10
$\begingroup$

If one starts with a fiber bundle $f: X \to Y$ so that fibers having trivial integral homology by using spectral sequence one can get the induced map $f_*: H_*(X;\mathbb{Z}) \to H_*(Y;\mathbb{Z})$ is an isomorphism.

My question is as follows:

Given a surjective submersion (I am not considering it to be proper) $f:X \to Y$ (connected smooth manifolds) such that all the fibers have trivial integral homology does this mean the induced map $f_*: H_*(X;\mathbb{Z}) \to H_*(Y;\mathbb{Z})$ is isomorphism?

I will be grateful to see some counterexamples or proof.

Thanks in advance!!

$\endgroup$
4
  • 1
    $\begingroup$ No, that is not true. This has come up in MO before. I will include a link in the next comment. $\endgroup$ Commented Feb 21, 2022 at 23:57
  • $\begingroup$ @JasonStarr that will be really helpful.. $\endgroup$
    – piper1967
    Commented Feb 22, 2022 at 0:06
  • $\begingroup$ Actually I was thinking of another example where all fibers are diffeomorphic, but the morphism is not a fiber bundle. However, in the example that I know, the fibers have nontrivial integral cohomology (but the Kuenneth theorem does not hold, not even locally on the base). $\endgroup$ Commented Feb 22, 2022 at 1:27
  • 1
    $\begingroup$ It's true under the assumption that each fibre has trivial homotopy, in which case $f$ is a fibration and a homotopy equivalence (this is due to G. Meigniez). This condition is sufficient but not necessary for $f$ to be a homology equivalence. $\endgroup$
    – Tyrone
    Commented Feb 22, 2022 at 16:05

1 Answer 1

1
$\begingroup$

This seem to be true, and here is perhaps an over-sophisticated proof. By Poincare duality, homology identifies with compactly supported hypercohomology with coefficients in the Verdier dualizing sheaf $\omega_X$ (resp. $\omega_Y$), which is in the smooth case a shift of the orientation local system. The map in homology corresponds via this identification to the integration map $$H^*_c(X,\omega_X)\to H^*_c(Y,\omega_Y)$$, so we want this integration map to be an isomorphism.

This map is obtained from the counit $f_!f^!\omega_Y \to \omega_Y$ of the adjunction between derived categories $f_!\colon D(X)\leftrightarrows D(Y):f^!$ by taking sections with proper support, so its enough to show that this counit is an isomorphism in $D(Y)$. Here, $D(-)$ is the derived category of sheaves of abelian groups. Since $f$ is a submersion, we have a natural isomorphsm $f^!A\simeq f^*A\otimes \omega_f$ which shows that $f^!$ satisfies base-change along arbitrary smooth map $i\colon Z\to Y$. Namely, for the projections $i'\colon X\times_Y Z\to Z$ and $f'\colon X\times_Y Z \to X$, we have $i^*f^! \simeq (f')^!(i')^*$.
In particular, taking $Z= \{y\}$ for $y\in Y$ and using also proper base-change, we see that the stalk of the map $f_!f^!\omega_Y\to \omega_Y$ at $y\in Y$ is the (derived) tensor product of the integration map $\int_{X_y}\colon C^*_c(X_y;\omega_{X_y})\to \mathbb{Z}$ for the fiber $X_y$ of $f$ over $Y$ with the identity map of the stalk $(\omega_Y)_y$. Finally, the assumption that all the fibers of $f$ are acyclic implies that the integration map $C^*_c(X_y;\omega_{X_y})\to \mathbb{Z}$ is an isomorphism in $D(Ab)$, by Poincare duality for the acyclic smooth manifold $X_y$.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .