11
$\begingroup$

Suppose that $M$ is a compact, connected three-manifold with boundary. Suppose that $\pi_1(M)$ is infinite. Suppose that $\tilde{M}$, the universal cover of $M$, has at least three boundary components. Does $\tilde{M}$ have infinitely many boundary components?

Obviously the answer is "yes". I believe I have a proof of this, but my proof is annoyingly long and perhaps too machinery-heavy. It is also possible that I have missed some subcase of a subcase... Is there a clean (short, elementary, clever..) proof?

Non-counter-examples to keep in mind (for $M$) include handlebodies and interval bundles over closed surfaces. Another non-counter-example is $S^3$ minus a disjoint collection of (open, round) three-balls.

$\endgroup$
5
  • 5
    $\begingroup$ Is there an approach through geometric group theory here? Specifically the Freudenthal-Hopf theorems that any finitely generated group either has $\leq 2$ ends or infinitely many. $\endgroup$ Commented Oct 12, 2023 at 17:56
  • 3
    $\begingroup$ Would you prefer a reference? I think I can dig out one. The relevant fact is that if $W$ is a compact manifold with boundary such that one boundary component carries the entire fundamental group, then $W$ is a compression body. $\endgroup$ Commented Oct 12, 2023 at 18:06
  • 1
    $\begingroup$ When the boundary has fundamental group isomorphic to that of $W$ then this is Theorem 10.2 in Hempel's book. But a quick skim of the usual suspects (Hempel, Hatcher, Jaco, Martelli) does not find the exact statement we want. So, yes, a citable reference would be much appreciated. $\endgroup$
    – Sam Nead
    Commented Oct 12, 2023 at 19:28
  • 1
    $\begingroup$ @SamNead You can reduce to that case. C.f. Lemma 2.1 in arxiv.org/pdf/1508.02993.pdf $\endgroup$
    – biringer
    Commented Oct 12, 2023 at 19:56
  • 1
    $\begingroup$ Or alternatively, the point is to take the characteristic comp body C and say that each interior boundary comp S is adjacent to its own component X of M - C (this uses pi1 surjectivity) and that S included into X is pi1 surjective, hence X is an interval bundle. $\endgroup$
    – biringer
    Commented Oct 12, 2023 at 20:31

1 Answer 1

12
$\begingroup$

Let $M$ be a compact connected 3-manifold with nonempty boundary such that $\pi_1(M)$ is infinite and the universal cover $\tilde{M}$ has finitely many boundary components. I will prove that $\tilde{M}$ has either one or two boundary components.

Note: After writing this up I fear that this is just as tedious as what you describe (though it doesn't use any real technology), so it probably isn't so far off from what you had in mind.

Passing to a double cover, we can assume that $M$ is orientable.

Consider a component $B$ of $\partial M$. If the image of $\pi_1(B) \rightarrow \pi_1(M)$ has infinite index, then the preimage of $B$ in $\tilde{M}$ has infinitely many components. Since $\tilde{M}$ has finitely many boundary components, we conclude that the image of $\pi_1(B) \rightarrow \pi_1(M)$ is finite-index. Passing to a finite cover, we can assume that the map $\pi_1(B) \rightarrow \pi_1(M)$ is surjective.

Passing to a further finite cover might make $B$ separate into several components, but each of those components is still $\pi_1$-surjective. We can thus perform the procedure in the previous paragraph repeatedly and ensure that for each component $A$ of $\partial M$, the map $\pi_1(A) \rightarrow \pi_1(M)$ is surjective.

The proof now breaks up into two cases.

Case 1: $M$ has at least two boundary components.

Let $A$ and $B$ be two of the boundary components.

If $A$ is compressible, then we can find a compressing disk $D$ with $\partial D \subset A$. The disk $D$ must be separating; indeed, if it is nonseparating then we can find a loop $\gamma$ in $M$ that intersects $D$ once with positive sign. Since $\gamma$ can be homotoped into $B$, however, it can also be homotoped to be disjoint from $D$, a contradiction. Since $B$ lies on a single side of $D$ and $\pi_1(B) \rightarrow \pi_1(A)$ is surjective, we see that $\pi_1$ of the component of $M \setminus D$ not containing $B$ must be trivial. But this contradicts the fact that $\partial D$ is an essential loop in $A$ (reflect upon half-lives half-dies).

We deduce that $A$ is incompressible, so $\pi_1(A) \cong \pi_1(M)$. But it now follows from results in the chapter on I-bundles in Hatcher that $M$ is an $I$-bundle, so $\tilde{M}$ has two boundary components.

Case 2: $M$ has exactly one boundary component.

We are in the situation where $M$ is a compact oriented $3$-manifold with connected boundary such that the map $\pi_1(\partial M) \rightarrow \pi_1(M)$ is surjective. A standard exercise using the loop theorem shows that this means that $M$ is a handlebody, so $\tilde{M}$ has connected boundary.

$\endgroup$
2
  • 1
    $\begingroup$ My (too long) proof worked directly with $M$. I used the canonical compression body (following Bonahon) and, if that was trivial, the canonical $I$-bundle as in JSJ theory. Your proof (plus Moishe's remark) is shorter. I think the following is good. Lift to get some boundary component $B$ to surject the fundamental group. The cover is thus a compression body. If it is a handlebody or a product, we have a contradiction. So it has an interior boundary component, $C$. The fundamental group of $C$ has infinite index, so we win. $\endgroup$
    – Sam Nead
    Commented Oct 12, 2023 at 18:51
  • 2
    $\begingroup$ @SamNead: Yes, I think that's a good way to do it! $\endgroup$ Commented Oct 12, 2023 at 18:54

You must log in to answer this question.

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