|
It is easy to see that if a graph has diameter (and has any cycles at all), its girth can be no more than . For
suppose and is a cycle of that minimum length . Take two vertices (nodes) A and B on that are steps apart along , one way round; they are
steps apart the other way round. Now either there is no shorter route between A and B (contradicting diameter ) or there is a shorter route of length creating a cycle of length (contradicting girth ).
Definition: A Moore graph is a connected graph with maximal girth for its diameter .
It can be shown that Moore graphs are regular, i.e. all vertices have the same valency. So a Moore graph is characterised by its diameter and valency.
Diameter 1 means every vertex (node) is adjacent to every other, that is, a complete graph. Indeed, complete graphs that have cycles ( ) have triangles, so the girth is 3 and they are Moore graphs. Every valency occurs ( has valency ).
This is the most interesting case. And the proof that every vertex has the same valency, say, and that the graph now has vertices in all, is easy here.
With some more work, it can be shown there are only 4 possible values for and :
The first three cases each have a unique solution. The existence or otherwise of the last case is still open. It has been shown that if it exists it has, unlike the first three, very little symmetry.
The Hoffman-Singleton graph is a bit hard to draw. Here's a unified description of the three known Moore graphs of , all indices (mod 5):
The automorphism group of the pentagon is the dihedral group with 10 elements. The one of the Petersen graph is isomorphic to , with 120 elements. And the one of the HS graph is isomorphic to
, with 252000 elements, a maximal subgroup of another HS, the Higman-Sims group.
In these cases, there are only Moore graphs with valency 2, graphs consisting of a single -gon cycle. This was proven independently by Bannai and Ito and by Damerell.
|