On main conjectures in non-commutative Iwasawa theory and related conjectures (Q486433): Difference between revisions

This is a very important paper on non-commutative main conjectures, a field which has seen rapid developments over the last years. The classical situation is roughly as follows: \(K/k\) is an abelian extension of totally real fields such that \(K\) is finite over the \(p\)-cyclotomic extension \(k_\infty\) of \(k\). One can write \(K=Fk_\infty\) with \(F/k\) finite abelian (we may and will assume that \(F\cap k_\infty=k\)), and one considers the standard \(p\)-ramified Iwasawa module \(\mathcal X(K)\), the projective limit of the \(p\)-part of the ray class groups with conductor \(p^\infty\) of \(Fk_n\). This is a torsion module over the Iwasawa algebra \(\Lambda(G) \cong \mathbb Z_p[[T]][\text{Gal}(F/k)]\), and the main conjecture MC, very roughly, describes \(\mathcal X(K)\) in terms of \(p\)-adic \(L\)-functions. The modern version of this is much more general. The module \(\mathcal X\) is replaced by a perfect complex associated to a suitable motive, and \(K\) (still totally real) may be a much larger extension of \(k\), in the following sense: \(G=\text{Gal}(K/k)\) can be a compact \(p\)-adic Lie group of arbitrary finite rank, but still \(K\) must contain \(k_\infty\). The subgroup \(H=\text{Gal}(K/k_\infty)\subset G\) plays an important role. The abelian rank 1 case for the Tate motive \(\mathbb Q(1)\) comes down to the classical situation. A few years ago, \textit{J. Ritter} and \textit{A. Weiss} [Indag. Math., New Ser. 15, No. 4, 549--572 (2004; Zbl 1142.11369); Homology Homotopy Appl. 7, No. 3, 155--171 (2005; Zbl 1154.11037); Math. Ann. 336, No. 1, 27--49 (2006; Zbl 1154.11038)] and \textit{M. Kakde} [J. Algebr. Geom. 20, No. 4, 631--683 (2011; Zbl 1242.11084); Invent. Math. 193, No. 3, 539--626 (2013; Zbl 1300.11112)] (independently) proved MC in the noncommutative rank one case for the classical Tate motive. Their proofs are rather different but came out roughly at the same time. Kakde's proof makes essential use of a beautiful idea of Burns how to reduce the finite rank case to the rank one case. This idea arose in 2006, and as Burns says himself, large parts of the paper under review are a ``long overdue'' update on this idea. Section 1 gives a formalization of main conjectures, which allows to state the reduction theorem 2.1 in a very abstract way. An ``abstract main conjecture'' MC takes as input a perfect complex \(C^\bullet\) of \(\Lambda(G)\)-modules, and a family of ``interpolation data'' \(\xi_\rho\in \mathbb Q_p^c\), indexed by characters \(\rho\) of \(G\). An element \(\xi \in K_1(\Lambda(G)_{S^*})\) is said to validate MC in this setting if two properties hold: for all \(\rho\) and almost all type W characters \(\omega\), \(\xi\) has the value \(\xi_{\rho\otimes\omega}\) at \(\rho\otimes \omega\), and the refined Euler characteristic of \(C^\bullet\) is the image of \(\xi\) under the connecting homomorphism from \(K_1(\Lambda(G)_{S^*})\) to the relative K-group \(K_0(\Lambda(G), \Lambda(G)_{S^*})\). (Note: Every reader will see that we have left many details unexplained for reasons of space, and experts will see that the above explanation is oversimplified and not totally precise.) Now Theorem 2.1 says that an abstract MC is valid for the group \(G\) iff it is valid for all quotients \(G/U\) with \(U\) open in \(H\). Note that all these quotients \(G/U\) are of rank one, since \(H/U\) is in fact finite. A further reduction in the abstract setting is achieved in Theorem 2.2: the validity of MC for rank one subquotients \(\mathcal A\) of \(G\) is replaced by a kind of ``validity character by character''; in some way the group \(\mathcal A\) is eliminated, or at least reduced to a free procyclic group \(\Gamma(\mathcal A)\), so that one can work over Iwasawa algebras of the form \(\Lambda_{\mathcal A,\phi} =\mathcal O(\phi)[[\Gamma(\mathcal A)]]\), which are regular local rings. But there is a very important side condition: one needs a strong interpolation property for the family \(\xi_\rho\), where \(\rho\) runs over the characters of any given rank one quotient \(\mathcal G\) of \(G\). Subsequently in Theorem 2.3, a certain diagonal map \(\Delta_{G,S}\) comes into play; it is defined on \(K_1(\Lambda(G)_S)\) and takes values in the product of the full quotient rings of \(\mathcal O(\phi)[[\Gamma(\mathcal A]]\), with \(\mathcal A\) ranging over certain rank one subquotients \(\mathcal A\) and \(\phi\) over all characters of \(\mathcal A\). Then MC is valid at level \(G\) (very roughly speaking) if and only if the family of interpolation data is in the image of the diagonal map. This gives a strategy for proving the ``concrete'' MC, where by ``concrete'' we mean that the interpolation data come from appropriate values (or leading terms) of \(L\)-functions. We are not going to make this precise. These abstract results have arithmetic applications. They are basically of two kinds. Firstly there is a major theorem 2.6 (and the related Theorem 9.1 from which it is deduced) on the validity of the concrete MC for the classical Tate motive. Secondly the paper contains affirmative results on many other important conjectures (Theorem 2.8 and Cor.~2.10). For lack of space let us focus on the first kind of application. As we said, the qualifier ``concrete'' means that the interpolation data in the input for MC come from \(L\)-functions. The complex \(C^{\bullet}\) also comes from arithmetic. In fact, complexes are considered modulo pseudo-isomorphism, and in the setting of Theorem 2.6 the complex \(C^{\bullet}\) is represented by a complex which only has one nontrivial term, sitting in degree 0. This nontrivial term is a ``standard Iwasawa module'': the Galois group \(X_\Sigma(K)\) of the maximal \(\Sigma\)-ramified pro-\(p\)-extension of the big field \(K\). In order to get a perfect complex, one has to assume that \(G\) has no elements of order \(p\). In fact the author also proves a more general result (9.1) which does not need this condition on \(G\), at the price of having a less explicit complex. For the sake of completeness we should add that in the results 2.6, 9.1, 2.8, 2.10 appropriate hypotheses on the vanishing of the \(\mu\)-invariant are needed, but this is usual practice in this field. There are close relations among the work presented here and other recent papers on the main conjecture. As said before, Kakde's proof of the main conjecture makes crucial use of the abstract results formulated in subsection 2.2 of the present paper. As Burns acknowledges himself, his initial ideas in this direction were taken up and improved by Kato, so that the approach may be called the ``Burns-Kato strategy for the proof of the general MC''. Kakde also gives a proof of Theorem 9.1 of the present paper which does not rely on the Ritter-Weiss result, while the proof of Theorem 9.1 given by the author uses both his abstract results of subsection 2.2, and the Ritter-Weiss result. As stated above, we will not say much about Theorem 2.8 and Cor.~2.10, which also contain strong and important results. Among other things, the author gives affirmative results concerning Chinburg's \(\Omega(1)\)- and \(\Omega(3)\)-conjectures for nonabelian extensions; apart from some quaternionic cases not much had been known before. There are many other results in the paper which we cannot go into; but one more thing really deserves to be pointed out. In Section 5, the author presents a very satisfactory version of the Weierstraß\ preparation theorem. It is formulated for a \(K_1\)-group instead of a unit group, and this eliminates a somewhat pathological non-commutativity phenomenon observed in an earlier paper by the author and the reviewer. The paper under review is long, complex and substantial. It is written in a very clear and careful way. Already the preprint version of 2007 was quite influential, and the published paper will now be a standard reference.