Height one specializations of Selmer groups (Q2421912)

Let \(F\) and \(G\) be two ordinary cuspidal Hida families, \(f\) a cuspidal \(p\)-stabilized eigenform obtained from \(F\) by specializing at some ideal of weight \(k\) and \(\mathbb{Q}_\infty/\mathbb{Q}\) the cyclotomic \(\mathbb{Z}_p\)-extension of \(\mathbb{Q}\) with Galois group \(\Gamma\). Let \(\Sigma\) be a finite set of primes including \(\infty\), \(p\), all primes dividing the levels of \(F\) and \(G\) and another finite prime \(\ell\neq p\): denote by \(G_\Sigma\) the Galois group of the maximal extension \(\mathbb{Q}_\Sigma/\mathbb{Q}\) unramified outside \(\Sigma\). Via the action on \(G_\Sigma\) of certain rank 2 modules \(L_F\) and \(L_G\) (defined over rings \(R_F\) and \(R_G\) associated to \(F\) and \(G\) respectively), \textit{H. Hida} [Invent. Math. 85, 545--613 (1986; Zbl 0612.10021)] defined \(2\)-dimensional Galois representations \(\rho_F\) and \(\rho_G\). Tensoring with \(R_{F,G}:=R_F\widehat{\otimes}R_G\) (completed tensor product) one obtains \(4\)-dimensional representations associated to the rank 4 modules: \begin{itemize} \item \(\Hom_{R_{F,G}}(L_F\otimes_{R_F} R_{F,G},L_G\otimes_{R_G} R_{F,G})\) \item\(\Hom_{R_{F,G}}(L_F\otimes_{R_F} R_{F,G},L_G\otimes_{R_G} R_{F,G})\otimes_{R_{F,G}}R_{F,G}[[\Gamma]](\kappa^{-1})\) (cyclotomic deformation of the previous one, with \(\kappa\) the cyclotomic character) \item \(\Hom_{R_{f,G}}(L_F\otimes_{R_F} R_{f,G},L_G\otimes_{R_G} R_{F,G})\) (specialization at \(f\)). \end{itemize} For these modules \textit{H. Hida} constructed, respectively, a 2, 3 and 1-variable \(p\)-adic \(L\)-function \(\theta_i\) [Elementary theory of \(L\)-functions and Eisenstein series. Cambridge: Cambridge University Press (1993; Zbl 0942.11024)], where \(i=2,3,1\) denotes the number of variables, and provided interpolation properties and specialization formulas between them. In particular, one has relations between \(\pi^3_2(\theta_3)\) and \(\theta_2\) (where \(\pi^3_2(\Gamma)=1\)) and between \(\pi^2_1(\theta_2)\) and \(\theta_1\). On the algebraic side one can define Selmer groups associated with the representations above: there are Main Conjectures relating the divisors of the \(\theta_i\) with the divisors of the Pontrjagin duals of the Selmer groups \(\operatorname{Sel}_i\) (plus some explicit correction terms in the appropriate rings \(R_{F,G}\), \(R_{F,G}[[\Gamma]]\) and \(R_{f,G}\)). The paper under review proves ``control theorems'' for the Selmer groups, providing relations between \(\operatorname{Div}(\operatorname{Sel}_3^\vee\otimes_{R_{F,G}[[\Gamma]]} R_{F,G})\) and \(\operatorname{Div}(\operatorname{Sel}_2^\vee)\) over \(R_{F,G}\) and between \(\operatorname{Div}(\operatorname{Sel}_2^\vee\otimes_{R_{F,G}} R_{f,G})\) and \(\operatorname{Div}(\operatorname{Sel}_1^\vee)\) over \(R_{f,G}\) (where \(^\vee\) denotes the Pontrjagin dual). The author shows that such relations are compatible with the various Main Conjectures and can be therefore interpreted as specialization formulas between Selmer groups analogous to the specialization formulas between the \(p\)-adic \(L\)-functions \(\theta_i\). The proofs (under some additional hypotheses, most notably the fact that all duals of Selmer groups are assumed to be torsion modules) rely heavily on previous work of the same author, see [Int. Math. Res. Not. 2018, No. 24, 7483--7554 (2018; Zbl 1459.11213)], and on cohomological computations which lead to a more precise description of the structure of Selmer groups. In particular, looking at localizations at height 2 primes to control pseudo-null submodules, the author verifies that the maps defining the Selmer groups are surjective and that \(\operatorname{Sel}_3^\vee\) and \(\operatorname{Sel}_2^\vee\) have no nontrivial pseudo-null submodules. Both results are crucial to obtain a ``closed'' formula for the divisors of the dual of Selmer groups involved in the control theorems.