3
$\begingroup$

I am reading Bowditch's notes on Geometric Group Theory and I am trying to understand a proof. He proves that if $F_p$, $F_q$ are free groups of rank $p,q\geq 2$, then they are commensurable, i.e. there are subgroups $G, G^{'}$ of $F_p$ and $F_q$, which are isomorphic, and have finite index.

In the proof, he defines $K_n$ as the graph obtained by taking the circle, $R/nZ$, and attaching a loop at each point of $Z/nZ$ — that is $n$ additional circles. He proves that $\pi_1(K_n)=F_{n+1}$ (this is clear), but then states that $K_{mn}$ covers $K_n$, which I don't understand. Then he claims that $K_r$ covers both $K_p$ and $K_q$, where $r=(p-1)(q-1)+1$ and because of compactness, $F_r$ is a finite index subgroup of $F_p$ and $F_q$.

Is there anyone who can help me understand the proof?

  • 0
    @LeeMosher I think that $K_{mn}$ does cover $K_n$. Treat $K_n$ as a subspace of $K_{mn}$ (with $m-1$ loops between loops in $K_n$) and then go counter clockwise (starting at some loop) and map each $m$ loops into starting loop and then proceed to the next one. Hard to explain. For example in case $2, 4$ we would have $f:K_4\to K_2$ induced by $z\mapsto z^2$ (if we look at $K_n$ as a unit circle in $\mathbb{C}$ with attached loops). But then again I don't see how $K_r$ covers $K_p$ and $K_q$ for such $r$. I mean $r$ does not have to be a multiple of $p$ and $q$ right? Can't we just pick $r=pq$? – 2017-02-13
  • 0
    I removed my earlier comment and wrote a careful description of the covering map $K_{mn} \to K_n$. I can also see the possibility of a typographical error: assuming $r = (p-1)(q-1)+1$ it follows that $r-1=(p-1)(q-1)$ and therefore $K_{r-1}$ covers both $K_{p-1}$ and $K_{q-1}$. – 2017-02-13
  • 0
    I can see why $K_{mn}$ covers $K_n$. It also makes sense what you say about the possible typo. – 2017-02-13

1 Answers 1

4

To describe the covering map $K_{mn} \mapsto K_n$, first map the main circle $R / (mnZ) \subset K_{mn}$ around the main circle $R / nZ \subset K_n$ by a covering map of degree $m$ (taking the point $x + mnZ$ to the point $x + nZ$ for each $x \in R$).

Next, the loop of $K_{mn}$ attached to the main circle at a vertex $p$ in $Z / (mnZ) \subset K_{mn}$ is mapped homeomorphically to the loop of $K_n$ attached to the main circle at the image of $p$ in $Z / nZ \subset K_n$.

The rest of the proof is standard covering space theory, explained in most algebraic topology textbooks. I recommend Hatcher's "Algebraic Topology", which has a beautiful page very early in the book showing many covering spaces of the graph $K_1$ whose fundamental group is $F_2$.