14
$\begingroup$

Let $X$ and $Y$ be two path-connected $n$-dimensional manifolds. Then one can construct their connected sum $X\# Y$ which can be visualised as removing a disc from $X$, removing a disc from $Y$, and gluing the boundaries together using a cylinder (a copy of $S^{n-1}\times[0,1]$).

There is a natural $(n-1)$-sphere in $X\# Y$ due to the cylinder. If $X$ or $Y$ is a sphere, then this $(n-1)$-sphere will bound. Is this the only case in which it bounds?

If $X\# Y$ is a non-trivial connected sum (i.e. neither $X$ nor $Y$ is a sphere), is the map $S^{n-1} \to X\# Y$ essential (i.e. not nullhomotopic)?

Oscar Randal-Williams' answer to this question shows that the inclusions of the sphere into $X\setminus D^{\circ}$ and $Y\setminus D^{\circ}$ are essential.

If the answer is yes, then $\pi_{n-1}(X\# Y) \neq 0$. I don't even know if this is true.

If $X \# Y$ is a non-trivial connected sum of $n$-dimensional manifolds, is $\pi_{n-1}(X\# Y)$ necessarily non-zero?

I believe the answer to both of these questions is yes, but I have not been able to completely convince myself.

$\endgroup$
2
  • 4
    $\begingroup$ At the level of generality you've asked your question, the answer is no. Let $X$ and $Y$ be inverse homotopy spheres, then $X \# Y$ is an $n$-sphere even though the summands are not. You can do this in any dimension where there are non-trivial homotopy spheres. $\endgroup$ Oct 4, 2017 at 5:13
  • 2
    $\begingroup$ @RyanBudney: the question does not seem to be about smooth manifolds. $\endgroup$ Oct 4, 2017 at 13:19

1 Answer 1

15
$\begingroup$

The answer to this is no, in a stronger sense than Ryan's comment above. There are two sources, Kahn, Peter J. Codimension-one imbedded spheres (Invent. Math. 10 1970 44–56), and an appendix Null-homotopic, codimension-one embedded spheres, to a paper of Terng-Thorbergsson (Taut Immersions into Complete Riemannian Manifolds in Tight and taut submanifolds. MSRI Publications, 32, 1997) that I wrote, not knowing Kahn's result. The main theorem of the appendix states:

Suppose that $\imath:S^{n-1} \to M^n$ is a smooth embedding which is null-homotopic. Then either $S$ bounds a homotopy ball on one side, or {1.} M is a rational homology sphere, and therefore X and Y are as well. {2.} The fundamental groups of both X and Y are finite, and at least one of them is trivial.

In the appendix, there are constructions of connected sums of rational homology spheres where the summing sphere is null-homotopic that show that this result is essentially best possible.

$\endgroup$
3
  • $\begingroup$ Do you know of an example where the second possibility occurs? $\endgroup$ Oct 4, 2017 at 18:46
  • $\begingroup$ All the examples I know are in the papers I cited; it's been a long time so I don't remember the constructions. The appendix that I wrote is on my web page people.brandeis.edu/%7eruberman/drpapers/taut.final.gz. I think Kahn's paper also gives concrete examples. I'm actually confused a bit by your question; I think that what I intended was that either there's a homotopy ball on one side, or (1 and 2 hold). $\endgroup$ Oct 4, 2017 at 19:48
  • $\begingroup$ My bad. I didn't realise that the alternative to bounding a homotopy ball on one side is 1 and 2. $\endgroup$ Oct 4, 2017 at 21:56

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.