Let $M$ be a connected manifold with precisely $k$ ends $\epsilon_1,...,\epsilon_k$. Choose a collection $(U_i)_{i=1}^k$ of pairwise disjoint open $\epsilon_i$-neighborhoods. Then I wonder how to prove that $N := M \setminus (\bigcup_{i=1}^k U_i)$ is a bounded subset of M, i.e, a set with compact closure.

Thoughts I had so far: It is clear that each component of $N$ is bounded by itself. However, it is quite possible that $N$ consists of infinitley many such components, so an argument on finite union doesnt work. Does anybody have an idea ?

Edit: As established in the comments, the problem lies in the fact that there at least two definitions for ends of manifolds that are non-equivalent in general:

I used a version of Siebemanns definition (1965), where an end of $X$ is an equivalence class of sequences $U_1 \supset U_2 \supset U_3...$ of connected open subsets with compact boundary, so that $\bigcap_{i=1}^\infty cl(U_i) = \emptyset$. Two such sequences $(U_i)$ and $(V_i)$ are equivalent if for each $n \in \mathbb N$ there exists $j \in \mathbb N$, so that $U_i \supset V_j$ and $V_i \supset U_j$. The set of corresponding ends is denoted by $\mathcal E_1$.

If $X$ admits a compact exhaustion $(K_i)_{i=1}^\infty$ (for example if $X$ is a manifold), i.e. a sequence of compact subsets $K_1 \subset K_2 \subset ...$ with $\bigcup_{i=1}^\infty int(K_i) = X$, then one can define another set of ends $\mathcal E_2$ as the inverse limit of the system $\pi_0(X\setminus K_1) \leftarrow \pi_0(X \setminus K_2) \leftarrow \pi_0(X \setminus K_2)....$.

Now my original question can be formulated as follows:

If $X$ is a connected manifold, is there a bijection of sets $f: \mathcal E_1 \to \mathcal E_2$ ?

  • $\begingroup$ By a bounded set, I mean a set with compact closure. This is independent of any particular metric and coincides with the usual definition of metric boundedness, provided I have chosen a complete metric on $M$ $\endgroup$
    – H1ghfiv3
    Nov 29 '15 at 9:20
  • $\begingroup$ Also, it does not directly follow from the definition of an end. For example, one could take the wedge sum of infinitley many closed intervals, along with attaching a half-open interval of the form [0,1) at 0. The resulting space has one end, but one can easily find an end neighborhood with unbounded complement $\endgroup$
    – H1ghfiv3
    Nov 29 '15 at 9:24
  • $\begingroup$ The infinite wedge of closed intervals is not a nice space. If you put the quotient topology on it, it is not first countable. Even if you put the "french railway metric" on it, the center does not have a compact neighbourhood. So, what would an end be in such a space? $\endgroup$ Nov 29 '15 at 11:35
  • $\begingroup$ I think the problem lies in the fact that there are at least two definitions for ends out there that are non-equivalent in general. I use the definition from Siebemann's PHD thesis (1965, Princeton), where an end is, essentially, an equivalence class of nested sequences $(U_i)_{i=1}^n$ of connected open sets with compact boundary with the property that $\bigcap_{i=1}^\infty cl(U_i) = \emptyset$. Two such sequences are equivalent if they "contain one another". There is also the (more modern) definition as the limit of a particular inverse system, found on the english wikipedia $\endgroup$
    – H1ghfiv3
    Nov 29 '15 at 11:47
  • 1
    $\begingroup$ @BerniWaterman: in resolving these foundational confusions you may benefit from reading arxiv.org/abs/1210.6741 especially section 3. $\endgroup$ Nov 29 '15 at 14:51

Assume that $M$ is a noncompact manifold. Then there exists a bijection between both sets of ends.

Start with $e\in\lim_{\longleftarrow}\pi_0(M\setminus K)$. For each exhausting sequence $K_1\subset K_2\subset\dots$ of compact subsets with $\bigcup_iK_i=M$, let $U_i$ be the connected component of $M\setminus K$ indexed by $e$. Then $\partial U_i=cl(U_i)\cap K_i$ is compact. One checks that different compact exhaustions of $M$ give equivalent sequences.

In the opposite direction, let $(U_i)_i$ be a Siebenmann end, and let $K_i$ be a compact exhaustion as above. For each $i$ there exists $j$ with $\partial U_i\subset K_j$, and we regard $V_i=U_i\setminus K_j$. If $V_i$ is connected, then $V_i$ defines a point $e(K_j)\in\pi_0(M\setminus K_j)$. If not, we find $k>i$ with $U_k\cap K_j=\emptyset$ because $\bigcap_kcl(U_k)=\emptyset$. Then $U_k$ is contained in one of the unbounded connected components of $V_i$, which we pick as $e(K_j)$. Because the sequence $K_i$ is an exhaustion, this extends to the whole inverse system.

Back to the original question. If there are only finitely many ends, choose one neighbourhood $U_i$ for each end and take $N$ as in the question. In particular, there exist compact sets $K_i$ such that $U_i$ contains one connected component of $M\setminus K_i$. Put $K=K_1\cup\dots\cup K_k$, then $K$ is compact and $N\subset K$.

  • $\begingroup$ The fact that the set $V_i$ that you defined contains an unbounded component is not so trivial in my eyes. It is true, however, so your proof works fine! $\endgroup$
    – H1ghfiv3
    Nov 29 '15 at 13:58

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.