Work

Positivity of Weil's quadratic form for Dirichlet L-functions on windows [−½ log μ, ½ log μ] from verification of the Generalized Riemann Hypothesis up to height H: μ ≤ (1 − o(1))(log qH)²

Year: 2026
Type: preprint
Abstract: This is a companion to a paper on the Riemann zeta function (doi:10.5281/zenodo.24360462). Let χ be a primitive Dirichlet character of conductor q ≥ 3, and let W_χ be Weil's quadratic form of L(s,χ), restricted to functions supported in (−½ log μ, ½ log μ) (the window μ). Let GRH_χ(H) be the statement that every nontrivial zero ρ of L(s,χ) with |Im ρ| ≤ H lies on the critical line (multiple zeros allowed). For every m ≥ e² we give an explicit height H_{m,q} and a constant A_{m,q} > 0 such that GRH_χ(H) with H ≥ H_{m,q} implies W_χ(u) ≥ ½A_{m,q}‖u‖² on every window μ ≤ m. Consequently GRH_χ(H) implies positivity on the windows μ ≤ (1 − ε)(log qH)² when log(qH) is large; for these windows no lower bound on H alone is needed, and for sufficiently large q positivity is unconditional on μ ≤ (1 − ε)(log q)². The finite-height comparisons of Sections 4 and 5 require h_χ(H) > κ^(q)(μ), where h_χ is the archimedean weight and κ^(q)(μ) is the norm of the prime part on the window; the phases of χ do not reduce this norm. With Platt's verification of GRH_χ for q ≤ 400 000, parameters checked by two separately written programs give, for example, positivity on all windows μ ≤ 74.8 for q = 3, and μ ≤ 61.2 for all primitive characters of eleven conductors up to 10007; unconditional windows, such as μ ≤ 138.3 for the odd characters of conductor 400 000, are also given. For most characters modulo a large prime q we prove positivity on the windows μ ≤ (36/25)(log q)², unconditionally. Finally, for ζ and for L(s,χ), we show that a zero off the critical line which is the only one, up to symmetry, gives a negative value of the form on explicit windows. No statement about the truth of the Riemann Hypothesis or of its generalization is made. The computer-assisted parts (the tables of windows, the examples for q = 3, and the constants in the proofs) can be rechecked with the interval-arithmetic code at doi:10.5281/zenodo.24375007, which contains two independently written programs (mpmath interval arithmetic and Arb ball arithmetic) and runs without any AI system. The finite-height windows depend on the computer-assisted verification of Platt (doi:10.1090/mcom/3077). Not yet reviewed by independent human experts. Generative AI tools were used in this work under the author's direction; see the section "Use of generative AI" in the paper. Version 2 (20 November 2026): corrections of wording after a complete rereading; no change to the theorems, proofs, tables or constants, and the code (doi:10.5281/zenodo.24375007) is unchanged. The introduction now says of the family theorem that the threshold for q is not explicit and that the proof supplies no numerical instance in the range q ≤ 400 000 (version 1 said that the threshold is very large, which is not proved). The description of which files of the first implementation are adapted from the code of the parent paper is completed. A cross-reference in the title of a remark, the order of two references, a series volume number and some sentences are corrected.
Author: Mori, Ryota
Language: English
Open Access status: green
OpenAlex ID: W7236843048
Cites: 2
Cited by:
Related to: 10