8
$\begingroup$

Let $R$ be the ring of integers of some $p$-adic field $K$ (finite over $\mathbb{Q}_p$) with uniformizer $\pi$ and residue field $k$. I'd like to understand the finite etale extensions of $R((x)) := R[[x]][x^{-1}]$.

Note that $R((x))$ is a non-local PID with uncountably many prime ideals, all of which are generated by elements of the form $\pi^n + r_1x + r_2x^2 + \cdots$

One special prime is just the prime $\pi$, at which the residue field is $k((x))$.

Question: Is every connected finite etale extension of $R((x))$ given by $R'((x^{1/e}))$ where $R'$ is the ring of integers in some finite unramified extension $K'$ of $K$ and $e$ is coprime to $p$? If not, is it at least dominated by such an extension?

I'd also appreciate any relevant references.

$\endgroup$

1 Answer 1

7
$\begingroup$

The answer is affirmative in the "dominated by" aspect, with $R$ any complete discretely valued field whose fraction field $K$ has characteristic 0 and residue field $k$ has characteristic $p \ge 0$ (using $e$ not divisible by $p$; i.e., $e \in k^{\times}$). One cannot do better since the absence of various roots of unity from $K$ can mess up the "easy" Galois theory description of intermediate extensions. This is a standard application of Abhyankar's Lemma, as follows.

In the ring $A = R[\![x]\!]$, the height-1 prime $P = (x)$ has residue field $K$ of characteristic 0 (and uniformizer $x$), so all ramification at $P$ in any finite extension $L'$ of $L := {\rm{Frac}}(A) = {\rm{Frac}}(R(\!(x)\!))$ is tame. Hence, if $A'$ is the (module-finite) normalization of $A$ in $L'/L$ and $A'[1/x]$ is etale over $A[1/x] = R(\!(x)\!)$ then there is only tame ramification over ${\rm{Spec}}(A)$. It follows by Abhyankar's Lemma (as proved in an appendix of the book of Freitag-Kiehl on etale cohomology, for example) that ${A'}^{\rm{sh}} = A^{\rm{sh}}[x^{1/e}]$ for an integer $e \ge 1$ with $e \in k^{\times}$. Thus, $A' \subset A^{\rm{sh}}[x^{1/e}]$.

Since $A$ is complete local noetherian, so $A^{\rm{sh}}$ is exhausted by local finite etale extensions of $A$, which in turn correspond to finite separable extensions of $k$, as we vary through the local finite etale extensions $R'$ of $R$ the resulting finite etale $A$-algebras $R'[\![x]\!]$ exhaust $A^{\rm{sh}}$. Hence, for big enough such $R'$ we see that $A' \subset R'[\![x]\!][x^{1/e}]$.

The extension of Dedekind domains $$R(\!(x)\!) = A[1/x] \rightarrow R'(\!(x^{1/e})\!)$$ is clearly finite etale, so any intermediate module-finite Dedekind domain is also finite etale over $A[1/x]$. Thus, the finite etale $R(\!(x)\!)$-algebra domains are precisely the module-finite subalgebras of such extensions $R'(\!(x^{1/e})\!)$ for $R'$ local finite etale over $R$ and $e \ge 1$ not divisible by ${\rm{char}}(k)$. That is the "dominated by" assertion we wanted to prove.

$\endgroup$
5
  • $\begingroup$ I guess the question is - why do the extensions $R'[[x]]$ exhaust $A^{\text{sh}} = R[[x]]^{\text{sh}}$? Is this obvious? $\endgroup$
    – Will Chen
    Oct 25, 2015 at 4:46
  • $\begingroup$ I guess another way to phrase it is: Why is $R[[x]]^{\text{sh}} = R^{\text{sh}}[[x]]$? $\endgroup$
    – Will Chen
    Oct 25, 2015 at 4:58
  • $\begingroup$ Another question is: What do you mean by $A'^{\text{sh}}$ and $A^{\text{sh}}$? After all, neither $A$ nor $A[1/x]$ is local. $\endgroup$
    – Will Chen
    Oct 25, 2015 at 5:20
  • 1
    $\begingroup$ I defined $A$ to be $R[\![x]\!]$, so $A$ and $A'$ are local. And $A^{\rm{sh}}$ is much smaller than $R^{\rm{sh}}[\![x]\!]$ when $[k_s:k]$ is infinite (as it usually is), so your "another way to phrase it" is incorrect. Since $A$ is henselian, its local finite etale extensions are in functorial equivalence with the finite separable extensions of its residue field; the same goes for $R$ which has the same residue field. Hence, it is then obvious that $A^{\rm{sh}}$ is the direct limit of the extensions $R'[\![x]\!]$ for $R'$ local finite etale over $R$. $\endgroup$
    – nfdc23
    Oct 25, 2015 at 11:15
  • $\begingroup$ AHH! I had assumed that since $R((x))$ wasn't local, that $R[[x]]$ also wouldn't be, but of course that's wrong. Great! $\endgroup$
    – Will Chen
    Oct 25, 2015 at 15:02

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.