6
$\begingroup$

Is there a simply-connected smooth closed 4-manifold with a characteristic class $x \in H_2(X; \mathbb{Z})$ such that $x$ can not be represented by a disjoint union of tori in $X$?

I would not know how to prove this without the characteristic hypothesis either so any thoughts on that would also be appreciated.

$\endgroup$
0

1 Answer 1

15
$\begingroup$

In $H_2(CP^2)$, every class $nH$ where $H$ is a generator and n is odd is characteristic. However, if $n >3$, then such a class is not represented by a torus. It is not represented by a disjoint union of tori, either. For non-zero classes in $H_2(CP^2)$ have non-zero intersection numbers. So if you had a disjoint union of tori, then at most one would be non-trivial in homology (and that one would represent $nH$, contradicting the previous step.)

$\endgroup$
4
  • $\begingroup$ Thanks for this example - this then shows that in $CP^2$ there are exactly two characteristic homology classes that can be represented by a disjoint union of tori. I wonder if there are any such 4-manifolds where no characteristic homology class can be represented by disjoint tori. $\endgroup$
    – user101010
    Feb 21, 2021 at 16:39
  • $\begingroup$ Such a manifold might be hard to find; you can find characteristic tori (indeed spheres) in the only known examples of definite manifolds (sums of $CP^2$). But if your manifold is indefinite, then you will have characteristic classes of negative square, and obstructions to being represented by a torus are harder to come by. $\endgroup$ Feb 21, 2021 at 18:54
  • $\begingroup$ @DannyRuberman Can you please expand the last implication: why must all but one be trivial? $\endgroup$
    – magicker72
    Feb 22, 2021 at 3:43
  • 1
    $\begingroup$ @magicker72 If two where nontrivial, then they would need to intersect (since any two homologically nontrivial surfaces intersect), therefore you would not have disjoint surfaces. $\endgroup$
    – user101010
    Feb 22, 2021 at 11:35

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.