Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Statement
Here is the largest number of lines through exactly three points of a -point set, as defined on the Theorem 1 page.
Theorem 3 (p. 409). For every ,
with for the print's , "the greatest integer ".
The graph (p. 408), used again in Theorems 4--8. For a -arrangement , has the points as nodes, two of them adjacent exactly when no line of contains both. A point on lines of has valence , so every valence has the parity opposite to , and has edges, that is (the paper's ). Remark (6) (p. 420) notes that Coxeter calls the Menger graph of .
Read depth. Claims checked: the statement and its derivation were read clause by clause on the page images of the print. Nothing here is independently reviewed.
Proof pointer
Pp. 408--409. From and , for all . For even every node has odd, hence positive, valence, so and . The theorem writes the two cases as one formula. The argument is purely combinatorial, which Theorem 9 uses to carry it to pseudolines.
Remark (7) (pp. 420--421) calls a purely combinatorial generalization of Theorem 3 the question of , the largest number of triples on symbols with no pair in two triples. It records Kirkman's and Schönheim's values of (listed in Table I, p. 399) and says that Theorem 3 strengthens to , with a pseudoline analogue of between them.
Source. S. A. Burr, B. Grünbaum and N. J. A. Sloane, The orchard problem, Geometriae Dedicata 2 (1974), 397--424, DOI 10.1007/BF00147569 (source card).
Bears on
- Problem 669: in the problem's notation the theorem bounds above by , which is the pair count for lines through exactly three points with a parity correction for even . It says nothing about , which also counts lines through more than three points.