The original statements in PROOF.md are formally verified in the exact dependency closure of LonelinessHyperplaneMain. The development proves the actual probability and counting statements from their definitions. It assumes no orbit-density, finite-subspace-cover, rational-fan, Ehrhart, or counting formula as an extra premise.
Normal, sample and maximum
The coordinate type is finite. PrimitiveNormal m means that the gcd of the absolute coordinates of the integer normal is one; nonzeroness and integer Bézout coefficients are proved from that condition. RealAdmissible m means that the real kernel contains a vector with every coordinate positive and with pairwise distinct coordinates.
For box size N, the sample consists of labelled functions into Fin N. Adding one to each coordinate gives precisely the ordered integer vectors in [1,N]^k. The sample is filtered by the exact integer relation and coordinate injectivity. Every admissible velocity is counted once. No primitive-velocity restriction is imposed.
torusVelocityValue is the actual supremum over circle time, and the development proves that this supremum is attained. The equivalent real-lift maximum is attained on [0,1]. The containing-torus profile is
The exact hyperplane maximum has an explicit attaining point. Every orthogonal integer velocity satisfies L(v,s) ≤ L_m(s) at every shift. The normal may have zero coordinates for these statements and for the main probability theorem.
Checked original conclusions
| Conclusion | Lean endpoint |
|---|---|
| Exact shifted hyperplane maximum and attainment | shifted_hyperplane_maximum, shifted_hyperplane_attainment |
| Upper bound for every orthogonal integer velocity | orthogonal_velocity_value_le |
| Uniform inverse-box-size probability estimate over all shifts | exists_torus_uniform_probability_rate |
| Eventual nonempty samples and a vanishing failure probability | exists_uniform_probability_rate, uniformFailureProbability_tendsto_zero |
| Fixed-shift limiting distribution at the exact profile | fixedShift_torus_cdf_tendsto |
| Uniform Haar distribution of the profile | haar_predictedValue_isUniform |
| Independent Haar-shift limiting distribution for the actual sampled maximum | independentShiftLaw_cdf_tendsto_uniform |
| Exact finite proper-rational-subspace description of every fixed subthreshold event | RealSubspace.hyperplane_closed_sublevel_finite_union |
| Exact denominator quasipolynomial | positiveBoxAdmissible_quasipolynomial |
Exact numerator quasipolynomial of degree at most k−2 for every positive box size |
positiveBoxBelow_progression_polynomials |
| Exact unshifted endpoint frequency for full-support normals | endpointFrequency_tendsto |
Sharp inverse-box-size example on v₁+v₂=v₃ |
sharp_uniform_failure_probability_two_sided |
These names are in Loneliness.Hyperplane, except for the displayed RealSubspace subnamespace.
For each positive tolerance, the probability-rate theorem supplies constants and a threshold beyond which the sample is nonempty and the probability of an error at least that tolerance, with a supremum over all shifts, is at most K/N. The finite probability definitions are totalized when a small box has no sample. Eventual nonemptiness is proved and suffices for all limiting statements.
The fixed-shift and independent-shift limits are expressed as convergence of cumulative distribution functions. The independent-shift law is the actual average of the Haar pushforward law for each sampled velocity; its limit is uniform on [1/2−1/(2A), 1/2], with A=∑|mᵢ|. The proof establishes the link from the sampled maxima to that law.
For fixed s and u<L_m(s), the numerator counts the literal event L(v,s)≤u. Its residue-class polynomials agree with that count for every N≥1 and have degree at most k−2. The denominator has an exact quasipolynomial of degree at most k−1, including N=0.
For the endpoint formula only, the shift is zero and every normal coordinate is nonzero. Set Q=2 if A is even and Q=A otherwise. The verified limit is
The theorem includes k=2. The residue argument proves the canonical point's order and primitivity, the exact hit criterion, the original-box residue frequencies and the required summable tails. The numerical 1/4 endpoint example is an instance of this general formula at the normal (1,1,−1). The sharpness development separately proves the profile 1/3, the exact value 1/4 for every positive multiple of (1,2,3), the attaining time 1/(4g), and matching inverse-box-size bounds.
Exact verification evidence
Run 44531509741 / job 135585736594 checked source commit ee1f018fef258cdd54f73af31cdaf6485fd151cb with Lean 4.35.0 and Mathlib 56a005855ea17e0b0958f73ec801ae140b0a1679.
All 72 project modules were freshly compiled with warnings treated as errors and no linter override. The audit checked every owned declaration, including raw serialized private declarations: 1,403 declarations, 1,149 theorems and 44 private declarations. All 106 required endpoints were checked to have theorem kind. Only propext, Classical.choice and Quot.sound were allowed as transitive axioms.
Every one of the 72 modules then passed a fresh process of the unchanged pinned stock checker as an exact singleton target. The receipt binds the source bytes, all emitted proof-object facets, compiler/checker binaries, verification helpers, commands, logs and declaration inventory. The preservation process authenticated the actual original archive bytes and every recorded evidence file; it ran no new Lean proof check.
- Unchanged raw receipt, SHA-256
1c0d6a6fbe2ea16bbcfa88808ec8d83f0e412217e76c03803780ff0737fa1c15. - Preservation record.
- Original artifact
14535236819, ZIP SHA-25627d65f3b5d667260dfad8d639f6d1182a3b5e54c032a2e2954d7764e060fe638. - Earlier strict/theorem-audit receipt, retained with its original scope.
The checked closure is registered in the accepted root, required endpoint audit and replay list. The repository-wide accepted-package check is a separate gate; this component receipt does not claim a new pass of that larger package.