Abstract \(\ell\)-adic 1-motives and Tate's canonical class for number fields (Q1656223)

From the authors' abstract: ``In [J. Algebr. Geom. 24, No. 4, 629--692 (2015; Zbl 1330.11070)], they constructed a new class of Iwasawa modules as \(\ell\)-adic realizations of what we called abstract \(\ell\)-adic 1-motives in the number field setting. The authors proved in [loc. cit.] that the new Iwasawa modules satisfy an equivariant main conjecture. In this paper they link the new modules to the \(\ell\)-adified Tate canonical class and give an explicit construction of (the minus part of) \(\ell\)-adic Tate sequences for any Galois CM extension \(K/k\) of an arbitrary totally real number field \(k\).'' Before going into more technical details, let us first introduce some notations. Throughout \(\ell\) is an odd prime number, \(G= \text{Gal}(K/k)\). Denote by \(K_\infty\) (resp. \(k_\infty\)) the cyclotomic \(\mathbb{Z}_\ell\)-extension of \(K\) (resp. \(k\)), and assume that the classical Iwasawa \(\mu\)-invariant associated to \(K_\infty\) and \(\ell\) vanishes. Fix two disjoint \(G\)-invariant sets of primes \(S\) and \(T\) in \(K\), such that \(S\) contains the ramification locus of \(K_\infty/k\) and \(T\) contains at least two primes of distinct residue characteristics. From the data \((K/k,S,T,\ell)\), the authors constructed in [loc. cit.] a canonical \(\ell\)-adic 1-motive \({\mathcal M}:={\mathcal M}^\ell_{S,T}(K/k)\), whose \(\ell\)-adic realization (Tate module) \(T_\ell({\mathcal M})\) is \(\mathbb{Z}_\ell\)-free of finite rank, naturally endowed with a \(\mathbb{Z}_\ell[[{\mathcal G}]]\)-module structure, where \({\mathcal G}= \text{Gal}(K_\infty/k)\). The complex conjugation \(j\) of the CM field \(K_\infty\) acts on \(T_\ell({\mathcal M})\) with eigenvalue \((-1)\), so \(T_\ell({\mathcal M})\) can be viewed as a module over the quotient ring \(\mathbb{Z}_\ell[[{\mathcal G}]]^-\). The main result of [loc. cit.] is that (1) \(pd_{\mathbb{Z}_\ell[[{\mathcal G}]]} T_\ell({\mathcal M})= 1\); (2) If \(G\) is abelian, the initial Fitting ideal \(Fit_{\mathbb{Z}[[{\mathcal G}]]^-}(T_\ell({\mathcal M}))\) is principal, generated by a certain equivariant \(\ell\)-adic \(L\)-function \(\Theta^\infty_{S,T}\in\mathbb{Z}_\ell[[{\mathcal G}]]^-\). For simplicity, suppose now that \(k_\infty/k\) and \(K/k\) are linearly disjoint (this hypothesis can be removed), so that \({\mathcal G}\cong G\times\Gamma\), where \(\Gamma:=\text{Gal}(k_\infty/k)\), and consequently \(\mathbb{Z}_\ell[[{\mathcal G}]]^-\cong\Lambda[G]^-\), where \(\Lambda\) is the usual Iwasawa algebra. It follows [loc. cit.] that \(T_\ell({\mathcal M})\) is projective over \(\mathbb{Z}_\ell[G]\) (also on \(\mathbb{Z}_\ell[G]^-\)), and one obtains a perfect complex of \(\mathbb{Z}_\ell[G]\)-modules \(C^*[T_\ell({\mathcal M})@>\gamma-1>> T_\ell({\mathcal M})]\) concentrated in degrees 0 and 1, where \(\gamma\) is a topological generator of \(\Gamma\). The main goal of the present article is first to describe the cohomology groups \(H^0(C^*)= T_\ell({\mathcal M})^\Gamma\) and \(H^1(C^*)= T_\ell({\mathcal M})_\Gamma\) in terms of arithmetic invariants associated to \((K/k,S,T)\), second to describe the class of \(C^*\) in an appropriate group \(\text{Ext}^2_{\mathbb{Z}_\ell[G]^-}(*,*)\). From now on, suppose that \(S\) is ``\(\ell\)-large'', i.e., that \((cl_T(K)\otimes\mathbb{Z}_\ell)^-\) is generated by ideal classes supported at primes in \(S\) (here \(cl_T(K)\) is the ray class group of \(K\) with conductor equal to the product of the prime ideals belonging to places in \(T(K)\)). Then: 1) Based on a concrete expression of \(T_\ell({\mathcal M})\) given in \S3 of [loc. cit.], it is not difficult to show the existence of a \(\mathbb{Z}_\ell[G]\)-module isomorphism \(T_\ell({\mathcal M})^\Gamma\cong (U_{S,T}\otimes\mathbb{Z}_\ell)^-\), where \(U_{S,T}\) is the group of \(S\)-units in \(K\) which are congruent to 1 modulo all primes in \(T\). 2). A \(\mathbb{Z}_\ell[G]\)-module isomorphism \(T_\ell({\mathcal M})^\Gamma\cong\mathbb{Z}_\ell[S]^-\) can be derived by simple homological algebra from an exact sequence \(0\to\mathbb{Z}_\ell[S_\ell]^-\to T_\ell({\mathcal M})_\Gamma\to \mathbb{Z}_\ell[S\setminus S_\ell]^-\to 0\). Unfortunately, both the existence and the splitting of such an exact sequence are rather laborious and not quite canonical \(K/k\), \(S\), \(T\). 3) Consider the Tate canonical class \(\tau= \tau_{S,K/k}\in \text{Ext}^2_{\mathbb{Z}[G]}(X_S, U_S)\), where \(U_S\) is the group of \(S\)-units of \(K\) and \(K_S\) the group of degree 0 divisors in \(K\) supported at \(S\). This class is represented by a Yoneda 2-extension of \(\mathbb{Z}[G]\)-modules \(0\to U_S\to A\to B\to X_S\to 0\), where \(A\) and \(B\) are finitely generated and cohomologically trivial. To apply this to \(T_\ell({\mathcal M})\), the authors have to \(\ell\)-adify, \(T\)-modify and take the minus part. They obtain the desired link between \(C^*\) and a modified Tate class \(\tau'= \tau_{K/k,S,T}\in \text{Ext}^2_{\mathbb{Z}[G]} (X_S, U_{S,T})\), via calculations in a certain derived category \(D(\mathbb{Z}_\ell[G])\), relying essentially upon deep results of \textit{D. Burns} [Doc. Math. 9, 357--399 (2004; Zbl 1077.11049)] and \textit{D. Burns} and \textit{M. Flach} [Am. J. Math. 120, No. 6, 1343--1397 (1998; Zbl 0929.11050)]. The main theorem states the existence of a canonical isomorphism \((\mathbb{Z}_\ell\otimes \tau')^-\cong C^*\) in \(D(\mathbb{Z}_\ell[G])\). Note that this agrees on the cohomological level with the previous isomorphism \(T_\ell({\mathcal M})_\Gamma\cong\mathbb{Z}_\ell[S]^-\) (which was less smooth, but definitely more explicit).