Multiple Dirichlet series predictions for moments of $L$-functions: unitary, symplectic and orthogonal examples
Siegfred Baluyot, Martin Čech
TL;DR
This work develops a multiple Dirichlet series (MDS) framework to predict shifted moments for four families of $L$-functions and systematically compares it to the Conrey–Farmer–Keating–Rubinstein–Snaith (CFKRS) recipe. For the unitary, symplectic, and orthogonal families, the authors demonstrate a precise one-to-one correspondence between residues of the MDS and the main terms in the CFKRS predictions, and they show that the MDS approach recovers the full CFKRS predictions with power-saving errors. In the elliptic-curve twist family, a root-number–coefficient correlation necessitates a modified version of the CFKRS recipe to align with the MDS predictions. The appendix confirms Shen’s first-moment result within this unified framework, and the paper thereby emphasizes that residues from MDS encapsulate the same information as CFKRS terms across these symmetry types, providing a unifying analytic perspective on moments of $L$-functions.
Abstract
We devise heuristics using multiple Dirichlet series to predict asymptotic formulas for shifted moments of (1) the family of Dirichlet $L$-functions of all even primitive characters of conductor $\leq Q$, with $Q$ a parameter tending to infinity, (2) the family of quadratic Dirichlet $L$-functions, (3) the family of quadratic twists of an $L$-function associated to a fixed Hecke eigencuspform for the full modular group, and (4) the family of quadratic twists of an $L$-function of a fixed arbitrary elliptic curve over $\mathbb{Q}$ that has a non-square conductor. For each of these families, the resulting predictions agree with the predictions of the recipe developed by Conrey, Farmer, Keating, Rubinstein, and Snaith, except for (4), where the recipe requires a slight modification due to a correlation between the Dirichlet coefficients and the root number of the corresponding $L$-functions. We find a one-to-one correspondence between the residues from the multiple Dirichlet series analysis and the terms from the recipe prediction.
