The main conjecture of Iwasawa theory for totally real fields (Q376217): Difference between revisions

Let \(p\) be an odd prime. The author proves the main conjecture of non-commutative Iwasawa theory for every admissible \(p\)-adic Lie-extension of totally real fields under the single assumption that Iwasawa's \(\mu\)-invariant vanishes. Here an admissible \(p\)-adic Lie-extension \(F_{\infty}\) of the number field \(F\) is a Galois extension such that (i) \(F_{\infty} / F\) is unramified outside a finite set \(\Sigma\) of primes of \(F\), (ii) \(F_{\infty}\) is totally real, (iii) the Galois group \(\mathcal G\) is a compact \(p\)-adic Lie-group and (iv) \(F_{\infty}\) contains the cyclotomic \(\mathbb Z_p\)-extension of \(F\). Note that there is an independent proof due to \textit{J. Ritter} and \textit{A. Weiss} [J. Am. Math. Soc. 24, No. 4, 1015--1050 (2011; Zbl 1228.11165)] when \(\mathcal G\) is a \(p\)-adic Lie-group of dimension \(1\). We briefly recall the formulation of the main conjecture. Let \(\Lambda(\mathcal G)\) be the Iwasawa algebra of \(\mathcal G\) and \(\Lambda(\mathcal G)_S\) its localization at the canonical Ore set \(S\). Then one has the following exact sequence of \(K\)-theory: \[ K_1(\Lambda(\mathcal G)) \rightarrow K_1(\Lambda(\mathcal G)_S) \rightarrow K_0(\Lambda(\mathcal G),\Lambda(\mathcal G)_S) \rightarrow 0. \] The canonical complex \[ C(F_{\infty}/F) := \mathrm{RHom}\left(R\Gamma_{\text{ét}}\left(\mathrm{Spec}\left(\mathcal O_{F_{\infty}}\left[\frac{1}{\Sigma}\right]\right), \mathbb Q_p / \mathbb Z_p\right), \mathbb Q_p / \mathbb Z_p\right) \] defines a class in the derived category of \(\Lambda(\mathcal G)\)-modules. It is perfect and, when an appropriate \(\mu\)-invariant vanishes, its cohomology is \(S\)-torsion. Using an appropriate description of the relative algebraic \(K\)-group one obtains a class \[ \left[ C(F_{\infty}/F) \right] \in K_0(\Lambda(\mathcal G),\Lambda(\mathcal G)_S). \] Let \(\kappa\) be the cyclotomic character, and for an Artin character \(\rho\) let \(L_{\Sigma}(\rho,s)\) be the \(\Sigma\)-truncated Artin \(L\)-function attached to \(\rho\). Now the main conjecture asserts that there is a (unique) pre-image \(\zeta(F_{\infty}/F)\) of \(-[C(F_{\infty}/F)]\) in \(K_1(\Lambda(\mathcal G)_S)\) such that for every Artin character \(\rho\) of \(\mathcal G\) and every positive integer \(r\) divisible by \(p-1\) one has \[ \zeta(F_{\infty}/F)(\rho \kappa^r) = L_{\Sigma}(\rho, 1-r). \] Here, \(K_1(\Lambda(\mathcal G)_S) \rightarrow \overline{\mathbb Q_p} \cup \left\{\infty\right\}\), \(x \mapsto x(\rho \kappa^r)\) is the evaluation map of \textit{J. Coates} et al. [Publ. Math., Inst. Hautes Étud. Sci. 101, 163--208 (2005; Zbl 1108.11081)]. Two remarks are in order: First, the uniqueness statement in the main result of the article under review is different: The author works with a variant of \(K_1\) which he calls \(K_1'\); this essentially is \(K_1\) modulo \(SK_1\). Second, the formulation of the main conjecture due to Ritter and Weiss does not require \(\mu=0\) (whereas their proof does), but is restricted to the \(1\)-dimensional case. In this case, however, both conjectures are equivalent by a result of the reviewer [Proc. Lond. Math. Soc. (3) 106, No. 6, 1223--1247 (2013; Zbl 1273.11155)]. For more on this topic the reader may also consult \textit{O. Venjakob}'s paper [Springer Proc. Math. Stat. 29, 159--182 (2013; Zbl 1270.11112)]. We now briefly outline the strategy of the proof. For abelian \(\mathcal G\) the conjecture follows from the work of {A. Wiles} [Ann. Math. (2) 131, No. 3, 493--540 (1990; Zbl 0719.11071)]. For non-abelian \(\mathcal G\) the author follows a strategy of Burns and Kato (see \textit{D. Burns} [J. Reine Angew. Math. (to appear); {\url{doi:10.1515/crelle-2013-0001}}]). More precisely, they showed that the main conjecture for \(\mathcal G\) is equivalent to a family of abelian main conjectures (corresponding to abelian subquotients of \(\mathcal G\)) together with a list of congruences between abelian \(p\)-adic \(L\)-functions. These congruences are then verified by using the \(q\)-expansion principle of \textit{P. Deligne} and \textit{K. A. Ribet} [Invent. Math. 59, 227--286 (1980; Zbl 0434.12009)] for Hilbert modular forms. First, the technique of Burns and Kato is used to reduce the main conjecture to the \(1\)-dimensional case. Then further reduction steps enable the author to assume that \(\mathcal G\) is the direct product of a pro-\(p\)-group and a finite cyclic group of order prime to \(p\). Note that these reduction steps (for \(1\)-dimensional \(\mathcal G\)) were already known by work of \textit{J. Ritter} and \textit{A. Weiss} [Homology Homotopy Appl. 7, No. 3, Conf., 155--171 (2005; Zbl 1154.11037)]. However, in the article under review the author systematically follows the strategy of Burns and Kato. In the heart of the paper, the author describes \(K_1(\Lambda(\mathcal G))\) (and variants of it) of a \(1\)-dimensional pro-\(p\), \(p\)-adic Lie-group in terms of the \(K_1\)'s of Iwasawa algebras of abelian subquotients. This makes heavily use of the integral group logarithm of Oliver and Taylor. The proof is therefore reduced to the verification of certain congruences between abelian \(p\)-adic \(L\)-functions. Finally, this is achieved via the Deligne-Ribet \(q\)-expansion principle, following the approach of \textit{J. Ritter} and \textit{A. Weiss} [Math. Res. Lett. 15, No. 4, 715--725 (2008; Zbl 1158.11047)]. We finally point out that the main conjecture (as the name suggests) is expected to have (or already has) far reaching consequences, in particular towards a proof of the equivariant Tamagawa number conjecture for Tate motives; the reader may consult an article of the reviewer [Compos. Math. 147, No. 4, 1179--1204 (2011; Zbl 1276.11174)] or of \textit{D. Burns} and \textit{O. Venjakob} [J. Inst. Math. Jussieu 10, No. 1, 59--118 (2011; Zbl 1213.11134)].