Wiki
Wiki

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

Updated


Statement

g2(n,m)g_2(n,m) is as on Theorem 3: the largest size of a set of edges, every two on a 44-cycle of the whole graph, that every graph with nn vertices and mm edges contains for all sufficiently large nn (p. 269).

Theorem 6 (p. 271). For each positive constant cc there exists a positive constant c1c_1 such that

g2(n,(n2)−cn3/2)≥c1(n2).g_2\Bigl(n,\binom n2-cn^{3/2}\Bigr)\ge c_1\binom n2.

After the proof the authors say the bound is essentially best possible (pp. 272--273): for each positive constant cc and sufficiently large nn there are a constant c2<1c_2<1 and a graph G(n,(n2)−cn3/2)G(n,\binom n2-cn^{3/2}) in which every C4C_4-connected set has at most c2(n2)c_2\binom n2 edges. They argue it from the case c=12c=\frac12, where deleting from KnK_n the edges of an Erdős--Rényi--Sós graph (their [7]: p2+p+1p^2+p+1 vertices, every two with exactly one common neighbour) allows c2=12+ϵc_2=\frac12+\epsilon for every ϵ>0\epsilon>0.

Source. Richard A. Duke, Paul Erdős and Vojtěch Rödl, Cycle-connected graphs, Discrete Math. 108 (1992), 261--278, doi:10.1016/0012-365X(92)90680-E; Theorem 6 on printed p. 271, its proof on p. 272 and the sharpness discussion on pp. 272--273, read on the page images of the publisher's scan. The edition read is identified in the source digest.

Read depth. Claims checked: the statement was read clause by clause on the page image. The proof and the sharpness argument were read for structure only.

Proof pointer

Page 272. For c≤c0=12(621)3/2c\le c_0=\frac12(\frac6{21})^{3/2}, Lemma 4 gives a set of at least 17(n2)(1+o(1))\frac17\binom n2(1+\mathrm o(1)) edges (display (11)). For c>c0c>c_0 the same bound is applied inside a random set of N=αnN=\alpha n vertices with α c∼c0\sqrt\alpha\,c\sim c_0, which gives a set of about c047c4(n2)\frac{c_0^4}{7c^4}\binom n2 edges (displays (12)--(13)).

Dependencies

Lemma 4 (p. 270), stated on Theorem 5.

Bears on

No problem page is reached by this theorem: it concerns 44-cycles in graphs missing cn3/2cn^{3/2} edges, and no problem the corpus records asks about them.