Double Grothendieck Polynomials for Symplectic and Odd Orthogonal Grassmannians

In [9], Kirillov and Naruse introduced the double Grothendieck polynomials of classical types in order to represent the $K$-theoretic Schubert classes for the corresponding flag varieties. In this paper, we study these functions for the odd orthogonal and symplectic Grassmannians. We first set up a combinatorial model for the ring formally spanned by the stable Schubert classes. This construction is based on previous work for the maximal isotropic Grassmannians ([8]). By utilizing our result in [5] and equivariant localization techniques ([10], [11]), we will explicitly specify the elements in the ring corresponding to the stable Schubert classes.

Fix integers $k$ and $n$ such that $0\leq k<n$. Let ${\mathbb{F}}$ be an algebraically closed field of characteristic zero. We equip the vector space ${\mathbb{F}}^{2n+1}$ with a non-degenerate symmetric bilinear form and ${\mathbb{F}}^{2n}$ with a symplectic form. We denote respectively by ${OG}^{k}(n)$ and $SG^{k}(n)$ the odd-orthogonal and symplectic Grassmannians of $(n-k)$-dimensional isotropic subspaces in ${\mathbb{F}}^{2n+1}$ and ${\mathbb{F}}^{2n}$. We denote both these isotropic Grassmannians by ${{IG}}^{k}(n)$. Each of such spaces can be realized as a homogeneous space $G/P_{k}$ where $P_{k}$ is a parabolic subgroup either of the special orthogonal group $G={{SO}}_{2n+1}({\mathbb{F}})$ or of the symplectic group $G={Sp}_{2n}({\mathbb{F}})$. Let $T_{n}$ be a maximal torus of $G$, $B$ a Borel subgroup of $G$ containing $T_{n}$, and $B_{-}$ a Borel subgroup opposite to $B$, i.e., $B\cap B_{-}=T_{n}.$ For both ${{SO}}_{2n+1}({\mathbb{F}})$ and ${Sp}_{2n}({\mathbb{F}})$, the Weyl group $W_{n}$ is isomorphic to the hyperocthedral group given by $\{-1\}^{n}\rtimes S_{n}$. Let $W_{n}^{(k)}$ be the set of minimal length coset representatives of $W_{n}/W_{n,k}$ where $W_{n,k}$ is the Weyl group of $P_{k}$. There is a natural bijection between $W_{n}^{(k)}$ and the set ${\mathcal{S}}{\mathcal{P}}^{k}(n)$ of $k$-strict partitions contained in the $(n-k,n+k)$ rectangle (cf. Buch–Kresch–Tamvakis [2]). Moveover, in both the odd-orthogonal and symplectic cases, $W_{n}^{(k)}$ is also in bijection with the set of $T_{n}$-fixed points in ${{IG}}^{k}(n)$. For each $\lambda\in{\mathcal{S}}{\mathcal{P}}^{k}(n)$, let $e_{\lambda}$ be the corresponding $T_{n}$-fixed point and define the Schubert variety $\Omega_{\lambda}$ associated to $\lambda$ as the closure of its $B_{-}$-orbit.

We will use the torus-equivariant connective $K$-theory studied by Krishna [11] (cf. Dai–Levine [3]). For any non-singular variety $X$ endowed with an action of an algebraic torus $T_{n}\cong({\mathbb{F}}^{\times})^{n}$, there is a ${\mathbb{Z}}$-graded algebra ${C\!K}_{T_{n}}(X)$ over ${C\!K}_{T_{n}}(pt)$ which can be identified with the graded formal power series ring ${\mathbb{Q}}[\beta][[b_{1},\ldots,b_{n}]]_{\operatorname{gr}}$ where $\deg(b_{i})=1$ and $\deg(\beta)=-1$. It interpolates the $T_{n}$-equivariant chow ring ${C\!H}_{T_{n}}^{*}(X)$ and the $K$-theory $K_{T_{n}}(X)$ of $T_{n}$-equivariant algebraic vector bundles on $X$. By this we mean that it specializes to ${C\!H}_{T_{n}}^{*}(X)$ for $\beta=0$ and to $K_{T_{n}}(X)$ for $\beta=-1$. Since throughout the paper we will work with rational coefficients, here ${C\!H}_{T_{n}}^{*}(X)$ stands for the equivariant chow ring with ${\mathbb{Q}}$ coefficients and $K_{T_{n}}(X)$ denotes the equivariant $K$-theory tensored with ${\mathbb{Q}}$.

In ${C\!K}_{T_{n}}({{IG}}^{k}(n))$, there exists a fundamental class $[\Omega_{\lambda}]_{T_{n}}$ associated to the Schubert variety $\Omega_{\lambda}$, which specializes to the corresponding $T_{n}$-equivariant Schubert class in ${C\!H}^{*}_{T_{n}}({{IG}}^{k}(n))$ and to the class of the structure sheaf of $\Omega_{\lambda}$ in $K_{T_{n}}({{IG}}^{k}(n))$. As we are interested in $[\Omega_{\lambda}]_{T}$, the limit of the classes $[\Omega_{\lambda}]_{T_{n}}$ in the graded projective limit of the rings ${C\!K}_{T_{n}}({{IG}}^{k}(n))$, we introduce ${\mathbb{K}}^{\operatorname{fin}}_{T}({{IG}}^{k})$ a subring which contains all of them (see Definition 4.5). More precisely, if we set

then ${\mathbb{K}}^{\operatorname{fin}}_{T}({{IG}}^{k})$ is the ${\mathcal{R}}_{b}$-module on the formal basis given by $[\Omega_{\lambda}]_{T}$ indexed by the all $k$-strict partitions $\lambda$.

In order to deal with ${\mathbb{K}}^{\operatorname{fin}}_{T}({{IG}}^{k})$ algebraically and combinatorially, we introduce the ring ${\mathcal{K}}_{\infty}$, which governs the equivariant $K$-theoretic Schubert calculus of the full-flag varieties of both types B and C. A cornerstone of the construction of ${\mathcal{K}}_{\infty}$ is the ring ${{G\Gamma}}$ introduced in [8]. It is defined as a subring of ${\mathbb{Q}}[\beta][[x]]_{\operatorname{gr}}$ with variables $x=(x_{1},x_{2},\dots)$ consisting of symmetric graded formal power series with a certain cancellation property (see §3.1). Then one sets

This construction is analogous to the one of the ring ${\mathcal{R}}_{\infty}$ introduced as an algebraic model for the torus equivariant cohomology of full-flag varieties of type B, C, and D in [7]. There are two (called left and right) commuting actions of $W_{\infty}=\bigcup_{n\geq 1}W_{n}$ on ${\mathcal{K}}_{\infty}$. The left action commutes with ${\mathcal{R}}_{a}$ while the right one commutes with ${\mathcal{R}}_{b}$. The ring ${\mathcal{K}}^{(k)}_{\infty}$ is then the invariant subring with respect to the right action of the group $W_{(k)}=\bigcup_{n>k}W_{n,k}$. In particular, the invariant ring ${\mathcal{K}}^{(0)}_{\infty}$ coincides with ${G\Gamma}\otimes_{{\mathbb{Q}}[\beta]}{\mathcal{R}}_{b}$.

The first main result is the following algebraic realization of ${\mathbb{K}}^{\operatorname{fin}}_{T}({{IG}}^{k})$, which we obtain by comparing the GKM descriptions of ${\mathcal{K}}^{(k)}_{\infty}$ and ${C\!K}_{T_{n}}({{IG}}^{k}(n))$.

The second main result is the description of the those elements in ${\mathcal{K}}_{\infty}^{(k)}$ corresponding to $[\Omega_{\lambda}]_{T}$ under the above isomorphisms. For each $k$-strict partition $\lambda$, we define the functions ${}_{k}{G\Theta^{\prime}}_{\!\lambda}$ and ${}_{k}{G\Theta}_{\lambda}$ in ${\mathcal{K}}_{\infty}^{(k)}$ via a Pfaffian formula (Definition 3.13) analogous to the one obtained in [5] for the Schubert classes for ${C\!K}_{T_{n}}({{IG}}^{k}(n))$ (cf.​ Theorem 4.7). This allows us to show that those functions correspond to the limits of the Schubert classes (Theorem 5.10). Note that the Pfaffian formula in [5] describes the Schubert classes in terms of the (equivariantly shifted) special classes which we also identify with certain functions in ${\mathcal{K}}_{\infty}^{(k)}$ (Lemma 5.9). It is worth mentioning that when $\beta=0$, the functions ${}_{k}{G\Theta}_{\lambda}$ specialize to Wilson’s double $\vartheta$-functions [16] which coincide with the type C double Schubert polynomials (see [6]).

For the maximal isotropic Grassmannians in type B and C, the isomorphisms in the above theorem were obtained in [8]. There have also been introduced the so-called ${GP}$- and ${GQ}$-functions in a combinatorially explicit formula, which represent the $K$-theory Schubert classes. Thus by our result we find that, in the case $k=0$, they coincide with ${}_{0}{G\Theta^{\prime}}_{\!\lambda}$ and ${}_{0}{G\Theta}_{\lambda}$ respectively, hence establishing their Pfaffian formula simultaneaously.

Kirillov–Naruse [9] constructed the double Grothendieck polynomials representing the equivariant $K$-theoretic Schubert classes for the full-flag varieties of type B and C. One can prove that our new functions consist of an appropriate subfamily of their polynomials, simultaneously establishing their explicit closed formula in the form of Pfaffians. We also remark that by their construction, those functions are polynomials in $x$, $a$, $b$ variables and $\beta$ with coefficients in ${\mathbb{Z}}_{\geq 0}$, once we choose a positive integer $n$ and set $x_{i}=0$ for all $i>n$.

Given the previous remark related to Kirillov–Naruse’s constructions, let us mention some combinatorial problems that still remain with regard to our ${G\Theta}$- and ${G\Theta^{\prime}}$-functions. When $k=0$, it was proved in [8] that ${GQ}$- and ${GP}$-functions are given in terms of shifted set-valued tableaux, generalizing the classical tableaux formula of Schur $Q$- and $P$-functions (cf. [13, §III.8]). When $\beta=0$ and $b_{i}=0$ for all $i$ but for general $k$, Tamvakis [14] obtained an analogous combinatorial formula for the theta polynomials. Thus it would be an interesting problem to find similar combinatorial formulas for our ${G\Theta}$- and ${G\Theta^{\prime}}$-functions. It is also worth mentioning that, again from Kirillov–Naruse’s construction, we can deduce the fact that ${G\Theta}$-functions (resp. ${G\Theta^{\prime}}$-functions) can be expanded into a linear combination of ${GQ}$-functions (resp. ${GP}$-functions) with coefficients in ${\mathbb{Z}}[\beta][a,b]$. It is an open problem to find an explicit formula for the coefficients of the expansions. Note that when $\beta=0$, such formula has been known in the work of Buch–Kresch–Tamvakis [2, Theorem 4] and Tamvakis–Wilson [15, Corollary 2].

Finally, we note that recently Anderson [1] proved a formula describing a larger family of $K$-theoretic Schubert classes including the case of type D isotropic Grassmannians, generalizing our results for type B and C in [5]. Given this, we expect that it is possible to define the functions analogous to ${G\Theta}$ and ${G\Theta^{\prime}}$ explicitly, extending the main results of this paper to type D.

This paper is organized as follows. In Section 2, we recall the combinatorics related to the Weyl group of type B and C. In Section 3, we define the rings ${G\Gamma}$, ${\mathcal{K}}_{\infty}$ and ${\mathcal{K}}_{\infty}^{(k)}$ and introduce the functions ${}_{k}{G\Theta^{\prime}}_{\!\lambda}$ and ${}_{k}{G\Theta}_{\lambda}$. In Section 4, we recall the definitions of the odd orthogonal and symplectic Grassmannians together with the torus equivariant $K$-theory Schubert classes. In Section 5, we compare the GKM descriptions of ${\mathcal{K}}_{\infty}^{(k)}$ and ${C\!K}_{T_{n}}({OG}^{k}(n))$ and define a map relating those two algebras. Finally we establish our main results, namely the isomorphisms in the above theorem, and the fact that ${}_{k}{G\Theta^{\prime}}_{\!\lambda}$ and ${}_{k}{G\Theta}_{\lambda}$ represent the Schubert classes.

We fix notation on Weyl groups, root systems, and $k$-strict partitions.

We first define the ring ${G\Gamma}$ and introduce its formal basis $\{{GP}_{\lambda}(x)\}_{\lambda\in{\mathcal{S}}{\mathcal{P}}^{0}}$. The ring ${\mathcal{K}}_{\infty}$ of double Grothendieck polynomials is then introduced. On this ring we have a natural (right) action of $W_{\infty}$. For each nonnegative integer $k$, we define the functions ${}_{k}{G\Theta}_{\lambda}$ and ${}_{k}{G\Theta^{\prime}}_{\!\!\lambda}$ associated to all $\lambda\in{\mathcal{S}}{\mathcal{P}}^{k}$ as elements in the invariant subring ${\mathcal{K}}^{(k)}_{\infty}$ with respect to the action of the subgroup $W_{\infty}^{(k)}$ of $W_{\infty}$.