A strictly henselian ring is final in the opposite category of etale neighborhood of its closed point.

293 Views Asked by At

In Etale cohomology theory, by Lei Fu, Page, 195, it is stated that given a strictly Henselian ring $A$ with closed point $s$, then $\mathrm{Spec}A$ is a final object in the opposite category of étale neighborhoods of $s$, so that for any étale sheaf (say, over $X$) $\mathscr{F}$, $\mathscr{F}_s$ is $\mathscr{F}(\mathrm{Spec}A)$.

The given proof is a reference to the following two facts:

  1. For an unramified separated morphism $f: X \to Y$ with $Y$ connected, a section of $f$ is uniquely determined by its value at a single point
  2. That for an étale morphism $f: X \to \mathrm{Spec}A$, any section $f_s: s \to X \times_A s$ is deduced by base change from a section $\mathrm{Spec} A \to X$ (in this fact, A simply has to be Henselian and not strictly Henselian).

There are several things I don't understand here: the first being, in what way is $\mathrm{Spec}A$ an étale $X$-scheme? Can we find an étale morphism $\mathrm{Spec}A \to X$ over $s$ because $A$ is strictly henselian? Is such morphism directly constructed or is it taken as a section of a morphism $X \to \mathrm{Spec}A$ (in this case, how to construct it?)

And finally, how does this imply that $\mathrm{Spec} A$ is final in the opposite category of étale neighborhoods of s, i.e that for any étale neighborhood $U$ over $s$, then one can find a morphism $\mathrm{Spec}A \to U$?

I suppose the construction I ask for in the first question would give an answer to that second question, as it would be enough to repeat it for $U$ instead of $X$, the question being etale local near $s$.


Edit:

By adapting an argument a few lines below, I have managed to reduce the case to show that it indeed suffice to show that $\mathrm{Spec}A$ is an étale $X$-scheme:

One has that $$\begin{align} \mathrm{Hom}_X(\mathrm{Spec}A, U) &\cong \mathrm{Hom}_{\mathrm{Spec}A}(\mathrm{Spec}A, \mathrm{Spec}A \times_X U)\\ \end{align} $$ The latter is canonically isomorphic to $\mathrm{Hom}_s(s, s \times_{\mathrm{Spec}A} \mathrm{Spec}A \times_X U)$ because $A$ is strictly henselian: because $U$ and $\mathrm{Spec}A$ are étale $X$-schemes,the morphisms in $\mathrm{Hom}_{\mathrm{Spec}A}(\mathrm{Spec}A, \mathrm{Spec}A \times_X U)$ are etales, and since $A$ Henselian, the sections are surjective on the set of s-morphisms $s \to s \times_{\mathrm{Spec}A}\mathrm{Spec}A \times_X U$, and since $A$ is strictly Henselian, and we can assume the morphism is also separated, by replacing U with a smaller neighborhood (affine for example), since we ultimately want to show that $\mathrm{Spec}A$ is final, it doesn't impact the proof (however Fu acts as if separation wasn't an issue to show that the map $\mathrm{Hom}_{\mathrm{Spec}A}(\mathrm{Spec}A, U) \to \mathrm{Hom}_s(s, s \times_{\mathrm{Spec}A} U)$ is injective, why that?).

As $\mathrm{Hom}_s(s, s \times_{\mathrm{Spec}A}\mathrm{Spec}A \times_X U)$ canonically isomorphic to $\mathrm{Hom}_s(s,s \times_X U)$, which is canonically isomorphic to $\mathrm{Hom}_X(s, U)$, we are done.

So it really boils down to showing that $\mathrm{Spec}A$ is an etale $X$-scheme. I haven't made any progress in that direction. Also, is the above correct?