Motivic cohomology over Dedekind rings (Q706118)

A long standing conjecture, the so-called Bloch-Kato conjecture, states that, for any field \(F\) and any integer \(m\) prime with the characteristic of \(F\), the symbol map from Milnor \(K\)-theory to Galois cohomology \(K^M_n(F)/m\to H^n_{\text{ét}} (F,\mu_m^{\otimes n})\) is surjective. For \(n=2\) the result is a celebrated theorem by \textit{A.S. Merkuriev} and \textit{A.A. Suslin} [Math. USSR, Izv. 21, 307--340 (1983; Zbl 0525.18008)]. The conjecture is known if \(m\) is a power of 2; Voevodsky has recently announced a proof for any \(m\). In this paper the author shows how to deduce from the Bloch-Kato conjecture a series of important results about the motivic cohomology groups of a smooth scheme \(X\) over a Dedekind ring \(R\). The following is a list of some of these results; here \(\mathbb{Z}(n)\) denotes Bloch's cycle complex and \(\varepsilon:X_{\text{èt}}\to X_{\text{Zar}}\) the change of site map. 1. (Beilinson-Lichtenbaum conjecture) There is an isomorphism \[ \mathbb{Z}(n)_{\text{Zar}} @>\simeq>> \tau_{n+1}R \varepsilon_*\mathbb{Z}(n)_{\text{ét}}, \] where \(\tau_r\) means good truncation of the complex of sheaves. 2. If \(m\) is invertible in \(R\) then there is a quasi-isomorphism of complexes of étale sheaves \(\mathbb{Z}/m(n)_{\text{ét}} \simeq\mu_m^{\otimes n} [0]\). 3. (Gersten resolution) For any \(m\) there is an exact sequence of Zariski sheaves on \(X\); \[ 0\to{\mathcal H}^s \bigl(\mathbb{Z}/m(n)_{\text{Zar}}\bigr)\to\coprod_{x\in X^0}(i_x)_*(H^s \bigl(k (x), \mathbb{Z}/m (n)\bigr)\to\coprod_{x\in X^1}(i_x)_*(H^{s-1}\bigl(k(x), \mathbb{Z}/m(n-1)\bigr)\to \cdots \] where \(X^i\) are the points of codimension \(i\) on \(X\). The paper also contains the proof of the existence of a Gersten resolution for the sheaves \({\mathcal H}^s(\mathbb{Z} (n)_{\text{Zar}})\) under the assumption that, for every discrete valuation ring \(V\) essentially of finite type over Spec\,\(R\), with quotient field \(K\), the map \(H^s(V,\mathbb{Z}(n)\to H^s(X,\mathbb{Z}(n)\) is injective.