6 0. Prequel
0.3. The ultimate duality theorem
Homology and cohomology are powerful tools in the study of embeddings.
One reason for their usefulness is the many duality theorems they satisfy that
relate the homology/cohomology of an embedded object to the cohomol-
ogy/homology of its complement. The duality theorems found in (Munkres,
1984), which apply to triangulable homology manifolds, are not quite gen-
eral enough for our purposes, so we state the theorems to be employed. The
following basic duality theorem can be found on page 342 of (Spanier, 1966).
Theorem 0.3.1 (Duality). Let G be an abelian group, and let (A, B) be a
pair of closed subsets of the G-orientable n-manifold M. Then for all p 0
there exists a natural isomorphism
Hn−p(M B, M A; G)


B; G).
The group on the right is called the pth Alexander (relative) cohomol-
ogy group of (A, B) with G-coefficients and compact supports. (Throughout
this section G will denote an Abelian group.) When A and B are com-
(A, B; G) equals the more usual pth Alexander cohomology of the
pair (again with G-coefficients), and when, in addition, both A and B are
themselves complexes, manifolds, or absolute neighborhood retracts (to be
defined in §0.6), it equals the usual singular cohomology of the pair. A
connected n-manifold M is G-orientable if Hc

= G; an arbitrary
n-manifold is G-orientable if each of its components is. Every n-manifold is
Z2-orientable; those that contain a copy of MB × In−2, where MB denotes
the obius Band, fail to be Z-orientable.
Say that a ∂-manifold W is G-orientable if Int W is.
Corollary 0.3.2 (Poincar´ e-Lefschetz Duality). If W is a G-orientable n-
dimensional ∂-manifold, then Hn−p(W ; G)

= Hc
(W, ∂W ; G) for every p.
Proof. It follows from Collaring Theorem 2.4.10, to be proved later, that
the manifold W = W ∂W × (−1, 0], obtained from the disjoint union of
W and ∂W × (−1, 0] by attaching the product ∂W × (−1, 0] to ∂W W
along ∂W × {0} in the obvious way, is homeomorphic to Int W and, thus, is
G-orientable. Apply the duality theorem in W with A = W and B = ∂W
to obtain
∂W ; G)

= Hn−p(W ∂W, W W ; G)
= Hn−p(Int W ∂W × (−1, 0),∂W × (−1, 0); G)

= Hn−p(Int W ; G)

= Hn−p(W ; G).
Previous Page Next Page