The \(\tau \)-invariant and elimination (Q615864)

A \textit{Rees algebra} over an algebraic variety \(V\) (smooth over a field \(k\)) is a finitely generated subalgebra of \({\mathcal O}_{V}[T]\) (with \(T\) an indeterminate). They are important in the theory of resolution of singularities. Indeed, one may define the singular locus of a Rees algebra, and if Rees algebras can be resolved in a suitable ``canonical'' or ``algorithmic'' way (by means of a sequence monoidal transformations with well determined smooth centers), there is a formalism leading to embedded resolution of a subvariety \(X\) of a smooth ambient one \(V\). There are indications, specially thanks to the efforts of O. Villamayor, that, working over perfect fields of positive characteristic, Rees algebras are natural objects to consider in resolution problems. As usual, to resolve Rees algebras at a crucial step one would like to use induction on the dimension of the base variety \(V\). In positive characteristic, as indicated by Giraud, Hironaka and others, it seems hopeless to use smooth hypersurfaces on \(V\) to define an auxiliary inductive algebra to implement the inductive procedure. Instead, Villamayor considers (locally defined) suitable ``generic'' projections \(\beta:V \to V'\), where \(\dim V' < \dim V\), and introduces a new Rees algebra \({\mathcal R}_{ {\mathcal G}, \beta}\) over \(V'\), called the \textit{elimination algebra} (defined using techniques inspired by classical elimination theory). Trying to compare \(\mathcal G\) and its elimination algebra, the so-called \(\tau\)-invariant, introduced by Hironaka long ago, seems to be very important. If \(\mathcal G\) is a Rees algebra over \(V\) and \(x \in V\) is a closed point, \({\tau}_{{\mathcal G},x}\) is defined as the codimension of a certain linear subspace (related to the tangent cone \(C_{{\mathcal G},x}\)) of the tangent space \(T_{V,x}\). In the reviewed paper, Benito studies properties of the \(\tau\)-invariant. Working over a perfect field \(k\) of any characteristic, her main results are: (1) A new proof of the fact that if \(\mathcal G\) and \({\mathcal G}'\) are Rees algebras over the variety \(V\) and both have the same integral closure (in \({\mathcal O}_V[T]\)), then for every closed point \(x \in V\) we have \({\tau}_{{\mathcal G},x}= {\tau}_{{\mathcal G '},x} \) (a result previously obtained, with other methods, by H. Kawanoue). (2) If \(\beta: V \to V'\) is a generic projection of algebraic varieties over \(k\), with \(\dim V' = \dim V -1\), \(x\) is a closed point of \(V\) and \(x'=\beta(x)\) and, moreover, \(\mathcal G\) is \textit{differential} (i.e., closed under application of differential operators) then \({\tau} _ { {\mathcal R}_{{\mathcal G},\beta} , x' } = {\tau}_{{\mathcal G},x} -1\). (3) If, in (2), we drop the assumption that \(\mathcal G\) is differential, then we have an inequality \({\tau} _ { {\mathcal R}_{{\mathcal G},\beta} , x' } \leq {\tau}_{{\mathcal G},x} -1\). Much of the the paper is devoted to a review of basic material on Rees and elimination algebras, necessary to give an understandable presentation of the main results.