Geometric mitosis (Q727707)

Positive presentations of Schubert cycles such as classical Schubert polynomials play a key role in the Schubert calculus. Ideas of toric geometry and theory of Newton (or moment) polytopes motivated search for positive presentations with a more convex geometric flavor. For instance, Schubert cycles on the complete flag variety for \(\mathrm{GL}_n\) were identified by various means with unions of faces of Gelfand-Zetlin polytopes. In the present paper, the author develops an algorithm for representing Schubert cycles by faces of convex polytopes in the case of complete flag varieties for arbitrary reductive groups. In this paper,first she defines geometric mitosis operations \(M_1,\ldots, M_r\) on faces of polytope \(P_{\lambda}\) as convex geometric counterparts of Demazure operators \(D_1, \ldots , D_r\). The definition of mitosis operations is elementary, and its main ingredient is mitosis on parallelepipeds introduced in [\textit{V. A. Kirichenko} et al., Russ. Math. Surv. 67, No. 4, 685--719 (2012); translation from Usp. Mat. Nauk 67, No. 4, 89--128 (2012; Zbl 1258.14055), Section 6]. Mitosis assigns to every face a collection (possibly empty) of faces of dimension one greater. Mitosis on parallelepipeds can be viewed as a convex geometric realization of the mitosis of Knutson-Miller restricted to two consecutive rows of pipe dreams. Then she uses mitosis on parallelepipeds as a building block for mitosis on more general polytopes \(P_{\lambda}\) (called parapolytopes) that admit \(r\) different fibrations by parallelepipeds. The combinatorics of mitosis depends significantly on the combinatorics of \(P_{\lambda}\). For instance, for Gelfand-Zetlin polytopes she gets mitosis on usual pipe dreams, and for a polytope associated with the cone of adapted strings in type \(C\) she gets different combinatorial objects that we call skew pipe dreams. For \(\mathrm{Sp}_4\) and a Newton-Okounkov polytope of the symplectic flag variety, the algorithm yields a new combinatorial rule that extends to \(\mathrm{Sp}_{2n}\).