTIOUS MULTIPLICATIVE FUNCTIONS AND THE PRIME NUMBER THEOREM FOR ARITHMETIC PROGRESSIONS by DIMITRIS KOUKOULOPOULOS, especially Footnote 2.



Not the answer you're looking for? Browse other questions tagged or ask your own question.