
Antichains in the cycle continuous order
If ,
are graphs, a function
is called cycle-continuous if the pre-image of every element of the (binary) cycle space of
is a member of the cycle space of
.




The definition of a cycle-continuous mapping is based on some work of Jaeger, and the most interesting question on this subject is undoubtedly Jaeger's Petersen coloring conjecture.
Let us define a relation on the set of all finite graphs with at least one edge by the rule if there is a cycle-continuous mapping from
to
. It is not difficult to verify that
is a quasi order (reflexive and transitive). In this order, every Eulerian graph dominates every other graph, and every graph with a cut edge is dominated by every other graph.
Let be the graph on two vertices with
parallel edges. Then
with all the inequalities strict, so this sequence is an infinite chain. Very little else seems to be known about this order. In particular, the problem highlighted above - does there exist an infinite antichain? remains open.
Bibliography
* indicates original appearance(s) of problem.