Derivative of the saddle-point phase
ProvedRSInt.saddle_hasDerivAtarda-rh65k-transfernumber-theoryriemann-zeta
Let and be real, and define
With the principal complex logarithm,
This derivative controls the oscillatory phase near the saddle point.
Preamble
import Definitions.Def_ArdaRH65K_RS_B2
import Definitions.Def_ArdaRH65K_RS_B3
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
/- RS_B3Bound.lean -- lane RS, brick B3 (the explicit Riemann-Siegel remainder), part 3: the
analytic estimates and THE B3 HEADLINE. Mathlib-only (imports `RS_B3`).
## The headline (no `sorry`, no `native_decide`, no new axioms)
`rs_remainder_C0` : for t >= 10000, 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) (t/2pi)^(-1/4) Psi(p)| <= 2 t^(-3/4),
where rsRem t is the EXACT Riemann-Siegel remainder of B2 and Psi = Gabcke's C0. With B2's
`completed_re_eq_rs_cos` this is the classical
Z(t) = 2 sum_{n<=N} n^(-1/2) cos(theta - t log n) + (-1)^(N-1) (t/2pi)^(-1/4) C0(p) + R,
|R| <= 2 t^(-3/4) (t >= 10000),
with an explicit, kernel-checked constant (Gabcke's sharper 0.127 is NOT claimed).
`rs_remainder_C0_alpha` is the same bound in the saddle variable a = sqrt(t/2pi):
|...| <= (2/5) a^(-3/2).
## The estimates behind it
* `integral_sq_mul_exp_neg_mul_sq` : int v^2 e^{-b v^2} dv = sqrt(pi/b) / (2b).
* `norm_rsD_ge_two_pi_im` : |rsD z| >= 2 pi |Im z|; `two_le_norm_rsD_midline` : |rsD| >= 2 on
the line Re x - Im x = 1/2 (sin^2 + sinh^2 >= 1).
* `norm_rsPhi_sub_one_le` : for |w| <= a/2 the saddle-point amplitude satisfies
|rsPhi a w - 1| <= e e^e, e = |w|/(2a) + |w|^2/(2a^2) + (4 pi/3)|w|^3/a
(complex Taylor bounds for log(1+u), Mathlib `norm_log_sub_logTaylor_le`).
* `saddle_arg_le` : along the steepest-descent line, 2 pi a^2 arg(1 + v(1+i)/a) - 2 pi (a+v) v
<= -pi v^2 for every real v (a derivative-sign argument: the saddle is a global maximum);
`norm_gauss_rsPhi_le` : |e^{2 pi i w^2} rsPhi a w| <= 2 e^{-pi v^2} on w = v(1+i).
* `rsJ_eq_mid`, `norm_rsJ_le` : the tau = 2 Mordell integral moves (strip Cauchy, no pole) to the
midline through 1/2 - p, where |rsJ p| <= 4 uniformly in p ∈ (0,1).
* `norm_E1_integrand_near` (|v| <= a/9) / `norm_E1_integrand_far` (|v| >= a/9), `norm_rsE1_le`,
`E1_const_le` : |rsE1 a p| <= (9/50)/a for a >= 39.
conjecture1_proved = False. An explicit remainder inequality for the Riemann-Siegel formula at
finite height; 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. The second Gaussian moment -/
/-! ## 2. Lower bound for the denominator -/
/-! ## 3. The amplitude near the saddle point -/
/-! ## 4. The saddle point is a global maximum along the steepest-descent line -/
Formal statement
theorem RSInt.saddle_hasDerivAt {a : ℝ} (ha : 0 < a) (v : ℝ) :
HasDerivAt (fun v : ℝ => 2 * π * a ^ 2 * (Complex.log (1 + v * (1 + I) / a)).im -
2 * π * (a + v) * v + π * v ^ 2)
(-(2 * π * v * (3 * a ^ 2 + 4 * a * v + 2 * v ^ 2) / ((a + v) ^ 2 + v ^ 2))) v := by sorry
Source