Semi-local units at \(p\) of a cyclotomic \(\mathbb{Z}_p\)-extension congruent to 1 modulo \(\zeta_p - 1\) (Q2634759)

For \( p \) a prime number, let \( K \) be an abelian number field such that \( p\nmid [K:\mathbb{Q}] \) and \( \zeta_p\in K, \) with \( \zeta_p \) a primitive \( p \)-th root of unity. Let \( K_{\infty}/K \) be the cyclotomic \( \mathbb{Z}_p \)-extension and \( K_n \) the \( n \)-th layer (with \( K_0=K \)). Let \( \mathcal{U}_n \) be the group of semi-local principal units of \( K_n \) at \( p, \) let \( \mathcal{U}_{\infty}=\varprojlim \mathcal{U}_n\) be the projective limit with respect to the relative norms \( K_m\to K_n \) (for \( m>n \)) and let \( \mathcal{V}_n \) be the image of the projection \( \mathcal{U}_{\infty}\to\mathcal{U}_n. \) Further, let \( \mathcal{U}_n^{(1)}=\{u\in \mathcal{U}_n:u\equiv1\pmod {\zeta_p-1} \} \) and \( \mathcal{V}_n^{(1)}=\mathcal{U}_n^{(1)}\cap\mathcal{V}_n. \) The groups \( \mathcal{U}_n,~\mathcal{V}_n,~\mathcal{U}_n^{(1)} \) and \( \mathcal{V}_n^{(1)} \) can be regarded as modules over the Galois groups \( \Delta=\mathrm{Gal}(K/\mathbb{Q}) \) and \( \Gamma=\mathrm{Gal}(K_{\infty}/K). \) The structure of \( \mathcal{U}_n \) as a Galois module is well understood. In this paper, the aim of the author is to determine the structure of \( \mathcal{U}_n^{(1)}. \) For \( \bar{\mathbb{Q}}_p \) a fixed algebraic closure of \( \mathbb{Q}_p, \) let \( \chi \) be a fixed \( \bar{\mathbb{Q}}_p \)-valued character of \( \Delta \) and let \( \mathcal{O}=\mathcal{O}_{\chi} \) be the subring of \( \bar{\mathbb{Q}}_p \) generated by the values of \( \chi \) over \( \mathbb{Z}_p. \) Upon choosing a generator \( \gamma \) of \( \Gamma \) and identifying the completed group ring \( \mathcal{O}[[\Gamma]] \) with the power series ring \(\Lambda=\Lambda_{\chi}=\mathcal{O}[[s]] \) (via \( \gamma\leftrightarrow 1+s), \) one can regard the \( \chi \)-parts \( \mathcal{U}_{\infty}(\chi),~\mathcal{U}_n(\chi) \) etc. as modules over \( \Lambda. \) The \( \Lambda \)-module structures of \( \mathcal U_n(\chi) \) and \( \mathcal V_n(\chi) \) are known, see \textit{R. Coleman} [Proc. Am. Math. Soc. 89, 1--7 (1983; Zbl 0528.12005)], \textit{R. Gillard} [Ann. Inst. Fourier 29, No. 4, 1--15 (1979; Zbl 0403.12006)], \textit{K. Iwasawa} [J. Math. Soc. Japan 16, 42--82 (1964; Zbl 0125.29207)]. Usually, \( \mathcal{U}_n(\chi)=\mathcal{V}_n(\chi) \) and \( \mathcal{V}_n(\chi)\cong\Lambda/w_n \) (as \( \Lambda \)-modules) where \( w_n=w_n(s)=(1+s)^{p^n}-1. \) Building up on the work in [J. Reine Angew. Math. 462, 169--184 (1995; Zbl 0815.11055)], where the author determined the ideal \( J_{n,\chi} \) of \( \Lambda \) corresponding to the submodule \( \mathcal{V}_n^{(1)}(\chi) \) under the above isomorphism in the case when \( p\geq 3 \) does not split in \( K \) and \( \chi \) is \textit{even}, i.e., \( \chi(-1)=1, \) the author describes, in the present paper, the \( \Lambda \)-module structure of \( \mathcal{V}_n^{(1)}(\chi) \) in the general situation when \( p\nmid [K:\mathbb{Q}] \) and \( \chi \) is not necessarily even, including the case \( p=2. \) More precisely, on regarding \( \chi \) and its dual \( \chi^*=\omega_{\tilde{p}}\chi^{-1} \) (where \( \tilde{p}=2p \) or \( p \) if \( p=2 \) or \( p\geq 3 \) respectively, and \( \omega_{\tilde{p}}\) is the Teichmüller character of conductor \( \tilde{p} \)) as primitive Dirichlet characters, one classifies the character \( \chi \) into three types: \[ \text{(A)}~\chi(p)\neq1\text{ and }\chi^*(p)\neq1,\quad \text{(B)}~\chi^*(p)=1,\quad\text{(C)}~\chi(p)=1. \] If \( \mathbb{T} \) denotes the \( \Lambda \)-torsion submodule of \( \mathcal{U}_{\infty}(\chi), \) it is known (see \textit{R. Gillard} [Ann. Inst. Fourier 29, No. 4, 1--15 (1979; Zbl 0403.12006)]) that, as \( \Lambda \)-modules, \[ \mathcal{U}_{\infty}(\chi)\cong\Lambda,\;\text{for type (A) or (C)\quad and\quad}\mathcal{U}_{\infty}(\chi)\cong \Lambda\oplus\mathbb{T},~\text{for type (B)} \] and also, if we set \( \tilde{\mathcal{V}}_n(\chi)=\mathcal{V}_n(\chi) \) for type (A) or (C) and \( \tilde{\mathcal{V}}_n(\chi)=\mathcal{V}_n(\chi)/\mathbb{T}_n\) for type (B), where \( \mathbb{T}_n \) is the projection of \( \mathbb T \) to \(\mathcal{U}_{n}(\chi),\) that \[ \tilde{\mathcal{V}}_n(\chi)\cong \Lambda/(w_n),~\text{for type (A) or (B)\quad and}\quad \tilde{\mathcal{V}}_n(\chi)\cong\Lambda/(w_n,2w_n/s),~\text{for type (C).} \] Let \(\tilde{\mathcal{V}}_n^{(1)}(\chi)=\mathcal{V}_n^{(1)}(\chi)\) for type (A) or (C) and \( \tilde{\mathcal{V}}_n^{(1)}(\chi)=\mathcal{V}_n^{(1)}(\chi)\mathbb{T}_n/\mathbb{T}_n \) for type (B). Define the ideal \( J_{n,\chi} \) of \( \Lambda \) containing \( w_n \) (resp. \( w_n \) and \( 2w_n/s \)) for type (A) or (B) (resp. type (C)), so that the above isomorphism induces \[ \tilde{\mathcal{V}}_n^{(1)}(\chi)\cong J_{n,\chi}/(w_n),~\text{for type (A) or (B)\quad and}\quad\tilde{\mathcal{V}}_n^{(1)}(\chi)\cong J_{n,\chi}/(w_n,2w_n/s),~\text{for type (C).} \] For \( n\geq 1 \) let the ideal \( I_{n,\chi} \) of \(\Lambda\) be given by \[ I_{n,\chi}=\langle p^n,p^{n-1-k}s^{p^k}\mid 0\leq k\leq n-1\rangle,~\text{for type (A) or (C)} \] and \[ I_{n,\chi}=\langle p^{n-1-k}s^{p^k-1}\mid 0\leq k\leq n-1\rangle,\;\text{for type (B)} \] and put \( I_{0,\chi}=\Lambda. \) The main result of the paper can now be formulated as follows: \[ \mathit{Under the above setting, we have}\;J_{n,\chi}=I_{n,\chi}\text\textit{for all }\;n\geq 0\text\textit{and }\;\chi. \] The two main steps of the proof are showing that \( I_{n,\chi}\subseteq J_{n,\chi} \) for each \( \chi \) associated to \( K \) and showing that the products \( \prod_{\chi}|\Lambda_{\chi}/I_{n,\chi}| \) and \( \prod_{\chi}|\Lambda_{\chi}/J_{n,\chi}| \) are equal. It is worth mentioning that this is done in a rather elementary way.