On the Gersten-Witt complex of an Azumaya algebra with involution (Q2142471): Difference between revisions

Let~\(R\) be a regular ring with fraction field~\(K\) of characteristic not~\(2\), and \((A,\tau)\) an Azumaya algebra with involution of the first- or second kind over~\(R\). Associated with this data is a so called \(\varepsilon\)-hermitian Gersten-Witt complex (\(\varepsilon\in\{\pm 1\}\)) \[ 0\longrightarrow\mathrm{W}_{\varepsilon}(A,\tau)\longrightarrow\bigoplus\limits_{\mathrm{ht} P=0}\mathrm{W}_{\varepsilon}(A(P),\tau(P))\longrightarrow \bigoplus\limits_{\mathrm{ht} P=1}\mathrm{W}_{\varepsilon}(A(P),\tau(P))\longrightarrow\ldots\, ,\tag{1} \] where \(W_{\varepsilon}(A,\tau)\) (respectively \(\mathrm{W}_{\varepsilon}(A(P),\tau(P))\)) denotes the \(\varepsilon\)-hermitian Witt group of \((A,\tau)\) (respectively of \(k(P)\otimes (A,\tau)\), \(k(P)\) the residue field of the prime ideal \(P\)). Adapting the construction of \textit{P. Balmer} and \textit{C. Walter} [Ann. Sci. Éc. Norm. Supér. (4) 35, No. 1, 127--152 (2002; Zbl 1012.19003)] in the symmetric case, which uses Balmer's derived Witt theory, such a complex has been constructed by the reviewer. Recently the author of the paper under review together with \textit{E. Bayer-Fluckiger} and \textit{R. Parimala} gave a new (less abstract?) construction of such a complex [``On the Grothendieck-Serre conjecture for classical groups'', J. Lond. Math. Soc. (to appear)], which avoids derived Witt theory. The aim of the article under review is to show that this new complex is isomorphic to the one, which arises from derived Witt theory. In the final section of the paper the author shows then that the hermitian Gersten Witt complex (1) is exact if \(R\) is a semilocal regular integral domain of Krull dimension \(\leq 3\), the index of \(A\) is \(\leq 2\), and \(\tau\) is either orthogonal or symplectic.