Higher relative class number formulae (Q1849476): Difference between revisions

Let \(E\) be a totally complex Abelian number field with maximal real subfield \(F\), \(\chi\) the nontrivial character of Gal\((E/F).\) Then \(\zeta_{E} (s) = \zeta_{F} (s) \ldotp L (E/F, \chi, s)\) and the well-known relative class number formula reads: \[ L(E/F, \chi, 0)= {2^{r_{1}} \over Q} {h^{-} \over w(E)}, \] where \(r_{1} = [ F : \mathbb Q],\) \(h^{-}\) is the relative class number, \(w(E)\) is the number of roots of unity in \(E\) and the \(Q\)-index is defined as \(Q = [{\mathcal O}^{\ast}_{E} : {\mathcal O}^{\ast}_{F} \ldotp \mu (E)] = 1\) or 2. Similar to the classical case \(n = 1,\) the special value \(L(E/F, \chi, 1-n)\) at \(1-n\) for odd \(n \geq 3\) is given here by a relative class number formula of the form: \[ L(E/F, \chi, 1-n) = \pm {2^{r_{1}+1} \over Q_{n}} \ldotp {h^{-}_{n} \over w_{n}(E)}, \] where \(w_{n} (E) = |H^{0} (E, \mathbb Q/\mathbb Z (n))|,\) \(Q_{n}\) is a higher \(Q\)-index, equal to 1 or 2, and \(h^{-}_{n} = {h_{n}(E) \over h_{n}(F)},\) \(h_{n}(L)\) denoting the order of the finite group \(H^{2} ({\mathcal O}_{L}, \mathbb Z(n)) := \Pi_{p} H^{2}_{\text{ét}}\) (spec \({\mathcal O}_{L} [1/p], \mathbb Z_{p}(n)).\) The proof consists in comparing the \(p\)-adic valuation on both sides for each prime \(p.\) For odd \(p,\) this is an immediate consequence of the Main Conjecture (= Wiles' theorem) of Iwasawa theory. For \(p = 2,\) there are two difficulties: the Main Conjecture is so far only proved for \(\mathbb Q\) (hence the restriction to absolutely Abelian number fields); in order to control contributions from real infinite primes, another ingredient, the so-called ``totally positive cohomology'' (Bruno Kahn), is needed. These difficulties suggest that in such statements on special values of \(L\)-functions as the Lichtenbaum conjectures, \(K\)-theory should be replaced by motivic cohomology.