Suppose $\mathcal{X}$ is a smooth quasi-projective variety over $\mathbb{C}$ (I apologize if these hypotheses have little to do with the question at hand). Let $\mathcal{F}$ be a coherent sheaf on $\mathcal{X}$. Is $H^i(X,\mathcal{F})$ finitely generated over $\Gamma(O_X)$ if $\mathcal{F}$ is coherent ? This statement is simple enough that I probably would have heard it if it were true.

If it makes a difference, the statement that I really would like to understand is whether $\Gamma(O_X) \to Ext^i(\mathcal{G},\mathcal{G})$ is module finite for any coherent sheaf.

Most of the schemes that I am "friendly with" are either projective or affine. The statements are correct in those particular cases, so probably I just need to learn more examples... Thanks!

  • $\begingroup$ Probably could have posted this on stackexchange since it's probably not a "research question". My apologies if it's not appropriate. $\endgroup$ Feb 25, 2012 at 10:06
  • 1
    $\begingroup$ Is the following true: If $X$ is a quasi-projective variety and all cohomology groups of coherent sheaves on $X$ are finitely generated, and $\Gamma(X)=k$, is then $X$ projective? $\endgroup$ Feb 25, 2012 at 13:18
  • 1
    $\begingroup$ @Martin: the case $\dim X=1$ is well-known, otherwise choose an embedding $X\subset\mathbb P^N$, suppose there is $x\in\bar X\setminus X$, choose a hyperplane $H$ containing $x$ such that $x\in\overline{H\cap X}$. By induction on the dimension, $H\cap X$ is projective, contradiction. $\endgroup$
    – user2035
    Feb 25, 2012 at 15:00
  • 2
    $\begingroup$ Daniel: see tea.mathoverflow.net/discussion/1300/1/… I don't know how others will feel about the appropriateness of your question, but I personally am glad you asked it here. $\endgroup$ Feb 25, 2012 at 15:46

2 Answers 2


This is false even for $\mathrm H^0$: take $X$ to be $\mathbb A^2 \smallsetminus \{0\}$, and as $F$ the structure sheaf of $L \smallsetminus \{0\}$, where $L$ is a line through $0$.


Counterexample: For $X=\mathbb A^2\setminus\{0\}$, one has $\Gamma(X,O_X)=\mathbb C[u,v]$, but $H^1(X,O_X)\cong\mathbb C[u^{\pm1},v^{\pm1}]/(\mathbb C[u,v^{\pm1}]+\mathbb C[u^{\pm1},v])$ is not finitely generated over $\mathbb C[u,v]$.


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.