8
$\begingroup$

Let $n \in \mathbb{Z}_{>0}$. I'm wondering about pairs $(X,V)$ where $V$ is a $2n$-dimensional vector bundle on the CW complex $X$, such that

  • The ring homomorphism $\mathbb{Q}[e] \to H^*(X;\mathbb{Q})$ induced by the Euler class $e = e(V) \in H^{2n}(X)$ is an isomorphism.
  • $X$ has cells only in dimension 0, $2n$, $4n$, $6n$, $\dots$.

The first property asserts a rational equivalence $X \to K(\mathbb{Z},2n)$, which is then to be factored over the map $BSO(2n) \to K(\mathbb{Z},2n)$ representing the Euler class. By an obstruction theory argument, it is not hard to prove existence of such a pair $(X,V)$, but the proof makes many choices and uses that the homotopy groups of $BSO(2n)$ are finite in all odd degrees. So I'm wondering:

How explicitly can one exhibit a pair $(X,V)$ with these two properties?

Sketch of existence, by induction on skeleta: Start with $X(2n) = S^{2n}$ and $V\vert X(2n)$ being the tangent bundle. Assume the skeleton $X(2n(k-1))$ and the restriction $V\vert X(2n(k-1))$ have been constructed, and choose a map $f': S^{2nk -1} \to X(2n(k-1))$ representing a generator of $\pi_{2nk-1}(X(2n(k-1)))\otimes \mathbb{Q} \cong \mathbb{Q}$. Then the CW complex $X' = X(2n(k-1)) \cup_{f'} D^{2nk}$ has rational cohomology ring $H^*(X';\mathbb{Q}) \cong \mathbb{Q}[x]/x^{k+1}$ for a class $x$ restricting to $e(V)$ on $X(2n(k-1))$. We are free to replace $f'$ by any non-zero multiple, and since $\pi_{2nk-1}(BO(2n))$ is a finite group, we can choose $f: S^{2nk-1} \to X(2n(k-1))$ such that the vector bundle admits an extension to $$X(2nk) = X(2n(k-1)) \cup_f D^{2nk},$$ and its euler class satisfies $e^k \neq 0 \in H^{2nn}(X(2nk)) = \mathbb{Z}$. Q.e.d.

I am curious whether one can instead just write down an explicit example without making choices. For $n=1$ we can use $\mathbb{C} P ^\infty$ and for $n=2$ we can use $\mathbb{H} P^\infty$, with their canonical bundles. For higher $n$, perhaps some other geometric construction works? (The space $X = \Omega S^{2n+1}$ has the right kind of cell structure, but I can't think of any interesting vector bundles on it.)

$\endgroup$

0

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.