Let $F$ be a finite field (for example I could take $\mathbb{Z}_2$) and $f:X\longrightarrow Y$ a continuous map between compact, orientable and connected manifolds of dimension $n$. Suppose I have an isomorphism $H^n(f,F):H^n(Y,F)\longrightarrow H^n(X,F)$. How do I prove that $H_n(f,F):H_n(X,F)\longrightarrow H_n(Y,F)$ or $H_n(f,\mathbb{Z}):H_n(X,\mathbb{Z})\longrightarrow H_n(Y,\mathbb{Z})$ is an isomorphism?

So far I have tried to UCT for cohomology the following way:

$H^n(f,F)$ isomorphism implies I have and isomorphism between $\text{Hom}(H_n(Y,\mathbb{Z}),F)$ and $\text{Hom}(H_n(X,\mathbb{Z}),F)$. I can't follow from here.


Let $M$ be an orientable (connected, closed) $n$-manifold. Then classically $H_n(M;F) \cong F$. But moreover, $H_{n-1}(M;\mathbb{Z})$ is torsion-free. This is Corollary 3.28 in Hatcher's book. The proof basically goes like this: if $H_{n-1}(M;\mathbb{Z})$ weren't torsion free, by the UCT then $H_n(M; \mathbb{F}_p)$ would be bigger than just $\mathbb{F}_p = H_n(M;\mathbb{Z}) \otimes \mathbb{F}_p$, which contradicts orientability.

It follows then that $H^n(M;F) \cong \hom_\mathbb{Z}(H_n(M;\mathbb{Z}), F)$, and moreover the isomorphism is functorial, i.e. if $f : X \to Y$ as in your question, then $f^* : H^n(Y;F) \to H^n(X;F)$ is the dual map of $f_* : H_n(X;\mathbb{Z}) \to H_n(Y;\mathbb{Z})$. This is because the short exact sequence in the UCT is functorial, and both $\operatorname{Ext}$-terms vanish.

Now $H_n(X;\mathbb{Z}) \cong H_n(Y;\mathbb{Z}) \cong \mathbb{Z}$, so the question boils down to this:

Let $F$ be a finite field. Given some group morphism $\phi : \mathbb{Z} \to \mathbb{Z}$, if $\phi^* : \hom_\mathbb{Z}(\mathbb{Z}, F) \to \hom_\mathbb{Z}(\mathbb{Z}, F)$ is an isomorphism, is $\phi$, respectively $\phi \otimes F$, an isomorphism?

(To apply this to your situation, let $\phi = f_* : H_n(X;\mathbb{Z}) \to H_n(Y;\mathbb{Z})$.)

The answer to the first question is actually "no". Indeed, suppose $\phi(k) = ak$, where $a$ is some number prime to $p$ (where $p$ is the characteristic of $p$). Then $\phi^* : F \to F$ is multiplication by $a$, which is an isomorphism! But $\phi$ is clearly not an isomorphism. So just choose a map of degree prime to the characteristic of your field to get a counterexample.

However $\phi \otimes F : \mathbb{Z} \otimes F \to \mathbb{Z} \otimes F$ will be an isomorphism, because multiplication by $a$ is an isomorphism.

tl;dr If $H^n(f,F)$ is an isomorphism, then $H_n(f,F)$ is an isomorphism, but $H_n(f,\mathbb{Z})$ may not be.

  • $\begingroup$ If you happen to know it for all finite fields then you can say it is an isomorphism $\endgroup$ – Spotty May 16 '16 at 9:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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