Three formulas for CSM classes of open quiver loci

Moriah Elkin \addressmark 1 · 2025

Read Three formulas for CSM classes of open quiver loci by Moriah Elkin \addressmark 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-17723.txt.

Introduction

Moriah Elkin works in pure mathematics on spaces tied to quivers, which are diagrams of points and arrows used to encode maps between vector spaces. Within one such space she defines open quiver loci by placing strict rank demands on those maps.

Elkin gives three formulas for a refined class attached to these loci, one based on geometry and two based on counting patterns. The last uses chained pipe dreams, a picture method adapted from earlier work. The classes refine older quiver polynomials, so the formulas add finer detail to a known algebraic story.

Catalog record

CREATORMoriah Elkin \addressmark 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE108 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-17723

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Three formulas for CSM classes of open quiver loci Moriah Elkin \addressmark 1 Abstract In the space of equioriented type A A quiver representations, we define subvarieties called "open quiver loci" by placing strict rank conditions on the maps within representations. The closures of these subvarieties are the quiver loci, whose equivariant cohomology classes are the quiver polynomials of Buch and Fulton. We present one geometric formula and two combinatorial formulas that compute equivariant Chern–Schwartz–MacPherson (CSM) classes of open quiver loci; these classes refine the data of the quiver polynomials. The second combinatorial formula is in terms of "chained generic pipe dreams," which modify the pipe dreams of Bergeron and Billey to more strongly resemble the lacing diagrams of Abeasis and Del Fra. We also present two new formulas for quiver polynomials; these are streamlined versions of known formulas due to Knutson, Miller, and Shimozono, in the sense that they contain fewer terms. keywords Schubert calculus, pipe dreams, quiver loci, equivariant cohomology 1 Introduction Historically, much attention has been paid to the "degeneracy locus" associated to a given map of manifolds: the set of points where the induced map on tangent bundles is below a certain rank. Thom gave a formula for the cohomology class Poincaré dual to a degeneracy locus, and Porteous [ 12 ] extended the theory to maps of arbitrary vector bundles. Buch and Fulton [ 6 ] gave a formula in the case of n -fold compositions of vector bundle maps over a manifold, where the degeneracy loci are given by forcing each composition of any number of the maps to have a fixed maximal rank. The corresponding cohomology classes are polynomials in the Chern roots of the given vector bundles, and Knutson, Miller, and Shimozono [ 7 ] gave four combinatorial formulas to compute these polynomials. We present refined versions of two of these formulas in § 3 . The language of equivariant cohomology enabled Knutson, Miller, and Shimozono to work with maps of vector spaces – type A equioriented quiver representations – rather than vector bundles. In this setting, degeneracy loci translate to quiver loci . Their equivariant cohomology classes, with respect to the group action that changes bases in each of the vector spaces, are called quiver polynomials . It is this "universal" perspective that we take here. Given a fixed sequence of vector spaces, the quiver loci are closed subvarieties of the corresponding space of quiver representations, and they are given by weak rank conditions: by restricting the maximal allowed rank of each composition of maps. So far, little attention has been paid to the finer objects given by strict rank conditions: by requiring that each composition of maps have some particular rank. We term these subvarieties open quiver loci . They are the orbits of the change-of-basis group action mentioned earlier, and their closures are exactly the quiver loci. The technology of equivariant Chern–Schwartz–MacPherson (CSM) classes ( [ 10 ] , [ 11 ] ) enables us to associate elements of equivariant cohomology to the open quiver loci in a natural way. These classes include the data of the quiver polynomials, as well as the topological Euler characteristic of the open quiver loci. The first main result we present is Theorem 12 in § 4 , which gives a formula for the CSM classes of open quiver loci in terms of (known) CSM classes of Schubert cells in the complete flag variety. We then translate this result into a second formula in terms of the pipe dreams of Bergeron and Billey [ 4 ] , Proposition 13 . To improve on the pipe dream formula, we define chained generic pipe dreams (CGPDs) , analogues of pipe dreams which closely resemble the lacing diagrams of Abeasis and Del Fra [ 1 ] . Our second main result is Theorem 18 , which applies a formula of Su [ 14 ] to express CSM classes of open quiver loci as sums of weights of CGPDs. We are currently working to extend this formula to K -theory using the analogous formula present in [ 13 ] . Finally, from Theorem 18 we deduce Corollary 20 , a CGPD formula for quiver polynomials. 2 Preliminaries 2.1 Quiver loci and quiver polynomials Consider the equioriented type A_{n+1} quiver \mathrel{\overset{\makebox[0.0pt]{\mbox{\tiny 0}}}{\bullet}}\to\mathrel{\overset{\makebox[0.0pt]{\mbox{\tiny 1}}}{\bullet}}\to\cdots\to\mathrel{\overset{\makebox[0.0pt]{\mbox{\tiny$n$}}}{\bullet}}. Fix a sequence of vector spaces V=(V_{0},V_{1},\dots,V_{n}) over \mathds{C} with \dim(V_{i})=r_{i} , and define the vector space Hom\coloneqq\textrm{Hom}(V_{0},V_{1})\times\cdots\times\textrm{Hom}(V_{n-1},V_{n})=\{V_{0}\xrightarrow{\phi_{1}}V_{1}\xrightarrow{\phi_{2}}\cdots\xrightarrow{\phi_{n}}V_{n}\}, of quiver representations of V . Given an array of nonnegative integers \mathbf{r}=(r_{ij})_{0\leq i\leq j\leq n} , the quiver locus \Omega_{\mathbf{r}} is the subset of quiver representations whose composite maps V_{i}\to V_{j} have rank at most r_{ij} for all i<j . The open quiver locus \Omega_{\mathbf{r}}^{\circ} is the set of quiver representations such that each map V_{i}\to V_{j} has rank exactly r_{ij} . 1 1 1 We use ”open quiver locus” in the sense of ”open Richardson variety,” to indicate that we are not taking the closure of the rank conditions. Only one of the open quiver loci is truly open in Hom , though all are locally closed. Fix bases for V_{0},\dots,V_{n} , and identify elements of Hom with tuples of matrices. We will consider the action on Hom of the group GL=GL(V_{0})\times\cdots\times GL(V_{n}) via change-of-basis in each vector space V_{i} : for each i , GL(V_{i}) acts by inverse multiplication on the right of \textrm{Hom}(V_{i-1},V_{i}) when i>0 , and by multiplication on the left of \textrm{Hom}(V_{i},V_{i+1}) when i<n . The goal of § 3 is to provide (novel) formulas for quiver polynomials , which we define to be the equivariant cohomology classes [\Omega_{\mathbf{r}}]\in H^{*}_{GL}(Hom) of the quiver loci. 2 2 2 Throughout, we implicitly use Poincaré duality to identify the equivariant Borel-Moore homology classes defined by GL -invariant subvarieties with the corresponding classes in equivariant cohomology. Two quiver representations are isomorphic if they are in the same GL -orbit, and these orbits are exactly the open quiver loci \Omega_{\mathbf{r}}^{\circ} . Combining this fact with a standard result in algebraic quiver theory yields the following proposition: Definition/Proposition 1 . Where 0\leq p\leq q\leq n , define a quiver representation I_{p,q}:~\mathrel{\overset{\scriptscriptstyle 0}{0}}\to\cdots\to\mathrel{\overset{\scriptscriptstyle p-1}{0}}\to\mathrel{\overset{\scriptscriptstyle p}{\raisebox{-2.5pt}{\includegraphics[scale]{C.png}}}}\xrightarrow{\sim}\cdots\xrightarrow{\sim}\mathrel{\overset{\scriptscriptstyle q}{\raisebox{-2.5pt}{\includegraphics[scale]{C.png}}}}\to\mathrel{\overset{\scriptscriptstyle q+1}{0}}\to\cdots\to\mathrel{\overset{\scriptscriptstyle n}{0}} to have copies of the field \mathds{C} at indices p,\dots,q with identity maps between them, and zero vector spaces elsewhere. Now the I_{p,q} are indecomposable, and every open quiver locus \Omega^{\circ}_{\mathbf{r}}\subseteq\textrm{Hom} contains exactly one representation of the form \bigoplus(I_{p,q})^{s_{pq}} for some \mathbf{s}=(s_{pq})_{0\leq p\leq q\leq n} . We call \mathbf{s} the lace array associated to \mathbf{r} . We can represent any lace array pictorially using a graph in the plane called a lacing diagram , defined in [ 1 ] and developed in [ 7 ] ; 3 3 3 There are two notable differences between our lacing diagrams and those discussed in [ 7 ] . First, we consider lacing diagrams only up to permutation within rows, where in [ 7 ] the order of the basis vectors is fixed at each vertex. Second, our lacing diagrams are read top-to-bottom rather than left-to-right. Theorem 18 in § 4 will take this perspective. The vertices of a lacing diagram are arranged in n+1 rows with r_{i} left-justified vertices in row i , labeled top-to-bottom. All edges must occur between two adjacent

Read the complete text