As I said previously, I have some problems in the theory of valuations and places.

Let L/K be a finite (say) Galois extension, F a place of L, and v a valuation of L. I denote by l and k the residue field of F and of the restriction of F to K resp. It is known that the map from the decomposition group D of F to the group of automorphisms of the residual field extension l/k is onto. This means that for every k-automorphism t of l/k, there exists an automorphism s of Gal(L/K) such that tF = Fs.
Also the inertia group of F is exactly the set of automorphisms s of Gal(L/K) such that Fs = F. But what is known about the set of automorphisms s of Gal(L/K) such that vs = v ? and is there a theorem according to which the map from the decomposition group D of v to the group of automorphisms t of the Abelian ordered group of values of v that fixes the group of values of the restriction of v to K, is onto ? (it is clear that an automorphism s of D induces such an automorphism via the formula tv(x) = v(sx), but is this homomorphism surjective) ? I have found not hint in Fried and jarden, nor in Bourbaki.


D consists of the automorphisms which preserve the valuation v of L, by definition of decomposition group. In the case of a discrete valuation, D has a filtration by "ramification groups" given by valuation of (sx/x - 1) for s in D (where x is a uniformizer for v). See Serre, "Local Fields".

| cite | improve this answer | |
  • $\begingroup$ I am not sure you have understood my question. D consists in the automorphisms of G that preserve de valuation v of L UP TO AN EQUIVALENCE, but I asked for the group of automorphisms of G that preserve exactly the valuation v (let us denote this group by J, by analogy with the inertia group I which is the set of automorphisms of G that preserves exactly a place F, not up to an equivalence). My second question is related to an analogue of Frobenius theorem. $\endgroup$ – MikeTeX Dec 11 '14 at 16:00
  • $\begingroup$ To be more explicit, D is by definition the group of automorphisms such that $\sigma O_v = O_v$. If $\tau$ is an automorphism of the ordered group of values of v, then the valuation v' defined by $v'(x) = \tau v(x)$ has the same valuation ring as v, but is not exactly equal to v, only equivalent to it. So, D is the group that preserves v up to an equivalence. $\endgroup$ – MikeTeX Dec 11 '14 at 16:23
  • $\begingroup$ @MikeTex: Any s in D preserves the local ring at v so it preserves valuations for all elements of value 0 (they are the units of the local ring). The value group of K is of finite index in the value group of the finite extension L, so if x is in L then there exist integer n and y in K such that v(x^n/y)=0. This implies that nv(x)-v(y)=v(x^n/y)=0=v(s(x^n/y))=v(s(x)^n/y)=nv(s(x))-v(y) and thus v(s(x))=v(x) because the value group is ordered and so torsion free. $\endgroup$ – David Lampert Dec 11 '14 at 18:24
  • $\begingroup$ David : I have just realized that I even not voted for your answers (this is done now). I appologize : I was entirely new in Mathoverflow and not used with the customs here. $\endgroup$ – MikeTeX Jan 26 '15 at 9:02

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.