Companion artifacts for Linear crossing families via dual levels by Tim Gehrunger. This archive contains the Lean formalization of the main theorem and the original ProofCouncil output from which the manuscript developed.
LinearCrossingFamiliesAll.lean: all project definitions and proofs in one file, importing only Mathlib.ProofCouncil/LinearCrossingFamilies.tex: the original ProofCouncil output of 18 September 2026, preserved byte for byte. This historical output is distinct from the revised article.lean-toolchain,lakefile.toml, andlake-manifest.json: the pinned Lean environment and dependencies needed to build the proof.LICENSE: Creative Commons Attribution 4.0 International.
There is an absolute constant c > 0 such that every finite set of n ≥ 2
points in the real plane, with no three collinear, spans at least c * n
segments with pairwise distinct endpoints that cross pairwise in their
relative interiors.
The final declaration is
LinearCrossingFamilies.linear_crossing_families.
Its constant is quantified before the point set, and the theorem has no
additional extraction or coordinate assumptions.
The file includes the geometric argument, the fixed-error near-avoidance
extraction (ε = 1/4) needed from the Pach–Rubin–Tardos framework, and the
required finite ε-net construction. The full parameter-dependent extraction
theorem and the article's applications to plane trees and spoke sets are
outside the scope of this formalization.
Install Lean's Elan toolchain manager and Git. From this repository's root, run:
lake exe cache get
lake buildThe versions are pinned to:
- Lean
leanprover/lean4:v4.34.0; - Mathlib
5ed2965256430c3649e86755f9576b54eca72435; - the transitive dependency revisions in
lake-manifest.json.
To inspect the proof interactively, open the repository folder in VS Code with the Lean 4 extension. To check the source and display the final theorem's axiom dependencies directly, run:
lake env lean LinearCrossingFamiliesAll.leanThe combined proof was built and audited on 24 September 2026. The final
theorem uses only propext, Classical.choice, and Quot.sound.
There are no proof placeholders or added axioms in the source.
An additional Comparator check passed statement and definition comparison against a separate reference, the restricted axiom check, and exported-proof replay in Lean's kernel. This check used Comparator's macOS development mode; Linux process sandboxing and an additional independent kernel were not part of that run. The reference used the reviewed geometric definitions and the main theorem's explicit statement. Mathematical interpretation of the definitions was also reviewed against the article.
SHA-256 of LinearCrossingFamiliesAll.lean:
84d389ca0fd1f6d1dbacd63e74f1624b50841033c037d5e845a8799ac8120bef
The mathematical argument originated in an updated version of ProofCouncil during preparations for the third batch of First Proof. The Lean formalization was prepared with Codex and subsequently audited.
The original workflow recorded 14 model calls and US$32.35794275 in model-API costs (US$32.36 rounded). This covers the original research, review, and rewrite workflow; it excludes subsequent manuscript editing, Lean formalization, and infrastructure costs.
SHA-256 of ProofCouncil/LinearCrossingFamilies.tex:
bf577bf90ccd308af043475cf502dd949c5849f6755be392bba28f3ea9fa391c
The materials distributed in this repository are licensed under
Creative Commons Attribution 4.0 International (CC BY 4.0).
See LICENSE for the full terms. For attribution, identify
Tim Gehrunger, the article Linear crossing families via dual levels, and
this repository. Preserve the identification of the original ProofCouncil
output when redistributing that file, and indicate any modifications.
Lean, Mathlib, and other fetched dependencies retain their own licenses and are not included in this repository.