Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Source. Karamanlis, published p. 4, Lemma 4 (canonical PDF). The corresponding arXiv v1–v3 result is Lemma 3.2.
Karamanlis identifies this as a reformulation of Frankl–Pach–Reiher–Rödl, Borsuk and Ramsey type questions in Euclidean space, Lemma 4.9, and includes its proof. The proof reconstructed here is that included argument; it is not an independent review of the earlier chapter.
Statement. Let and let be a real symmetric array with and for . Put . Suppose
Then this array is the distance array of an affinely independent set of points in a product of at most regular simplices. Arrays and their realizations satisfying (1) are called almost regular.
Proof. Define
Take a regular simplex with vertices and edge length . For each pair with , take a regular simplex with vertices and edge length . For there are no such pairs, because its sole distance is maximal. Thus no positive edge length is being assigned to a one-point factor.
Label the vertices of by . In the factor , label its vertices by the classes of the partition of whose only nonsingleton class is . For each label , let have base coordinate and, in each pair factor, the class containing . This defines all simultaneously in
For distinct labels , the base coordinates are distinct. The coordinates in agree exactly when . With when needed, their squared distance is therefore
Diagonal distances are zero directly. Projection onto the base factor sends the labels bijectively to the affinely independent vertices of . Any affine relation among the would project to one among those vertices, so all its coefficients vanish.
At least one pair attains and consequently has . There are at most nonzero pair factors; together with the base factor this gives at most factors.
Source precision. The printed distance calculation is introduced for all ; its displayed off-diagonal formula applies to . The diagonal case is handled separately above. The partition labels also make explicit that no extra pair is identified in a pair factor. These are compilation explanations, not an author-issued erratum.
Use. Proposition 5 and Proposition 11.