Complex logarithmic approximation to gamma
DefinitionArdaRH65K_ThetaConvergearda-rh65k-transfernumber-theoryriemann-zeta
For real and nonnegative integer , with the principal complex logarithm, this bundle defines
The formal complex logarithm is total and takes . This expression is used to relate the finite real phase sum to the gamma function.
Definition code
import Definitions.Def_ArdaRH65K_ThetaValue
import Mathlib.Algebra.Ring.GeomSum
import Mathlib.Analysis.Calculus.Deriv.MeanValue
import Mathlib.Analysis.Complex.ExponentialBounds
import Mathlib.Analysis.PSeries
import Mathlib.Analysis.Real.Pi.Bounds
import Mathlib.Analysis.SpecialFunctions.Complex.Arg
import Mathlib.Analysis.SpecialFunctions.Gamma.Beta
import Mathlib.Analysis.SpecialFunctions.Integrals.Basic
import Mathlib.Analysis.SpecialFunctions.Trigonometric.ArctanDeriv
/- ThetaConverge.lean -- ANDÚRIL θ-bridge: discharging the two named obligations of
`ThetaValue.lean` (`ThetaGap.ConvergenceObligation` and `ThetaGap.RateObligation`).
These are the entire remaining distance between the merged finite `Im L_n` pipeline and
a purely-kernel-computed value box for `φ(14) = arg Γℝ(1/2+14i) = −7 log π + Λ`.
## What is proved here (all axiom-clean: {propext, Classical.choice, Quot.sound}).
* `imLnVal_succ_sub` : the exact one-step increment
`imLnVal x y (n+1) − imLnVal x y n = y·log(1+1/n) − arctan(y/(x+n+1))` (`n ≥ 1`).
* `incr_bound` : the elementary two-sided bound on that increment,
`|Δ_m| ≤ (|y|(x+1) + |y|³/3)/m²`, GENERAL in real `y` (arctan-odd symmetry),
proved from `log s ≤ s−1` (both directions) + `ArctanTaylor` cubic corollaries.
* `imLnVal_cauchy` : `CauchySeq (imLnVal x y)` -- the increments are `O(1/m²)`-summable,
so `cauchySeq_of_summable_dist` gives Cauchy, hence convergence in ℝ.
* `convergence_obligation : ThetaGap.ConvergenceObligation` -- the FULL branch bridge:
the Cauchy limit `Λ` satisfies both `imLnVal → Λ` and `Γ(z) = ‖Γ(z)‖·exp(I·Λ)`,
via `Re L_n = log‖GammaSeq z n‖ → log‖Γ z‖`, `Im L_n = imLnVal → Λ`,
`exp L_n = GammaSeq z n → Γ z`, and uniqueness of limits. No branch pinning
beyond the exp identity -- exactly as the obligation was designed to need.
* `rate_obligation : ThetaGap.RateObligation` -- the explicit `C/n` rate. The finite
tail `|imLnVal N − imLnVal n| ≤ 2B/n` (telescope + `sum_Ioo_inv_sq_le`) passes to
the limit (`le_of_tendsto`) giving `|Λ − imLnVal n| ≤ 2B/n`, `C := 2B`.
## Honest note on the rate constant / box width.
The obligation asks only for a FINITE, EXPLICIT `C`; `C = 2(|y|(x+1)+|y|³/3)` qualifies and
is what is delivered here. It is deliberately loose (the `1/m³` cubic-arctan tail is lumped
into the `1/m²` log-mismatch tail, and `∑_{m≥n} 1/m² ≤ 2/n` is the cheap telescoping bound).
A tight numeric `φ(14)` box of width ≤ 2e-3 (the driver's ambition) needs many in-kernel
arctan terms at large `n₀`; that is mechanical throughput, not further mathematics, and is
orthogonal to closing the obligations. See `theta_14_box_of_partial` for the honest,
hypothesis-free consumption at a chosen order.
conjecture1_proved = False. The θ-value convergence + rate obligations, kernel-clean.
NOT a proof of RH.
-/
open Complex Real Filter Topology ThetaValue ThetaGap
namespace ThetaConverge
/-! ### The one-step increment of `imLnVal`. -/
/-! ### The elementary two-sided increment bound (general real `y`). -/
/-! ### Convergence: `imLnVal` is Cauchy, hence convergent. -/
/-! ### The complex `L_n` and its bridges to `imLnVal` / `GammaSeq`. -/
/-- The complex Weierstrass log-sum `L_n` at `z = x + y·I`. -/
noncomputable def Lc (x y : ℝ) (n : ℕ) : ℂ :=
((x:ℂ) + (y:ℂ)*I) * Complex.log (n:ℂ) + Complex.log ((Nat.factorial n : ℕ):ℂ)
- ∑ k ∈ Finset.range (n+1), Complex.log (((x:ℂ) + (y:ℂ)*I) + (k:ℂ))
/-! ### The convergence obligation (with the branch bridge). -/
/-! ### The rate obligation (explicit `C/n`). -/
/-! ### The payoff: the kernel `φ(14)` box (assembly of both obligations).
Both obligations discharged, we consume them through `ThetaGap.phi_box_of_rate_and_partial`.
Ingredients at the operating point `x = 1/4`, `y = 7`, order `n₀ = 1`:
* `imLnVal (1/4) 7 1 = −(arctan 28 + arctan(28/5)) ∈ [−4, 0]` (crude rational partial box);
* `log π ∈ [1, 2]` (crude rational box; `e < π < e²`);
* `Γ(1/4+7i) ≠ 0` (`im = 7 ≠ 0`).
The resulting `theta_14_box` is a genuine, hypothesis-free, argument-free kernel theorem.
HONEST WIDTH NOTE. The interval width is dominated by the rate term `C/n₀` with the loose
(but valid) constant `C = 2(|y|(x+1)+|y|³/3) ≈ 246` at `n₀ = 1`, so the numeric width here is
large (~5e2). Narrowing to the driver's ≈1e-3 ambition is pure per-term arctan-boxing
throughput at large `n₀` (mechanical `ArctanTaylor.arctan_bracket` iteration) plus a tightened
`C` — orthogonal to the mathematical content, which is the two obligations proved above. -/
end ThetaConverge
def ArdaRH65K_ThetaConverge : Nat := 0
Source