How many hypersurfaces does it take to cut out a Segre class? (Q342856): Difference between revisions

It is known that, in order to compute the multiplicity of a variety \(Y\) along a subscheme \(Z\) (at one of its irreducible components), the ideal defining the subvariety can be replaced by an ideal generated by a precise number of equations (in fact the codimension). On the other hand this multiplicity can also be computed by means of Segre classes. Then it is natural to ask if the number of equations of the Segre clasess can also be controlled. The main result of the paper (see Theorem 1.1 for a precise statement) states that the Segre class of \(Z\) in \(Y\) can be cut out by \(\dim(Y)+1\) hypersurfaces in a linear system cutting out \(Z\) and by one less in a neighborhood of \(Z\). Moreover the theorem is refined to describe the \(k\)-dimensional components of the Segre class. The proof of this result is based on a study of Segre classes of zero-schemes of sections of vector bundles, concretely (see Section 2), it is a consequence of the fact that the Segre class is essentially preserved by taking suitable quotients of the vector bundle. Another application of this study is a Bertini type theorem. In fact, the set-theoretically equality between the singular locus of a general hyperplane section of a hypersurface (of a smooth projective variety) and the hyperplane section of its singular locus lifts to an equality of Segre classes (although not to the level of schemes), see Theorem 1.3 for a precise statement.