Let $X$ be a nice enough topological space, perhaps a complex algebraic variety with its analytic topology. I'm hoping someone could help me prove that the category $\text{Constr}(X)$ of constructible sheaves on $X$ is abelian.

I think one could use that subsheaves and quotients of local systems are local systems. However, conceptually what is confusing me is that given two constructible sheaves $\mathcal{F}, \mathcal{G}$, they might be constructible with respect to completely different Whitney stratifications of X, right?

So given a morphism $\mathcal{F} \to \mathcal{G}$ of constructible sheaves, can you always "refine" the two Whitney stratifications to get one such that the morphism restricts to a morphism of local systems?

Or should I be thinking along another lines?

  • 2
    $\begingroup$ There is always a stratification that refines any two given stratifications. Isn’t this almost obvious from the definition? $\endgroup$
    – Will Sawin
    Jan 23, 2020 at 1:29
  • $\begingroup$ @WillSawin I suppose, but is it clear the refinement will be Whitney? Or is it not true that a constructible sheaf must be locally constant over a Whitney stratification? I've been confused whether you need Whitney in the definition of constructible, because I've seen it both ways. $\endgroup$
    – Benighted
    Jan 23, 2020 at 2:04
  • $\begingroup$ I think every stratification of a reasonable space has a refinement that is a Whitney stratification. This would make the two definitions nicely equivalent, and answer this question as well. $\endgroup$
    – Will Sawin
    Jan 23, 2020 at 2:14
  • $\begingroup$ For instance Theorem 2.2 of Verdier's paper seems to be the desired statement eudml.org/doc/142424 $\endgroup$
    – Will Sawin
    Jan 23, 2020 at 2:17
  • $\begingroup$ @WillSawin Great that makes sense. Thanks a lot. If it matters, I would accept it if you wrote this into an answer. $\endgroup$
    – Benighted
    Jan 23, 2020 at 2:27

1 Answer 1


Verdier has proved (for multiple classes of spaces, including in particular complex varieties) that for any finite set of analytic subsets, there exists a Whitney stratification for which all these analytic subsets are unions of strata.

In particular, given two constructible sheaves, and a stratification on which they are lisse, there exists a Whitney stratification that refines both of them.

This suffices to check that kernels and cokernels of maps of constructible sheaves are constructible, as well as direct sums of constructible sheaves.


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.