4
$\begingroup$

How to use Serre spectral sequence to compute cup product structures?

Let $F\to E\to B$ be a fibration. Suppose all the differentials of the corresponding Serre spectral sequence of cohomology are zero. Can we obtain that for any field $k$, $$ H^*(E;k)\cong H^*(F;k)\otimes H^*(B,k) $$ as rings?

For example, how to solve the following problem?

enter image description here

$\endgroup$

2 Answers 2

8
$\begingroup$

Maybe the simplest example is the following. There are two fiber bundles with base and fiber both $S^2$. One is the product $S^2\times S^2$ and the other consists of two copies of the mapping cylinder of the Hopf map $S^3 \to S^2$ glued together via the identity map between their "source" ends $S^3$. The resulting space is the connected sum of two copies of ${\mathbb C}P^2$ (with opposite orientations). In terms of clutching functions this is the $S^2$-bundle associated to the nontrivial element of $\pi_1SO(3)$. The cup product structures in the two total spaces are different because in the nontrivial bundle obtained from Hopf maps there are classes in $H^2$ that square to a generator of $H^4$, as one can see from the quotient map collapsing one of the mapping cylinders to a point, with quotient space ${\mathbb C}P^2$, while for $S^2\times S^2$ the square of any class in $H^2$ is always an even multiple of a generator of $H^4$.

$\endgroup$
7
$\begingroup$

A good example showing the answer to question 1 is no is the case of the projectivization of a complex vector bundle $V$ over $B$. This has $F=\mathbb C P^{n-1}$, and it is a theorem (see e.g. Hatcher, Vector bundles and K-Theory) that all differentials vanish or equivalently $H^*E$ is a free $H^*B$-module on generators $1,x,\dots x^{n-1}$, where $x$ restricts to a generator of $H^2E$. However, one has the equation $$\sum_{i=0}^n c_i(V)x^{n-i}=0,$$ where $c_i(V)$ are the Chern classes, not $x^n=0$.

So in general: the thing one knows is the Leray-Hirsch theorem: $H^*E$ is a free $H^*B$-module on generators $x_i$ which map to generators $y_i$ of $H^*F$.

If $y_iy_j=y_k$, then one only knows that $x_ix_j=x_k$ modulo terms of higher filtration.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.