Wiki
Wiki

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

Updated

Furedi 1983 graphs without quadrilaterals

../

lemma_p188: A quadrilateral-free graph on q^2+q+1 vertices whose maximum degree is at least q+2 has at most q(q+1)^2/2 edges.

proposition_p190: For every even q, a quadrilateral-free graph on q^2+q+1 vertices has at most q(q+1)^2/2 edges.

theorem: For q a power of 2 (q = 2^k with k at least 1), the quadrilateral-free extremal number at q^2+q+1 vertices is q(q+1)^2/2.


Zoltán Füredi, Graphs without Quadrilaterals, J. Combin. Theory Ser. B 34 (1983), 187-190.

Source and version. The copy read for this card is the four-page published article, including its note added in proof. Printed pp. 187-190 correspond to PDF pp. 1-4. The PDF identifies the publication item as 0095-8956(83)90018-7, giving DOI 10.1016/0095-8956(83)90018-7. The inherited source link is the author's publication site: https://users.renyi.hu/~furedi/. The file prints "Copyright © 1983 by Academic Press, Inc. All rights of reproduction in any form reserved.", every other right reserved.

Write f(n)=ex⁡(n,C4)f(n)=\operatorname{ex}(n,C_4) for the maximum number of edges of a finite simple graph on nn vertices containing no four-cycle as a subgraph. On printed p. 187, the Kővári-Sós-Turán/Reiman bound is

f(n)≤n4(1+4n−3),f(n)\leq\frac n4(1+\sqrt{4n-3}),

and at n=q2+q+1n=q^2+q+1 its displayed specialization (1) is

f(q2+q+1)≤12(q2+q+1)(q+1).f(q^2+q+1)\leq\frac12(q^2+q+1)(q+1).

This real upper bound exceeds q(q+1)2/2q(q+1)^2/2 by (q+1)/2(q+1)/2.

The polarity graph described on pp. 187-188 gives the lower bound q(q+1)2/2q(q+1)^2/2 for prime powers qq, attributed to Erdős-Rényi-Sós and independently Brown. The degree count is q+1q+1 vertices of degree qq and q2q^2 vertices of degree q+1q+1. The graph from the 1966 source is recorded as Theorem 1.

The unnumbered Theorem in Section 2 (p. 188) proves f(q2+q+1)=q(q+1)2/2f(q^2+q+1)=q(q+1)^2/2 for qq a power of 22, which the abstract writes as q=2kq=2^k; the statement concerns k≥1k\geq1. The proof separates a maximum-degree neighborhood, using the unnumbered Lemma on p. 188, whose proof is on p. 189. The Lemma treats maximum degree at least q+2q+2; the remaining case uses the evenness of qq. The Proposition on p. 190 records the resulting upper bound for every even integer qq, while equality in the Theorem also uses the finite-field construction.

Note added in proof. On p. 190, Füredi announces that the upper bound holds for all qq and that equality graphs are the Erdős-Rényi graphs, with a further publication promised. Those stronger arguments are absent from this article. The same page's preceding equality discussion also says its proof is omitted. These source announcements are distinct from the power-of-two theorem proved in the body. Ma and Yang's later introduction reports the upper bound for every integer q≥14q\geq14, citing this article and Füredi's 1996 paper; that later proof has not been read here.

Reading and proof scope. All four complete rendered pages were inspected, including the exact statement, elementary degree count, proof organization, Proposition, note added in proof and references. No full proof reconstruction or independent proof review is claimed. The missing proofs of the stronger announcements and the original external upper/lower-bound sources remain outside this reading coverage.

Bears on. #765: the Theorem gives the exact value of ex⁡(n;C4)\operatorname{ex}(n;C_4) at the orders n=q2+q+1n=q^2+q+1 with q=2kq=2^k, k≥1k\geq1; the Proposition gives the upper bound q(q+1)2/2q(q+1)^2/2 at those orders for every even qq, and the Lemma gives it for graphs of maximum degree at least q+2q+2. These concern the orders q2+q+1q^2+q+1 only; the problem asks for an asymptotic formula for all nn, and the paper does not address the proposed second-order formula for general nn.

No file of this source is held: no license on record permits its redistribution, and the card cites the edition it names above.