Talk:S-duality I (Ex)

From Manifold Atlas
Revision as of 00:34, 30 May 2012 by Martin Palmer (Talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Exercise 0.1 (first part). Any finite CW complex X has an S^N-dual, for N sufficiently large.

Proof. (tentative: see remarks below) Let X be a finite pointed CW complex with basepoint x_0. Embed X\hookrightarrow S^N with regular neighbourhood M, so that

\displaystyle  X \hookrightarrow M \hookrightarrow S^N

with X\hookrightarrow M a homotopy-equivalence and M an N-dimensional manifold-with-boundary embedded in S^N.

Now let

\displaystyle  c\colon S^N \longrightarrow S^N / (S^N \setminus \mathring{M}) = M / \partial M

be the collapse map,

\displaystyle  \Delta\colon M / \partial M \longrightarrow (M\times M) / ((\partial M \times M)\cup (M\times \lbrace x_0 \rbrace)) \;\cong\; (M / \partial M) \wedge M

be the map induced by

\displaystyle  x\mapsto (x,x)\colon (M,\partial M) \longrightarrow (M\times M, (\partial M \times M)\cup (M\times \lbrace x_0 \rbrace)),

and let \alpha be the composite

\displaystyle  S^N \xrightarrow{c} M / \partial M \xrightarrow{\Delta} (M / \partial M) \wedge M \xrightarrow{1\wedge r} (M / \partial M) \wedge X ,

where r\colon M\to X is a chosen homotopy-inverse for X\hookrightarrow M.

The map c is degree-1 by construction, so c_*([S^N]) = [M] \in \widetilde{H}_N(M / \partial M) = H_N(M,\partial M).

For any x\in \widetilde{H}^i(M), we have the identity

\displaystyle  x\setminus \Delta_*([M]) \;=\; x\cap [M] \;\in\; \widetilde{H}_{N-i}(M / \partial M) = H_{N-i}(M,\partial M),

so by Poincare-Lefschetz duality,

\displaystyle  -\setminus (\Delta\circ c)_*([S^N]) \;\colon\; \widetilde{H}^i(M) \longrightarrow \widetilde{H}_{N-i}(M / \partial M)

is an isomorphism for all i.

In general for any X\wedge Y \xrightarrow{1\wedge f} X\wedge Z and a\in \widetilde{H}_N(X\wedge Y),

\displaystyle  f^*(-)\setminus a \;=\; -\setminus (1\wedge f)_*(a) \;\colon\; \widetilde{H}^i(Z) \longrightarrow \widetilde{H}_{N-i}(X).

So in our case we have

\displaystyle  -\setminus \alpha_*([S^N]) \;=\; r^*(-)\setminus (\Delta\circ c)_*([S^N]) \;\colon\; \widetilde{H}^i(X) \longrightarrow \widetilde{H}_{N-i}(M / \partial M)

which is an isomorphism by above and the fact that r is a homotopy-equivalence. This witnesses that M / \partial M is an S^N-dual of X.

\square

Remark 0.2. This uses reduced homology instead of unreduced homology in the definition of S-duality, which is correct (?) as we are using smash products rather than direct products.

Remark 0.3. It doesn't seem to be immediately clear why the map \Delta defined above is the correct geometric `diagonal' map to use so that the claimed identity x\setminus \Delta_*(z) = x\cap z holds. It seems more natural to use the map

\displaystyle  M / \partial M \longrightarrow M\times M / (\partial M \times M),

but this would witness that M / \partial M is an S^N-dual of X_+, rather than X\ldots

Personal tools
Namespaces
Variants
Actions
Navigation
Interaction
Toolbox