Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Source. M. Grinsztajn, A 63-dimensional counterexample to Borsuk's conjecture, unpublished note, May 2026, as described on the source card. Lemma 1 ("finite verified facts") is on p. 2; the graph it concerns is defined in Section 2, pp. 1--2.
Setting
Section 2 (pp. 1--2) uses Brouwer's projective model of the graph. Over , with the Hermitian form on the projective plane , the vertices of are the unordered triples of pairwise orthogonal non-isotropic projective points. For such , is the set of 15 isotropic points lying on the three lines spanned by pairs of points of , and are adjacent when .
Statement
The note attributes each of the following to the verification script in its accompanying repository (reference [4], p. 6).
- has 416 vertices and is strongly regular with parameters ; hence its nontrivial adjacency eigenvalues are and , with multiplicities and .
- : the script finds a 5-clique and verifies that no 6-clique exists.
- Let be the first isotropic point in the script's deterministic order, let be the set of vertices containing a non-isotropic point orthogonal to , and let . Then and , and the graph induced on has three connected components , each of size 32.
- Writing for the neighborhood in : for ; for and ; for ; for and ; for .
Proof pointer
The note gives no hand proof. Section 7 (p. 6) says the script rebuilds the graph from , checks the strongly regular parameters, builds , and checks the degree data used in Lemma 3 and the clique obstruction used in Lemma 6, with exact finite-field arithmetic and integer bitsets. The eigenvalues and multiplicities in item 1 follow from the parameters by the standard formulas for strongly regular graphs.
Dependencies and read depth
External: Brouwer's description of the graph (reference [3]) and the repository's script (reference [4]). Read depth: claims checked; the statement was read clause by clause on p. 2. The script has not been run and no certificate audited here, so these finite facts are the note's computational claims, not facts verified in this corpus.
Bears on. E0505: all four items are finite inputs on which the note's dimension-63 claim (Theorem 1) rests: item 1 through Lemma 2, items 3 and 4 through Lemmas 3 to 5, and item 2 through Lemma 6.