8
$\begingroup$

For constructible sheaves $\mathcal F$ on real analytic manifolds $X$, there is a notion of the singular support $SS(\mathcal F)$ which is a radially invariant singular Lagrangian subset of the cotangent bundle $T^\ast X$.

For constructible $\ell$-adic (or $\mathbb F_\ell$ if you prefer) sheaves $\mathcal F$ on the etale site of a smooth variety $X$ in positive characteristic, Beilinson defined an analogous notion of singular support $SS(\mathcal F)\subseteq T^\ast X$. In positive characteristic, the singular support is always half-dimensional (every irreducible component is of the same dimension as $X$), but it need not be Lagrangian!

What is the simplest example of such a sheaf $\mathcal F$ in positive characteristic with non-Lagrangian singular support? I am led to believe that the sheaf $j_!\mathbb Q_\ell$ for $$j:\{(x,y,t)\in\mathbb A^3_{x,y,t}:t^2-t=xy\}\to\mathbb P^2_{x,y}$$ in characteristic $2$ is such an example (this example was communicated to me by Takeshi Saito as an example of something else). Is this example correct, and is there a simpler one?

$\endgroup$
2
  • 1
    $\begingroup$ I think the point is that in characteristic $p$, $t^p - t = xy^{p-1}$ is fiercely ramified at infinity (purely inseparable extension of the residue field). For $p=2$ this is your example. It's probably correct but I don't remember the computation. It also appears in my paper "Wild ramification and K(pi, 1) spaces" section 7.1 as an example for something closely related, and you should be able to deduce what you want combining it with Remark 3.3 in the same paper. By the way, I think Deligne showed that for surfaces you can get any closed two-dim conical subset you want as the singular support. $\endgroup$ Commented Sep 18, 2017 at 19:19
  • 1
    $\begingroup$ Also no I don't think there is a simpler example than an Artin-Schreier sheaf on $\mathbb{A}^2$ ;) Of course your sheaf has generically rank $p$, but it splits into rank one sheaves (Artin-Schreier sheaves for characters $\mathbb{F}_p\to \mathbb{F}_\ell^\times$) and you can take each of those except for the constant one. $\endgroup$ Commented Sep 18, 2017 at 19:20

0

You must log in to answer this question.