3
$\begingroup$

Let $E\to M$ be a holomorphic vector bundle ($M$ is compact) and assume $s\colon M\to E$ is a non-trivial section transverse to the zero section.

Is it be possible that $s^{-1}(0)\neq \emptyset$, yet $c_{top}(E)=0$?

I exect the answer to be NO. If $M$ is Kahler, integrating the Kahler form on $s^{-1}(0)$ shows that it is homologically non-trivial. But how do you prove it for general complex manifold or without using the Kahler form.

$\endgroup$
  • 4
    $\begingroup$ The 'diagonal section of the trivial line bundle on $\mathbb C$ gives an example, no? $\endgroup$ – Mariano Suárez-Álvarez Dec 18 '13 at 2:01
  • $\begingroup$ I should have said M is compact. $\endgroup$ – Mohammad F. Tehrani Dec 18 '13 at 4:13
  • 2
    $\begingroup$ I suggest looking at the Hironaka twist, an example of a non-projective smooth proper threefold (Hartshorne B 3.4.1). It has two smooth disjoint curves $M_1$ and $M_2$ with the property that $M_1+M_2$ is numerically trivial. It could be possible to construct a rank 2 vector bundle $E$ with a section $s$ whose zero scheme is $M_1+M_2$. $\endgroup$ – Piotr Achinger Dec 18 '13 at 4:43
  • 1
    $\begingroup$ @user20497: Aren't the Chern classes $c_k$ defined in $H^{2k}(X,\mathbb{Z})$? Maybe the class of $D$ is a torsion class in $H^2(X,\mathbb{Z})$. $\endgroup$ – Jason Starr Dec 19 '13 at 13:48
  • 1
    $\begingroup$ @Jason You're right. But here is another example, hope it works: it's a primary Hopf surface $X$, Harvey and Lawson call it of class $0$ in their 1983 Inventiones paper. There exists a smooth elliptic curve $T\subset X$ such that $[T]$ (the current induced by integrating on $T$) is $d$-exact. The Hopf surface is $S^3\times S^1$ from the differential point of view, so there is no torsion in $H^*(X,{\mathbb Z})$. The line bundle $E={\mathcal O}_X(T)$ associated to $T$ satisfies $c_1(E)=0\in H^2(X,{\mathbb Z})$. Probably the non-Kahler elliptic surfaces have the same property. $\endgroup$ – user20497 Dec 19 '13 at 16:24
3
$\begingroup$

Take a Hopf surface $H$, projected to ${\Bbb C}P^1$ with fibers elliptic curves, and let $L=\pi^* O(1)$ be a pullback of $O(1)$ from ${\Bbb C}P^1$ to $H$. Since $H=S^3\times S^1$, all line bundles are topologically trivial, but $L$ has sections (pulled back from ${\Bbb C} P^1$) vanishing somewhere on $H$.

$\endgroup$

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.