No label defined (Q4940667)

The authors consider \(G\)-extensions \(K/k\) of \(p\)-adic fields (\(p>2\)) where \(G\) is an abelian \(p\)-group, and ideals of \(O_K\) as \(O_k[G]\)-modules. In a previous article [with \textit{I. B. Zhukov}, Algebra Anal. 9, No.~4, 28-46 (1997; Zbl 0902.11050)], a criterion was established for the existence of a decomposable ideal: this happens iff the ramification index of \(K/k\) divides the different \(D_{K/k}\). (This is the case for instance if \(K=\mathbb{Q}(\zeta_{p^m})\) and \(k=\mathbb{Q}(\zeta_p)\); actually there one has equality.) The present article is a sequel; it determines in this case just which ideals in \(K\) are decomposable. The precise result is as follows. Write the ideal \(I\) as \(\kappa\)-th power of the maximal ideal; then \(I\) is decomposable if either \(p\) divides \(\kappa\) or one among the coefficients \(\kappa_0,\ldots, \kappa_{m-1}\) in the \(p\)-adic representation of the integer \(\kappa=\sum\kappa_ip^i\) exceeds \(\bar h\), the residue modulo \(p\) of the maximal ramification jump \(h\) of \(K/k\). Example: If \(k=\mathbb{Q}(\zeta_p)\) and \(K=k(\pi^{1/p})\) with \(\pi\) a parameter of \(k\), then \(\bar h=0\), and \textit{all} ideals of \(K\) are decomposable in this sense. The method of proof is, essentially, the following: the case \(G\) of prime order is treated by explicit calculation, and the general case by induction. It is very helpful that direct sum decompositions of an ideal \(I\) of \(K\) as \(O_k[G]\)-module stay this way when intersected down to an intermediate field \(k\subset F\subset K\); the reason being simply that intersecting with \(F\) can also be described as taking invariants under Gal\((K/F)\).