Witt equivalence of semilocal Dedekind domains in global fields (Q2479800)

Harrison's criterion says that the Witt rings of two fields \(K\) and \(L\) of characteristic not \(2\) are isomorphic iff there exists an isomorphism of square-class groups \(t:\dot{K}/\dot{K}^2\to \dot{L}/\dot{L}^2\) such that \(t(-1)=-1\) and such that the diagonal binary form \(\langle a,b\rangle\) (\(a,b\in \dot{K}\)) represents \(1\) over \(K\) iff \(\langle ta,tb\rangle\) represents \(1\) over \(L\). The aim of the present paper is to generalize this criterion to Witt rings of nondegenerate bilinear forms over semilocal Dedekind domains, in particular to such domains inside global fields. Let \(A\), \(B\) be semilocal Dedekind domains with fraction fields \(K\) and \(L\) (of characteristic not \(2\)), respectively. Let \(\Omega (A)\) denote the set of nonzero prime ideals of \(A\) and \(E(A)=\{ x\in \dot{K}\mid \text{ord}_{{\mathfrak p}}x\equiv 0\bmod 2\quad\forall {\mathfrak p}\in\Omega (A)\}\). The first main result of the first section states that if there is an isomorphism of Witt rings \(\phi:W(A)\to W(B)\), then there exists a map \(t\) with the same properties as in Harrison's criterion above, except that the square-class groups are replaced by \(E(A)/\dot{K}^2\) and \(E(B)/\dot{L}^2\), and with an additional technical property on the behavior of Pfister forms under the map \(\phi\). If the isomorphism of Witt rings sends classes of unary forms to classes of unary forms, then the underlying rings \(A\) and \(B\) are said to be strongly Witt equivalent. The authors show that \(A\) and \(B\) are strongly Witt equivalent iff there exists an isomorphism \(t:E(A)/\dot{K}^2\to E(B)/\dot{L}^2\) with \(t(-1)=-1\) and \(\langle a_1,a_2,a_3,a_4\rangle\cong\langle b_1,b_2,b_3,b_4\rangle\) (\(a_i,b_i\in E(A)\)) over \(A\) iff \(\langle ta_1,ta_2,ta_3,ta_4\rangle\cong\langle tb_1,tb_2,tb_3,tb_4\rangle\) over \(B\). In the second section, the authors turn their attention to semilocal Dedekind domains in global fields and the Hilbert-symbol equivalence (a notion introduced by \textit{R. Perlis, K. Szymiczek, P. E. Conner} and \textit{R. Litherland} in [Contemp. Math. 155, 365--387 (1994; Zbl 0807.11024)] and called reciprocity equivalence there). Let \({\mathcal P}\) be a finite set of places in a global field \(K\), and let \({\mathcal O}_{{\mathcal P}}=\bigcap_{{\mathfrak p}\in {\mathcal P}}{\mathcal O}_{{\mathfrak p}}\). \(\Omega_{{\mathcal P}}(K)\) denotes all noncomplex places of \(K\) except those that are nondyadic and in \({\mathcal P}\). If \({\mathcal Q}\) is another finite set of places in another global field \(L\), then \({\mathcal O}_{{\mathcal P}}\) and \({\mathcal O}_{{\mathcal Q}}\) are said to be Hilbert-symbol equivalent if there exists an isomorphism \(t:E({\mathcal O}_{\mathcal P})/\dot{K}^2\to E({\mathcal O}_{\mathcal Q})/\dot{L}^2\) and a bijection \(T:\Omega_{{\mathcal P}}(K)\to \Omega_{{\mathcal Q}}(L)\) such that \(t(-1)=-1\) and \((a,b)_{{\mathfrak p}}=(ta,tb)_{T{\mathfrak p}}\) for all \(a,b\in E({\mathcal O}_{\mathcal P})\) and all \({\mathfrak p}\in \Omega_{{\mathcal P}}(K)\) (here, \((a,b)_{{\mathfrak p}}\) denotes the local Hilbert symbol taking as values \(\pm 1\)). It is shown that for rings \({\mathcal O}_{\mathcal P}\) and \({\mathcal O}_{\mathcal Q}\) as above, Witt equivalence and strong Witt equivalence are equivalent, and these are in turn equivalent to having a Hilbert-symbol equivalence mapping the real places of \(K\) to the real places of \(L\). In the fourth section, it is shown that all these statements are equivalent to \(K\) and \(L\) satisfying three properties: \(-1\in \dot{K}^2\) iff \(-1\in\dot{L}^2\), the numbers of real places in \(K\) and in \(L\) are equal, and a third slightly technical property involving the dyadic places of \(K\) and \(L\).