Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

A Gaussian-moment bound for the saddle remainder

Proved
RSInt.norm_rsE1_le_prime

by Johan Mercedes · Oct 28, 2026 · Mathlib 0df444a (Lean v4.33.1)

arda-rh65k-transfernumber-theoryriemann-zeta

Let a0,a,pa_0,a,pa0​,a,p be real with 9≤a0≤a9\le a_0\le a9≤a0​≤a. Put D(z)=eπiz−e−πizD(z)=e^{\pi iz}-e^{-\pi iz}D(z)=eπiz−e−πiz, Φa(w)=e2πiaw−πiw2(1+w/a)−1/2−2πia2\Phi_a(w)=e^{2\pi iaw-\pi iw^2}(1+w/a)^{-1/2-2\pi ia^2}Φa​(w)=e2πiaw−πiw2(1+w/a)−1/2−2πia2, and

E1(a,p)=∫Re2πiw2(Φa(w)−1)D(p+w)(1+i) dv,w=v(1+i),Mb(j)=∫R∣v∣je−bv2 dv.E_1(a,p)=\int_{\mathbb R}\frac{e^{2\pi iw^2}(\Phi_a(w)-1)}{D(p+w)}(1+i)\,dv, \quad w=v(1+i),\quad M_b(j)=\int_{\mathbb R}|v|^j e^{-bv^2}\,dv.E1​(a,p)=∫R​D(p+w)e2πiw2(Φa​(w)−1)​(1+i)dv,w=v(1+i),Mb​(j)=∫R​∣v∣je−bv2dv.

Using the coefficient functions cmain,ctwo,cSc_{\rm main},c_{\rm two},c_Scmain​,ctwo​,cS​ from the accompanying definitions, the principal complex power, and formal division at zero,

∣E1(a,p)∣≤12πa∑j=04cmain(j)M4π(j)+22πa2∑j=03ctwo(j)M4π(j)+2e1/(2a02)2πa2∑j=07cS(a0,j)M4π−37/10(j)+5e−πa2/18ππ/2.\begin{aligned} |E_1(a,p)|&\le\frac1{2\pi a}\sum_{j=0}^{4}c_{\rm main}(j)M_{4\pi}(j) +\frac{\sqrt2}{2\pi a^2}\sum_{j=0}^{3}c_{\rm two}(j)M_{4\pi}(j)\\ &\quad+\frac{\sqrt2e^{1/(2a_0^2)}}{2\pi a^2}\sum_{j=0}^{7}c_S(a_0,j)M_{4\pi-37/10}(j) +5e^{-\pi a^2/18}\sqrt{\frac{\pi}{\pi/2}}. \end{aligned}∣E1​(a,p)∣​≤2πa1​j=0∑4​cmain​(j)M4π​(j)+2πa22​​j=0∑3​ctwo​(j)M4π​(j)+2πa22​e1/(2a02​)​j=0∑7​cS​(a0​,j)M4π−37/10​(j)+5e−πa2/18π/2π​​.​

This integrated bound controls the saddle-point error.

Preamble
import Definitions.Def_ArdaRH65K_RS_B2
import Definitions.Def_ArdaRH65K_RS_B3
import Definitions.Def_ArdaRH65K_RS_B3Sharp
import Definitions.Def_ArdaRH65K_RS_C0
import Definitions.Def_ArdaRH65K_RS_Entire
import Definitions.Def_ArdaRH65K_RS_Fold
import Definitions.Def_ArdaRH65K_RS_Kernel
import Definitions.Def_ArdaRH65K_RS_Mellin
import Definitions.Def_ArdaRH65K_RS_Strip
import Mathlib
import Theorems.Thm_RSInt_norm_E1_integrand_near_prime
import Theorems.Thm_RSSeam_norm_E1_integrand_far
namespace RSInt
end RSInt
/-  RS_B3Sharp.lean -- lane RS, brick B3 SHARPENED (RSC0 Phase 2): the same C0-corrected
    Riemann-Siegel remainder as `RS_B3Bound` / `RSSeam_B3Bound`, with smaller kernel-checked constants.
    New theorems only; imports `RSSeam_B3Bound` (hence `RS_B3Bound`) read-only and edits nothing.

    ## The headlines (no `sorry`, no `native_decide`, no new axioms)

      `RSInt.rs_remainder_C0_alpha'` : for t >= 4250 (a = sqrt(t/2pi) >= 26), 0 < p = rsFrac t,
      cos(2 pi p) ≠ 0, and every phase phi with |phi - rsThetaMain t| <= 1/t,
          |2 Re(e^{i phi} rsRem t) - (-1)^(N+1) a^(-1/2) Psi(p)| <= (3/20) a^(-3/2);
      `RSInt.rs_remainder_C0'` : the same in t, <= (3/5) t^(-3/4).
      `RSSeam.rs_remainder_C0_alpha_seam'` : for t >= 509, <= (21/100) a^(-3/2);
      `RSSeam.rs_remainder_C0_seam'` : the same in t, <= (21/25) t^(-3/4).
      The old statements (2/5 and 2 at t >= 10000; 13/20 and 13/5 at t >= 509) are re-derived from
      the new ones as `*_of_sharp` corollaries at the end of the file.

    ## What is sharper than `RS_B3Bound` (same architecture, elementary steps)

      * the saddle exponent E is split as E = E1 + E2 with E1 = -2 pi i w^3/(3a) - w/(2a) EXACT and
        |E2| <= pi |w|^4/a^2 + |w|^2/(2a^2) (Mathlib's log-Taylor remainders of orders 4 and 2), and
        |e^E - 1| <= |E| + |E|^2 e^{|E|} (`norm_rsPhi_sub_one_le'`);
      * on the steepest-descent line |E1| = (sqrt2/2)|v| |1 - 8 pi v^2/3| / a EXACTLY (`rsEone_line`):
        the linear and cubic terms have OPPOSITE signs, and |1 - x| <= 41/40 - (5/4)x + (5/8)x^2;
        the main term then integrates against the FULL Gaussian e^{-4 pi v^2} to (49/120)/(2 pi a);
      * only the second-order remainder pays for e^{|E|}, with e^{-(4 pi - 37/10) v^2};
      * absolute Gaussian moments int |v|^j e^{-b v^2} = b^{-(j+1)/2} Gamma((j+1)/2) (`gaussMom_eq`);
      * one near-zone lemma (|v| <= a/3) for every a >= a0 >= 9, the far zone reused verbatim from
        `RSSeam.norm_E1_integrand_far`;
      * |rsJ p| <= 4/5 by completing the square on the midline (was 4) (`norm_rsJ_le'`).
      Budget: |rsE1| <= (7/100)/a (a >= 26) and (9/100)/a (a >= 9); phase 4/(5 pi a^2) <= (1/100)/a
      resp. (3/100)/a; 2 (7/100) + 1/100 = 3/20 and 2 (9/100) + 3/100 = 21/100; (2 pi)^(3/4) <= 4.

    The numeric lemmas bound a|rsE1| by 0.0689 (a >= 26) and 0.0821 (a >= 9).  Untrusted design
    numerics (closed forms 0.0688 and 0.0770): ~/rh-team/rsc0/out/phase2/lean_numerics.py.

    Not yet `lean_lib`s (lakefile.toml untouched): compile with `lake env lean`; a landing PR adds
    append-only `[[lean_lib]]` entries for RS_B3Sharp and RS_BridgeSharp (and RS_AxiomGuardSharp).

    conjecture1_proved = False.  An explicit finite-height remainder inequality; nothing here bears on
    the Riemann Hypothesis.
-/


open Complex MeasureTheory Filter Topology Set
open scoped Real ComplexConjugate

noncomputable section

section
namespace RSInt
end RSInt
open RSInt

/-! ## 1. Absolute Gaussian moments -/





































/-! ## 2. The saddle exponent split into its exact linear-plus-cubic part and a remainder -/











/-! ## 3. The near-zone majorant (|v| <= a/3, every a >= a0 >= 9) -/





































/-! ## 4. The error integral, sharpened -/









Formal statement

theorem RSInt.norm_rsE1_le_prime {a0 a p : ℝ} (h9 : 9 ≤ a0) (ha : a0 ≤ a) :
    ‖rsE1 a p‖ ≤ 1 / (2 * π * a) * (∑ j ∈ Finset.range 5, cMain j * gaussMom (4 * π) j) +
      Real.sqrt 2 / (2 * π * a ^ 2) * (∑ j ∈ Finset.range 4, cEtwo j * gaussMom (4 * π) j) +
      Real.sqrt 2 * Real.exp (1 / (2 * a0 ^ 2)) / (2 * π * a ^ 2) *
        (∑ j ∈ Finset.range 8, cS a0 j * gaussMom (4 * π - 37 / 10) j) +
      5 * Real.exp (-(π * a ^ 2 / 18)) * Real.sqrt (π / (π / 2)) := by sorry
Source
https://github.com/DrMurphyIsIn/Arda/blob/af4fde8972eb6d4e78c7d9a95933b18b81adf52d/telperion/examples/zeta_reflection/lean/RS_B3Sharp.lean#L598-L671

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me