Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Source. Theorem 8, p. 12, with its proof on pp. 12–14, of David Ellis, Irredundant families of subcubes, arXiv:1003.2960v1 (2010), published in Mathematical Proceedings of the Cambridge Philosophical Society 150(2) (2011), 257–272, as identified on the source card. Labels and pages are those of arXiv:1003.2960v1.
Statement
Write and (p. 2); subcubes and irredundance are as on the Theorem 4 page.
Theorem 8 (p. 12). If is an irredundant family of -subcubes of which contain or , then .
The extremal families are not unique (p. 14): besides the principal families and of all -subcubes through , respectively , any family that contains, for each middle-layer vertex , exactly one of the subcube between and and the subcube between and attains the bound.
Proof pointer
Pages 12–14, a linear-algebra argument. Take maximal. Each middle-layer vertex is the meeting point of the subcube from up to and the subcube from up to ; sort the middle layer by whether both, one or neither of these lies in . Then , where holds the vertices with both and those with neither, and it remains to show . Private vertices just below and just above each vertex of give sets in whose intersection sizes satisfy (8), p. 13, and the case of Lemma 9, p. 13 (a mod- inner-product count for a prime : pairs with for and force ) gives .
Dependencies
Lemma 9, stated on p. 13 and proved on p. 14. Theorem 8 is the base case of Corollary 10.
Read depth. Claims checked: the statement on p. 12, Lemma 9 on p. 13 and the non-uniqueness remark on p. 14 were read clause by clause. The proof was read but not checked step by step.
Bears on
No Erdős problem.