2
$\begingroup$

Let $P$ be a finite polyhedron and $N$ be a normal subgroup of $G=\pi_1 (P)$. It is known that there exists a covering space $(\tilde{P},p)$ so that $p_* \pi_1 (\tilde{P})=N$. It follows that for the finite polyhedron $\tilde{P}$ (which is related to $P$), we have $\pi_1 (\tilde{P})\cong N$.

My question is that:

Is there any finite polyhedron $Q$ (connected to $P$) so that $\pi_1 (Q)\cong \frac{G}{N}$?

Thanks in advance.

$\endgroup$
11
  • 4
    $\begingroup$ Seems implausible, consider the case where $P$ is a graph. The quotient group can be an arbitrary finitely generated group, and while that is the fundamental group of a two-complex, it seems unlikely that this complex would have any intimate connection with the graph you started with. $\endgroup$
    – Igor Rivin
    Oct 19, 2017 at 4:17
  • $\begingroup$ Note that $\tilde{P}$ is only finite when $N$ has finite index in $G$. $\endgroup$
    – HJRW
    Oct 19, 2017 at 12:05
  • $\begingroup$ Let $\{r_i\}$ normally generate $N$. Represent (the conjugacy class of) each $r_i$ by a loop in the 1-skeleton of $P$. We can now define $Q$ by gluing a disc to $P$ for each $r_i$, using the corresponding loop as the attaching map. This defines a cell complex that contains $P$. Of course, it can be taken to be finite if and only if $N$ is finitely generated as a normal subgroup. It's unclear to me whether it's "connected to $P$" in the way that you want. (@IgorRivin thinks not, but I prefer to leave it up to the OP to decide.) $\endgroup$
    – HJRW
    Oct 19, 2017 at 12:11
  • $\begingroup$ Note that if, as you seem to imply in your question, you are actually interested in the case that $N$ has finite index in $G$, then there is a canonical finite choice: take the mapping cylinder of the covering map $\tilde{P}\to P$, and then cone off the end isomorphic to $\tilde{P}$. $\endgroup$
    – HJRW
    Oct 19, 2017 at 12:12
  • $\begingroup$ @IgorRivin Thank you for the comment. You are right. This may happen. I check that. $\endgroup$
    – MHenry
    Oct 19, 2017 at 14:01

1 Answer 1

3
$\begingroup$

For a finite polyhedron $P$ and finite-index normal subgroup $N$ of $G=\pi_1P$, there is a canonical finite polyhedron $Q$ with $\pi_1Q\cong G/N$ constructed as follows. Let $\tilde{P}\stackrel{p}{\to} P$ be the covering map corresponding to $N$, as in the question.

We now construct the polyhedron $Q$ as follows:

$Q= ((\tilde{P}\times [0,1]) \sqcup P) / \sim$

where $(\tilde{x},0)\sim (\tilde{y},0)$ and $(\tilde{x},1)\sim p(\tilde{x})$, for all $\tilde{x},\tilde{y}\in\tilde{P}$.

That is, $Q$ is constructed from the mapping cylinder of the covering map $p$ by crushing the canonical copy of $\tilde{P}$ to a point. Alternatively, as I said in comments, one can think of this as obtained by gluing the cone on $\tilde{P}$ to the mapping cylinder of $p$.

The Seifert-van Kampen theorem then tells us that $\pi_1Q\cong \pi_1P/p_*\pi_1\tilde{P}\cong G/N$, as required.

Clearly, this construction can be performed for any subgroup $N$ of $G$, but only gives a finite polyhedron in the case when $N$ has finite index. By making a choice, one may construct a suitable (non-canonical) polyhedron whenever $G/N$ is finitely presentable.

$\endgroup$
1
  • $\begingroup$ Thank you so much for your great answer. This is exactly what I want and expect. That's perfect. Thank you for taking the trouble to help me. I do appreciate you. $\endgroup$
    – MHenry
    Oct 19, 2017 at 15:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.