Grothendieck classes of quiver cycles as iterated residues (Q2255328)

Suppose some complex veector bundles are given over a manifold \(M\), together with some (sufficiently generic) vector bundle maps among them. Over a point in \(M\) then one has a quiver: some vector spaces with linear maps among them. Quivers may degenerate -- e.g. dimensions of intersections of images, kernels may drop, and more general degenerations also happen. A quiver degeneracy locus is the set a points in \(M\) over which the quiver degenerates in a particular way. It is known that the fundamental class of quiver degeneracy loci can be expressed by a universal polynomial (the quiver polynomial) in the charactersitic classes of the bundles involved, both in cohomology and in \(K\)-theory. The paper under review presents such universal polynomials in \(K\)-theory if the quiver is of Dynkin type. The formula presented is a non-conventional generating function (named Iterated Residue generating function, pioneered by Bérczi-Szenes, Kazarian, Feher-Rimányi). Advantages of the presented formula for quiver loci include that stabilization properties are explicit, and the expansion in terms of Grothendieck polynomials is well motivated. The paper ends with interesting comments comparing \(K\)-theory and cohomology Iterated Residue formulas, as well as remarks on Buch's positivity conjecture on the coefficients of the Grothendieck polynomial expansion.