4
$\begingroup$

I'am try to work with Chern class of the coherent sheaves, in this sense. If I have a vector bundle $E$ of rank $r$ and $L$ a line bundle we have the Chern class property

$$c_{r}(E\otimes L) = \sum_{i = 0}^{r}c_{i}(E)c_{1}(L)^{r-i}.$$

I need a similar result for coherent sheaf.

If I have a coherent sheaf $\mathcal{F}$ over a smooth projective variety X we consider a locally free resolution

$$0 \longrightarrow E_{n} \longrightarrow \cdots \longrightarrow E_{0} \longrightarrow \mathcal{F} \longrightarrow 0.$$

We can define the total Chern class of $\mathcal{F}$ by

$$c(\mathcal{F}) = \displaystyle\Pi_{i = 0}^{n} c(E_{i})^{(-1)^{i}}.$$

My question: Is true that $c_{n}(\mathcal{F}\otimes L) = \sum_{i = 0}^{n}c_{i}(\mathcal{F})c_{1}(L)^{n-i},$ for the coherent sheaf $\mathcal{F}$? If this does not exist, we can have the similar relationship?

Thank you !!

$\endgroup$
3
$\begingroup$

No, this is false. Take the simple example $\mathcal{F}=\mathcal{O}_Y$, where $Y$ is a hypersurface in $X$. Using the resolution $0\rightarrow L(-Y)\rightarrow L\rightarrow L\otimes \mathcal{O}_Y\rightarrow 0$, one gets $c_1(L\otimes \mathcal{O}_{Y})=c_1(L)-c_1(L(-Y))= [Y]$, so it is independent of $L$, while your formula gives $c_1(L\otimes \mathcal{O}_{Y})=c_1(L)+ c_1(\mathcal{O}_{Y})=c_1(L)+[Y]$.

To get a general formula it is much easier to use the Chern character: we have $\operatorname{ch}(\mathcal{F}\otimes L)=\operatorname{ch}(\mathcal{F})\operatorname{ch}(L) $. Indeed the formula is classical if $\mathcal{F}$ is a vector bundle, and it extends readily to coherent sheaves using a finite locally free resolution.

$\endgroup$
  • $\begingroup$ abx, thank you! It's great, very well. Thus if I need to calculate the Chern class of a coherent sheaf the most coerent is to use a locally free resolution. $\endgroup$ – Student85 Jun 19 '18 at 23:20

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.