Upload
maud-terry
View
220
Download
2
Embed Size (px)
Citation preview
Planar graphs with no 5-cycles, 6-cycles or intersecting triangles are 3-colorableCarl Yerger, Davidson College
Clemson Mini-Conference 2012
A Historical Problem
Grotzsch’s Theorem (1959): Any planar graph that contains no 3-cycles is 3-colorable.
Thomassen found several short proofs of Grotzsch’s theorem as a consequence of some of his other results.
Are there other classes of graphs that exclude certain length cycles that are 3-colorable?
Steinberg’s Conjecture
Steinberg conjectured (1976) that any planar graph without 4- or 5- cycles is 3-colorable.
The conditions of Steinberg are necessary:
Erdos suggested a method of attack in 1991 by asking what is the minimum k such that if G excludes cycles of length 4 up to length k, G is 3-colorable?
Progress on Steinberg - Coloring
Results for k = 11, 9, 8, 7. The most recent result within the paradigm of Erdos is that of Borodin et al (2005) who show that k ≤ 7.
Proof Technique for k = 7: Show that every proper 3-coloring of the vertices of any single face of length 8 to 11 in a connected graph G that excludes cycles of length 4 through 7 extends to a proper 3-coloring of G.
The proof of this theorem uses discharging.
Discharging Euler’s Formula:
Give all vertices charge Give all faces charge except a
precolored face, which is given Goal: We want all vertices and faces to
have nonnegative charge (positive charge for precolored face).
How do we distribute the charges?
Complications
One Forbidden Structure: A Tetrad
If G were a minimum counterexample and contained a tetrad, identify vertices x and t. Delete vertices .
If the resulting graph is 3-colorable, then the original graph is 3-colorable.
Excluding tetrads gives faces more charge after discharging, which allows us to prove that such a minimal counterexample does not exist.
The k = 7 proof
Consider a minimal counterexample. Begin with a graph G having a 3-colored face, D.
The aim is to prove structural properties. No separating cycles of length at most 11. G is 2-connected, vertices of degree 2 must be
part of D. No cycle of length at most 13 has a non-
triangular chord, nor does D.
Forbidden Structures in the k = 7 Proof
Steinberg’s Conjecture and Higher Surfaces
The condition given by Erdos holds for arbitrary surfaces: Zhao showed that for every surface Σ, there exists some k such that if G is a graph embedded on Σ, and G has no cycles of length 4 through k, then G is 3-colorable.
Unfortunately, for surfaces with nonpositive Euler characteristic the bound for k is given by .
Euler characteristic of some simple surfaces are plane = 2, torus = 0, Klein bottle = 0, double torus = -2, etc.
Example: Double torus, k = 35.
A Topological Result
We prove that k = 10 with a linear size of minimal exceptions.
Theorem [Thomas, CY]: Let Σ be a surface of Euler genus g. If G is 4-critical and has no cycles of length four through ten, then
A similar construction
[Dvorak, Thomas] If G is a 4-critical, triangle free graph drawn on a torus, then
Tightness of this bound
On the plane, when cycles of length 4 through 9 are excluded, no 4-critical graphs exist by the result of Borodin et al.
On higher surfaces 4-critical graphs do exist since there are graphs with arbitrarily large girth and chromatic number.
The Hajos construction allows us to construct a sequence of critical graphs with size linear in the genus of G.
Intersecting Triangles By including triangles spaced far
apart, we can reduce the cycle restrictions dramatically.
2010: Borodin and Glebov showed that every planar graph excluding 5-cycles and with minimum distance between triangles at least two is 3-colorable.
The (Strong) Bordeaux Coloring Conjecture: (Borodin and Raspaud) Can we reduce the restriction to only (adjacent) intersecting triangles and 5-cycles?
Eared cycles Whalen showed that planar graphs with no
cycles of length 4, 5, 6 or eared cycles of length 7 are 3-colorable.
Methodology
Whalen seeks a minimum counterexample and proves the customary pre-colored cycle analogue:
Theorem: Let G be a plane graph with outer cycle of length at most 11. Suppose G excludes cycles of length 4, 5, 6 and eared cycles of length 7. Suppose G does not contain a vertex adjacent to three vertices on the outer cycle. Then any proper 3-coloring of the outer cycle extends to a proper 3-coloring of G.
Coloring with an Independent Set and a Forest
Can we partition the vertices of a graph into an independent set and a set that induces a forest?
How does this relate to Steinberg? Coloring with an independent set and a
forest is stronger than 3-coloring. Does there exist a planar graph excluding
4-cycles and 5-cycles that is not colorable with an independent set and a forest?
Suggests a possible extension of Steinberg’s conjecuture.
Borodin’s Technique A common technique for
solving problems related to Steinberg’s conjecture is to examine what happens inside a precolored cycle.
Colorings inside and outside the cycle do not affect each other.
What about for an independent set and a forest?
An extension of Grotzsch’s Theorem
Kawarabayashi and Thomassen (2008) proved the following: Let G be a plane graph. Assume that every triangle of G has a vertex v which is on the outer face boundary and such that v is contained in no 4-cycle. Assume also that the distance between any two triangles is at least 5. Then G has chromatic number at most 3.
This is proved with the help of a structural theorem in coloring with an independent set and a forest.
Our Theorem
Theorem: Let G be a planar graph that excludes 5-cycles, 6-cycles and intersecting triangles. Then G is 3-colorable.
Proof uses a discharging technique and strategies similar to the proof of Borodin and Glebov.
General Strategy Assume that there exists a pre-colored cycle of
length 7 through 10 and other basic structural properties hold.
Describe a set of discharging rules. Exclude various structures, including tetrads, most
4-cycles and some 7-cycles and 8-cycles. Show that every vertex and face has nonnegative
charge, contradicting Euler’s formula.
The Discharging Rules, I
The Discharging Rules II
Excluding a Tetrad
Excluding most 4-cycles
An 8-cycle argument
An 8-cycle argument, continued
Reducible Configurations
More Reducible Configurations
Wrap-up
To finish the proof, we must show each of the previous configurations is reducible.
All remaining face combinations must have charge at least zero via the discharging rules alone.
We get a contradiction via Euler’s formula.
Future directions
Cycles of length seven are the more challenging case for Steinberg.
No known counterexample for the Steinberg conditions for an independent set and a set that induces a forest.
Need another idea to push discharging to work for Steinberg’s conjecture.
Other types of coloring?
Thank you for your attention!
Any questions?