enter image description here

I have some difficulty understanding the above proof.

Here $C(\mu)$ denotes the category of semi-stable vector bundles (locally free sheaf) of slope $\mu$. Here the sheaves are considered over a smooth connected algebraic curve $X$.

A locally free sheaf $F$ is called semi-stable if for every subsheaf $G$, the slope of $G$, $\mu(G)$, is less than equal to slope of $F$, $\mu(F)$. The slope of a locally free sheaf is defined as $deg(F)/rank(F)$.

I've two questions about the above proof

  1. It says that the $ker(f)$ and $coker(f)$ are locally-free sheaves. Well, the $ker(f)$ is a subsheaf of $E$, which is a locally-free sheaf, so $ker(f)$ is also a locally free sheaf ( because these are sheaves on a smooth curve and therefore stalks are pid, and a submodule of a free module over a pid is again a free module). But I've absolutely no idea how it says $coker(f)$ is also a locally-free sheaf. Why is this true?

  2. In the proof of $ker(f)$ is semistable, it says if it's not semistable it would have a sub-bundle of slope of slope greater than $\mu$. So, somehow there are assuming slope of $ker(f)$ is also $\mu$. So, I guess since slope of $E$ and $F$ is $\mu$ and they are semistable then it is forcing the slope of both $ker(f)$ and $coker(f)$ to be $\mu$. Is this correct?


1) If $Coker(f)$ is not locally free, it has a summand $T$ which is a torsion sheaf. The kernel $G$ of the composition $F \to Coker(f) \to T$ is a locally free sheaf, whose slope is less than $\mu$. The image of $f$ is contained in $G$, so, we obtain a morphism $E \to G$ from a semistable sheaf to a sheaf of smaller slope, which is impossible.

2) Consider the image $G = Im(f)$. We have an epimorphism $E \to G$, hence all quotients of $G$ have slope greater or equal than $\mu$. On the other hand, we have a monomorphism $G \to F$, hence all subsheaves in $G$ have slope less or equal than $\mu$. It follows that in the Harder-Narasimhan filtration of $G$ all factors have slope $\mu$, hence $G$ is semistable of slope $\mu$. Therefore $Ker(f) = Ker(E \to G)$ and $Coker(f) = Coker(G \to F)$ also have slope $\mu$.

  • $\begingroup$ Thank you very much for the answer! I don't understand why the slope of $G$ should be less than $\mu$. Why can't it be equal to $\mu$? $\endgroup$ – grok Feb 10 '18 at 17:22
  • $\begingroup$ @grok: Because its rank equals the rank of $F$, and its degree is equal to the degree of $F$ minus the length of $T$, i.e., less than the degree of $F$. $\endgroup$ – Sasha Feb 11 '18 at 8:15
  • $\begingroup$ Thanks for the reply! What exactly do you mean by the term length of $T$? $\endgroup$ – grok Feb 12 '18 at 4:31
  • 1
    $\begingroup$ @grok: A torsion sheaf on a curve is an iterated extension of structure sheaves of points. Its length is the number of factors in such iterated extension. $\endgroup$ – Sasha Feb 12 '18 at 6:53

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.