Weak pentagon problem

Recomm. for undergrads: no
Posted by: Robert Samal
on: July 13th, 2007
Conjecture   If $ G $ is a cubic graph not containing a triangle, then it is possible to color the edges of $ G $ by five colors, so that the complement of every color class is a bipartite graph.

This conjecture has several reformulations: the conclusion of the conjecture can be replaced by either of the following:

    \item $ G $ has a homomorphism to the Clebsch graph. \item there is a cut-continuous mapping from $ G $ to $ C_5 $.

For the latter variant, few definitions are in place. A cut-continuous mapping from a graph~$ G $ to a graph~$ H $ is a mapping $ f : E(G) \to E(H) $ such that the preimage of every cut in~$ H $ is a cut in~$ G $. Here, by a cut in~$ H $ we mean the edge-set of a spanning bipartite subgraph of~$ H $---less succinctly, it is the set of all edges leaving some subset of vertices of~$ H $.

Cut-continuous mappings are closely related with graph homomorphisms (see [DNR], [S]). In particular, every homomorphism from~$ G $ to~$ H $ naturally induces a cut-continuous mapping from~$ G $ to~$ H $; thus, the presented conjecture can be thought of as a weaker version of Nesetril's Pentagon problem.

We mention a generalization of the conjecture, that deals with longer cycles/larger number of colors. The $ n $-dimensional projective cube, denoted $ PQ_n $, is the simple graph obtained from the $ (n+1) $-dimensional cube~$ Q_{n+1} $ by identifying pairs of antipodal vertices (vertices that differ in all coordinates). Note that $ PQ_4 $ is the Clebsch graph.

Question   What is the largest integer $ k $ with the property that all cubic graphs of sufficiently high girth have a homomorphism to $ PQ_{2k} $?

Again, the question has several reformulations due to the following simple proposition.

Proposition   For every graph $ G $ and nonnegative integer $ k $, the following properties are equivalent.
    \item There exists a coloring of~$ E(G) $ by $ 2k+1 $ colors so that the complement of every color class is a bipartite graph. \item $ G $ has a homomorphism to $ PQ_{2k} $ \item $ G $ has a cut-continuous mapping to~$ C_{2k+1} $

There are high-girth cubic graphs with the largest cut of size less then $ 0.94\cdot |E| $. Such graphs do not admit a homomorphism to $ PQ_{2k} $ for any $ k \ge 8 $, so there is indeed some largest integer~$ k $ in the above question. To bound this largest~$ k $ from below, recall that every cubic graph maps homomorphically to $ K_4 = PQ_2 $. Moreover, it is known [DS] that cubic graphs of girth at least 17 admit a homomorphism to $ PQ_4 $ (the Clebsch graph). This shows $ k\ge 2 $ (and also provides a support for the main conjecture).

Bibliography

[DNR] Matt DeVos, Jaroslav Nesetril and Andre Raspaud: On edge-maps whose inverse preserves flows and tensions, \MRref{MR2279171}

*[DS] Matt Devos, Robert Samal: \arXiv[High Girth Cubic Graphs Map to the Clebsch Graph}{math.CO/0602580}

[S] Robert Samal, On XY mappings, PhD thesis, Charles University 2006, tech. report


* indicates original appearance(s) of problem.