This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: MATH 682 Notes Combinatorics and Graph Theory II 1 Ramsey Theory 1.1 Classical Ramsey numbers Furthermore, there is a beautiful recurrence to give bounds on Ramsey numbers, but we will start with a simple but distinctly nontrivial example, to set the stage: Proposition 1. R (3 , 4) 10 . Proof. Consider a vertex v in a coloring of a K 10 . v has 9 incident edges, in some variety of colors. It must either have at least 6 incident blue edges, or 4 incident red edges (since otherwise it would have fewer than 8 incident edges total). If v is incident on 6 blue edges, consider the endpoints { v 1 ,...,v 6 } of those edges. Among these 6 vertices, a K 6 is colored red and blue; by our previous proposition, this K 6 contains either a red K 3 or a blue K 3 . A blue K 3 among vertices { v i ,v j ,v k } , together with the blue edges to v from each of these, will form a blue K 4 . Thus, if v has 6 blue edges incident on it, we are guaranteed either a blue K 4 or a red K 3 . On the other hand, if v is incident on 4 red edges, then either there is a red edge between the endpoints of these edges, producing a red K 3 , or the edges among all four of them are blue, producing a blue K 4 . We can actually improve this result slightly by maknig use of a parity argument. Corollary 1. R (3 , 4) 9 . Proof. The above argument demonstrated that if any vertex is incident on 6 blue edges or 4 red edges, then the graph must contain a red K 3 or a blue K 4 . In order to prevent this from happening in a coloring of K 9 , in which each vertex is incident on 8 edges, every vertex must be incident specifically on 5 blue edges and 3 red edges. However, the total number of pairs ( v,e ) of vertices and incident blue edges would then be 9 5 = 45, but the number of such pairs must be even, since for each blue edge e , there are exactly two vertices which are its endpoints. Thus, a coloring with the incidence properties described above is impossible, and some vertex has either 6 incident blue edges or 4 incident blue edges; thus any coloring of K 9 contains a red K 3 or a blue K 4 . This bound is in fact sharp, as can be seen by this example of a coloring of a K 8 without a red K 3 or blue K 4 . The above argument doesnt really use anything special about R (3 , 4), and can in fact be generalized to give a nice upper bound on R ( k, ): Page 1 of 6 April 1, 2010 MATH 682 Notes Combinatorics and Graph Theory II Theorem 1. R ( k, ) R ( k 1 , ) + R ( k, 1) . Furthermore, if R ( k 1 , ) and R ( k, 1) are both even, this bound can be reduced by 1. Proof. Let n = R ( k 1 , ) + R ( k, 1), for brevity. Consider a vertex v in a coloring of a K n . v has n 1 incident edges, and must either have at least R ( k 1 , ) incident red edges, or R ( k, 1) incident blue edges (since otherwise it would have fewer than R ( k 1 , )+ R ( k, 1) incident edges total)....
View Full
Document
 Spring '09
 WILDSTROM
 Combinatorics, Graph Theory

Click to edit the document details