Positivity of Weil's quadratic form on windows [−½ log μ, ½ log μ] from verification of the Riemann Hypothesis up to height H: μ ≤ (1 − o(1))(log H)², and μ ≤ 243.8 at H = 3·10¹²
Weil's criterion states that the Riemann Hypothesis (RH) holds if and only if Weil's quadratic form W is nonnegative on all test functions. For μ > 1 we restrict W to functions supported in (−½ log μ, ½ log μ) (the window μ). Let PT(H) be the statement that every nontrivial zero ρ of the Riemann zeta function with |Im ρ| ≤ H lies on the critical line (multiple zeros allowed). We prove that PT(H) implies the positivity of W on a range of windows growing with H. For every m ≥ e² we give an explicit height H_m and an explicit constant A_m > 0 such that PT(H) with H ≥ H_m implies W(u) ≥ ½A_m‖u‖² on every window μ ≤ m. The supremum of the admissible m is asymptotic to (log H)²; in particular the windows μ ≤ (1 − ε)(log H)² are covered for every fixed ε > 0 and all large H, and an elementary version covers μ ≤ (¼ log H)² as soon as log H ≥ 10⁶. The leading constant 1 cannot be improved by comparisons of the kind used in the proof; the largest window on which positivity follows from PT(H) is not determined. For H = 3·10¹², for which PT(H) is the result of the computer-assisted verification of Platt and Trudgian, parameters checked in interval arithmetic give W(u) ≥ exp(−69176589181)‖u‖² on every window 1 < μ ≤ 243.8 for smooth test functions, and for C² functions supported in [−½ log 243.8, ½ log 243.8]. The proof splits the sum over zeros by an entire multiplier; the zeros up to height H contribute a nonnegative amount, and the contribution of the higher zeros is compared, frequency by frequency, with the positive part of the geometric side. A second proof, with a multiplier whose inverse Fourier transform has compact support, gives the same leading constant and W(u) ≥ exp(−63000000000)‖u‖² on the windows 1 < μ ≤ 240. A model of point sets, real up to height H, has a negative form on a window (1 + o(1))(log H)²; it is not a statement about the zeta function. No statement about the truth of RH is made. The computer-assisted parts (the weighted Schur bounds, the error budgets of Theorems 1.4 and 1.5 and Example 6.1, Table 1, and the numerical constants of the model) can be rechecked with the interval-arithmetic code at doi:10.5281/zenodo.24360460 (version 1.1.0), which runs without any AI system. The numerical case H = 3·10¹² depends on the computer-assisted verification of Platt and Trudgian (doi:10.1112/blms.12460). 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 3 (19 November 2026): the leading constant in the main theorem is improved from 4/9 to 1, and the window at H = 3·10¹² from μ ≤ 97.8474 to μ ≤ 243.8. The contribution of the zeros above height H is now compared locally in frequency with the positive part of the geometric side, which gives the smaller height H_m = 3KV⁴√(s + log K). A second proof with a compactly supported convolution kernel is added (Section 7). The statement for C² functions now includes the closed window itself (Lemma 2.10(iii)). The proposition of versions 1 and 2 on the failure of the earlier comparison for μ ≥ 97.9 concerns that comparison only and is removed; the model section now states the threshold (1 + o(1))(log H)². The positivity results of versions 1 and 2 are contained in those of version 3.
Authors
- Ryota Mori
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-19
- DOI
- https://doi.org/10.5281/zenodo.24346371
- Primary Topic
- Analytic Number Theory Research
- Type
- preprint