4

Let $* \to \mathbb A^1$ be the embedding of the point defined by $t=0$, where $t$ is the coordinate for $\mathbb A^1$. Let $X$ be a smooth projective variety (say $\mathbb P^1$), and $i\colon X \to X \times \mathbb A^1$ be the embedding of the central fibre. Let ${\mathrm D}(-)$ denote the bounded derived category of coherent sheaves. My question is:

Assume $F\in{\mathrm D}(X \times \mathbb A^1)$ such that $F \xrightarrow{t} F$ is zero map (in ${\mathrm D}(X \times \mathbb A^1)$), where $t$ denotes the morphism which is acted by $t$ in each degree. Then, is it true that there exists a $E\in {\mathrm D}(*)$, such that $F= i_*E$?

In the non-derived setting (i.e. replace $\mathrm {D}$ by $\mathrm{Coh}$) this is true by definition. I would like to know if this is true also in the derived setting.


My attempt. Basically, what I want is to find a representative of $F$ such that $F \xrightarrow{t} F$ is homotopic to zero (then we can define such a $E$). However, I guess the following is not true in general (unless $Y$ is affine):

Let $f\colon F\to F$ be a chain map of bounded complex of locally free sheaves on $Y$. Assume $f$ is zero in the derived category. Then, is $f$ homotopic to zero?

(Edit: Indeed, for example, see this counter-example.) However, one may take advantage of the form $F \xrightarrow{t} F$ for $Y=X \times \mathbb A^1$, so I am not sure if my first question is true or false.

Thanks in advance!

Cyrist
  • 71
  • If you’re willing to replace the derived category by its DG enhancement then I believe this is indeed true. Basically, since $Y=X \times \mathbb{A}^1$ is affine over $X$, the derived category of the former is modules for the pushforward of $\mathcal{O}_Y$ in $D(X)$, ie for $\mathcal{O}_X[t]$. Modules over $\mathcal{O}_X[t]/(t)$ in $D(X)$ should then be the derived category of the central fiber – Exit path Sep 25 '24 at 21:48
  • @Exitpath Thank you for the comment. Yes, I agree with what you said, but I am not sure it is indeed an $\mathcal O_X[t]/(t)$-module. Is $F \xrightarrow{t} F$ being zero in the derived category already enough to conclude it is an $\mathcal O_X[t]/(t)$-module? Is there a reference on this? – Cyrist Sep 25 '24 at 23:46
  • @Exitpath What I thought before is, if $F \xrightarrow{t} F$ is homotopic to zero, then $F$ is an $\left[\mathcal O_X[t] \xrightarrow{t} \mathcal O_X[t]\right]$-module (the null-homotopy gives precisely the action of the first term) and hence equivalent to an $\mathcal O_X[t]/(t)$-module. This is why I will need it to be null-homotopic. I would happy to see this is detoured. – Cyrist Sep 25 '24 at 23:54

0 Answers0