Tamagawa defect of Euler systems (Q999718)

For a \(p\)-adic Galois module \(T\) attached to a motif over a number field, Kolyvagin's machinery can be used to provide an upper bound for the size of the \(p^n\)-Selmer group (for all \(n)\) associated to the dual \(T^\ast = \Hom\;(T, \mu_{p^\infty}).\) Starting from an Euler system (ES for short), the machinery produces by descent a collection of cohomology classes called ``derivative classes'' from which the desired bound is obtained. \textit{B. Mazur} and \textit{K. Rubin} [Mem. Am. Math. Soc. 168, No. 799 (2004; Zbl 1055.11041)] observed that the strong interrelations satisfied by the derivative classes can be used to define so-called Kolyvagin systems (KS for short) possessing a more ``rigid'' structure than the ES, which in many ways resembles (an enriched version of) the ``leading term'' of an \(L\)-function. Thus the KS can provide sharper bounds, and in many cases of interest, they can even yield the structure of the dual Selmer group. But although KS exist for representations for which no ES is known, all the cases so far where the bound provided by a KS can be made explicit are those where the KS comes from an ES via Kolyvagin's descent. This map from the collection of ES to the collection of KS will be referred to as the \(E-K\) map (see op. cit., thm. 3.2.4, for a precise definition). In this paper, the author studies the size of the cokernel of the \(E-K\) map in terms of the local Tamagawa numbers of \(T\). His main results are: (Thm. A): Let \(p\neq 2\) and let \(\mathcal O\) be the ring of integers of a finite extension of \(\mathbb Q_p\), \(\mathcal M = (\pi)\) the maximal ideal of \(\mathcal O\). Let \(T\) be a free \(\mathcal O\)-module of finite rank on which \(G_{\mathbb Q}\) acts continuously, and the action of \(G_{\mathbb Q}\) on \(T\) is unramified outside a finite number of places. Suppose that \(\pi^n\) divides a Tamagawa number of \(T\). Under suitable hypothesis, the image of the \(E-K\) map is contained in \({\mathcal M}^n \text{KS}(T)\), where \(\text{KS}(T)\) denotes the \({\mathcal O}\)-module of KS for \(T\). In particular, the hypotheses of Thm. A hold when \(T\) is the Tate module \(T_p(E)\) of an elliptic curve \(E/{\mathbb Q}\) with conductor \(N\). Let \(\kappa\) (Kato) denote the Kolyvagin system constructed from Kato's \(K\)-theoretic Euler system for \(E\). Let \(c_\ell\) be the Tamagawa number of \(E\) at \(\ell\), and suppose \(p^n \mid c_\ell\). Then: (Thm. B): \[ \kappa (\text{Kato}) \in p^n \text{KS}(T). \] As a corollary, the bound obtained by \textit{K. Kato} [Astérisque 295, 117--290 (2004; Zbl 1142.11336)], length \((\text{Ш}_E [p^\infty]) \leq \text{ord}_p (L_N(E, 1)/\Omega_E),\) can be improved as follows: (Thm. C): \[ \text{length}\,(\text{Ш}_E [p^\infty]) \leq \text{ord}_p\;(L_N (E,1)/c_\ell \;\Omega_E). \]