1
$\begingroup$

Let $R$ be a commutative ring and $X$ a topological space. Define a sheafy cohomology theory (see here) to be a collection of functors $\mathrm{H}^q:\mathrm{Sh}(X;R\mathrm{Mod})\to R\mathrm{Mod}$ such that the following conditions are satisfied:

  • If $q<0$, then $\mathrm{H}^q(\mathscr{F})=0$. If $q=0$, then there is an isomorphism $\alpha_\mathscr{F}:\mathrm{H}^0(\mathscr{F})\to\mathscr{F}(X)$.

  • If $\mathscr{F}$ is flasque or fine, then $\mathrm{H}^q(\mathscr{F})=0$ for $q>0$.

  • If $0\to\mathscr{F}\to\mathscr{G}\to\mathscr{H}\to0$ is a short exact sequence of sheaves, then there is a long exact sequence: $$\cdots\to\mathrm{H}^q(\mathscr{F})\to\mathrm{H}^q(\mathscr{G})\to\mathrm{H}^q(\mathscr{H})\to\mathrm{H}^{q+1}(\mathscr{F})\to\cdots$$

My question is as follows: is ordinary sheaf cohomology $\mathrm{H}^q(X,\mathscr{F})$ the only sheafy cohomology theory (removing the flasque sheaf requirement)? If not, are there any other examples of a sheafy cohomology theory?

$\endgroup$
5
  • 6
    $\begingroup$ Yes, use that every sheaf embeds in a flasque sheaf and induct on q. $\endgroup$ Commented Feb 20, 2015 at 1:21
  • 3
    $\begingroup$ The uniqueness theorem can be found for example in Lang's "Algebra" XX.7 $\endgroup$ Commented Feb 20, 2015 at 1:31
  • $\begingroup$ @Daniel I used a simple diagram chasing argument with the short five lemma and got the required isomorphism. Thanks! $\endgroup$
    – user62675
    Commented Feb 20, 2015 at 3:40
  • $\begingroup$ This was a silly question on my part; could it be deleted/closed? $\endgroup$
    – user62675
    Commented Feb 20, 2015 at 14:42
  • $\begingroup$ You should ask Tyler. It would be unfair for me to delete his contribution without his consent. $\endgroup$
    – S. Carnahan
    Commented Feb 20, 2015 at 14:57

1 Answer 1

5
$\begingroup$

With respect to the updated question: yes, there are other cohomology theories. For example, if $H$ is ordinary sheaf cohomology, then we can define a new sheaf cohomology theory $K$ by $K^q({\cal F}) = H^q({\cal F}) \times H^{q-1}({\cal F})$. (This is a special instance of a hypercohomology construction which is genuinely important in some areas.)

With respect to the original question: A. Grothendieck, "Sur quelques points d’algèbre homologique" (the Tohoku paper); Cartan-Eilenberg, "Homological algebra".

$\endgroup$
4
  • $\begingroup$ That's a pretty neat example. Thanks! $\endgroup$
    – user62675
    Commented Feb 20, 2015 at 3:11
  • $\begingroup$ Me and Elden had a discussion on facebook - isn't $\mathrm{H}^q(\mathcal{F})=\prod^q_{i=n}\mathrm{H}^i(X,\mathcal{F})$ also a sheafy cohomology theory in the revised definition? $\endgroup$
    – user62675
    Commented Feb 20, 2015 at 3:31
  • $\begingroup$ @SanathDevalapurkar I'm a little bit confused by your suggestion, because it seems that such a definition would cut off some of the long exact sequence if $n > 0$, and if $n < 0$ it would mean the same as if $n=0$ (where it is true). $\endgroup$ Commented Feb 20, 2015 at 14:14
  • $\begingroup$ Yeah, I see that my suggestion wasn't right. Thanks! $\endgroup$
    – user62675
    Commented Feb 20, 2015 at 15:40

You must log in to answer this question.