Let $X$ be a $k$-scheme and $\mathcal{L},\mathcal{N}$ two invertible sheaves on $X$. Assume that there exist a morphism $\mathcal{L} \to \mathcal{N}$ of $\mathcal{O}_X$-modules. Assume that $X$ has for every coherent sheaf $Coh(X)$ finite dimensional cohomology groups. (for example this holds when $X$ is projective or proper)

(*) My question is why does this already imply that $deg(\mathcal{L}) \le deg(\mathcal{N})$.

Remark: The degree of an invertible sheaf $\mathcal{L}$ is defined by $$deg(\mathcal{L}):= \chi(\mathcal{L}) - \chi(\mathcal{O}_X)$$

with Euler-characteristics

$$\chi(\mathcal{L}) := \sum _{i \ge 0} (-1)^i dim_k H^i(X, \mathcal{L})$$

My ideas:

The Euler -cahracteristic is additive in the sense that if

$$0 \to \mathcal{L'} \to \mathcal{L} \to \mathcal{L''} \to 0$$

is a short exact sequence then $\chi(\mathcal{L})= \chi(\mathcal{L'})+\chi(\mathcal{L''})$.

therefore it suffice to show that $\mathcal{L} \to \mathcal{N}$ is injective. is it always true? Since it is a local problem and invertible sheaves are locally trivial we can assume that

-$X= Spec(R)$


So the problem boils down to show that any $R$-module morphism $\phi: R \to R$ is injective. Everything is told on the image of $1$. But if $R$ isn't an integral domain $\phi$ might be not injective so my considerations fail.

Does anybody see an argument why (*) nevertheless must hold? Or do we need an extra assumption for $X$ like beeing integral or smooth (to garantee that local sections are integral). In the original script $X$ was a ruled surface (as defined in Hartshorne).

Or is there another way to verify (*)?

  • $\begingroup$ For arbitrary $k$-schemes (for example when $X$ is affine), your definition of degree is meaningless, since $\chi(L),\chi(\mathcal{O}_X)$ are infinite dimensional. $\endgroup$ – Mohan Apr 11 '19 at 0:11
  • $\begingroup$ @Mohan: yes that's a good point. I assume that all cohomology groups are finite dimensional. This holds for example when $X$ projective or proper. $\endgroup$ – KarlPeter Apr 11 '19 at 0:32
  • $\begingroup$ Try $X$ to be $2E\subset X'\to\mathbb{P}^2$, the blowing up of a point and $E$ the exceptional divisor. Then we have $L=O_{2E}(-E)\to O_{E}(-E)\to O_{2E}=L'$ a non-zero map, and you can check that the degree inequality does not hold. $\endgroup$ – Mohan Apr 11 '19 at 0:43

you have maps$\mathcal{L}\to \phi(\mathcal{L})\to \mathcal{L'}$ the first one is surjective the last one is injective and now you can use short exact sequence($\phi(\mathcal{L})$is an invertible sheaf)

  • $\begingroup$ in this case we obtain $deg(\phi(\mathcal{L}) ) \le deg(\mathcal{L})$ and $deg(\phi(\mathcal{L}) ) \le deg(\mathcal{L'})$ (with obvious exact seqences). But this doesen't imply $deg(\mathcal{L} ) \le deg(\mathcal{L'})$. Or do I oversee something? $\endgroup$ – KarlPeter Apr 10 '19 at 21:34

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.