Piecewise algebraic geometry, \(k\)-equivalence and motives (Q1943957)

Two smooth, connected, proper varieties \(X\) and \(Y\) over an algebraically closed field \(k\) of characteristic 0 are \(K\)-\textit{equivalent} if there exists a smooth, connected, proper variety \(Z\) over \(k\) and birational morphisms \(X \overset{f}\longleftarrow Z \overset{g}\longrightarrow Y\) such that the relative canonical divisors \(K_{Z/X}\) and \(K_{Z/Y}\) coincide. A theorem of M. Kontsevich asserts that, in this case, \(X\) and \(Y\) have the same Hodge numbers. It is natural to ask whether \(X\) and \(Y\) have isomorphic Chow motives or, at least, whether they have isomorphic Chow groups. In the paper under review, the authors propose a (conjectural) strategy for attacking this question by using the Grothendieck ring of varieties and piecewise algebraic geometry. Let \({\mathbb Z}[\text{Var}_k]\) be the free abelian group generated by the isomorphism classes \(\{X\}\) of \(k\)-varieties \(X\). The Grothendieck ring \(K_0(\text{Var}_k)\) is the quotient of \({\mathbb Z}[\text{Var}_k]\) by the subgroup generated by the elements of the form \(\{X\} - \{Z\} - \{X\setminus Z\}\), \(Z\) closed subvariety of \(X\), with the multiplicative structure defined by \([X]\cdot [Y] := [X\times_kY]\) (\([X]\) denoting the class of the variety \(X\) in \(K_0(\text{Var}_k)\)). The unit of \(K_0(\text{Var}_k)\) is \([\text{Spec}\, k]\). Let \(\mathbb L\) denote \([{\mathbb A}_k^1]\) and let \({\mathcal M}_k := K_0(\text{Var}_k)[{\mathbb L}^{-1}]\). A theorem of M. Kontsevich asserts that if \(X\) and \(Y\) are \(K\)-equivalent varieties as at the beginning of this review then \([X]\) and \([Y]\) have the same image into the completion \({\widehat {\mathcal M}}_k\) of \({\mathcal M}_k\) with respect to a certain filtration defined in terms of the dimension of the varieties. On the other hand, two \(k\)-schemes of finite type \(X\) and \(Y\) are \textit{piecewise isomorphic} if they admit partitions \((X_i)_{0\leq i \leq n}\), \((Y_i)_{0\leq i \leq n}\) such that \((X_i)_{\text{red}} \simeq (Y_i)_{\text{red}}\), \(i = 0, \dots , n\). Two piecewise isomorphic \(k\)-varieties have the same class in \(K_0(\text{Var}_k)\). \textit{M. Larsen} and \textit{V. A. Lunts} [Moscow Math. J. 3, No. 1, 85--95 (2003; Zbl 1056.14015)] asked the difficult question whether the converse of this statement is true. The strategy proposed by the authors of the paper under review for the question formulated at the beginning consists in splitting the question into two parts: 1) If \(X\) and \(Y\) are \(K\)-equivalent does it follow that \([X] = [Y]\) in \(K_0(\text{Var}_k)\)?; 2) If \([X] = [Y]\) in \(K_0(\text{Var}_k)\) do \(X\) and \(Y\) have isomorphic Chow motives? Step 1 splits into two substeps: 1a) Is the morphism \({\mathcal M}_k \rightarrow {\widehat {\mathcal M}}_k\) injective?; 1b) Is the morphism \(K_0(\text{Var}_k) \rightarrow {\mathcal M}_k\) injective? \textit{J. Sebag} [Proc. Am. Math. Soc. 138, No. 4, 1231--1242 (2010; Zbl 1192.14012)] related 1b) to the question of Larsen and Lunts. Using results of \textit{U. Jannsen} [Invent. Math. 197, No. 3, 447--452 (1992; Zbl 0762.14003)] and of \textit{S.-I. Kimura} [Math. Ann. 331, No. 1, 173--201 (2005; Zbl 1067.14006)], the authors notice that the question formulated in Step 2 would have a positive answer if the conjecture asserting that every Chow motive is ``finite dimensional'' were true. The main contribution of the authors of the paper under review consists in answering some of the above questions for ``small dimensional'' varieties. They show that if two smooth, projective surfaces have the same class in \(K_0(\text{Var}_k)\) then they have isomorphic Chow motives. They also show that if two connected, smooth, projective varieties have the same class in \(K_0(\text{Var}_k)\) then they have isomorphic Néron-Severi groups. Finally, the authors show that if \(X\) and \(Y\) are two \(K\)-equivalent, smooth, projective varieties admitting a \(K\)-exceptional locus \((C_X,\, C_Y)\) with \(\dim C_X \leq 2\) then \(X\) and \(Y\) are piecewise isomorphic. As a corollary, two Calabi-Yau varieties of dimension \(\leq 4\) are piecewise isomorphic if and only if they are birationally equivalent.