MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Compact Kahler manifolds have the property that surjective maps induce injections on cohomology with coefficents in $\mathbb{Q}$ (That is, if $X,Y$ compact Kahler, then a surjective map $\phi: X \rightarrow Y$ induces injections $\phi^*: H^i(Y, \mathbb{Q}) \rightarrow H^i(X, \mathbb{Q})$ for all $i$, [Voisen, Hodge Theory I, p 177]).

Question: I'm wondering if I should think of this as a property of compact Kahler manifolds, or as an instance of something more general. For example, can the Kahler condition be replaced with a more general class of manifolds (not dropping the compactness hypothesis). Perhaps one that includes not just complex manifolds but maybe a few manifolds of odd (topological) dimension? I know that if we require that $\dim X = \dim Y$ then the fact above is true more generally just for compact oriented manifolds, for formal reasons.

share|cite|improve this question
up vote 15 down vote accepted

In order to have this property it is sufficient to require that $\phi^{-1}(y)$ is a non-zero cycle in $H_*(X,\mathbb Q)$, where $y$ is a generic point in $Y$. This holds indeed when $X$ and $Y$ are Kahler.

Let me give an example showing that this does not work when $X$ and $Y$ are just complex.

Example. Let $X$ be a Hopf surface $X=(\mathbb C^2\setminus 0)/\mathbb Z$ where $\mathbb Z$ is acting on $\mathbb C^2$ by multiplication by (say) $2$. Then there is a fibration $\phi\colon X\to \mathbb CP^1=Y$. This fibration comes from the standard $\mathbb C^*$-action on $\mathbb C^2$. Now $\phi^{-1}(y)$ is null-homlogous.

share|cite|improve this answer
Nice criteria! Is it easy to see that $\phi^{-1}(y)$ is a non-zero cycle in $H_*(X,\mathbb{Q})$ in the Kahler case? – LMN Oct 1 '12 at 2:57
Sure, the preimage of a point is a complex submanifold and the appropriate power of Kahler form of $X$ is a volume form on it. – Dmitri Oct 1 '12 at 3:00
Great! Can you give me a hint (or reference) of how you would prove this criteria implies that the maps on cohomology with $\mathbb{Q}$ coefficents are injective? – LMN Oct 1 '12 at 3:04
This follows form Poincare duality. So let $Z\in Y$ be a non-trivial cycle (i.e $[Z]\ne 0 \in H_*(Y,\mathbb Q)$). Let us prove that $\phi^{-1}(Z)$ is a non-trivial cycle in $X$. Chose $Z'$ in $Y$ Poincare dual to $Z$. And let us make so that $Z$ intersects $Z'$ transversally. Then $\phi^{-1}Z\cap \phi^{-1}\Z'$ is a collection of fibers. So the intersection is non-zero in homology of $X$, so both preimages in non-zero in homology of $X$. – Dmitri Oct 1 '12 at 3:13
Remark: your condition is necessary and sufficient, since it is equivalent to $\phi^\ast:H^n(Y)\hookrightarrow H^n(X)$, where $n=dim Y$. – Ian Agol Oct 1 '12 at 14:13

Suppose one has compact symplectic manifolds $(X,\omega), (Y,\sigma)$, and a map $f:X\to Y$ such that $f^\ast\sigma=\omega$ (if $f$ is also a diffeomorphism, then this map is a symplectomorphism, but I'm not sure the terminology in this case). Then $f^\ast: H^*(Y;\mathbb{Q})\to H^\ast(X;\mathbb{Q})$ will be injective. If $dim Y=2k$, then $\sigma^k\neq 0\in H^{2k}(Y)$ is a fundamental class, and $dim X=2n \geq 2k$, then $\omega^n\neq 0\in H^{2n}(X)$. So $f^\ast(\sigma^k)= \omega^k \neq 0 \in H^{2k}(X)$.

Now, consider $\alpha \in H^j(Y)$. By Poincare duality, there exists $\beta\in H^{2k-j}(Y)$ such that $\alpha\cup \beta = \sigma^k$. Then $f^\ast(\alpha\cup \beta)=\omega^k\neq 0$, so $f^\ast \alpha\neq 0$.

share|cite|improve this answer
In fact $Y$ is not required to be symplectic. It suffices that $f$ is surjective and $X$ has a closed $2$ form which is symplectic on a generic fiber (so that dimensions have equal parity). When $f$ is a submersion, it is called a symplectic fibration, but there are important examples with critical points, the so-called (symplectic) Lefschetz fibrations. – BS. Oct 1 '12 at 11:34

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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