4
$\begingroup$

My question is about the proof of Example 8 in section 9.2 of the book "Neron models." There we have a semistable curve $X$ over an algebraically closed field $K$ and we let $\pi\colon \widetilde{X} \rightarrow X$ be its normalization. We have established the surjectivity of the natural pullback map $\pi^*: \mathrm{Pic}^0_{X/K} \rightarrow \mathrm{Pic}^0_{\widetilde{X}/K}$ of $K$-group schemes (the target being an abelian variety) and we would like to investigate the kernel $T := \mathrm{Ker}(\pi^*)$. In particular, we would like to show that $T$ is a torus. My question is: why is $T$ a torus?

The argument given for this in loc. cit. seems to be that $T$ being a torus somehow follows from an exact sequence of the form $$1 \rightarrow K^* \rightarrow \prod_{i = 1}^r K^* \rightarrow \prod_{j = 1}^N K^* \rightarrow \mathrm{Pic}(X) \rightarrow \mathrm{Pic}(\widetilde{X}) \rightarrow 1. $$ I understand the proof of this sequence, but this sequence seems to be entirely about abelian groups (at least with the level of justification given in the cited proof), so I fail to understand how it manages to imply the structure of $T$ as an algebraic group.

I would be very grateful if someone could spell out the argument or provide any clarifying comments.

$\endgroup$
  • 2
    $\begingroup$ What is written there is not at all a rigorous proof. But it contains nearly all of the key ideas. One has to think carefully about the functorial meaning of the entire calculation in terms of ${\rm{R}}^i(f_{\ast})$'s for the structure map $f$ between fppf sites to turn it into a rigorous justification (and if $K$ isn't algebraically closed then one has to use the hard fact that the non-smooth points are $K$-etale to get Weil-restricted GL$_1$'s attached to non-smooth points for describing the relevant tori over the ground field). The proof of Prop. 9 gives ample inspiration (read ahead!). $\endgroup$ – user74230 Jan 16 '15 at 4:00
  • $\begingroup$ Would you mind writing out (an outline of) the details as an answer? That would be extremely helpful. I've read the proof of Prop. 9 but that is more of the same: start out by pretending to work with Zariski sheaves, then move to other sites with no justification (or mentioning to which ones: big etale or small etale?); as a bonus put in some cryptic references to Serre's book (written in pre-EGA terminology). Here are the points of Ex. 8 that seem most obscure to me: (1) If we are to carry out the argument in the fppf site, then why is (*) exact (how to deal with arbitrary base change)? ... $\endgroup$ – Question Mark Jan 16 '15 at 6:35
  • $\begingroup$ ... (2) How is $\pi_* \mathbb{G}_m$ representable on the fppf site ($\pi$ is only finite, not locally free); same for the quotient? (I guess your helpful remark about the residue fields at non-smooth points is important here; on this subject, one should refer to III.2.7 of Freitag-Kiehl for a proof of the related remark preceding Thm. 7). (3) If one works in the fppf site, why is $R^1f_* \mathscr{Q} = 0$ (again how to deal with arbitrary base change)? $\endgroup$ – Question Mark Jan 16 '15 at 6:41
  • 1
    $\begingroup$ @QuestionMark: There are definitely other references for this, as well. If memory serves (and I often get these things wrong), Altman-Kleiman prove this in their paper on compactifications of Picard schemes of semistable curves. $\endgroup$ – Jason Starr Jan 16 '15 at 12:34
  • 2
    $\begingroup$ @user74230: I have already put in a number of hours into this (hence the decision to ask on MO), and I don't feel that it is right that the reader should be expected to redo some of the proofs. Of course, the struggle is useful in the end (but also very time-consuming). $\endgroup$ – Question Mark Jan 16 '15 at 16:48
2
$\begingroup$

One may consult Corollary 12.5 in Oda, Seshadri "Compactifications of the generalized Jacobian variety" for a seemingly more complete argument.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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