2
$\begingroup$

Let $T$ be the universal covering tree of some finite, connected, non-tree graph, and let $n_0(T)$ be the smallest positive integer such that there exists a graph $G$ (loops and multiple edges allowed) on $n=n_0(T)$ vertices which is covered by $T$.

If $T$ is the $k$-regular tree for some $k\ge 2$, then it's easy to check that $$ n_0(T)=\begin{cases} 1 & \text{if $k$ is even,} \\ 2 & \text{if $k$ is odd.}\end{cases} $$

If $T$ is the $(c,d)$-biregular tree, with $2\le c<d$, then one clearly has $n_0(T)\le c+d$ (because $K_{c,d}$ is covered by $T$). If $(c,d)=1$, then one can also conclude from Sunada's gap labeling theorem, and the fact that $\mu_T(\{0\})=(d-c)/(c+d)$, that $n_0(T)=c+d$. What can we say in the case $(c,d)>1$?

Are there any general results in the literature concerning this quantity $n_0(T)$ (other than immediate consequences of Sunada's gap labeling theorem)?

$\endgroup$
8
  • $\begingroup$ I may have problems with the definition of covering, but can't every tree (or even graph) cover a single node with a loop? $\endgroup$
    – M. Winter
    Commented Nov 3, 2020 at 12:02
  • $\begingroup$ @M.Winter A covering, like here is meant, is a surjection which is also a local bijection, that is which preserve the (local) degree structure. $\endgroup$ Commented Nov 3, 2020 at 12:18
  • $\begingroup$ Then how can $n_0(T)=1$ for a, say, 4-regular tree? $\endgroup$
    – M. Winter
    Commented Nov 3, 2020 at 12:23
  • 1
    $\begingroup$ Then how is this "locally bijective"? Not every edge of the tree is mapped onto a unique loop. And why can't I do the same with a 3-regular tree. My guess: this is what you mean with "degree structure" and you need to map the tree's edges somehow uniformly to the edges of the covered graph? Is this right? I would not call this bijective though. Is this standard? $\endgroup$
    – M. Winter
    Commented Nov 3, 2020 at 12:29
  • 1
    $\begingroup$ $n_0(T)\leq (c+d)/(c,d)$, as $K_{c,d}$ covers a graph with $(c+d)/(c,d)$ vertices. I do not know what Sunada's gap labeling theorem is, but I guess that this is optimal. $\endgroup$ Commented Nov 3, 2020 at 14:32

1 Answer 1

3
$\begingroup$

I don't know about results in the literature, but here is a hands-on solution for the $(c,d)$-biregular tree $T$ when $(c,d)$ is arbitrary.

First note that $n_0(T) \leq \frac{c+d}{\gcd(c,d)}$; indeed $T$ covers the graph obtained from $K_{\frac c{\gcd(c,d)},\frac d{\gcd(c,d)}}$ by replacing each edge by $\gcd(c,d)$ many parallel edges.

Now let $G$ be any finite graph covered by $T$. It is not hard to see that $G$ must be bipartite with bipartition $V(G) = A \uplus B$, where each vertex in $A$ is incident to $c$ edges, and each vertex in $B$ is incident to $d$ edges. So $|E| = c \cdot |A| = d \cdot |B|,$ and dividing by $\gcd(c,d)$ we get that $$\frac c{\gcd(c,d)} \cdot |A| = \frac d {\gcd(c,d)} \cdot |B|.$$ Since $\frac c{\gcd(c,d)}$ and $\frac d {\gcd(c,d)}$ are relatively prime, this implies that $\frac d {\gcd(c,d)}$ divides $|A|$ and consequently $\frac d {\gcd(c,d)} \leq |A|$, and similarly $\frac c {\gcd(c,d)} \leq |B|$. So $|V(G)| = |A|+|B| \geq \frac{c+d}{\gcd(c,d)}$ thus showing that $$n_0(T) = \frac{c+d}{\gcd(c,d)}.$$

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .