This is the part I struggled for the whole afternoon and I could not figure out obvious reason.

Let $L/K$ be a finite galois extension with $K=Frac(O)$ where $K$ has a finite residue field.($O$ is dedekind domain. Then $p\in Spec(O)-0$, $|\frac{O}{p}|<\infty$.) Assume $K$ is complete. Since $K$ is complete, $O$ has a single prime and $O_L$ has a single prime $p$ corresponding to previous prime. Let $\Delta_0$ be the inertia group of $L/K$ extension. Let $\Delta_1$ be the 1st ramification group.

$\Delta_1=\{\sigma\in\Delta_0\vert \forall a\in O_L, \sigma(a)-a\in p^2\}$.

Now $\frac{\Delta_0}{\Delta_1}\to k_L^\star$ is embedding where $k_L$ is the residue field associated to $L$. This yields $p\nmid |\frac{\Delta_0}{\Delta_1}|$. The book starts trying to show $|\Delta_1|=p^s$ for some $s\in Z_{\geq 0}$. Prove by contradiction.

"Since $\Delta_1$ is normal in $\Delta_0$ by $\Delta_1$ kernel of $\Delta_0\to k_L^\star$, it suffices to argue $\Delta_1$ is a $p-$group. Assume $r\mid|\Delta_1|$ s.t. $(r,p)=1$. Then pick out the $p-$group in $r$. Pick out an element $\sigma\in\Delta_1$ s.t. $\sigma^p\neq 1$ and $\sigma^r=1$. Take uniformizer $\pi\in L$ as $L$ is local. Then $\sigma(\pi)-\pi\in p^2$. Hence $\sigma(\pi)=\pi(1+a)$ where $a\in p^x-p^{x+1}$ for some $x\geq 1$. Consider $\sigma^p(\pi)=\pi\prod_{0\leq i\leq p-1}(1+\sigma^i(a))$. This reduction by $p^{x+1}$ yields $\sigma^p(\pi)\equiv\pi(\mod p^{x+1})$ via $\sigma^i(a)\equiv a(\mod p^{x+1})$.

Then the book concludes for any integer $s$, $\sigma^{ps}(\pi)=\pi(1+\beta)$ s.t. $\beta\in p^{x+1}$." enter image description here

$\textbf{Q:}$ How did the book deduce $\beta\in p^{x+1}$?

$\textbf{Q':}$ I guess $\sum_{0\leq i\leq p-1}$ does not annhilate everything. Is this correct?

Ref. Algebraic Number Theory by Taylor, Frohlich Chpt 3 Section 4.


Although I did not read the book you mentioned, I modified the proof a bit so that it works.

Assume $\sigma \in \Delta_1$ with order $r>1, p\nmid r$. Choose $\lambda$ a uniformizer of $\mathcal{O}_L$. (I guess this is what meant by $\lambda$ in the image). Then $\sigma(\lambda) = \lambda(1+\alpha)$ for some $\alpha \in \mathfrak{P}^t$, $\alpha \notin \mathfrak{P}^{t+1}$. Then it can be deduced that $\sigma^i(\alpha) \equiv \alpha \pmod{\mathfrak{P}^{t+1}}$. I will show that $$\tag{1} \sigma^p(\lambda) \equiv \lambda \pmod{\mathfrak{P}^{\color{red}{t+2}}}$$

Firstly, we have $$\prod_{i=0}^{p-1}(1+\sigma^i(\alpha)) \equiv (1+\alpha)^p \pmod{\mathfrak{P}^{t+1}}$$ therefore $$\sigma^p(\lambda) = \lambda \prod_{i=0}^{p-1}(1+\sigma^i(\alpha)) \equiv \lambda(1+\alpha)^p \pmod{\mathfrak{P}^{t+2}}$$ Now $$\lambda(1+\alpha)^p = \lambda + \binom{p}{1}\lambda \alpha + \cdots + \lambda \alpha^p$$ since $p\geq 2$, we have $\lambda\alpha^p \in \mathfrak{P}^{2t+1} \subset \mathfrak{P}^{t+2}$. Also, since $p\mid \binom{p}{i}$ for $0<i<p$ and $p\in \mathfrak{P}$. Hence $\lambda(1+\alpha)^p - \lambda \in \mathfrak{P}^{t+2}$, this established $(1)$.

Thus $$\sigma^{sp}(\lambda) \equiv \lambda \pmod{\mathfrak{P}^{t+2}}$$ so $\sigma^{sp}(\lambda) = \lambda(1+\beta_s)$ for some $\beta_s \in \mathfrak{P}^{t+1}$. Then the proceed as in the picture.

In fact, for $i\geq 1$, quotients of higher ramification groups $\Delta_i / \Delta_{i+1}$ can be embedded into the additive group of $\mathcal{O}_L/\mathfrak{P}$. Hence $\Delta_1$ must be a $p$-group.

  • $\begingroup$ Thanks. I think the important observation is $p\in B$ due to $char(O/B)=p$. And may be this is a dumb question. I run into the proof in Cassels' local fields that one of step requires showing $\cap_i\Delta_i=\{1\}$ in conjunction with $\Delta_i/\Delta_{i+1}$ embedding into $k$ for $i>1$. Cassels said one can find $A$ s.t. $\sigma(A)\neq A$(which is true by fixed fied.) Then he said $\sigma\not\in\Delta_i$ if $\sigma(A)-A\not\in B^{i+1}$. How to achieve the last step directly picking $A$ here? I did by contradiction if $\sigma\in\Delta_i$ for all $i$, then $\sigma$ fixes everything. $\endgroup$ – user45765 Jul 28 '18 at 12:33
  • $\begingroup$ Hence $\sigma=Id_K$ $\endgroup$ – user45765 Jul 28 '18 at 12:33

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.