I've been trying to understand the following proof in Kobayashi's book. Let me state some relevent definitions.

Let $X$ be a complex manifold and $\mathscr{O}_X$ its structure sheaf. A coherent sheaf $\mathscr{E}$ is torsion-free if the map $\mathscr{E}\to\mathscr{E}^{**}$ is injective. It is a torsion sheaf if the above map is a zero map. A torsion-free sheaf $\mathscr{E}$ of rank $r$ has a determinant line bundle $\det\mathscr{E}=(\wedge^r\mathscr{E})^{**}$.

Every monomorphism $\mathscr{F}\to\mathscr{F'}$ between torsion-free coherent sheaves of the same rank induces a sheaf monomorphism $\det\mathscr{F}\to\det\mathscr{F'}$.

The proof goes like this:

Outside their singular sets (where coherent sheaves are not locally free), the map $\mathscr{F}\to\mathscr{F'}$ and the induced map $\det\mathscr{F}\to\det\mathscr{F'}$ are isomorphisms. Hence, the $\ker(\det\mathscr{F}\to\det{F'})$ is a torsion sheaf. Since it is a subsheaf of a torsion-free sheaf, it must be zero.

My questions are in the following:

  1. Outside their singular sets, $\mathscr{F}$ and $\mathscr{F'}$ are locally free. Since they have the same rank, the map $\mathscr{F}\to\mathscr{F'}$ has a cokernel which has rank 0 and hence is a torsion sheaf. But why is it necessarily zero so that $\mathscr{F}\to\mathscr{F'}$ is an isomorphism?

  2. If $\mathscr{F}\to\mathscr{F'}$ is an isomorphism outside the singular sets, then $\ker(\det\mathscr{F}\to\det{F'})=0$ outside the singular set. Then, how do I conclude that it must be a torsion sheaf? I don't know its stalks inside the singular set. To show it is a torsion sheaf, I need to make sure every stalk is a torsion module.


  1. I don't see that this is true without other assumptions. Let $X=\Bbb C$ and $\mathscr{O}$ be the structure sheaf. Then $\mathscr{O}\stackrel{\cdot z}{\to}\mathscr{O}$ is a perfectly good morphism of free sheaves with cokernel the skyscraper sheaf supported at $0$ with sections $\Bbb C$.

Perhaps there should be a modification of the proof here - since the cokernel is torsion, it's supported on a closed submanifold of strictly smaller dimension, and we can throw that in with the singular set as a locus we can ignore - we will still get that the map on determinant bundles is an isomorphism on a dense open subset, which is really the thing that makes the proof work.

  1. Any sheaf supported on a submanifold of strictly smaller dimension is torsion: consider the action given by multiplying by a nonzero function which vanishes on this submanifold.
  • $\begingroup$ Hi, Thank you for your reply. Your second bullet point is very useful. So, your definition of a torsion sheaf is that in every stalk every element is a torsion element, correct? $\endgroup$ – YYF Oct 4 '18 at 21:42
  • $\begingroup$ Yes. Did something I write suggest something else? $\endgroup$ – KReiser Oct 4 '18 at 22:10
  • $\begingroup$ Then, maybe I am just tired today but I don't follow your second point. Let's say $\mathscr{F}$ is supported in a proper analytic subvariety $A$ of the base manifold $X$ and $s_x\in\mathscr{F}$ where $x\in A$. Then, take some holomorphic function $f$ that vanishes on $A$, then why is $(fs)_x=0$? It is true that $(fs)_y=0$ for every $y\notin A$. It is also true that $(fs)|_A=0$. But I must've missed something obvious. $\endgroup$ – YYF Oct 4 '18 at 22:23
  • $\begingroup$ One more thing in my last comment: $(fs)_|A=0$ does not make much sense, since this is a section in an abstract sheaf and is not a function. $\endgroup$ – YYF Oct 4 '18 at 22:40
  • $\begingroup$ But the problem is $U\cap A$ is not open. Therefore, even if I am tempted to say $(fs)|_{U\cap A}=f|_{U\cap A}s|_{U\cap A}=0$, I couldn't. $\endgroup$ – YYF Oct 5 '18 at 1:57

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.