# Restriction of Ext sheaves on closed subschemes

Let $$f:X\rightarrow C$$ be a morphism, where $$C$$ is a smooth curve. For $$t\in C$$ let $$i_t:X_t = f^{-1}(t)\rightarrow X$$ be the inclusion of the fiber of $$f$$ over $$t$$, and let $$\mathcal{F}$$ a coherent sheaf on $$X$$ that is flat over $$C$$.

Does there exist an isomorphism $$i_t^{*}\mathcal{E}xt^1(\mathcal{F},\mathcal{O}_X)\cong \mathcal{E}xt^1(i_t^{*}\mathcal{F},\mathcal{O}_{X_t})$$ ?

• Take $f =Id_C$ and $\mathcal{F} = \mathcal{O}_t$. The RHS vanishes whereas $\mathcal{E}xt^1(\mathcal{O}_t, \mathcal{O}_X) \simeq \mathcal{O}_t$. So there's no isomorphsm in general. – HYL Jan 13 at 5:16
• We are assuming that $\mathcal{F}$ is a coherent sheaf on $X$ that is flat over $C$. – gxg Jan 13 at 11:27

## 1 Answer

Yes. Here's some extra words necessary for MO to allow this as an answer.

• I may have been a bit too hasty in my response, but I believe the statement in your setup should be true, see e.g. mathoverflow.net/questions/37889/… . Usually some flatness assumption of i_t would be necessary (which does not hold here), or as in loc. cit. flatness of the sheaf (which is one of your assumptions), or some regularity of the embedding which'd ensure finiteness of tor dimension (you haven't specified if this is true). – Frank Jan 12 at 16:45
• It is seems that in order to use Lemma 15.87.2 the flatness of the morphism is fundamental. In my case the inclusion $i_t:X_t\rightarrow X$ is not flat. – gxg Jan 12 at 22:02