I am trying to prove the above as an exercise in the topic of connectivity. I have tried to do so using ear decompositions, as odd degree vertices may be characterized as end points of ears, but to no avail. Any recommendations are appreciated. Thanks
Asked
Active
Viewed 469 times
1 Answers
11
The statement is false. Take the following $5$-regular graph (inspired by the graph in this MathOverflow answer, which being $4$-regular didn't quite do the trick):
In this graph, every degree is odd, so we are looking for a Hamiltonian path. However, to visit each of the five parts around the sides, we would have to go through the middle vertices multiple times, so this is impossible.
For a slightly more formal argument: if a graph $G$ has a Hamiltonian path, it has a path $P_n$ as a subgraph. Deleting two vertices from $P_n$ leaves at most $3$ components, so the same must be true of $G$ (which is $P_n$ with extra edges). But in the graph above, deleting the two middle vertices leaves $5$ components, so it can't have a Hamiltonian path.
Misha Lavrov
- 159,700
-
Maybe I’m missing something, but it looks like every vertex has degree four, and the question as it appears now is about paths that go through every odd-degree vertex. There are no such vertices in your graph. The Thomassen graphs here might be the example needed: http://mathworld.wolfram.com/ThomassenGraphs.html (oops, not, I think now...) – Steve Kass May 06 '19 at 00:23
-
1Whoops - the MathOverflow question was about all $k$-regular graphs for $k=3$, so naturally this graph uses $k=4$ rather than $k=5$... I will fix this. – Misha Lavrov May 06 '19 at 00:27
-
1(+1) Because the claim doesn't assume regularity, you can get a smaller (and planar!) counterexample by taking five $K_4$s rather than five $K_6$s. – hmakholm left over Monica May 06 '19 at 01:06
-
@HenningMakholm We could even take only four $K_4$s rather than five; then we wouldn't need to visit both cut vertices, but that's not the obstacle to begin with... – Misha Lavrov May 06 '19 at 01:10
