2
$\begingroup$

Let $(E,h)\to X$ be a holomorphic Hermitian vector bundle over a compact Kähler manifold. Denote by $F_h$ the curvature of its Chern connection. Can we know a priori the sign of the quantity $$\int_X\operatorname{Tr}(F_h^2)\wedge \omega^{n-2}$$
or if $X$ is a surface the sign of $$\int_X\operatorname{Tr}(F_h^2)?$$

$\endgroup$

1 Answer 1

1
$\begingroup$

The short answer: No.

The longer answer: The $(2,2)$-form $\operatorname{Tr}(F_h^2)$ represents the second Chern class of $E$, or $c_2(E)$. In general, the integral of that class over the manifold doesn't have a sign independent of the bundle. What we do have in general is the pointwise identity $$ \left| \frac{i}{2\pi} F_h \right|^2 \omega^{n}/n! = (2 c_2(E,h) - c_1(E,h)^2) \wedge \omega^{n-2} / (n-2)! + \left| \operatorname{Tr}_{\omega} \frac{i}{2\pi} F_h \right|^2 \omega^n / n!, $$ where $c_k(E,h)$ is the representative of the $k$-th Chern class of $E$ defined by the curvature form of the metric. We don't need $\omega$ to be Kahler for this to hold, nor $X$ to be compact; this holds on general Hermitian manifolds. (This is known, and usually implicit somewhere in any differential-geometric proof of the Kobayashi-Lubke inequality. See here for one proof.)

To be able to say anything more I think we have to assume that $\omega$ is Kahler and $E = T_X$, in which case $\operatorname{Tr}_\omega \frac{i}{2\pi} F_h = c_1(T_X,\omega)$. The above then reduces to $$ \left| \frac{i}{2\pi} F_h \right|^2 \omega^{n}/n! = 2 c_2(T_X,\omega) \wedge \omega^{n-2} / (n-2)! + s(\omega)^2 \omega^n / n!, $$ where $s(\omega)$ is the scalar curvature of $\omega$. This still doesn't force $c_2(T_X,\omega) \wedge \omega^{n-2}$ to have a definite sign unless we assume the scalar curvature of $\omega$ is constantly zero, in which case $\omega$ will have to be a Ricci-flat Kahler-Einstein metric. In that case, yes, $c_2(T_X,\omega) \wedge \omega^{n-2} \geq 0$ with strict inequality unless the metric is flat.

$\endgroup$
1
  • $\begingroup$ Shouldn't it be $Tr(F^2_h)=c_1^2(E)-c_2(E)$? $\endgroup$
    – AG learner
    Commented Dec 29, 2021 at 5:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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