Actions of Cremona groups on CAT\((0)\) cube complexes (Q2073273)

Groups of birational transformations of algebraic varieties are huge groups that are usually complicated to study. In dimension two, some breakthroughs have been made in recent years with the help of the action of the group of birational transformations of a surface on some hyperbolic space. Birational transformations of surfaces can be thus understood with the help of hyperbolic geometry. In this spirit of hyperbolic geometry, the first author attempted in her previous work, e.g.\ [\textit{A. Lonjou}, Publ. Mat., Barc. 63, No. 2, 521--599 (2019; Zbl 1447.20013)], to construct new spaces with hyperbolic geometry which are more combinatorial and which the group of birational transformations of a surface still acts on. The paper under review is about the construction of such a space, not only for surfaces, but for varieties of any dimension. A birational map between two algebraic varieties is called a \(l\)-pseudoisomorphism if it is an isomorphism in codimension \(l\). For any \(l\) and any algebraic variety \(X\), the authors constructed a simply connected \(CAT(0)\) cube complex \(C^l(X)\) on which \(\operatorname{PseudoAut}^l(X)\) acts. Roughly speaking, vertices of \(C^l(X)\) are (equivalence classes of) couples \((Y,\phi)\) where \(\phi\) is a \(l\)-pseudoisomorphism from \(Y\) to \(X\) and there is an edge between \((Y,\phi)\) and \((Z,\psi)\) if \((Y,\phi)\) is the restriction of \((Z,\psi)\) to the complement of an irreducible suvariety of codimension \(l\). Acutally, as the authors mentioned, the construction is soft and can lead to different but similar spaces. Especially in dimension two, the author gives a variation of the construction which leads to more precise answers to some specific questions concerning surfaces. A subgroup of \(\operatorname{PseudoAut}^l(X)\) is called (pseudo)regularizable if it is conjugate to a subgroup of \(\operatorname{PseudoAut}^{l+1}(Y)\) for some \(Y\) by a \(l\)-pseudoisomorphism. It means basically that this subgroup fixes a vertex in \(C^l(X)\). Therefore the construction in this paper is particularly suitable for the study of when a birational transformation or a group of birational transformations is regularizable. For example one of the main theorems in this paper is that groups with property FW are always regularizable. Of course the cube complexe \(C^l(X)\) can also be used for other purposes. For example one expects that a general birational transformation has almost cyclic centralizer and the authors prove by using their construction that this is indeed the case if the birational transformation is not pseudo-regularizable and does not preserve any rational fibration.