2
$\begingroup$

We start with product manifold $ X := S_g \times (0,1)$, where $S_g$ is a closed, orientable surface of genus $g \geq 1$. Since this has $\mathbb R^3$ as its universal cover, $X$ is irreducible. Now $X$ can also be seen as the interior of $\bar{X} = S_g \times [0,1]$. We want to manipulate $X$ a bit in order to create a bordism between $S_g$ and $S_{g+1}$, whose interior is ,hopefully, also an irreducible manifold.

Consider a boundary component $R := S_g \times \{0\}$ and fix a 2-disc $D \subset R$. This disc has an open subset $U \subset \bar{X}$ that is diffeomorphic to $D \times [0,\epsilon)$, such that $U \cap R = D$ (This follows since $R$ has a collar). Choose two distinct points $p,q \in D$ and connect them via a simple (injective) smooth arc $\gamma \subset U$ with the property that $\gamma \cap R = \partial \gamma = \{p,q\}$. Moreover, to ensure that $\gamma$ is as "simple" as possible, we additionally require that there exists a simple path $\gamma' \subset D$ between $p$ and $q$ such that the simple closed curve $\gamma \cup \gamma'$ bounds a disc $D' \subset U$, intersectig $D$ only in its boundary arc $\gamma'$. Drawing a picture isn't hard and really helps to make this paragraph more clear.

Now consider a tubular neighborhood $N \subset U$ of $\gamma$, such that $N \cong D_2 \times [0,1]$, where $D_2$ denotes the standart open unit disc of $\mathbb R^2$ and define $\bar{Y} := \bar{X} \setminus N$. Effectively, we have removed an open half-tube from $\bar{X}$ and simultaneously added a handle to the boundary component $R$, thereby increasing its genus by $1$. Hence, $\bar{Y}$ describes a bordism between $S_g$ and $S_{g+1}$ and, up to diffeomorphism, it doesnt depend on the exact choices of $D,p,q,U$ and $\gamma $. If $Y$ denotes its interior, is it clear that $Y$ is still irreducible ?

Remarks:

  1. If $\tilde{Y}$ denotes the universal cover of $Y$, then $Y$ is irreducible if and only if $H_2(\tilde{Y}) = 0$, which follows from Hurewicz and the sphere theorem.

  2. Knowing that $\mathbb R^3$ covers $X$, what can be said about $\tilde{Y}$? Since $Y$ can be regarded as an open subset of $X$ and $\mathbb R^3$ covers $X$ via some map $p$, we know that each component of $p^{-1}(Y)$ is a covering space of $Y$. What can be said about those components ?

$\endgroup$
  • $\begingroup$ In the first line you want $X := S_g \times (0,1)$. $\endgroup$ – Lee Mosher Nov 4 '15 at 18:45
3
$\begingroup$

Yes, $Y$ is still irreducible, and this holds by a simple connectivity argument.

Suppose that $\Sigma \subset Y \subset X$ is a smoothly embedded 2-sphere. Since $X$ is irreducible, $\Sigma$ bounds a smoothly embedded ball $B \subset X$. Note that $\bar X - \Sigma$ has two components: the inside of $\Sigma$ which equals the interior of $B$, and the outside of $\Sigma$ which equals the complement of $B$. In $\bar X$, clearly $\Sigma$ is disjoint from $N$. Since $N \cup (S_g \times \{0\})$ is a connected subset of $\bar X - \Sigma$, and since $S_g \times \{0\}$ is in the outside of $\Sigma$, it follows that $N \cup (S_g \times \{0\})$ is in the outside of $\Sigma$. Thus $B \subset Y$.

$\endgroup$
  • $\begingroup$ Nice and easy. I overcomplicated the situation by passing to the universal cover. Thank you very much! $\endgroup$ – Berni Waterman Nov 4 '15 at 19:15
2
$\begingroup$

To prove irreducibility it suffices to prove $\pi_2=0$. Assume $\pi_2Y\not=0$, then the sphere theorem gives you an embedded sphere representing a nontrivial element in $\pi_2Y$. we can assume that it doesn't hit the boundary of $Y$

Because of $\pi_2X=0$, this sphere bounds some map $f\colon B^3\to X$. After a small homotopy $f$ will be transversal to $\partial Y$, so $f^{-1}(X\setminus Y)$ is a properly embedded submanifold with boundary $f^{-1}(\partial Y)$.

But $f^{-1}(\partial Y)\cap S^2=\emptyset$, so the image of $f$ has to be contained in $Y$ and so the sphere does not represent a nontrivial element in $\pi_2Y$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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