Let λ_0(c) > λ_1(c) > ⋯ be the eigenvalues of the time and frequency limiting operator with kernel sin c(x−y)/(π(x−y)) on [−1,1]. Fuchs' asymptotic formula 1 − λ_n(c) ~ 4√π 8^n cn+1/2 e⁻²ᶜ/n! describes the defect of the top eigenvalues as c → ∞. Rigorous non-asymptotic bounds in the literature give the decay rate e⁻ᶜ. We prove that for 0 ≤ n ≤ 4 and every c ≥ 10π, 1 − λ_n(c) ≤ C_n cn+1/2 e⁻²ᶜ, with explicit constants C_n. The exponent e⁻²ᶜ and the power cn+1/2 are those of Fuchs' formula. The proof combines Slepian's identity ∂_c log λ_n = 2ψ_n(1)²/c with a sharp upper bound for the mass of the band-limited extension of ψ_n outside [−1,1]: this mass is at most (ψ_n(1)²/c)(2/π)K(χ_n/c²)(1+15/c²), where K is the complete elliptic integral. The bound comes from a Bessel-type Liouville transformation at the regular singular point t = 1, in which the first-order error cancels. All constants are established with ball arithmetic. In a companion paper, the case n = 4 is used to bound the bottom of windowed Weil quadratic forms, which are related to the Riemann Hypothesis through Weil's criterion. - **Notes:** Written with extensive use of generative AI systems (Anthropic Claude, Google Gemini, OpenAI GPT) under the direction of the author; see the section "Use of generative AI" in the paper. Not yet reviewed by independent human experts. **Reserved DOIs (2026-10-02):** paper I 10.5281/zenodo.23092454, paper II 10.5281/zenodo.23092542. Final PDFs: release v1.0.1 of prolate-weil-bounds (paper-I/main.pdf md5 e3ebfa3387f6e59fd5946903aa0e2d09, paper-II/main.pdf md5 b1d41435d6eb19c36359a879ac70045a).
No takes yet. Share an insight, caveat, or question.
Ryota Mori (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: