A question on an extension of a local field by adding $p^n$-th roots of a fixed uniformizer.

71 Views Asked by At

Let $K$ be a local field of characteristic $0$ with a perfect residue field of characteristic $p$. Fix a uniformizer $\pi$ and an algebraic closure $\overline{K}$. Now fix a choice of a system of $p^n$-th roots of $\pi$ in $\overline{K}$ for all $n\in \mathbb{N}$, noted $\xi=\{ \pi^{(n)} \}_{n\geq 0}$, $\ i.e.\ (\pi^{(n+1)})^p=\pi^{(n)}$ and $\pi^{(0)}=\pi$. Denote $K_{\xi}:=\cup_{n\in \mathbb{N}} K(\pi^{(n)})$.

Now I have some elementary question on the "relation" of $\xi$ with the usual cyclotomic extension $K^{cycl}:=\cup_{n\in \mathbb{N}} K(\zeta_{n})$, where $\zeta_m$ is a primitive $p^m$-th root unity for all $m\in \mathbb{N}$. I feel the situation should be simple at least if we assume $\zeta_n\not\in K, n\geq n_0$ for some fixed $n_0\in \mathbb{N}$, yet I can not describe it precisely. To make things more clear, I start by asking some precise questions as follows, based on the assumption just mentioned (I am also happy to be informed if you feel the assumption is not necessary).

  1. Is it true that $\pi^{(n+1)}\not \in K^{cycl}(\pi^{(n)})$ when $n\geq n_0$, and why?
  2. More or less the same question: What is $K^{cycl}\cap K_{\xi}$? Is it just $K$?

Noticing that when we make a different choice of system of $p^n$-th roots of $\pi$, say $\xi^{'}=\{ \pi^{'(n)} \}_{n\geq 0}$, we then know $\pi^{(m)}=\zeta_m^{c(m)}\pi^{'(m)}$ for some $c(m)\in \mathbb{Z}/p^m\mathbb{Z},\ m\in \mathbb{N}$.

  1. More or less the same question again: Is it true that $K_{\xi}\cap K_{\xi^{'}}=K(\pi^{(m)})$, where $m=\max \{n\in \mathbb{N};\ \pi^{(n)}=\pi^{'(n)} \}$?

Although I feel this question should be an elementary Galois theory exercise, I cannot write down a rigorous proof. Thanks in advance for clarifications or rigorous proofs.

1

There are 1 best solutions below

1
On

Local field : finite extension of $\Bbb{Q}_p$ or $\Bbb{F}_p((t))$.

$F(\zeta_{p^\infty-1})/F$ is unramified whereas $F(\pi_F^{1/p})/F$ is not.

Thus can replace $F$ by $F(\zeta_{p^\infty-1})$ the problem will stay the same.

It remains to check if $F(\pi_F^{1/p})=F(\zeta_{p^k})$ where $k$ is the least possible, thus $[F(\zeta_{p^k}):F]=p$.

Comparing the degrees and Galoisness we find $char(F)=0,k\ge 2,\zeta_p\in F$. The Galois groups have generators $\sigma(\pi_F^{1/p})=\zeta_p \pi_F^{1/p}$ and $\sigma'(\zeta_{p^k})=\zeta_{p^r}\zeta_{p^k}$, it must be $r=1$ because $[F(\zeta_{p^k}):F]=p$.

If $F(\pi_F^{1/p})=F(\zeta_{p^k})$ then $\sigma=(\sigma')^a,p\nmid a$ so $\sigma(\zeta_{p^k})=\zeta_p^a \zeta_{p^k},\sigma(\pi_F^{-a/p}\zeta_{p^k})=\pi_F^{-a/p}\zeta_{p^k}$ and $\pi_F^{-a/p}\zeta_{p^k}\in F$, a contradiction because $\pi_F^{-a/p}$ would have the same valuation as an element of $F$.