On the highest multiplicity locus of algebraic varieties and Rees algebras (Q493766)

The paper under review studies some properties of the notion of multiplicity of a point of an algebraic variety. First, it reviews some basic results, including proofs of some facts that often are mentioned without verification. For instance, if \(X\) is an algebraic variety over a field \(k\), or more generally an equidimensional algebraic scheme over \(k\) (i.e., a scheme of finite type over \(k\)), then the function \({\text{mult}}_X:X \to \mathbb N\), where for \(x \in X\) \({\text{mult}}_X(x)\) denotes the multiplicity of the local ring \({\mathcal {O}}_{X,x}\), is upper-semicontinuous. Hence \(X\) (or its underlying topological space) can be stratified, each (locally closed) stratum consisting of points where \({\text{mult}}_X\) is constant. This function is nonincreasing if we blow-up \(X\) with a regular center along which \({\text{mult}}_X\) is constant. He also proves that \({\text{mult}}_X\) and \({\text{mult}}_{X_{\mathrm{red}}}\) yield the same stratification of the common underlying topological space. But the main results have to do with the notion of \textit{representation}. Namely, if \(X\) is an equidimensional algebraic scheme over a field \(k\) we say that \(X\) is \textit{globally representable} if there is a closed immersion \(X \subset V\), where \(V\) is a regular algebraic \(k\)-scheme, and a Rees algebra \({\mathcal G}_X\) on \(V\) such that \(\mathrm{Sing} ({\mathcal G}_X)=\mathrm{Max}({\text{mult}}_X)\) (where \(\mathrm{Sing} ({\mathcal G}_X)\) and \(\mathrm{Max}({\text{mult}}_X)\) denote the singular locus of \({\mathcal G}_X\) and the set of points of maximum multiplicity of \(X\) respectively); moreover it is required that similar equalities remain valid when we take permissible transformations so that the maximum value of the multiplicity does not drop. If the base field is perfect and \(X\) does not have embedded components a result of Villamayor says that each point of \(X\) has an étale neighborhood \(U\) such that \(U\) is globally representable. After reviewing these results, Abad proves his main result. It says that if \(X\) is globally representable, say via a closed immersion \(X \subset V\) and a Rees algebra \({\mathcal G}_V\), then the Rees algebra (over \(X\)) \({\mathcal G}_X := {\mathcal G}_V | X\) is intrinsic to \(X\). That is, if \(X\) is also represented via \(X\subset V'\) and \({\mathcal G}_{V'}\), then \({\mathcal G}_V | X = {\mathcal G}_{V'} | X\) have the same integral closure. In this result, \(X\) may be singular. There is much interesting work on Rees algebras, but most of it is done under the assumption that they are defined over nonsingular schemes. To better evaluate Abad's theorem, it would be interesting to study how, and if, these results extend to the singular case. Finally, the author shows that if \(\phi:X' \to X\) is a finite morphism of algebraic varieties over a perfect field \(k\) of generic degree \(n\), with \(m\), \(m'\) denoting the maximum value of \({\text{mult}}_X\) and \({\text{mult}} _{X'}\) respectively, we always have \(m' \leq nm\). Moreover, if \(m'=nm\) then: (i) \(\phi(\mathrm{Max}({\text{mult}}_{X'})) \subseteq \mathrm{Max}({\text{mult}} _X)\), (ii) if \(F\) is a closed set in \(X'\) then the induced mapping \(F \to \phi(F)\) is a homeomorphism and \(F\) is regular if and only if \(\phi(F)\) is regular, (iii) if the Rees algebras \({\mathcal G}_X\) and \({\mathcal G}_{X'}\) (discussed above) are defined, then \({\phi}^{*}({\mathcal G}_{X}) \subseteq {\mathcal G}_{X'}\). There are some illustrative examples.