The motivic nearby fiber and degeneration of stable rationality (Q2315186)

In the article under review are investigated the specialization properties of stable rationality in regular families over a field of characteristic 0. For such a field \(k\), define \(R = k[[t]], K = k((t)), K(\infty) = \bigcup_n k((t^{\frac{1}{n}}))\). Let \(\hat{\mu}\) be the projective limit of the groups of roots of unity. For a proper scheme \(X/K\), an \(R\)-model of \(X\) is flat and proper scheme \(\mathcal{X}/R\) with an isomorphism \(\mathcal{X}_K \simeq X\). For smooth \(X\), \(\mathcal{X}\) is called strict normal crossings (snc) if it is regular, and the special fiber is snc divisor, so it can be represented as \(\mathcal{X}_k = \sum_i N_{i}E_i\). \(\mathbf{K}(\mathrm{Var}_K)\) is the Grothendieck ring of varieties over \(K\), and \(\mathbf{K}^{\hat{\mu}}(\mathrm{Var}_K)\) is the Grothendieck ring of varieties over \(K\) with \(\hat{\mu}\)-action. It is generated by the classes of \(\hat{\mu}\)-equivariant isomorphisms of schemes of finite type with a good \(\hat{\mu}\)-action. The first main result relies on the theorems of \textit{F. Bittner} [Compos. Math. 140, No. 4, 1011--1032 (2004; Zbl 1086.14016)] and \textit{M. Larsen} and \textit{V. A. Lunts} [Mosc. Math. J. 3, No. 1, 85--95 (2003; Zbl 1056.14015)], and it claims that stable rationality specializes in regular families whose fibers have at worst rational double points as singularities. Aiming to control stable rationality in families, two specialization morphisms are defined: the motivic volume \(\mathrm{Vol}_K: \mathbf{K}^G(\mathrm{Var}_K) \rightarrow \mathbf{K}^{\hat{\mu}}(\mathrm{Var}_k) \), inspired by [\textit{J. Denef} and \textit{F. Loeser}, Prog. Math. 201, 327--348 (2001; Zbl 1079.14003); \textit{E. Hrushovski} and \textit{D. Kazhdan}, Prog. Math. 253, 261--405 (2006; Zbl 1136.03025)], and the motivic reduction \(MR: \mathbf{K}(\mathrm{Var}_K) \rightarrow \mathbf{K}(Var_k)\). For both are obtained formulas for given snc model of proper and smooth scheme. As application is shown that for proper and smooth \(K(\infty)\)-schemes \(X, Y\), which are stably birational, \(\mathrm{Vol}(X) \equiv\mathrm{Vol}(Y)\bmod \mathbb{L}\). From it follows that having \(\mathcal{X}, \mathcal{Y}\) proper smooth \(R\)-schemes with \(\mathcal{X}_{K(\infty)}\), \(\mathcal{Y}_{K(\infty)} \) stably birational, then \(\mathcal{X}_k \) is stably birational to \(\mathcal{Y}_k \). This is generalized then to strict normal crossings degenerations. As a corollary, over uncountable algebraically closed \(k\) and \(S\) of finite type, given proper and smooth morphism \(f: X \rightarrow S\) whose very general closed fiber is stably rational then every closed fiber of \(f\) is stably rational. This is generalized further to regular \(R\)-models \(\mathcal{X}\) whose special fiber is reduced with at worst ordinary double points as singularities. It is proved that \(\mathcal{X}\) is then \(\mathbb{L}\)-faithful, that is, \([\mathcal{X}_k] \equiv \mathrm{Vol}(X_{K(\infty)})\pmod \mathbb{L}\). The second main result in the article is about proper flat families \(f:\mathcal{X} \rightarrow C\) of connected smooth schemes over a curve \(C\), with the geometric generic fiber of \(f\) stably rational. It asserts that any geometric fiber with at most ordinary double points has stably rational component. As an application is shown that a very general smooth quartic threefolds are not stably rational (see [\textit{J.-L. Colliot-Thélène} and \textit{A. Pirutka}, Ann. Sci. Éc. Norm. Supér. (4) 49, No. 2, 371--397 (2016; Zbl 1371.14028)]). In the appendix are proved the existence of the motivic volume with a formula in terms of smooth logarithmic models, generalizing the formula in the main text, and a refined version of the existence of the motivic reduction. The arguments used in the article under review are strengthened further in [\textit{M. Kontsevich} and \textit{Yu. Tschinkel}, ``Specialization of birational types'', Preprint, \url{arXiv:1708.05699}]. There are constructed specialization homomorphisms analogous to motivic volume and motivic reduction, with the Grothendieck ring replaced by the Burnside ring of \(k\), which is a new invariant. As a result is obtained an important generalization of the first main result above from stable rationality to rationality. Nevertheless the method in the present paper remains of independent interest.