English

Bilinear forms with Kloosterman fractions and applications

Number Theory 2026-01-07 v2

Abstract

We establish improved bounds for bilinear forms with Kloosterman fractions of the form m,nαmβne(am/(bn)){\sum\sum}_{m,n} \alpha_m \beta_n e(a\overline{m}/(bn)) with M<m2MM<m\le 2M, N<n2NN < n \le 2N and (m,n)=1(m,n)=1. Our approach works directly with arbitrary coefficient sequences (αm),(βn)C(\alpha_m), (\beta_n) \in \mathbb{C}, avoiding the temporary restriction to squarefree support used in prior work. While this requires handling additional arithmetic complexity, it yields strictly stronger bounds that improve upon the estimates of Duke, Friedlander, and Iwaniec \cite{DFI} and Bettin-Chandee \cite{BC}; in the balanced case MNM \approx N, the new saving over the trivial bound is 1/121/12%, compared to 1/481/48 in \cite{DFI} . As an application, we prove a generalized asymptotic formula for the twisted second moment of the Riemann zeta-function with Dirichlet polynomials of length T1/2+δT^{1/2+\delta} for δ=1/46\delta = 1/46, extending beyond the previously limiting θ=1/2\theta = 1/2 barrier established by Bettin, Chandee, and Radziwi{\l}{\l} \cite{BCR}. We also establish bounds for related Hermitian sums involving Sali\'{e}-type exponential phases and develop techniques for more general bilinear forms with Kloosterman fractions.

Keywords

Cite

@article{arxiv.2601.00292,
  title  = {Bilinear forms with Kloosterman fractions and applications},
  author = {Anji Dong and Nicolas Robles and Dirk Zeindler},
  journal= {arXiv preprint arXiv:2601.00292},
  year   = {2026}
}

Comments

We accidentally missed a factor of L^2 in equation (2.53), which turns L^5 into L^7. The rest of the argument is still valid, but does not lead to an improved bound as claimed. We acknowlede Alexandru Pascadi for discovering this error so fast

R2 v1 2026-07-01T08:47:46.107Z