Coefficient and elimination algebras in resolution of singularities (Q408456): Difference between revisions

A Rees algebra over a ring \(R\) is a finitely generated graded \(R\)-subalgebra \(\bigoplus _i I_iT^i\) of the polynomial ring \(R[T]= \bigoplus _i R T^i\), \(i \in \mathbb N = \{ 0, 1, \dots \}\). A Rees algebra over an algebraic variety \(W\) is a graded sheaf \(\mathcal I\) of \({\mathcal O}_W\)-algebras such that for an affine cover \(\{U_j ={\text{Spec}}(R_j)\}\) of \(W\) we have that the restriction \({\mathcal I}|U_j\) is a Rees algebra over \(R_j\), for all \(j\). If moreover \(I_{j+1} \subset I_j\) for all \(j\), the authors talk about an \({\mathbb N}\)- Rees algebra. Essentially as above, but substituting the index set \(\mathbb N\) by \({\mathbb Q}_0\) (the non-negative integers) they define the notion of \({\mathbb Q}\)-Rees algebra. Kawanoue introduced a similar concept, that of \textit{idealistic filtration} (IF). An idealistic filtration over a variety \(W\) is defined by an indexed family of \({\mathcal O}_W\)-ideals \(\{J_a\}\), \(a \in {\mathbb R}_0\) (the non-negative real numbers), such that \(J_0={\mathcal O}_W\), \((J_a J_b \subset J{a+b}\) and \(J_b \subset J_a\) if \(a \leq b\). When the IF is locally generated by a finite collection of elements \(f_1, \ldots, f_r\), where \(f_i \in I_{a_i}\) and \(a_i\) is rational for all \(i\) we talk about a \textit{rationally and finitely generated idealistic filtration} (RFIF). These filtrations might be a fundamental tool to resolve singularities in positive characteristic, see \textit{H. Kawanoue} and \textit{K. Matsuki} [Publ. Res. Inst. Math. Sci. 46, No. 2, 359--422 (2010; Zbl 1235.14017)]. The concepts mentioned above are closely related. For instance, it is possible to associate to a \(\mathbb Q\)-Rees \(\mathcal I\) algebra a RFIF \(\mathcal I '\), and for any RFIF \(\mathcal F\) there is a \(\mathbb Q\)-Rees algebra \(\mathcal I\) such that \(\mathcal F\) and \(\mathcal I '\) have the same integral closure. Similarly, up to integral closure we may compare Rees algebras and \(\mathbb N\)-Rees algebras as well as \(\mathbb N\)-Rees algebras and \(\mathbb Q\)-Rees algebras. Since the important properties of these objects depend on the integral closure only, we may say, for instance, that \(\mathbb Q\)-Rees algebras and RFIF's carry the same information. The authors extend to \(\mathbb Q\)-Rees algebras many notions and properties previously studied in the context of Rees algebras. In particular, they study singular loci as well as coefficient and elimination algebras. Given a \(\mathbb Q\)-Rees algebra on a smooth variety \(W\) (satisfying suitable properties) and a smooth hypersurface \(Z\) in \(W\), one may define an induced \(\mathbb Q\)-Rees algebra on \(Z\), called the coefficient algebra. Similarly given a smooth morphism from \(W\) to a smooth variety \(W'\) (of dimension \(\dim W -1\)), following ideals of Villamayor one may introduce a \(\mathbb Q\)-Rees algebra on \(W'\), called the elimination algebra. They show that in characteristic zero both concepts are essentially equivalent, but in characteristic \(p\) neither, as presented so far, looks too promising. Indeed, the right type of hypersurfaces (those of ``maximal contact'') are not always available in positive characteristic, while (as indicated in an example in the article under review) the elimination algebra done not have the desired properties). Finally, they describe an algorithm to resolve singularities of \(\mathbb Q\)-Rees algebras in characteristic zero. Namely, a notion of permissible transformation is introduced, the algorithm shows how to choose centers so that after successively transforming with these centers we obtain a \(\mathbb{Q}\)-Rees algebra whose singular locus is empty. From this result an algorithm to principalize a sheaf of ideals in as suitable way (the so-called log-principalization) can be obtained.