3
$\begingroup$

I am trying to understand the proof of Lemma 3.0.15 of this paper (Ben-Zvi, Nadler, Preygel - Integral transforms for coherent sheaves).

The context is of two triangulated categories $C,D$ with t-structures and a functor $F\colon C \to D$ which is left and right t-exact (up to shifts).

Recall that $F$ is left exact up to a shift if $F(C^{\geq 0}) \subset D^{\geq l}$. Similarly for right exact (call $r$ the error term).

They claim that given an object $c \in C$ there exist natural maps

$$ F(\tau_{>N} c ) \to \tau_{>N-r}F(c) $$ and $$ \tau_{\leq N+l}F(c) \to F(\tau_{\leq N}c) $$ whose cones have good properties.

How are these maps defined?

They moreover assume C and D to be enhanced (ie they work with stable $\infty$-categories), complete and cocomplete; and they assume F to be conservative and (homotopy) limit and colimit preserving (although this might actually not be important for my question on its own).

$\endgroup$
3
  • 2
    $\begingroup$ Please include the bibliographic data in the post. Moreover it is better not to link to the pdf but rather to /abs/ $\endgroup$
    – user9072
    Feb 15, 2016 at 21:34
  • $\begingroup$ @quid done. also, why are direct links to pdfs bad? $\endgroup$ Feb 15, 2016 at 22:23
  • 4
    $\begingroup$ Thanks. For example, on a mobile or otherwise limited connection it may be undesirable to download it, and some mobile devices it may be impossible/inconvenient to open it; especially when it is just to find out what article it is. Some might also prefer to retrieve a format other than pdf, for one reason or another. $\endgroup$
    – user9072
    Feb 15, 2016 at 23:25

1 Answer 1

1
$\begingroup$

Are you sure about your definition of right and left t-exact up to a shift? I think the names are swapped. Let me answer assuming left t-exact up to a shift means $F(C_{\le 0}) \subset D_{\le l}$. (Wait! You write the t-structure with superscripts instead of subscripts: is that related to me feeling the definitions are swapped? Maybe by $C^{\le n}$ you mean what I would call $C_{\ge -n}$? --in which case we are in agreement.)

Here's how the maps are defined: apply $\tau_{\le N+l} \circ F$ to the canonical morphism $c \to \tau_{\le N} c$ to get a morphism $\tau_{\le N+l} F(c) \to \tau_{\le N+l}F(\tau_{\le N} c)$. Since $F$ satisfies $F(C_{\le 0}) \subset D_{\le l}$ and commutes with shifts (you said it is assumed to preserve finite limits and colimits), $F(C_{\le N}) \subset D_{\le N+l}$. So we have $F(\tau_{\le N} c)\in D_{\le N+l}$, and thus $\tau_{\le N+l}F(\tau_{\le N} c) \cong F(\tau_{\le N} c)$. The other map is defined similarly.

I took a look at the paper to see what is claimed about the fiber of that map, and I am not sure I get it. They say the fiber is $F(\tau_{>N}c)$ and this lies in $D_{\ge N-r}$, so that the map becomes an isomorphism when you apply $\tau_{\le N-r}$. It seems to me that the fiber is not $F(\tau_{>N}c)$, but that the conclusion about $\tau_{\le N-r}$ is true.

Let's try to show that. Start from the exact triangle $\tau_{>N}c \to c \to \tau_{\le N}c$. Applying $F$ results in an exact triangle (again because we assume $F$ commutes with finite limits). So the fiber of $F(c) \to F(\tau_{\le N} c)$ is $F(\tau_{>N}c)$, but then this means it won't also be the fiber of the map $\tau_{\le N+l} F(c) \to F(\tau_{\ge N}c)$ unless $F(c) \in D_{\le N+l}$ (when the maps are basically the same).

On the other hand, the claim about $\tau_{\le N-r}$ seems correct: $\tau_{\le N-r}$ will send an exact triangle in $D$ to a cofibre sequence in $D_{\le N-r}$ (because it is left adjoint to the inclusion $D_{\le N-r} \hookrightarrow D$). In the resulting cofibre sequence, $\tau_{\le N-r} F(\tau_{>N}c) \to \tau_{\le N-r} F(c) \to \tau_{\le N-r} F(\tau_{\le n} c)$, the first object is 0 because $F$ is right t-exact up to a shift of $-r$, which shows that the second map is an equivalence.

$\endgroup$
1
  • $\begingroup$ thanks! I'm going to go over your answer. It seems to me this is related (if not implied by) Proposition 8.1.15 in the Hotta-Takeuchi-Tanisaki book. Regarding notation: I think BZFP use homological indexing in subscripts, so to get the standard (but nonstandard for topologists) cohomological indexing one should put a minus everywhere and reverse inequalities. In any case, I may have gotten my definition the wrong way round... $\endgroup$ Feb 16, 2016 at 19:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.