# yaml-language-server: $schema=https://raw.githubusercontent.com/mathlib-initiative/formalization.yaml/main/schema/formalization.schema.json version: "v0.4" project: name: "The radian nine-gon is Euclidean Ramsey iff e^(i theta) is algebraic" description: >- Theorem D: for real theta, the nine points (cos k theta, sin k theta), k = 0, ..., 8, of the Euclidean plane are distinct, lie on the unit circle and are not Euclidean Ramsey when tan(theta/2) is transcendental; and whenever they are distinct, they are Ramsey if and only if cos theta and sin theta are algebraic (that is, iff e^(i theta) is algebraic), if and only if tan(theta/2) is algebraic. This is an iff for a whole one-parameter family, where earlier explicit examples with a transcendental parameter (OpenAI's Proposition 7.7 and Corollary 7.5) settle single sets: the same nine-point pattern is Ramsey or not exactly as e^(i theta) is algebraic or transcendental. Nine is the exact threshold (eight consecutive multiples are Ramsey for every theta; by hand, in spherical-ramsey), and at theta = e the question becomes whether e^(i e) is algebraic, which appears to be open. Ramsey is OpenAI's definition from result 172 (for every r >= 2 there is D such that every r-colouring of R^D contains a monochromatic congruent copy). The statements are given in OpenAI's vocabulary plus the two-line definition of the nine points, and again with no definitions beyond Mathlib. The non-Ramsey direction mirrors OpenAI's nine-point proof (their Proposition 7.6): the coordinate field is Q(tan(theta/2)), its tensor square embeds in Q(T, U), and the denominator-cleared 9x9 evaluation matrix is nonsingular, certified at (T, U) = (2, 3) by an explicit integer triangular factorization checked by decide +kernel; OpenAI's linear obstruction concludes. The converse uses the theorem that finite spherical sets with algebraic coordinates are Ramsey (prior work of JD-Jones-ASES, Rogerhu12 and atimics; proved again here in Lean on OpenAI's development). The proof builds on OpenAI's complete Lean proof of the classification (openai/math fd4aeeb2), whose 35 files used here are included with header-only changes for the Lean module system. This repository is a module-system port, with no statement or proof line changed apart from a namespace rename, of Lean already public as rope AV3 of spherical-ramsey v1.1.0 (doi:10.5281/zenodo.23396387), packaged so that Theorem D can be checked on its own. All theorems depend on propext, Classical.choice and Quot.sound only. authors: - name: "Will Strinz" github: "wstrinz" orcid: "0009-0007-7682-8284" license: "MIT" responsible_maintainers: - "Will Strinz" classification: arxiv: ["math.CO", "math.MG", "math.NT"] msc2020: ["05D10", "52C10", "11J81"] sources: - title: "Two Möbius groups suffice: Euclidean Ramsey sets on spheres of every dimension, after the tensor classification (note in spherical-ramsey v1.1.0)" type: original-proof authors: - "Will Strinz" id: "doi:10.5281/zenodo.23396387" location: "note/spherical-ramsey.tex, Section 'Radian progressions and transcendence', Theorem D and the corollary after it; ropes/av3 (the Lean proof at Lean v4.34.1)" relationship: other note: >- The release this result comes from. Its note states Theorem D with a proof sketch, the theta = 1 corollary (by Lindemann-Weierstrass, not in Lean) and the sharpness remarks; its rope AV3 is this Lean proof at OpenAI's pin (Lean v4.34.1, Mathlib d13f23b7), checked there by stock Comparator on a fresh runner, a non-builder AI fidelity review and a fresh AI reproduction. This repository is that proof ported to the Lean module system (docs/PORT-LOG.md). No prior statement of Theorem D was found (docs/PRIOR-WORK-SEARCH.md). author_endorsement: "n/a" - title: "A classification of finite Euclidean Ramsey configurations (OpenAI math release, result 172)" authors: - "OpenAI" id: "https://github.com/openai/math/tree/fd4aeeb2ee4fc729c18d98444fed42fd0529eeeb" type: paper location: "preprints/A-classification-of-finite-Euclidean-Ramsey-configurations-September-23-2026/ (Theorem 1.1, Proposition 7.6); lean/OAI/Combinatorics/EuclideanRamsey/ (Main.lean import closure, and the nine-point and circle modules NineLinearObstruction, NineGeometry, IndependentTensor, CircleExtension, CircleGeometry, CircleInterpolation, CircleSpan)" relationship: other note: >- The foundation. The non-Ramsey direction of Theorem D is the argument of Proposition 7.6 (nine points with algebraically independent parameters) for a single parameter, and its Lean proof mirrors and imports OpenAI's EuclideanRamseyNine development (NineLinearObstruction.not_ramsey_of_det_ne_zero, IndependentTensor, CircleConsequence.circle_five_affineSpan); six lines of local instances and one proof pattern are adapted from NineField.lean, and RadianAlgebra.lean follows the clear-denominators-and-specialize structure of NineRationalDeterminant.lean, with attribution (NOTICE). The 35 files used are included, ported to the module system by header-only changes (docs/PORT-LOG.md). The definitions Space, Congruent and Ramsey are OpenAI's, verbatim. Proposition 7.7 (a 12-point three-squares set with a Liouville parameter, not Ramsey) and Corollary 7.5 (kites with a transcendental parameter, Ramsey) are earlier explicit examples with a transcendental parameter; Theorem D instead gives an iff for a whole one-parameter family. This work trusts OpenAI's Lean proof, which is replayed by the kernels here. license: "Apache-2.0" author_endorsement: "not-contacted" - title: "Möbius invariance of the Ramsey property for finite spherical sets (mobius-ramsey v1.0.0)" authors: - "Will Strinz" id: "doi:10.5281/zenodo.23339523" type: other location: "Palomar entry PALOMAR-2026-10-14-000008" relationship: background note: >- Sibling package from the same project: the same module-system port of OpenAI's files (byte-identical for the 28 shared modules) and the same Comparator check. Its ramsey_of_algebraic is reused here unchanged. author_endorsement: "n/a" - title: "Spherical sets with algebraic coordinates are Ramsey (used for the converse)" authors: - "JD-Jones-ASES" - "Rogerhu12" - "atimics" id: "https://github.com/JD-Jones-ASES/six-concyclic-lean; https://github.com/Rogerhu12/openai-math-extensions; https://github.com/atimics/math2" type: other location: "six-concyclic-lean root commit 0703d7d; openai-math-extensions commit f8a2bb2, papers/172-euclidean-ramsey/algebraic_spherical_configurations.tex; math2 commit 8ab51fa, EndToEnd/AlgebraicRamsey.lean" relationship: background note: >- Prior statements and proofs of the algebraic-coordinates theorem: criterion level in Lean (Ramsey conclusion conditional on hsuff), in TeX, and end to end in Lean for algebraic configurations with full affine span (atimics). This repository's ramsey_of_algebraic is a further Lean proof on OpenAI's development, not a first one; no code was copied. author_endorsement: "not-contacted" related_formalizations: - id: "https://github.com/openai/math/tree/fd4aeeb2ee4fc729c18d98444fed42fd0529eeeb/lean/OAI/Combinatorics/EuclideanRamsey" relationship: builds-on note: "OpenAI's Lean proof of the classification and of the nine-point example; the 35 modules used are included (Apache-2.0)." - id: "https://doi.org/10.5281/zenodo.23396387" relationship: builds-on note: "spherical-ramsey v1.1.0, ropes/av3: this proof at Lean v4.34.1 / Mathlib d13f23b7, before the port; Lean-checked there." - id: "https://doi.org/10.5281/zenodo.23339523" relationship: builds-on note: "mobius-ramsey v1.0.0 (Theorem C, Möbius invariance; Palomar PALOMAR-2026-10-14-000008), the sibling package: RadianRamsey/AlgebraicRamsey.lean (ramsey_of_algebraic) and the 28 ported OpenAI files of Main's import closure are taken from it unchanged." - id: "https://github.com/atimics/math2" relationship: independent note: "The algebraic-coordinates theorem end to end in Lean for configurations with full affine span (commit 8ab51fa); used here only as credited prior work." - id: "https://github.com/JD-Jones-ASES/six-concyclic-lean" relationship: independent note: "The algebraic-coordinates theorem at criterion level in Lean (root commit 0703d7d)." status: scope: >- Formalized: Theorem D for every real theta, both directions (radian_configuration, radian_not_ramsey, radian_ramsey_iff, radian_ramsey_iff_tan), in OpenAI's vocabulary and in Mathlib vocabulary (radian_configuration_mathlib, radian_ramsey_iff_mathlib, radian_ramsey_iff_tan_mathlib). Not formalized: any explicit theta with tan(theta/2) transcendental (theta = 1 needs Lindemann-Weierstrass, not in Mathlib; a Liouville witness theta = 2 arctan L is possible in principle but not done); the complex form "e^(i theta) is algebraic" of the iff (a one-line hand step); the extension from the normalized position (unit circle about 0, starting at (1, 0)) to congruent copies. sorry_count: 0 sorry_in_definitions: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] main_results: - declaration: "OAI.EuclideanRamsey.Radian.radian_configuration" file: "RadianRamsey/Radian.lean" sorry_count: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] comparator_config: "comparator.json" - declaration: "OAI.EuclideanRamsey.Radian.radian_not_ramsey" file: "RadianRamsey/Radian.lean" sorry_count: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] comparator_config: "comparator.json" - declaration: "OAI.EuclideanRamsey.Radian.radian_ramsey_iff" file: "RadianRamsey/Radian.lean" sorry_count: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] comparator_config: "comparator.json" - declaration: "OAI.EuclideanRamsey.Radian.radian_ramsey_iff_tan" file: "RadianRamsey/Radian.lean" sorry_count: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] comparator_config: "comparator.json" - declaration: "RadianRamsey.radian_configuration_mathlib" file: "RadianRamsey/MathlibStatement.lean" sorry_count: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] comparator_config: "comparator.json" - declaration: "RadianRamsey.radian_ramsey_iff_mathlib" file: "RadianRamsey/MathlibStatement.lean" sorry_count: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] comparator_config: "comparator.json" - declaration: "RadianRamsey.radian_ramsey_iff_tan_mathlib" file: "RadianRamsey/MathlibStatement.lean" sorry_count: 0 axioms: ["propext", "Classical.choice", "Quot.sound"] comparator_config: "comparator.json" automation: methods: - method: "agent" models: ["claude-opus-5-5"] framework: "Claude Code" tool_setup: >- A coordinating Claude Opus 5.5 session dispatched bounded Claude Opus 5.5 agents, each in its own git worktree: the Lean proof on top of OpenAI's development (Lean 4, Mathlib, lake build, #print axioms), first at Lean v4.34.1 with stock Comparator, then the module-system port and this package (lake comparator with three kernels, negative controls, GitHub Actions CI), the fidelity notes and the documentation. Exact computer-algebra checks (Python, SymPy) were used while exploring. cost: wall_time: "the proof within a multi-day project; a few hours for the port and this package" spend_usd: "subscription-based" hardware: "one desktop machine (Windows and WSL Ubuntu) and GitHub-hosted runners" prompting_notes: >- Written task packets with explicit status tags, time boxes and hard rules (no weakened statements, no added hypotheses, standard axioms only, no sorry or native_decide). - method: "agent" models: ["claude-sonnet-5-5"] framework: "Claude Code" tool_setup: "Read-only agents for the prior-work searches (arXiv API, OpenAlex, zbMATH, GitHub, the Palomar registry, MathOverflow) and for a fresh reproduction of the Lean v4.34.1 version of this proof from its README alone." - method: "other" models: [] framework: "Claude (chat)" tool_setup: "The exploratory conversation (claude.ai, Claude Opus 5.5) from which the project started." spend_usd: "subscription-based" notes: >- Will Strinz directed the work and made every release decision. The proofs, the port and the checks were produced by AI agents. See DISCLOSURE.md. fidelity: divergences: >- Ramsey, Congruent and Space are OpenAI's definitions, verbatim; the only further definition is radianConfig (two lines), the nine points (cos k theta, sin k theta) in EuclideanSpace R (Fin 2). The Mathlib-vocabulary statements have no definitions beyond Mathlib. Mathlib's tan has the junk value 0 at odd multiples of pi/2, so tan(theta/2) = 0 at theta = pi (mod 2 pi); there the transcendence hypothesis fails and the points are not distinct, so no statement is affected. No explicit theta with transcendental tan(theta/2) is exhibited in Lean. The points are in the normalized position (unit circle about 0, first point (1, 0)); congruent copies follow by hand. Item-by-item table: docs/FIDELITY.md. review: status: "agent-reviewed" notes: >- Statements checked by lake comparator against a Mathlib-only Challenge.lean, with five negative controls refused (sorry, redefined Ramsey, redefined radianConfig, radian_ramsey_iff without distinctness, a weaker Mathlib-vocabulary statement). The same statements at Lean v4.34.1 (spherical-ramsey rope AV3) had a fidelity review by a Claude Opus 5.5 agent that did not build them (faithful; documentation fixes applied) and a fresh reproduction by a Claude Sonnet 5.5 agent. The port's statements and documentation were reviewed by a further Claude Opus 5.5 agent that did not build them (statements identical to the rope's; documentation fixes applied). No human expert review. alignment: namespace: "OAI.EuclideanRamsey.Radian" statements: - source: "spherical-ramsey note, Theorem D (non-Ramsey direction)" lean: "OAI.EuclideanRamsey.Radian.radian_configuration, OAI.EuclideanRamsey.Radian.radian_not_ramsey, RadianRamsey.radian_configuration_mathlib" module: "RadianRamsey.Radian, RadianRamsey.MathlibStatement" status: "proved" note: "Hypothesis: tan(theta/2) transcendental. Distinctness is a conclusion." - source: "spherical-ramsey note, Theorem D (the iff)" lean: "OAI.EuclideanRamsey.Radian.radian_ramsey_iff, OAI.EuclideanRamsey.Radian.radian_ramsey_iff_tan, RadianRamsey.radian_ramsey_iff_mathlib, RadianRamsey.radian_ramsey_iff_tan_mathlib" module: "RadianRamsey.Radian, RadianRamsey.MathlibStatement" status: "proved" note: "Stated with cos theta and sin theta algebraic, and with tan(theta/2) algebraic; the e^(i theta) form is a hand step." - source: "spherical-ramsey note, corollary to Theorem D (theta = 1 radian)" lean: "" module: "" status: "not formalized" note: "Needs Lindemann-Weierstrass, which is not in Mathlib." acknowledgements: >- OpenAI (the classification of finite Euclidean Ramsey configurations, the nine-point example and their Lean proofs); JD-Jones-ASES, Rogerhu12 and atimics (prior work on algebraic coordinates); Mathlib; the Lean Comparator, NanoDa and con-ron kernels as bundled with the Lean toolchain; the PalomarTemplate scripts.