4
$\begingroup$

Let $M$ be a compact $3$-manifold such that no component of $\partial M$ is $S^2$ and one component $F$ of $\partial M$ is the projective plane.

If $i_*:\pi_1(F) \to \pi_1(M)$ is an isomorphism, can we prove that $M$ is homeomorphic to $F \times [0,1]$?

$\endgroup$

1 Answer 1

3
$\begingroup$

Yes, the result follows from a theorem of Livesay and the Poincaré Conjecture that if $M$ is a compact connected non-orientable 3-manifold with $\pi_1(M)$ finite, then $M$ is homeomorphic to $P^2\times I$ minus a collection of disjoint open 3-balls.


The first answer to the original question that I wrote here was flawed. Here is a corrected version, thanks to the comments below.

First note that $M$ is non-orientable since it contains a 2-sided projective plane. Let $N$ be the orientation double cover of $M$, and let $p:N\to M$ be the covering map. Then $N$ is simply-connected since $\pi_1(M)\cong\mathbb{Z}/2\mathbb{Z}$.

If some component of $\partial M$ had non-positive Euler characteristic, then the same would be true of some component of $\partial N$, but this cannot happen since $N$ is simply-connected. By assumption, no component of $\partial M$ is a 2-sphere. Hence $\partial M$ is a collection of projective planes. Let $k$ be the number of components of $\partial M$.

By the Poincaré Conjecture, $N$ is homeomorphic to a 3-sphere with the interiors of $k$ disjoint 3-balls removed. Let $q:N\to N$ be the covering transformation, an orientation reversing involution. Now $H_2(N;\mathbb{Q})\cong\mathbb{Q}^{k-1}$ and $H_2(N;\mathbb{Q})$ is generated by any $k-1$ components of $\partial N$. Since $q$ acts on each component of $\partial N$ as an orientation reversing homeomorphism, it follows that the Lefschetz number $\Lambda_q$ is given by $$\Lambda_q=1-0+(-(k-1))=2-k.$$ But $q$ has no fixed points, so $\Lambda_q=0$ and $k=2$. See Epstein for the original (and more general) proof of this step.

Therefore $N$ is homeomorphic to $S^2\times I$, and it follows by Livesay or Rubinstein that $M$ is homeomorphic to $P^2\times I$. The argument I had in mind here involves doubling $N$ along its boundary to get $S^2\times S^1$. The covering map $p$ also extends to the double. Then the classification of Seifert fibered spaces covered by $S^2\times\mathbb{R}$ tells us that the only compact quotient of $S^2\times S^1$ which contains a 2-sided projective plane is $P^2\times S^1$. Hence $M$ is homeomorphic to $P^2\times I$.

$\endgroup$
6
  • 1
    $\begingroup$ The fact that $k=2$ follows from the Lefschetz fixed point theorem. Indeed, since the deck transformation of the covering map has no fixed point, its Lefschetz number must be $0$. By direct calculation, $k=2$. What I do not understand is how to deduce $M=P^2 \times I$ if we know $N=S^2 \times I$? $\endgroup$
    – Adterram
    Commented Aug 29, 2021 at 6:13
  • 3
    $\begingroup$ It is a theorem of Livesay (1963 Annals of Math., pp. 582-593) that $M=P^2\times I$ if $N=S^2\times I$. A shorter proof was later published by Rubinstein in the 1976 Proceedings of the AMS, pp. 317-320. $\endgroup$ Commented Aug 29, 2021 at 13:07
  • $\begingroup$ If this new argument is now correct, the comment stating that there is a gap/mistake should probably be deleted or modified to avoid future confusion. $\endgroup$ Commented Sep 1, 2021 at 20:53
  • $\begingroup$ Surely the bottom line is that the desired result is in fact Epstein's theorem (plus the Poincare conjecture)? From the MR review of Epstein's paper: "In the second chapter the author shows, surprisingly, that any compact non-orientable 3-manifold with finite fundamental group has to be (modulo the Poincaré conjecture) homeomorphic to $P^2×I$ minus a number of 3-balls". $\endgroup$
    – HJRW
    Commented Sep 4, 2021 at 13:50
  • 2
    $\begingroup$ Epstein's result is that such a manifold is homotopy equivalent to $P^2\times I$ minus some balls. See the comments on p477 of his paper, where he states that to upgrade to homeomorphism, one needs to prove two conjectures: one is the Poincaré Conjecture, the other is the result of Livesay. $\endgroup$
    – Josh Howie
    Commented Sep 4, 2021 at 18:05

You must log in to answer this question.

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