Formal verification of the existence theory from:
N. Spence, "The Creative Determinant: Autopoietic Closure as a Nonlinear Elliptic Boundary Value Problem with Lean 4-Verified Existence Conditions," 2026.
Fifteen theorems and lemmas proved with zero sorry, organized in five dependency tiers:
| Result | Declaration | File | Topic |
|---|---|---|---|
| Spectral characterization (1D) | spectral_characterization_1d |
Theorems |
Algebra |
| Scaling algebraic contradiction | scaling_algebraic_contradiction |
Theorems |
Algebra |
| v^{p−1} ≤ b/c from PDE inequality | rpow_le_of_mul_rpow_le |
LinftyAlgebraic |
Real analysis |
| v ≤ (b/c)^{1/(p−1)} | linfty_bound_algebraic |
LinftyAlgebraic |
Real analysis |
| nextFixed ≤ super-fixed point | OrderHom.nextFixed_le_of_le |
MonotoneFixedPoint |
Order theory |
| Fixed point between sub/super | monotone_fixed_point_between |
MonotoneFixedPoint |
Order theory |
| Result | Declaration | File |
|---|---|---|
| Laplacian of zero | laplacian_zero |
OperatorLemmas |
| Laplacian linearity | laplacian_linear |
OperatorLemmas |
| Gradient norm of zero | gradNorm_zero |
OperatorLemmas |
| Result | Declaration | File |
|---|---|---|
| a(x) nonneg | SemioticContext.a_nonneg |
CoefficientLemmas |
| a(x) ≤ 1 | SemioticContext.a_le_one |
CoefficientLemmas |
| p − 1 > 0 | SemioticContext.p_sub_one_pos |
CoefficientLemmas |
| Result | Declaration | File | Dependencies |
|---|---|---|---|
| Scaling uniqueness (kΦ impossible for k > 1) | scaling_uniqueness |
ScalingUniqueness |
SemioticOperators axioms |
| Existence of weak coherent configurations | SemioticBVP.exists_isWeakCoherentConfiguration |
Theorems |
PDEInfra |
| Nontrivial configurations | SemioticBVP.exists_pos_isWeakCoherentConfiguration |
Theorems |
PDEInfra |
All definitions (semiotic manifold, BVP, operators, weak coherent configuration) are machine-checked against Mathlib.
The monotone fixed-point theorems depend only on [propext, Quot.sound] — no Classical.choice — making them candidates for constructive upstream contribution.
The PDEInfra typeclass in CdFormal/Axioms.lean packages five classical PDE results not yet in Mathlib:
| Axiom | Classical source | Mathlib status |
|---|---|---|
T_compact |
Schauder estimates + Arzelà–Ascoli | Uses IsVonNBounded + IsCompact (bornological typing, credit: Aaron Lin) |
linfty_bound |
Maximum principle (Gilbarg–Trudinger) | No max. principle for manifolds |
schaefer |
Schaefer 1955 | Not in Mathlib (draft issue) |
fixed_point_nonneg |
Strong maximum principle | No max. principle for manifolds |
monotone_iteration |
Amann 1976 | No sub-/super-solution theory |
The existence theorems explicitly carry [PDEInfra bvp solOp] so the axiom surface is visible to Lean's kernel. Run #print axioms in CdFormal/Verify.lean to confirm no sorryAx — all theorems depend only on [propext, Classical.choice, Quot.sound] (monotone fixed-point theorems use only [propext, Quot.sound]).
Requires Lean 4 (v4.28.0) and Mathlib.
lake build
lake build --wfail # fail on any sorry or warningCdFormal/
Basic.lean — Definitions (manifold, coefficients, operators, BVP)
Axioms.lean — PDEInfra typeclass (explicit axiom surface)
Theorems.lean — Existence and algebraic theorems
OperatorLemmas.lean — Laplacian/gradient-norm derived properties
CoefficientLemmas.lean — Bounds from [0,1] field constraints
ScalingUniqueness.lean — PDE-level scaling impossibility
MonotoneFixedPoint.lean — Knaster-Tarski fixed point between sub/super
LinftyAlgebraic.lean — L∞ bound algebraic core (Paper Lemma 3.10)
Verify.lean — #print axioms dashboard (17 declarations)
artifacts/
aristotle/ — Proved outputs from the Aristotle theorem prover
drafts/ — Mathlib issue drafts + Lean proof sketches
mathlib_issue_schaefer.md, BornologyBridge_sorry.lean
MonotoneFixedPoint_sorry.lean, LinftyAlgebraic.lean, OperatorLemmas.lean
ScalingUniqueness.lean, ScalingUniqueness_v2.lean, ScalingUniqueness_v3.lean
What the author did: The original equations, proof strategy, and formalization
architecture — choosing to axiomatize via PDEInfra, identifying which five
classical PDE results to package, designing the typeclass hierarchy, and structuring
the Schaefer → L∞ bound → existence → sub/super-solution → nontriviality proof
chain — are the core intellectual contribution. These are mathematical architecture
decisions that require understanding where the real difficulty lies. The underlying
equations and theory are documented in the paper.
What AI tools did: Claude Opus assisted with Lean 4 syntax, Mathlib API
navigation, and proof term synthesis. Aristotle (Harmonic.fun) automated proving
of standalone algebraic lemmas. These roles are analogous to omega, aesop, and
other proof automation — the strategy is human, the term-level search is machine-assisted.
Verification: The final arbiter is the Lean compiler, not trust:
lake build --wfail # type-checks or it doesn't — zero sorryRun #print axioms in CdFormal/Verify.lean to confirm the axiom surface.
Every assumption is explicit in PDEInfra. Nothing is hidden.
Copyright 2026 Nelson Spence. Licensed under Apache 2.0.