Content is user-generated and unverified.

PROOF_STRUCTURE_LEAN4.md — Lean 4 Formalization of APO-RH Proven Results

Created: April 8, 2026
Last updated: April 14, 2026 (v8.0 — 0 sorries, 20+ verified theorems)
Parent: APO Initiative — PROOF_STRUCTURE_RH.md
Working document: Sandbox/ApoRhSpec.lean (v8.0)
Companion: PROOF_STRUCTURE_RH.md (mathematical proof chain), ApoRhSpec.lean (logical specification)
Lean version: 4.29.0
Mathlib: current as of April 2026


Purpose

This document tracks the Lean 4 formalization status of results labeled PROVEN in APO-RH Paper I ("The Arithmetic Recognition Operator"). Only PROVEN results are formalized; ARGUED/CONJECTURED results are excluded by design.


Verification Summary

MetricCount
sorry warnings0
Theorems verified (standard axioms only)17
Theorems verified (modulo Schwarz axiom A1)+4
Custom axioms3 (1 closeable, 2 blocked)
Axiomatized zeta functions0 (all from Mathlib)

Axiom Inventory

#AxiomWhat it encodesCloseable?
A1completedRiemannZeta₀_conj_on_rightDirichlet series has real coefficients → Λ₀(s̄) = Λ̄₀(s) on Re s > 1Yes — finite computation via Dirichlet series term-by-term conjugation
A2solomonoffMeasureSpectral theory for integral operatorsBlocked by Mathlib
A3kraft_inequalityBound on A2Blocked by Mathlib

Key distinction: A1 is a verifiable computation. A2–A3 are genuine theory gaps. The analytic continuation from A1 to global Schwarz reflection is fully verified via the identity theorem.


Master Status Table

★ Verified — depends only on [propext, Classical.choice, Quot.sound]

#ResultPaper RefProof Method
L1recognitionOp_symm — ⊙(s,s') = ⊙(s',s)Thm III.2(1)rw [add_comm, mul_comm]
L2recognitionOp_self — ⊙(s,s) = 1Thm III.2(2)Real.sqrt_sq + div_self
L3completedObs_Z2_invariant — O_ξ(s,s') = O_ξ(1−s,1−s')Thm III.5ring + 3× completedRiemannZeta_one_sub
L4even_function_deriv_zero — f even ⇒ f'(0) = 0Prop III.7HasDerivAt chain rule + .unique + linarith
L5riemannZeta_re_pos — ζ(s) > 0 for real s > 1InfrastructureDirichlet series + Complex.re_tsum + tsum_pos
L6–L8Series summability + cast lemmasInfrastructureStandard Mathlib transfers
L9zetaDistribution_sum_one — Σ P_s(n) = 1Eq. 1tsum_div_const + div_self
L10sqrtEmbedding_norm_one — ‖ψ_s‖² = 1Eq. 8rpow_natCast + rpow_mul + convert from L9
L11recognitionOp_eq_tsum — ⊙ = Σ ψ_s ψ_{s'}Def. III.1div_mul_div_comm + sqrt_mul + rpow_add
L12recognitionOp_bounded — 0 < ⊙(s,s') ≤ 1Thm III.2(3)AM-GM + Summable.tsum_le_tsum + L10
L13–L15Summability + nonnegativity helpersInfrastructureComparison tests
L16differentiableAt_conjCompConjInfrastructureCR equations through conjCLE
L17eq_of_eqOn_open — identity theoremInfrastructureAnalyticOnNhd.eqOn_of_preconnected_of_frequently_eq

★★ Verified modulo A1

#ResultPaper RefProof Method
L18completedRiemannZeta_conj — Λ(s̄) = Λ̄(s)InfrastructureA1 + identity theorem (L17) + Λ₀→Λ lift
L19completedObs_conj — O_ξ(s̄,s̄') = Ō_ξ(s,s')InfrastructureL18 + ring hom lemmas
L20completedObs_norm_conjInfrastructureL19 + RCLike.norm_conj
L21recognition_Z2_indistinguishableProp IV.2 ★★★Z₂ (L3) + conjugate ID + Schwarz (L20)

Dependency Graph

Mathlib riemannZeta API
  └─→ L6 (series summable)
       └─→ L7 (cast) → L8 (real summable)
            └─→ L5 (ζ(s).re > 0) ★
                 └─→ L9 (Σ P_s = 1) ★
                      └─→ L10 (‖ψ_s‖² = 1) ★
                           └─→ L12 (0 < ⊙ ≤ 1) ★★★
                 └─→ L2 (⊙(s,s) = 1) ★

  └─→ L1 (⊙ symmetric) ★

Mathlib completedRiemannZeta API
  └─→ L3 (O_ξ Z₂ invariant) ★★★
       │
       ├── A1 [AXIOM: Dirichlet series conjugation]
       │    └─→ L16 (conjCompConj holomorphic) ★
       │    └─→ L17 (identity theorem) ★
       │         └─→ L18 (Schwarz reflection) ★★
       │              └─→ L19–L20 (O_ξ conjugation/norm) ★★
       │
       └─→ L21 (Z₂ indistinguishability) ★★★

L4 (even deriv zero) ★ — standalone

Session Log

April 8–12, 2026 — v1 through v7.9

See previous entries. Summary: built from 0 to 15 verified theorems, 1 sorry.

Multi-model contributions: Perplexity (API discovery), GPT-4o (scalar-series architecture, normalization), Claude (AM-GM design, error diagnosis), Gemini (Lean 4.29.0 compat), Sonnet (tactical iteration v7.3→v7.9).

April 14, 2026 — v8.0: Zero Sorries (Sonnet)

Closed the last sorry by factoring Schwarz reflection into verified analytic machinery plus one axiom:

  1. differentiableAt_conjCompConj — PROVEN. Hardest new lemma: CR equations for reflected function G(z) = f(z̄)̄, chain rule through three conjCLE compositions.
  2. eq_of_eqOn_open — PROVEN. Identity theorem for entire functions agreeing on {Re s > 1}.
  3. completedRiemannZeta_conj — PROVEN from A1. Lifts from Λ₀ (entire) to Λ via completedRiemannZeta_eq, handles pole corrections algebraically.
  4. recognition_Z2_indistinguishable — PROVEN. The proof uses two symmetries: Z₂ sends (σ+it, ½+it') → ((1-σ)-it, ½-it'), then conjugate identification restores imaginary parts, and ‖conj z‖ = ‖z‖ gives norm equality. Neither symmetry alone suffices — you need both.

Result: 0 sorries, 3 axioms. All core Paper I theorems machine-checked.


Next Steps

Priority 1: Close axiom A1

  • Prove completedRiemannZeta₀_conj_on_right from Mathlib's Dirichlet series
  • Would make all 21 results depend only on standard axioms

Priority 2: Merge conditional chain

  • Port Parts 3–5 of ApoRhSpec.lean onto v8.0's Mathlib-native definitions

Priority 3: lean4checker --fresh

  • Strongest kernel-level validation
Content is user-generated and unverified.
    Lean 4 Formalization of APO-RH Proven Results | Claude