Let $K$ be a number field which is Galois over $\mathbb{Q}$ with Galois group $G$ and let $p$ be a rational prime. For $\mathfrak{p}$ a prime of $K$ which sits above $p$, we denote by $U_{1,\mathfrak{p}}$ the group of local units in $K_{\mathfrak{p}}$ which are $1$ modulo $\mathfrak{p}$ and let $U_1=\prod\limits_{\mathfrak{p}\mid p} U_{1,\mathfrak{p}}$. It is a standard result in class field theory that $U_1\cong (finite)\times \mathbb{Z}_p^{[K:\mathbb{Q}]}$.

I would like to know if it's true that there exists an element $u \in U_1$ such that $\mathbb{Z}_p[G]u$ has finite index in $U_1$ and when this is the case, whether there is a connection betwen this result and the one I mentioned above.


I stick to your notations, adding only $G_v=Gal(K_v / \mathbf Q_p)$. Notice that:

1) Because $K/\mathbf Q$ is Galois, $G$ permutes transitively the places $v$ above $p$ and all the $G_v$'s are isomorphic. The semi-local $\mathbf Z_p [G]$-representation $U_1$ is induced from the local $\mathbf Z_p [G_v]$- representation $U_{1,v}$ (for a chosen $v$)

2) Then your question is equivalent to a local one: "Does there exists a cyclic $\mathbf Z_p [G_v]$-submodule $V$ of $U_{1,v}$ with finite index ?". If such a $V$ exists, it must be $\mathbf Z_p [G_v]$-free because of a question of rank

The answer to this local question is "yes", although somewhat hidden in the literature. In Cassels-Fröhlich's book, chap.6, Serre proves the following propos.3 : " $U_{1,v}$ contains an open sub-module which is $G_v$- cohomologically trivial". But if you look carefully at his first proof (he gives two), he actually shows that $V$ is free. I briefly recap the argument: the idea is to compare the additive-multiplicative structures via the exp-log maps. There is a normal (additive) basis of $K_v / \mathbf Q_p$, i.e. a basis consisting of the Galois conjugates of a single element $a$, which we can take in the local ring of integers. Define $A$ = the additive (free) $\mathbf Z_p [G_v]$-module generated by $a$. Take $v(a)$ large enough to ensure that the exp series converges in a neigbourhood of $0$. Then $V=exp(A)$ does the job.

NB: The Galois structure of the global units is much harder, because much deeper, because related, via not yet fully proved conjectures, to special values of L-functions.

  • $\begingroup$ Thank you very much for this very beautiful and clear answer, it is exactly the kind of argument I was looking for :) $\endgroup$ – Jake Haider Nov 26 '16 at 12:11

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.