MPS degeneration formula for quiver moduli and refined GW/Kronecker correspondence (Q691549): Difference between revisions

Manshot, Pioline and Sen (MPS) discovered a new formula reducing the Poincaré polynomial of smooth compact moduli spaces of some non-abelian stable quiver representations to the abelian case (thin dimension vectors). This is a big improvement on the previously known Harder-Narasimhan type recursions, and unlike them the MPS formula specializes to a similar formula for the quiver Euler characteristics. The main result of this paper is a motivic generalization of the MPS formula to arbitrary quivers, dimension vectors and stabilities. The motivic formula expresses the quotient stack (by the base change group) of the locus of semistable representations in terms of similar motives for thin dimension vectors of a covering quiver. A dual formula is also proved, it uses a smaller covering quiver at the expense of more general dimension vectors. Via the Gromov-Witten/Kronecker correspondence the Euler characteristic of certain quiver moduli (e.g. for coprime dimension vectors in complete bipartite quivers) can be computed as a Gromov-Witten invariant in a weighted projective plane. Translating their motivic formula into this context the authors find that it reduces to a standard degeneration formula expressing Gromov-Witten invariants in terms of relative ones with tangency conditions along divisors. The proof relies on identifying certain quiver Euler characteristics with counts of tropical curves. On the other hand, since the dimension vectors after applying the MPS formula are thin the moduli spaces corresponding to torus-fixed representations are isolated points, so localization leads to purely combinatorial expressions for the quiver Euler characteristics as sums over trees. Both the tropical curves and the localization data can be built recursively by gluing together smaller pieces, which leads to an explicit correspondence between them in some cases. Two families of such examples are described in the paper.