Residual intersection theory with reducible schemes (Q1879096)

The author proves a formula, given as Theorem 5.2, which under some conditions expresses top Chern classes of vector bundles on the vanishing locus \(V(s)\) of a section of the bundle. This formula applies to the case when \(V(s)\) is the union of local complete intersections giving the individual contribution of each component and their mutual intersections. The article concludes with applications to the enumeration of rational curves in complete intersections in projective space. Specifically, given the vector bundle \(E\) of rank \(n\) on a scheme \(X\) of dimension \(n\). If \(E\) is globally generated, then a general section \(s\) of \(E\) vanishes in \(\int_X c_n(E)\) points. This gives an easy way to compute the top Chern number of \(E\). If \(E\) is not globally generated or if there is just one section of \(E\) with a higher dimensional vanishing locus \(V (s)\), then this Chern number on the vanishing locus is computed in \textit{W. Fulton}'s classical book [``Intersection Theory'', Ergebnisse der Mathematik und ihrer Grenzgebiete, 3. Folge, Bd. 2. (Berlin etc.: Springer-Verlag) (1984; Zbl 0541.14005)]. This has a particularly nice form if \(V (s)\) is a local complete intersection (see Proposition 9.1.1 in [loc. cit.]. The author treats the case when \(V(s) = \bigcup _{i=1}^M Z_i\) is a union of local complete intersections, which intersect mutually in a ``nice'' way. The main result is Theorem 5.2 where the contribution of each \(Z_{i_1}\cap Z_{i_2}\cap \ldots Z_{i_k}\) to the top Chern number of \(E\) is computed. In Section 6 the author outlines how this result may be used to count rational curves on complete intersections. The problem is reduced to computations on moduli spaces of rational curves in \({\mathbb P}^n\) and to show that the corresponding vanishing cycle satisfies his assumption 4.1. Finally the author gives two examples showing how his theorem may be used to compute some Chern numbers. In the first example he computes the number of lines on a smooth cubic surface in \({\mathbb P}^3\). Here a point, which is the intersection of two irreducible components, gives a contribution of -6. In the second example he computes the number of lines on a general quintic threefold in \({\mathbb P}^4\) by linearizing the problem. To compute these numbers Kontsevich's idea of using a \({\mathbb C}\)*-action and Bott's fix point formula gives a more efficient way to compute these numbers along smooth subvarieties. However, when there is no \({\mathbb C}\)*-action the present technique might be useful to reduce the dimension. The review is based on the author's introduction and summary.