Cohomological Mackey functors in number theory (Q597123): Difference between revisions

For a finite Galois extension \(L/K\) of number fields with Galois group \(G\), certain number-theoretic invariants (e.g. \(p\)-class groups, Iwasawa modules, \(p\)-Shafarevich-Tate groups\dots) attached to subextensions \(M/K\) are not independent as \(M\) varies between \(K\) and \(L\). In each special case, the relations of dependence may give the impression that they are of an intrinsic arithmetic nature, but actually, their ubiquity is controlled by the partially ordered set of subgroups of \(G\), no matter which of the above invariants is considered. The present article gives a systematic and general approach to such dependencies in the framework of (cohomological) Mackey functors. Roughly speaking, a Mackey system \((\mathcal C,\mathcal O, G)\) consists of a set \(\mathcal C\) of subgroups of \(G\) and, after setting \(\mathcal C(H) = \{U\leq H\), \(U\in\mathcal C\}\), a family \(\mathcal O(H) \subset\mathcal C(H)\) for each \(H\in\mathcal C\) satisfying a list of natural conditions. A Mackey functor \(M\) over a commutative ring \(k\) on a Mackey system \((\mathcal C, \mathcal O, G)\) associates to each subgroup \(H\in\mathcal C\) a \(k\)-module \(M(H)\) together with various structure maps between then (induction, restriction, and particularly a technical ''Mackey formula'' involving double cosets). The arithmetic invariants cited above are examples of Mackey functors over \(\mathbb Z_p\). For a given \(G\), there is a kind of ``universal'' Mackey functor going to the so-called Burnside ring \(b(G)\), which in turn maps canonically to a representation ring \(a(kG)\): This means intuitively that the relations between the values of a Mackey functor are controlled by relations between permutation modules. By studying the kernel of the canonical map \(b(G)\to a(kG)\), the authors prove a general theorem which gives a relation between values of a Mackey functor \(M\) at groups between two fixed groups \(I\triangleleft H\). In characteristic zero (resp. \(p\)) one has to assume that \(H/I\) is not cyclic (resp. has no normal \(p\)-subgroup with cyclic quotient). Applying this ``universal'' relation to arithmetic examples, one gets relations (which bear a formal resemblance) on \(\mu\) and \(\lambda\)-invariants, on zeta values, on regulators etc. which contain many known results scattered in the literature.