Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Statement

Notation (printed p. 246): Kn∗K_n^* is the complete symmetric loopless digraph of order nn, LmL_m the transitive tournament of order mm, and r(Kn∗,Lm)r(K_n^*,L_m) "the smallest order pp so that every digraph on a set of pp vertices either has an independent set of nn vertices (no arcs in either direction between vertices) or includes a transitive tournament LmL_m of order mm". A "free" set in the proof is an independent set.

Proposition 3.1 (printed p. 248). "r(K4∗,L3)>13r(K_4^*,L_3)>13."

Proof as printed (p. 248). The proof is a table of in-neighborhoods N−(i)N^-(i) and out-neighborhoods N+(i)N^+(i) of a digraph F=(V,A)F=(V,A) on 13 nodes, which the authors say has no free subset of size 4 and no transitive tournament of size 3; they add that whatever motivated the digraph has been forgotten. The out-neighborhoods, transcribed from the page image:

iiN+(i)N^+(i)iiN+(i)N^+(i)iiN+(i)N^+(i)
01, 5, 953, 6, 1294, 7, 10
12, 8, 1160, 7, 8100, 11, 12
20, 3, 472, 5, 11113, 6, 9
31, 7, 1084, 5, 10122, 8, 9
41, 6, 12

The printed in-neighborhood column agrees with these except at i=0i=0, where it prints "2, 6, 8" while the out-neighborhoods put 00 in N+(2)N^+(2), N+(6)N^+(6) and N+(10)N^+(10); see the filing observations below. The two checks are left to the reader: for the absence of L3L_3, that no vertex jj is the middle point of an L3L_3, which holds when N+(i)∩N+(j)=∅N^+(i)\cap N^+(j)=\emptyset for every i∈N−(j)i\in N^-(j); and, for the free sets, a case analysis on the least element ii of a free set FF using P(i)={i+1,…,12}∖N(i)P(i)=\{i+1,\ldots,12\}\setminus N(i), with the instruction to "check that the largest free set has at most 2 vertices" in P(i)P(i).

In the problem's notation. r(Kn∗,Lm)=k(n,m)r(K_n^*,L_m)=k(n,m) in the letters of Problem 112, so the proposition is k(4,3)≥14k(4,3)\ge14. With the corollary r(K4∗,L3)≤16r(K_4^*,L_3)\le16 of Lemma 4.2 (p. 248) the paper brackets 14≤k(4,3)≤1614\le k(4,3)\le16, which its table of small values (p. 247) prints as "14−1614-16".

Source. J. A. Larson and W. J. Mitchell, On a Problem of Erdős and Rado, Ann. Comb. 1 (1997), 245--252; Proposition 3.1 with its proof on printed p. 248 (PDF p. 4 of the publisher scan), read on the page image and on a higher-resolution rendering of the table; the notation on p. 246 (PDF p. 2), read on the page image. The artifact is identified in the source digest.

Read depth. Claims checked: the statement, the proof paragraph and the table were read clause by clause on the page images. The two checks the printed proof leaves to the reader were carried out here by computer on the digraph the out-neighborhood columns define and are recorded below as filing checks, not review verdicts. Nothing here is independently reviewed.

Proof pointer

Page 248. The proposition rests entirely on the table. Filing checks made here on the digraph A={(i,j):j∈N+(i)}A=\{(i,j):j\in N^+(i)\} of the out-neighborhood columns: it has 39 arcs and no pair of opposite arcs, so it is an oriented graph in which every vertex has in-degree and out-degree 3; no triple x,y,zx,y,z has all three arcs x→yx\to y, y→zy\to z, x→zx\to z (the paper's middle-point test, N+(i)∩N+(j)=∅N^+(i)\cap N^+(j)=\emptyset for every arc i→ji\to j, holds at every arc); and no 4 vertices are pairwise non-adjacent (the largest independent sets have 3 vertices, and there are 29 of them). So FF has no L3L_3 as a subgraph and no independent set of size 4, and r(K4∗,L3)≥14r(K_4^*,L_3)\ge14. Filing observations: the printed N−(0)N^-(0), "2, 6, 8", disagrees with the out-neighborhood columns in one entry (88 for 1010); the digraph obtained by taking the printed in-neighborhood column as the arc set instead has the arc 8→08\to0 in place of 10→010\to0, and it contains two transitive triples and an independent set of 4 vertices, so the out-neighborhood columns are the ones to read. The sentence "no free sets of size greater than 4" is read as "of size 4", the property the proposition needs and the one the printed case analysis ("at most 2 vertices" in P(i)P(i), hence at most 3 with ii) establishes.

Dependencies

None within the paper; the witness is self-contained. The upper half of the bracket is Lemma 4.2 at n=4n=4.

Bears on

  • Problem 112: k(4,3)≥14k(4,3)\ge14, the lower half of the paper's 14≤k(4,3)≤1614\le k(4,3)\le16; the page had "k(4,3)>13k(4,3)>13" second-hand from the 2021 paper of Ihringer, Rajendraprasad and Weinert, whose Theorem 1.1 closes the bracket at k(4,3)=15k(4,3)=15 with a 14-vertex construction.