# Tensor product of structure sheaves

Let $\iota_A:A\hookrightarrow X$ and $\iota_B:B\hookrightarrow X$ be subschemes of a smooth ambient variety $X$. Then the derived tensor product $$\mathcal O_A\stackrel{L}{\otimes}\mathcal O_B\in D^b(X)$$ is a pushforward from $A$ (of $L\iota_A^*\mathcal O_B$), and a pushforward from $B$ (of $L\iota_B^*\mathcal O_A$).

Is it also a pushforward from $A\cap B$ ?

I assume not, which is presumably the need for all the homotopy complications of derived algebraic geometry ? (By this I mean that if it is not such a pushforward, then one has to carry round some of the information of the embedding of $X$, which in derived algebraic geometry is a non-canonical local choice; finding a category in which these choices "glue" is then the nasty bit I presume, though I know $\le0$ about it.)

-
@Moosbrugger: If $A = B$ then the tensor product being a pushforward from $A$ is a pushforward from $A \cap B = A$ as well. – Sasha Nov 5 '11 at 19:49
As Sasha points out, that's not a counterexample. Does someone have a simple one ? – Richard Thomas Nov 7 '11 at 20:44
Hi Richard, I'm not sure if anyone still cares about this question. But, I think a lot of counterexamples should come from derived critical loci. For example, if you consider functions like $uv^2$ or take a generic family of $K_3$ surfaces, $t:X \to A^1$ and consider the function $t^2$. I say I think because I haven't worked out rigorous proofs, but this would be consistent with other things I know about these examples. This more of an idea for where to look for counterexamples for people smarter than me. – Daniel Pomerleano Aug 30 '13 at 13:00