Mutations of puzzles and equivariant cohomology of two-step flag varieties (Q2352908)

\textit{A. Knutson} [``A conjectural rule for \(\mathrm{GL}_n\) Schubert calculus'', unpublished manuscript (1999)] conjectured that any Schubert structure constant of the cohomology ring of a partial flag variety \(X =\mathrm{GL}(n)/P\) can be expressed as the number of puzzles that can be created using a list of triangular puzzle pieces with matching side labels. This conjecture was proved in the special case of a Grassmann variety [\textit{A. Knutson} and \textit{T. Tao}, Duke Math. J. 119, No. 2, 221--260 (2003; Zbl 1064.14063)] and [\textit{A. Knutson} et al., J. Am. Math. Soc. 17, No. 1, 19--48 (2004; Zbl 1043.05111)]. Moreover, Knutson and Tao proved an equivariant generalization of the puzzle rule for Grassmannians that makes following Graham's positivity result explicit. The Schubert structure constants of the equivariant cohomology ring \(H_T^*(X,\mathbb{Z})\) are elements of \( H^*_T(\mathrm{pt}, \mathbb{Z})\cong \mathbb{Z}[y_1,\dots,y_n]\). Graham proved the equivariant Schubert structure constants are polynomials with positive coefficients in the differences \(y_{i+1} - y_i\) [\textit{W. Graham}, Duke Math. J. 109, No. 3, 599--614 (2001; Zbl 1069.14055)]. However, Knutson discovered counterexamples to his general conjecture. Later the author et al. [J. Am. Math. Soc. 16, No. 4, 901--915 (2003; Zbl 1063.53090)] proved that the (3 point, genus zero) Gromov-Witten invariants of Grassmannians are special cases of the Schubert structure constants of two-step flag varieties \(Fl(a, b; n)\). It was conjectured in the previous paper and proved in [the author et al., J. Algebr. Comb. 44, No. 4, 973--1007 (2016; Zbl 1356.05154)] that Knutson's conjecture predicts the Schubert structure constants of two-step flag varieties. A different positive combinatorial formula for the cohomological structure constants of two-step flag varieties had earlier been proved by \textit{I. Coskun} [Invent. Math. 176, No. 2, 325--395 (2009; Zbl 1213.14088)]. In [\textit{I. Coskun} and \textit{R. Vakil}, Proc. Symp. Pure Math. 80, 77--124 (2009; Zbl 1184.14080)] was published a conjectured Graham-positive formula for the equivariant Schubert structure constants of any two-step flag variety \(Fl(a, b; n)\), which generalizes both the equivariant puzzle rule for Grassmannians and the cohomological puzzle rule for two-step flag varieties. The main result in the present paper is a proof of the conjectured equivariant puzzle formula for two-step flag varieties. In [the author and \textit{L. C. Mihalcea}, Duke Math. J. 156, No. 3, 501--538 (2011; Zbl 1213.14103)] it was proved that the equivariant Gromov-Witten invariants of Grassmannians are special cases of the equivariant Schubert structure constants of two-step flag varieties, thus generalizing the ``quantum equals classical'' result from [the author et al., J. Am. Math. Soc. 16, No. 4, 901--915 (2003; Zbl 1063.53090)]. The author's puzzle formula therefore specializes to a Littlewood-Richardson rule for the equivariant quantum cohomology of Grassmannians that accords with \textit{L. C. Mihalcea}'s result [Am. J. Math. 128, No. 3, 787--803 (2006; Zbl 1099.14047)] that the equivariant Gromov-Witten invariants satisfy Graham positivity. While different formulas for equivariant Gromov-Witten invariants are known positive formulas have not been available earlier for either the equivariant cohomology of two-step flag varieties or the equivariant quantum cohomology of Grassmannians. The main combinatorial construction in our paper is an algorithm called mutation of puzzles, which is analogous to Schützenberger's jeu de taquin algorithm for semistandard Young tableaux. Recall that the jeu de taquin algorithm operates on Young tableaux that contain a flaw in the form of an empty box, and works by making natural changes to move the empty box to a different location in the tableau. The mutation algorithm similarly operates on flawed puzzles.