Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Digamma-log asymptotic along positive integers

Proved
ZeroFreeBridge.tendsto_digamma_sub_log_one_add_nat

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

arda-rh65k-transfernumber-theoryriemann-zeta

The complex digamma function and principal logarithm satisfy

lim⁡N→∞(ψ(1+N)−log⁡(1+N))=0,\lim_{N\to\infty}\bigl(\psi(1+N)-\log(1+N)\bigr)=0,N→∞lim​(ψ(1+N)−log(1+N))=0,

where NNN ranges over the natural numbers.

Preamble
import Definitions.Def_ArdaRH65K_Turing121_DlvpTheta
import Mathlib
/- TURING RUNG T2, brick 1: the Riemann–Siegel theta function, defined BRANCH-CUT-FREE as the
   integral of an explicit real integrand (TURING_METHOD_SCOPING §3, rung T2).

   Classically `θ(t) = Im logΓ(1/4 + it/2) − (t/2)·log π` with a continuously-tracked branch of
   `logΓ` — awkward to formalize.  Instead we DEFINE

     `θ(t) := ∫_0^t ( (1/2)·Re ψ(1/4 + i·u/2) − (1/2)·log π ) du`,

   where `ψ = Complex.digamma = logDeriv Γ` (Mathlib).  This agrees with the classical `θ`
   (both vanish at `0` and have the same derivative `θ'(u) = (1/2)·Re ψ(1/4 + iu/2) − (1/2)·log π`)
   and needs NO branch tracking: the integrand is a plain continuous real function of `u`.

   This brick: the ray `1/4 + (u/2)·i` avoids all Γ-poles (`Re = 1/4 > 0`), Γ is ANALYTIC there
   (differentiable on the open right half-plane ⟹ analytic), `ψ` is continuous on the ray
   (`deriv Γ / Γ`, with `Γ ≠ 0` by `Gamma_ne_zero_of_re_pos`), hence the integrand is continuous,
   `θ` is well-defined with `θ(0) = 0` and `θ' = integrand` (FTC for continuous integrands).

   Later bricks (T2 cont.): effective two-sided bounds `θ(t) = (t/2)log(t/2π) − t/2 − π/8 +
   O*(explicit/t)` via Binet-type digamma estimates; then T3 wires `θ` into the rectangle
   argument identity `N(T) = θ(T)/π + 1 + S(T)`.

   conjecture1_proved = False. -/


open Complex

section
namespace ZeroFreeBridge
end ZeroFreeBridge
open ZeroFreeBridge























/-! ## Brick 2: the Archimedean bridge — `θ'` IS the `Gammaℝ` log-derivative on the critical line.

`Gammaℝ s = π^(-s/2)·Γ(s/2)` is the Archimedean factor of the completed zeta
(`Λ = Gammaℝ · ζ` up to the `s(s-1)` polynomial), the object the T3 rectangle identity
splits off of `logDeriv Λ`.  We compute `logDeriv Gammaℝ s = -(log π)/2 + (1/2)·ψ(s/2)`
on `Re(s/2) > 0` and specialize to the critical line `s = 1/2 + iu`, where `s/2 = thetaRay u`:
the θ integrand is EXACTLY `Re (logDeriv Gammaℝ (1/2 + iu))`.  This hard-wires the
branch-cut-free `θ` into the counting formula's Archimedean term. -/















/-! ## Brick 3a: the θ-asymptotics skeleton (T2 rung, first sub-brick).

Target (brick 3 complete): `θ(t) = (t/2)·log(t/(2π)) − t/2 − π/8 + O*(explicit/t)`.
This sub-brick ships the three UNCONDITIONAL pieces every derivation route needs:

  * `digamma_shift` — the iterated recursion `ψ(s+N) = ψ(s) + Σ_{k<N} (s+k)⁻¹` (from Mathlib's
    single-step `digamma_apply_add_one`), which moves the evaluation point right, where the
    Stirling comparison is easy;
  * `norm_inv_sub_log_one_add_inv_le` — the telescoping step bound
    `‖s⁻¹ − log(1+s⁻¹)‖ ≤ ‖s⁻¹‖²·(1−‖s⁻¹‖)⁻¹/2` (Mathlib `norm_log_one_add_sub_self_le`):
    the summand of the Binet-series `ψ(s) − log s = −Σ_{k≥0} [(s+k)⁻¹ − log(1+(s+k)⁻¹)] + …`
    is quadratically small, so the telescoped series converges with an explicit `O(1/‖s‖)` tail;
  * `thetaMain` + `thetaMain_hasDerivAt` — the main term and its derivative
    `(log(t/(2π)))/2`, matching `θ'`'s leading behaviour, so brick 3c can compare the two via
    FTC on `[t₀, t]`.

Remaining sub-bricks: 3b = sum the telescoped series (Summable + tail bound ⟹ effective
`|ψ(s) − log s| ≤ C/‖s‖` on `Re s ≥ 1`, plus the anchor `ψ(s+N) − log(s+N) → 0`); 3c = integrate
`θ' − thetaMain'` and fix the `−π/8` constant.  conjecture1_proved = False. -/













/-! ## Brick 3b: the telescoped Binet series — `|ψ(s) − log s| ≤ 1/(Re s − 1)` modulo the anchor.

Telescoping `d(s) = ψ(s) − log s` with `d(s+1) − d(s) = s⁻¹ − log(1+s⁻¹)` (branch-safe on
`Re s > 0` since `|arg| < π/2` on the right half-plane) gives the FINITE identity
`d(s) = d(s+N) − Σ_{k<N} [(s+k)⁻¹ − log(1+(s+k)⁻¹)]`; the 3a step bound + an elementary
inline telescoping estimate `Σ_{k<N} 1/(x+k)² ≤ 1/(x−1)` bound the sum UNIFORMLY in `N`;
letting `N → ∞` against the anchor `d(s+N) → 0` (the ONE remaining analytic input, carried as
an explicit hypothesis — brick 3b′ discharges it) yields the effective Binet bound.
conjecture1_proved = False. -/













/-! ## Brick 3b′-i: the INTEGER anchor — `ψ(1+N) − log(1+N) → 0`.

The first of the anchor's three sub-bricks (3b′ plan: (i) integer anchor via the Euler–Mascheroni
limit, (ii) real-axis sandwich via `ψ` monotonicity (log-convexity of `Γ`) killing the 1-periodic
ambiguity `L(s+1) = L(s)` on a full real interval, (iii) identity theorem
(`eqOn_of_preconnected_of_frequently_eq`, the corpus `DlvpTransfer` pattern) extending `L ≡ 0`
to all `Re s > 1`).  Here: `ψ(1+N) = −γ + H_N` (our `digamma_shift` at `s = 1` + Mathlib
`digamma_one`), so `ψ(1+N) − log(1+N) = −γ + (H_N − log(N+1)) → −γ + γ = 0` by Mathlib's
`Real.tendsto_harmonic_sub_log_add_one`.  conjecture1_proved = False. -/

Formal statement

theorem ZeroFreeBridge.tendsto_digamma_sub_log_one_add_nat :
    Filter.Tendsto (fun N : ℕ => Complex.digamma (1 + N) - Complex.log (1 + N))
      Filter.atTop (nhds 0) := by sorry
Source
https://github.com/DrMurphyIsIn/Arda/blob/af4fde8972eb6d4e78c7d9a95933b18b81adf52d/telperion/examples/zero_free_bridge/lean/DlvpTheta.lean#L449-L474

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