Homology of complement of compact submanifold

From Topospaces
Revision as of 18:02, 3 December 2007 by Vipul (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Statement

Suppose M is a manifold and K is a compact connected manifold embedded as a submanifold inside M. The general problem asks for computation of the homology of MK, without any prior information about how K has been embedded inside M.

Particular cases

In general, the answer is not determined, and depends on the way that K sits inside M. However, for certain special cases, the answer can be determined; for instance, when K is a single point, is the problem of point-deletion inclusion, where the answer is completely determined in terms of the homology of M, except at dimensions n and n1.

When the submanifold is a sphere

Suppose M has dimension n and K=Sm has dimension m. Then the Alexander duality theorem yields that:

Hn(M,MK)=Hnm(M,MK)=Z

and all other relative homologies are 0. This puts significant constraints on the homologies of MK; nonetheless these homologies are not determined uniquely by the data.

When the manifold is compact connected orientable

If M itself is a compact connected orientable manifold, and K is any compact connected manifold, then we have:

Hn(M)Hn(M,MK)Hn(M,Mp)

All three groups are Z and the composite is an isomorphism; this forces the map from Hn(M) to Hn(M,MK) to be zero. Writing down the long exact sequence of homology of a pair (M,MK), we see that Hn(MK)=0, and the long exact sequence can be truncated to begin at:

0Hn1(MK)Hn1(M)

When the manifold and submanifold are spheres

A special case of both the above occurs when M=Sn and K is homeomorphic to Sm, m<n. In this case, the homologies of the complement of K in M can be determined completely, using the two observations above, and the fact that all lower reduced homologies of M are zero.

What we deduce is that the homologies of the complement are Z at nm1 and 0 if mn1, and ZZ at 0 if m=n1. Homologies are 0 elsewhere.