8
$\begingroup$

Suppose $f$ is a cusp form of half integral weight $k$ w.r.t. the group $\Gamma_0(4)$ ($k$ is not very low, can assume $k \ge 11/2$), and $a_n$ is its Fourier coefficient. The Linnik bound says that if $f$ is square-free then we have the estimate $$ |a_n| \ll_\epsilon n^{k/2-2/7+\epsilon} $$

My question is, is there a nontrivial bound for $\left| \sum_{n=N}^{N+H} a_n \right|$? Here nontrivial means better than simply adding up the Linnik bound individually. Also will averaging take care of the square-free issue in the Linnik bound?

An equivalent problem is to estimate $$ \sum_{n=N}^{N+H} \sum_{4|c} \frac{K(m,n,c)}{c} $$

Note: $m$ is assumed to be fixed and is not averaged. To give some order of magnitude, $H$ is supposed to be around $N^{1/5}$, so it is a rather short sum.

Improve this question
$\endgroup$
5
  • 5
    $\begingroup$ I think it is misleading to use the term "Linnik bound" as it was proved by Iwaniec in 1987, while Linnik died in 1972. This bound was improved several times, e.g. by Blomer-Harcos to $n^{k/2-5/16+\epsilon}$, see Corollary 2 in J. reine angew. Math. 621 (2008), 53-79. The bound is true as long as the primes dividing the level have bounded exponents in $n$ (which includes all square-free $n$'s, but many more). $\endgroup$
    – GH from MO
    Commented Jul 15, 2013 at 9:07
  • 3
    $\begingroup$ I don't have time to work on this, but the convexity bound for the Dirichlet series $\sum_n a_n/n^s$ (which is entire and satisfies a functional equation) coupled with Mellin transform techniques should yield better cancellation for your sum than the "Linnik bound" coupled with the triangle inequality. On the other hand, this will only work for some $H>N^{1/2}$, I am afraid. $\endgroup$
    – GH from MO
    Commented Jul 15, 2013 at 9:28
  • $\begingroup$ Thanks to the above. Just FYI, I'm actually balancing this estimate with another estimate, with $H$ an adjustable parameter. As per the bound in Blomer-Harcos, $H$ is now $N^{11/52}$. The better the bound is, the larger the $H$, so I'm still curious whether averaging in $n$ could take advantage of that. $\endgroup$
    – Fan Zheng
    Commented Jul 15, 2013 at 15:32
  • $\begingroup$ I am glad I could help. If you have a preprint (later in the future), please post the link here, as the topic is interesting. $\endgroup$
    – GH from MO
    Commented Jul 15, 2013 at 15:41
  • $\begingroup$ So, if we assume $H \gg N^{1/2}$, what kind of result will we get? I'm trying to decide if it will actually pushes $H$ that large, so that it is self-fulfilling. $\endgroup$
    – Fan Zheng
    Commented Jul 18, 2013 at 17:07

1 Answer 1

Reset to default
3
$\begingroup$

The paper of Kohnen-Zagier, "Values of L-series of Modular Forms at the Center of the Critical Strip" (http://gdz.sub.uni-goettingen.de/dms/load/img/?PPN=GDZPPN002097249&IDDOC=172193), Invent. math. 64 (1981) 175-198 discusses the (weighted) average of the square absolute value of Fourier coefficients of half-integral weight cusp forms. See for example Corollary 6.

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .