Structure of hypothetical 41–43 point kissing configurations in R^5
A kissing configuration in R^5 is a set of unit vectors with pairwise inner products at most 1/2. It is known that 40 ≤ τ5 ≤ 44 (Mittelmann and Vallentin 2010), and a 2026 preprint of T. S. Tomar gives τ5 ≤ 43 (doi:10.5281/zenodo.24093811); priority for τ5 ≤ 43 and for A(4, arccos 1/3) = 14 (doi:10.5281/zenodo.23603435) belongs to Tomar. This note collects exact structure facts that any kissing configuration in R^5 with 41, 42 or 43 points would have to satisfy: (1) a 41-point configuration has at most 18 antipodal pairs and a 42-point configuration at most 17; (2) a sharp robust cross-polytope lemma: five near-antipodal, near-orthogonal pairs force at most 42 points; (3) exact two-point counts of far-free and isolated points at N = 41 and 42; (4) every point of an N-point configuration has at least N − 15 others within 120 degrees, with an exact polynomial certificate; (5) larger radii for vertex-cover facts about inner products far from the alphabets of the known 40-point codes (conditional on two earlier, not yet published lemmas, stated as hypotheses); (6) new in version 2: the antipodal 40-point configuration D5 is locally rigid (an antipodal 40-point kissing configuration with all inner products within 49/10000 of {-1, -1/2, 0, 1/2} is D5), and an antipodal 40-point configuration that is not D5 has at least 4 line pairs far from D5; uniqueness of D5 among antipodal 40-point configurations remains open (we found no published proof); (7) a list of what these and earlier facts say together. Nothing in this note lowers τ5: it does not exclude 41, 42 or 43 points. The supplement contains all certificates, the checking programs, the logs and the reports of a second check. AI use: the derivations, certificates, programs and text were produced with Claude (Anthropic) under the author's direction; the lift argument behind result (1) and the frame-error lemma used in (5) came from sessions with ChatGPT (OpenAI). The second checks are by separate sessions of the same AI system, not independent human review. Version 2 (16 November 2026) adds Section 7; an earlier unpublished draft of its counting result claimed 5 line pairs because one finite check was run on the wrong pair; the corrected number is 4. Not yet peer reviewed. Contact: [email protected].
Authors
- Учрал Ням-Осор
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24243223
- Primary Topic
- Advanced Combinatorial Mathematics
- Type
- preprint