B-Bordism

From Manifold Atlas
Revision as of 18:58, 21 January 2010 by Crowley (Talk | contribs)
Jump to: navigation, search

This page has not been refereed. The information given here might be incomplete or provisional.

Contents

1 Introduction

On this page we recall the definition of the bordism groups of closed manifolds, with extra topological structure: orientation, spin-structure, weak complex structure etc. The ideas here date back to [Lashof1965]. There is a deatiled treatment in [Stong1968, Chapter II] and a summary in [Kreck1999, Section 1] as well as [Kreck&Lück2005, 18.10].

We specify extra topological structure universally by means of a fibration \gamma : B \to BO where BO denotes the classifying space of the stable orthogonal group and B is homotopy equivalent to a CW complex of finite type. Abusing notation, one writes B for the fibraion \gamma. A B-manifold is a compact manifold M together with a lift of a classifying map for the stable normal bundle of M to B:

\displaystyle  \xymatrix{  & B \ar[d]^{\gamma} \\ W \ar[r]_{\nu_W} \ar[ur]^{\bar \nu} & BO.}

The n-dimensional B-bordism group is defined to be the set of closed B-manifolds up modulo the relation of B-bordism and addition given by disjoint union

\displaystyle  \Omega_n^B := \{ (M, \bar \nu) \}/\simeq.

Alternative notations are \Omega_n(B) and also \Omega_n^G when B \to BO = BG \to BO for G \to O a stable represenation of a topological group G. Details of the definition and some important theorems for computing \Omega_n^B follow.

2 B-structures

In this section we give a compressed accont of [Stong1968, Chapter II]. Let G_{r, m} denote the Grassman manifold of unoriented r-planes in \Rr^n and let BO(r) = lim_{m \to \infty} G_{r, m} be the infinite Grassman and fix a fibration \gamma_r : B_r \to BO(r).

Definition 2.1. Let \xi: E \to X be a rank r vector bundle classified by \xi : X \to BO(r). A B_r-structure on \xi is a vertical homotopy class of maps \bar \xi : X \to B_r such that \gamma_r \circ \bar \xi = \xi.

Note that if \xi_0 and \xi_1 are isomorphic vector bundles over X then the sets of B_r-structures on each are in bijective equivalence. However B_r-structures are defined on specific bundles, not isomorphism classes of bundles: an specific isomorphism, up to appropriate equivalence, is required to give a map between the set of B_r structures. Happily this is the case for the normal bundle as we now explain.

Let M be a compact manifold and let i_0 : M \to \Rr^{n+r} be an embedding. Equipping \Rr^{n+r} with the standard metric, the normal bundle of i_0 is a rank r vector bundle over classified by its normal Gauss map \nu(i_0) : M \to G_{r, n+r} \subset BO(r). If i_1 is another such embedding and r >> n, then i_1 is regularly homotopic to i_0 and all regular homotopies are regularly homotopic relative to their endpoints. A regular homotopy H defines an isomorphism \alpha_H :\nu(i_0) \cong \nu(i_1) and a regular homotopy of regular homotopies gives a homotopy between these isomorphisms. Taking care one proves the following

Lemma 2.2 [Stong1968, p 15]. For r sufficiently large, (depending only on n) there is a 1-1 correspondence between the B_r structures of the normal bundles of any two embeddings i_0, i_1 : M \to \Rr^{n+r}.

Now let (B_r, \gamma_r) be a sequence of fibrations over BO(r) with maps g_r : B_r \to B_{r+1} fitting into the following commutative diagram

\displaystyle  \xymatrix{  B_r \ar[r]^{g_r} \ar[d]^{\gamma_r} & B_{r+1} \ar[d]^{\gamma_{r+1}} \\ BO(r) \ar[r]^{j_r} & BO(r+1) }

where j_r is the standard inclusion and let B = lim_{r \to \infty}(B_r). A B_r-structure on the normal bundle of an embedding i: M \to \Rr^{n+r} defines a unique B_{r+1}-structure on the composition of i with the standard inclusion \Rr^{n+r} \to \Rr^{n+r+1}.

Defition 2.3. A B-structure on M is an equivalence class of B_r-structure on M where two such structures are equivalent if they become equivalent for r sufficiently large. A B-manifold is a pair (M, \bar \nu) where M is a compact manifold and \bar \nu is a B-structure on M.

If W is a compact manifold with boundary \partial W then by choosing the inward-pointing normal vector along \partial W, a B-structure on W restricts to a B-structure on \partial W. In particular, if (M, \bar \nu_M) is a closed B manifold then W = M \times [0, 1] has a canonical B-structure \bar \nu_{M \times [0, 1]} such that restricting to (M, \bar \nu_M) on M \times \{ 0 \}. The restriction of this B-structure to M \times \{ 1 \} is denoted -\bar \nu: by construction (M \sqcup M, \bar \nu \sqcup - \bar \nu) is the boundary of (M \times [0, 1], \bar \nu_{M \times [0, 1]}.

Definition 2.4. Closed B-manifolds (M_0, \bar \nu_0) and (M_1, \bar \nu_1) are B-bordant if there is a compact B-manifold (W, \bar \nu) such that \partial(W, \bar \nu) = (M_0 \sqcup M_1, \bar \nu_0 \sqcup -\bar \nu_1). We write [M, \bar \nu] for the bordism class of (M, \bar \nu).

Proposition 2.5. The set of B-borism class of closed n-manifolds with B-structure,

\displaystyle  \Omega_n^B := \{ [M, \bar \nu ] \},

forms an abelian group under the operation of disjoint union with inverse -[M,\bar \nu] = [M, -\bar \nu].

3 The Pontrjagin Thom isomorphism

If E is a vector bundle, let T(E) denote its Thom space. Recall that B = (B_r, \gamma_r, g_r) is a sequence of fibrations \gamma_r : B_r \to BO(r) with compatible maps g_r : B_r \to B_{r+1}. These data give rise to a stable vector bundle as defined in [Kreck&Lück2005, 18.10] with E_r \to B_r equal to the pullback bundle \gamma_r^*(EO(r)) where EO(r). This stable vector bundle defines a Thom spectrum which we denote MB. The r-th space of MB is T(E_r).

By definition a B-manifold (M, \bar \nu) is an equivalence class of B_r structures on \nu(i), the normal bundle of an embedding i : M \to \Rr^{n+r}. Hence (M, \bar \nu) gives rise to the collapse map S^{n+r} \to T(E_r) where identify S^{n+r} with the one-point compatificiation of \Rr^{n+r}, we map via \bar \nu_r on a tubular neighbourhood of i(M) \subset \Rr^{n+r} and we map all other points to the base-point of T(E_r). As r increase these maps are compatibly related by suspension and the structure maps of the spectrum MB. Hence we obtain a homotopy class PT((M, \bar \mu)) \in lim_{r \to \infty}(\pi_{n+r}(T(E_r)) = \pi_n(MB). The celebrated theorem of Pontrjagin and Thom states in part that PT(M, \bar \nu) depends only on the bordism class [M, \bar].

Theorem 3.1. There is an isomorphism of abelian groups

\displaystyle  PT : \Omega_n^B \cong \pi_n^S(MB), ~~~[M, \bar \nu] \longmapsto PT([M, \bar \nu]).

4 Spectral sequences

\displaystyle  H_p(B; \pi_q^S) \Longrightarrow \Omega_{p+q}^B


5 References

Personal tools
Namespaces
Variants
Actions
Navigation
Interaction
Toolbox