The Chow-Witt ring (Q2372901)

Introduction: Let \(A\) be a commutative Noetherian ring of Krull dimension \(n\) and \(P\) a projective \(A\)-module of rank \(d\). One can ask the following question: does \(P\) admit a free factor of rank one? Serre proved a long time ago that the answer is always positive when \(d> n\). So in fact the first interesting case is when \(P\) is projective of rank equal to the dimension of \(A\). Suppose now that \(X\) is an integral smooth scheme over a field \(k\) of characteristic not 2. To deal with the above question, \textit{J. Barge} and \textit{F Morel} [C. R. Acad. Sci., Paris, Sér 5, Math. 330, 287--290 (2000; Zbl 1017.14001)] introduced the Chow-Witt groups \(\widetilde{\text{CH}}^j(X)\) of \(X\) and associated to each vector bundle \(E\) of rank \(n\) an Euler class \(\widetilde c_n(E)\) in \(\widetilde{\text{CH}}^n(X)\). It was proved recently that if \(X= \text{Spec}(A)\) we have \(\widetilde c_n(P)= 0\) if and only if \(P\simeq Q\oplus A\). It is therefore important to provide more tools, such as a ring structure and a pull-back for regular embeddings, to compute the Chow-Witt groups and the Euler classes. To define \(\widetilde{\text{CH}}^p(X)\) consider the fibre product of the complex in Milnor \(K\)-theory \[ 0\to K^M_p(k(X))\to \bigoplus_{x_1\in X^{(1)}} K^M_{p-1} (k(x_1))\to\cdots \to \bigoplus_{x_n\in X^{(n)}} K^M_{p-n} (k(x_n))\to 0 \] and the Gersten-Witt complex restricted to the fundamental ideals \[ 0\to I^p(k(X))\to \bigoplus_{x_1\in X^{(1)}} I^{p-1}({\mathcal O}_{X,x_1})\to\cdots \to\bigoplus_{x_n\in X^{(n)}} I^{p-n}({\mathcal O}_{X,x_n})\to 0 \] over the quotient complex \[ 0\to I^p/I^{p+1}(k(X))\to\cdots\to \bigoplus_{x_n\in X^{(n)}} I^{p-n}/I^{p+1-n}({\mathcal O}_{X, x_n})\to 0. \] The group \(\widetilde{\text{CH}}^p(X)\) is defined as the \(p\)th cohomology group of this fibre product. Roughly speaking, an element of \(\widetilde{\text{CH}}^p(X)\) is the class of a sum of varieties of codimension \(p\) with a quadratic form defined on each variety. We oviously have a map \(\widetilde{\text{CH}}^p(X)\to\text{CH}^p(X)\). Using the functoriality of the two complexes we see that the Chow-Witt groups satisfy good functorial properties (see [\textit{J. Fasel}, Groupes de Chow-Witt, www.math.ethZ.chi/\(\sim\)jfasel/grupes.chow-witt:pdj]). Using this functorial behaviour, it is possible to produce a good intersection theory. This is what we do in this paper using the classical strategy. First we define an exterior product \[ \widetilde{\text{CH}}^j(X)\times \widetilde{\text{CH}}^i(Y)\to \widetilde{\text{CH}}^{i+j}(X\times Y) \] and then a Gysin-like homomorphism \(i^!: \widetilde{\text{CH}}^d(X)\to \widetilde{\text{CH}}^d(Y)\) associated to a closed embedding \(i: Y\to X\) of smooth schemes. The product is then defined as the composition \[ \widetilde{\text{CH}}^j(X)\times \widetilde{\text{CH}}^i(X)\to \widetilde{\text{CH}}^{i+j}(X\times X)@>\Delta^!>>\widetilde{\text{CH}}^{i+j}(X), \] where \(\Delta: X\to X\times X\) is the diagonal embedding. To define the exterior product, we first note that \textit{M. Rost} already defined an exterior product on the homology of the complex in Milnor \(K\)-theory [Doc. Math., J. DMV 1, 319--393 (1996; Zbl 0864.14002)]. Thus it is enough to define an exterior product on the homology of the Gersten-Witt complex and show that both exterior products coincide over the quotient complex. We use the usual product on derived Witt groups and show that this product passes to homology using the Leibnitz rule proved by \textit{P. Balmer} [Compos. Math. 141, 1374--1404 (2005; Zbl 1087.18008)]. The definition of the Gysin-like map is done by follwwing the ideas of \textit{M. Rost} [loc. cit.]. It uses the deformation to the normal cone to modifiy any closed embedding to a nicer closed embedding and uses also the long exact sequence associated to a triple \((Z,X,U)\) where \(Z\) is a closed subset of \(X\) and \(U= X\setminus Z\). The product that we obtain has the meaning of intersecting varieties with quadratic forms defined on them. It is therefore not a surprise that the natural map \(\widetilde{\text{CH}}^{\text{tot}}(X)\to \text{CH}^{\text{tot}}(X)\) turns out to be a ring homomorphism. There is however a surprise: the product that we obtain is a priori neither commutative nor anticommutative. This comes from the fact that the product of triangulated Grothendieck-Witt groups \(\text{GW}^i\times \text{GW}^j\to \text{GW}^{i+j}\) does not satisfy any commutativitY property.