# topological type of smooth manifolds with prescribed homotopy type and pontryagin class

Can someone help explain the following result:

If the dimension is at least five, there are at most finitely many different smooth manifolds with given homotopy type and Pontryagin classes.

Thank you so much!

-
Diffeomorphism classes of smooth manifolds... but I guess people would know what you meant. How much detail do you need? –  David Roberts Jan 18 '11 at 0:27

In the $1$-connected case, one may argue as follows:

Let $X$ be a closed $1$-connected smooth $n$-manifold, $n \ge 5$. The theory of the Spivak fibration shows that any homotopy equivalence $f: M^n \to X$ with $M$ smooth is covered by a stable fiber homotopy equivalence of underlying stable tangent spherical fibrations of $M$ and $X$. Call $f$ stably tangential if this equivalence of stable spherical fibrations lifts to an isomorphism of stable tangent vector bundles.

Then the surgery exact sequence shows that any stable tangential homotopy equivalence $f: M \to X$ is homotopic to a diffeomorphism $f': M \sharp \Sigma \to X$, where $\Sigma$ is a homotopy sphere, and $\sharp$ means connected sum.

(You can either quote here Corollary II.3.8 of Browder's book, or you can deduce it directly from the surgery exact sequence. The point is that connected sum gives an action of the homotopy $n$-spheres on the the structure set of $X$, and one can compare the surgery exact sequence for $M$ and the sphere to deduce the above statement.)

To finish the proof of what you want, notice:

1. Kervaire and Milnor showed that there are only finitely many homotopy spheres in each dimension $\ge 5$.

2. The obstruction to the homotopy equivalence being stably tangential is given by its normal invariant, which is an element of $[X,\text{G/O}]$ (it's given by the "difference" between the stable tangent bundles of $M$ and $X$, i.e., $(f^{-1})^*\tau_M - \tau_X$, appropriately interpreted).

3. The map $\text{G/O} \to B\text{O}$ is a rational homotopy equivalence (by Serre). The image of the normal invariant in $[X,B\text{O}]$ is rationally detected by the difference of the Pontryagin classes of $M$ and $X$ using the fact that $H^*(B\text{O}; \Bbb Q)$ is a polynomial algebra on the $p_i$.

-
I should also mention that $G$ is the topological monoid whose classifying space classifies stable spherical fibrations. Its homotopy groups coincide with the homotopy groups of spheres in positive degrees. So it has trivial rational homotopy. –  John Klein Jan 18 '11 at 2:51
Without assuming $1$-connectedness I doubt that the statement is true. Can't there be infinitely many manifolds all h-cobordant but not diffeomorphic? –  Tom Goodwillie Jan 18 '11 at 3:20
Yes, I suspected as much... –  John Klein Jan 18 '11 at 3:25
@Goodwillie: Is there a reference for that fact? @Klein: great answer! –  Dr Shello Jan 18 '11 at 7:33