3
$\begingroup$

Let $K$ be a number field containing $n$-th roots of unity. The usual Kummer theory provides a correspondence between between abelian subgroups $A \subset K^*/(K^*)^n$ and abelian extensions of K of exponent dividing $n$ explicitly by $A\mapsto K(A^{\frac{1}{n}})$.

Let $L=K(A^{\frac{1}{n}})$ be such Kummer extension, $\frak{P}$ $ \subset \mathcal{O}_L$ a prime ideal lying over a prime ideal $\frak{p}$ $ \subset \mathcal{O}_K$ of $K$.

The usual criterion to decide over ramification behavior of $\frak{P}$ over $\frak{p}$ is to take a uniformizer $p$ of $\frak{p}$ and evaluate it under the normalized valuation associated with $\frak{P}$. It's unramified iff it's valuation equals one.

Question: Does in setting of Kummer extension $L=K(A^{\frac{1}{n}})$ exist a criterion for ramification behavior of a prime $\frak{P}$ over $\frak{p}$ depending only on properties of the abelian group $A$ "generating" $L$.

That is, "how much" information about ramification behavior of $L$ over $K$ can be more less directly read up from $A$?

The question is motivated by a constrution in this cript on descent on elliptic curves, where on page $5$ is constructed certain Kummer extension $L(U^{\frac{1}{n}})$ for

$$ U := \{a \in L^× \ \vert n \vert v(a) \text{ for all valuations associated to places } v \in S_L \}/(L^* )^n $$

for $S_L$ certain finite set of places/ primes. And the question is why this extension is unramified outside place in $S_L$ and those divisible by $n$.

Does it come from certain ramification criterion specific for Kummer extensions?

$\endgroup$

1 Answer 1

5
$\begingroup$

$K( A^{1/n})$ is unramified if and only if the image of $A$ in $K_{\mathfrak p}^* / (K_{\mathfrak p}^*)^n$ lies in a certain cyclic subgroup of order $n$, which is the cyclic subgroup generated by a primitive $|\mathcal O_K/\mathfrak p|-1$th root of unity if $\mathfrak p\nmid n$ or is something harder to describe if $\mathfrak p\mid n$.

This is just because the ramification depends only on the induced extension of $K_{\mathfrak p}$, which is also generated by the $n$th roots of elements of $A$ and thus is the Kummer extension of $K_{\mathfrak p}$ defined by the image of $A$ in $K_{\mathfrak p}^* / (K_{\mathfrak p}^*)^n$. Every local field has a unique unramified extension of exponent n, whose Galois group is $\mathbb Z/n$, and thus corresponds to a cyclic subgroup of order $n$ inside $K_{\mathfrak p}^* / (K_{\mathfrak p}^*)^n$ under the Kummer duality between $K_{\mathfrak p}^* / (K_{\mathfrak p}^*)^n$ and the Galois group. It remains to identify this subgroup. If $\mathfrak p\nmid n$ then $|\mathcal O_K/\mathfrak p|-1$ is a multiple of $n$ because the field $\mathcal O_K/\mathfrak p$ contains the $n$th root of unity and this implies the generator of the $|\mathcal O_K/\mathfrak p|-1$th roots of unity generates a cyclic subgroup of order $n$. Adjoining the $n$th root of this generator gives an unramified extension by Hensel's lemma.


If $A$ consists of $a$ such that $n\mid v(a)$ for all valuations associated to places not in $S_L$ then clearly for $\mathfrak p\notin S_L$ the image of $A$ consists of elements of $K_{\mathfrak p}^*$ with valuation a multiple of $n$. If also $\mathfrak p$ does not lie over $n$ then $K_{\mathfrak p}^* /(K_{\mathfrak p}^*)^n \cong (\mathbb Z/n \times \mathbb Z/n)$ with the projection to the first $\mathbb Z/n$ given by the valuation mod $n$ and the second $\mathbb Z/n$ is the cyclic subgroup generated by the roots of unity. So indeed the image of $A$ lies in this cyclic subgroup.

$\endgroup$
6
  • $\begingroup$ How do you obtain the iso $K_{\mathfrak p}^* /(K_{\mathfrak p}^*)^n \cong (\mathbb Z/n \times \mathbb Z/n) $ in case $\mathfrak p$ lies not over $n$? $\endgroup$
    – user267839
    Commented Feb 9 at 16:33
  • $\begingroup$ @user267839 From the iso $K_{\mathfrak p}^* \cong \mathbb Z\times \mathcal O_{K_\mathfrak p} \times \mu(K_{\mathfrak p}) $ obtainable from the logarithm map. $\endgroup$
    – Will Sawin
    Commented Feb 9 at 16:35
  • $\begingroup$ Right, that hold in full generality for local fields, at least when residue field is finite if I'm not confusing something:) Another point, to which extend the explicit construction of the subgroup we are interested in becomes harder if we consider the case where $\frak{k} $ $ \vert n$? How far the conceptional steps from the case $\mathfrak p\nmid n$ you explained above should be modified? Do I understnd it correctly that it goes in direction of general Kummer theory where the "container group" $K_{\mathfrak p}^* /(K_{\mathfrak p}^*)^n$ is going to be replaced by another object, $\endgroup$
    – user267839
    Commented Feb 9 at 16:59
  • $\begingroup$ the Witt module containing subgroups we are interested in giving rise to abelian ext's with exponent $n$; but in all the rest of analysis from taht point is more less similar modulo complexity of this new module? Is in that case explicit description also possible of the cyclic group of interest which detects the unramifiedness? $\endgroup$
    – user267839
    Commented Feb 9 at 17:06
  • $\begingroup$ @user267839 Well, one is supposed to take the next step of modding out each factor by $n$. Modding out $\mathbb Z$ by $n$ produces $\mathbb Z/n$ in each case. Modding out $\mu(K_p)$ by $n$ produces $\mathbb Z/n$ in each case (since we have the $n$th roots of unity in the field). But modding out $\mathcal O_{K_{\mathfrak p}}$ by $n$ is zero if $\mathfrak p \nmid n$ and nonzero otherwise, explaining the difference. $\endgroup$
    – Will Sawin
    Commented Feb 9 at 17:06

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .