Erdos-Gallai Theorem
If we want the Extremal Graph number for Path (path with edges, or vertices), then it is at most .
This bound is almost tight. It will be tight if is a multiple of . We can have be a union of complete graphs . If that happens, the as all the connected components of size , we will never have a connected subgraph with vertices. Then by Handshake Lemma and that is -regular, there are edges.
Also recall that the extremal graph need not be connected. We are simply trying to maximize the number of edges.
Proof:
Let be an vertex graph with at least edges. We prove by strong induction on that contains unless every component of is .
If is disconnected, then we can examine each piece individually. Suppose some component of contains a path of length or equals . We remove it and get a graph with vertices with the following edges:
edges. By induction, that graph is a union of or contains a path of length , and we are done.
Therefore we can assume is connected. If , then a single edge forms . If there are no edges, then every component is . Let . If contains a vertex degree less than , then has at least
edges remaining. Since still has enough edges, by induction it must contain a path or be a union of . Since , then the same holds for .
Thus, . By induction, contains a path of length . Let this be from . All , otherwise we could just extend the path and be done. By lemma1, we have a cycle of length containing .
If there is a vertex , and as is connected, there must be a path from , and in particular, adding edge with , we get a path of length ending with . Thus and so and as required.
This proof is almost entirely copied verbatim from the Verstraete textbook.
Erdos-Gallai Conjecture
If is any Tree with edges, then
The above Erdos-Gallai Theorem was used to prove this for paths specifically.
Footnotes
-
This is also known as Dirac’s Theorem. ↩