Integrability of an upper-kernel product on a diagonal line
ProvedRSInt.integrable_lineUp_rsG_mularda-rh65k-transfernumber-theoryriemann-zeta
Let be real, with not an integer, and put and . Suppose is continuous at every point of the strip , and real constants satisfy throughout that strip. Then
is integrable over .
This justifies the upper-contour integral under a mild growth bound.
Preamble
import Definitions.Def_ArdaRH65K_RS_Kernel
import Definitions.Def_ArdaRH65K_RS_Strip
import Mathlib
import Theorems.Thm_RSInt_integrable_of_continuous_of_tail
import Theorems.Thm_RSInt_norm_rsG_mul_le
/- RS_Kernel.lean -- lane RS, ANDURIL brick B2 (the Riemann-Siegel integral), part 2:
Riemann's kernels, their difference equations, Gaussian bounds, the Fresnel line integral,
the pole-crossing lemmas and the Mordell closed form.
Mathlib-only (imports `RS_Strip` only).
## The kernels
rsD x = e^{i pi x} - e^{-i pi x} (= 2 i sin(pi x); zeros exactly at the integers)
rsG x = e^{ i pi x^2} / rsD x (Riemann's kernel of the zeta integral)
rsH x = e^{-i pi x^2} / rsD x (the kernel of the chi-side integral)
## Contents (no `sorry`, no `native_decide`, no new axioms)
* `rsD_eq_zero_iff`, `rsD_add_one`, `rsG_add_one`, `rsH_add_one` : zeros and the difference
equations rsG(x+1) = rsG x + e^{i pi (x^2+x)}, rsH(x+1) = rsH x - e^{-i pi (x^2+x)}.
* `norm_rsG_le`, `norm_rsH_le` : on a slanted strip, away from the real axis (|Im x| >= 1), the
kernels are bounded by exp(-2 pi (Im x)^2 + 2 pi D |Im x|) (Gaussian decay).
* `lineUp_fresnel`, `lineUp_gauss` : the Gaussian line integrals, e.g.
lineUp c (x |-> e^{i pi (x^2+x) - x y}) = e^{y/2 + i y^2/(4 pi)} (every c, every y).
* `lineUp_same_gap`, `lineUp_cross` : moving the slope-(+1) line inside a gap (n, n+1) changes
nothing; moving it across the pole at the integer n adds phi(n):
lineUp c2 (rsG * phi) = lineUp c1 (rsG * phi) + phi n, n-1 < c1 < n < c2 < n+1,
for every phi holomorphic on the strip with at most exponential growth. No residue theorem
is used: the pole is removed by `dslope`, and the jump of the bare kernel is computed from
the difference equation and the Fresnel integral.
* `lineDn_same_gap`, `lineDn_cross` : the slope-(-1) analogues for rsH (by conjugation).
* `lineUp_mordell` : the Mordell closed form, for 0 < c < 1 and every y:
lineUp c (x |-> rsG x e^{-x y}) (e^{-y/2} - e^{y/2}) = e^{-y/2} - e^{i y^2/(4 pi)}.
conjecture1_proved = False. Contour identities for one explicit kernel; nothing here bears
on RH.
-/
open Complex MeasureTheory Filter Topology Set
open scoped Real
noncomputable section
section
namespace RSInt
end RSInt
open RSInt
/-! ## 1. The kernels and their algebra -/
/-! ## 2. Norm bounds -/
/-! ## 3. The Fresnel line integral -/
/-! ## 4. Gaussian tails, line integrability, and the strip wrappers -/
/-! ## 5. Linearity and translation of line integrals -/
/-! ## 6. Holomorphy of the kernels off the integers -/
/-! ## 7. The Fresnel/Gaussian integrands are integrable along the lines -/
/-! ## 8. Moving the line inside a gap, and across a pole -/
Formal statement
theorem RSInt.integrable_lineUp_rsG_mul {φ : ℂ → ℂ} {c₁ c₂ c : ℝ} (hc : c ∈ Icc c₁ c₂)
(hcl : ∀ m : ℤ, (m : ℝ) ≠ c)
(hφ : ∀ x : ℂ, x.re - x.im ∈ Icc c₁ c₂ → ContinuousAt φ x)
{A B : ℝ} (hgrow : ∀ x : ℂ, x.re - x.im ∈ Icc c₁ c₂ → ‖φ x‖ ≤ A * Real.exp (B * ‖x‖)) :
Integrable (fun v : ℝ => rsG (c + v * (1 + I)) * φ (c + v * (1 + I))) := by sorry
Source