Let $f\colon X' \to X$ be an étale morphism of degree $>1$ between two complex projective manifolds. Suppose $X'$ and $X$ are diffeomorphic to each other and $f$ induces an isomorphism of $\mathbb{Q}$-Hodge structures of $X'$ and $X$. Does $X$ admit a positive degree self-covering, i.e., an étale cover $\phi\colon X\to X$ such that $\operatorname{deg} \phi>1$ (just like the abelian varieties)?
1 Answer
Here is a counterexample. Let $E$ be an elliptic curve. It helps if we choose it in such a way that it does not have complex multiplication. Let $L$ be a line bundle on $E$, of degree $0$, corresponding to a divisor class $D$ of infinite order. Let $\pi:X=P(L\oplus 1)\to E$ be the projectivized bundle, a fiber bundle whose fibers are projective lines.
Suppose for contradiction that there is an etale cover $\phi:X\to X$ of degree $n>1$. The map $\phi$ must take each fiber $\pi^{-1}(e)$ of the bundle $\pi$ into some fiber $\pi^{-1}(e')$, since a holomorphic map from a projective line to $E$ must be constant. The resulting map $e\mapsto e'$ must be an etale cover $\psi:E\to E$ of degree $n$, with $\phi$ mapping the fiber $\pi^{-1}(e)$ isomorphically to $\pi^{-1}(\psi(e))$. Necessarily $\psi$ is given by $e\mapsto e_0+ me$ for some integer $m$, and $n=m^2$.
So $\phi$ gives an isomorphism of bundles between $P(\psi^\ast L\oplus 1)=\psi^\ast P(L\oplus 1)\to E$ and $P(L\oplus 1)\to E$. This implies that the divisor $D$ is equal to $\psi^\ast D= mD$, a contradiction.
On the other hand, we have plenty of etale covers $X'\to X$ with $X'$ diffeomorphic to $X$. In fact, the degree zero line bundle $L$ is trivial topologically, so that $X$ is diffeomorphic to $E\times P^1=S^1\times S^1\times S^2$; and the same is true of $X'$ if $X'$ is the fiber product $E'\times_E X$ for any etale cover $E'\to E$. And of course each of these maps $X'\to X$ gives an isomorphism in rational homology.
-
2$\begingroup$ A notation remark: for a standard algebraic geometer, what you call $P(L)$ is $\mathbb{P}(\mathscr{O}_E\oplus L)$ (or $\mathbb{P}(\mathscr{O}_E\oplus L^{-1})$, it doesn't matter). $\endgroup$– abxCommented Dec 31, 2022 at 7:46
-
$\begingroup$ Thank you for the counterexample. $\endgroup$ Commented Dec 31, 2022 at 9:58
-
$\begingroup$ @abx Yes, I did not write what I meant. I've changed it to $P(L\oplus 1)$ now. $\endgroup$ Commented Dec 31, 2022 at 12:01
\operatorname
automatically supplies operator spacing, so, for example,\operatorname{deg} \phi > 1
should be preferred to\text{deg}\ \phi > 1
. I have edited accordingly. $\endgroup$