2
$\begingroup$

Let $\mathcal T = D^b(\mathbb P^1)$, the bounded derived category of $\mathbb P^1$ over a field $k$. It is well known that $\mathcal T$ admits a strong and full exceptional sequence $\{\mathcal O, \mathcal O(1) \}$, such that $\mathrm{Hom}(\mathcal O, \mathcal O(1))$ is 2-dimensional over $k$. Now, let $\mathcal T'$ be another triangulated category, and let $F, G : \mathcal T \to \mathcal T'$ be exact functors. Write $i : \{ \mathcal O, \mathcal O(1) \} \hookrightarrow \mathcal T$ for the inclusion functor. At least under some reasonable assumptions, I can prove that a natural tranformation $\varphi : F \circ i \to G \circ i$ can be extended to a natural transformation $F \to G$ (this is true when $F$ and $G$ are enhanceable to dg-functors, and the proof I know involves those enhancements); what if I try to extend a natural transformation defined on the full subcategory $\{ \mathcal O, \mathcal O(1), C(x_0) \}$, where $C(x_0)$ is the cone of the morphism $x_0 : \mathcal O \to \mathcal O(1)$? Formally: let $j : \{ \mathcal O, \mathcal O(1), C(x_0) \} \hookrightarrow \mathcal T$ be the inclusion, and let $\varphi : F \circ j \to G \circ j$ be a natural transformation. Does an extension $\widetilde{\varphi} : F \to G$ of $\varphi$ (such that $\widetilde{\varphi} \circ j = \varphi$) exist? How can I construct it? If necessary, I can also assume that $F$ and $G$ come from dg-functors between suitable enhancements.

$\endgroup$
5
  • $\begingroup$ What about restricting and then extending? $\endgroup$ Commented Jan 28, 2015 at 21:08
  • $\begingroup$ I don't think it should work: I want $\widetilde{\varphi}$ to be equal to $\varphi$ on the specified full subcategory ($\widetilde{\varphi} \circ j = \varphi$). If I restrict $\varphi$ to $\{\mathcal O, \mathcal O(1) \}$ and then extend, I don't know if the obtained extension equals $\varphi$ on $C(x_0)$. $\endgroup$ Commented Jan 29, 2015 at 10:01
  • $\begingroup$ But you should have a way of controlling that, right? The third object is not far from the first two ones, and you say you can extend the nat. trans. from those two ones. $\endgroup$ Commented Jan 29, 2015 at 10:40
  • $\begingroup$ Actually, "controlling" that is a subtle point. The method I employ to extend the transformation from the two objects $\mathcal O, \mathcal O(1)$ involves dg-enhancements and functorial cones... $\endgroup$ Commented Jan 29, 2015 at 12:50
  • $\begingroup$ That method looks like a pro rather than a con, since the third object is a cone of the first two ones. Can you maybe explain it in your question? $\endgroup$ Commented Jan 30, 2015 at 11:53

0

You must log in to answer this question.