1. Introduction
1.1 Real–symmetric correspondence
 Let 
 $G_{\mathbb R}$
 be a real form of a connected complex reductive group
$G_{\mathbb R}$
 be a real form of a connected complex reductive group 
 $G$
. Let
$G$
. Let 
 $X=G/K$
 be the associated symmetric variety under Cartan’s bijection, where
$X=G/K$
 be the associated symmetric variety under Cartan’s bijection, where 
 $K$
 is the complexification of a maximal compact subgroup
$K$
 is the complexification of a maximal compact subgroup 
 $K_{\mathbb R}\subset G_{\mathbb R}$
.
$K_{\mathbb R}\subset G_{\mathbb R}$
.
 A fundamental feature of the representation theory of the real group 
 $G_{\mathbb R}$
 is that many results of an analytic nature have equivalent purely algebraic formulations in terms of the corresponding symmetric variety
$G_{\mathbb R}$
 is that many results of an analytic nature have equivalent purely algebraic formulations in terms of the corresponding symmetric variety 
 $X$
. We will call this broad phenomenon the real–symmetric correspondence. It allows one to use algebraic tools on the symmetric side to study questions on the real side and, conversely, to use analytic tools on the real side to study questions on the symmetric side. Famous examples include Harish-Chandra’s reformulation of admissible representations of real groups in terms of
$X$
. We will call this broad phenomenon the real–symmetric correspondence. It allows one to use algebraic tools on the symmetric side to study questions on the real side and, conversely, to use analytic tools on the real side to study questions on the symmetric side. Famous examples include Harish-Chandra’s reformulation of admissible representations of real groups in terms of 
 $({\mathfrak g},K)$
-modules, the Kostant–Sekiguchi correspondence between real and symmetric nilpotent orbits, and the Matsuki correspondence between
$({\mathfrak g},K)$
-modules, the Kostant–Sekiguchi correspondence between real and symmetric nilpotent orbits, and the Matsuki correspondence between 
 $G_{\mathbb R}$
- and
$G_{\mathbb R}$
- and 
 $K$
-orbits on the flag manifold of
$K$
-orbits on the flag manifold of 
 $G$
.
$G$
.
 In [Reference Chen and NadlerCN18], the first and third authors established a real–symmetric correspondence relating the dg derived category of spherical constructible sheaves on the real affine Grassmannian 
 ${\textrm{Gr}}_{G_{\mathbb R}}$
 of
${\textrm{Gr}}_{G_{\mathbb R}}$
 of 
 $G_{\mathbb R}$
 and the dg derived category of spherical constructible sheaves on the loop space
$G_{\mathbb R}$
 and the dg derived category of spherical constructible sheaves on the loop space 
 $\mathfrak{L}X$
 of
$\mathfrak{L}X$
 of 
 $X$
. We are interested in applying this real–symmetric correspondence to study questions in the real and relative geometric Langlands programs.
$X$
. We are interested in applying this real–symmetric correspondence to study questions in the real and relative geometric Langlands programs.
 In the present paper, we consider the question of a geometric Satake equivalence for real groups and symmetric varieties. We focus on the case where the real group is the quaternionic group 
 $G_{\mathbb R}={\textrm{GL}}_n({\mathbb H})$
 with associated symmetric variety
$G_{\mathbb R}={\textrm{GL}}_n({\mathbb H})$
 with associated symmetric variety 
 $X={\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
. We prove the derived geometric Satake equivalence for
$X={\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
. We prove the derived geometric Satake equivalence for 
 ${\textrm{GL}}_n({\mathbb H})$
 relating the dg constructible derived category of the quaternionic affine Grassmannian with the dg coherent derived category of a quotient stack associated to the Gaitsgory–Nadler dual group
${\textrm{GL}}_n({\mathbb H})$
 relating the dg constructible derived category of the quaternionic affine Grassmannian with the dg coherent derived category of a quotient stack associated to the Gaitsgory–Nadler dual group 
 $\check G_X$
 of
$\check G_X$
 of 
 $X$
 (which is
$X$
 (which is 
 $\check G_X={\textrm{GL}}_n$
 in this case). Via the real–symmetric correspondence, we obtain a derived geometric Satake equivalence for the symmetric variety
$\check G_X={\textrm{GL}}_n$
 in this case). Via the real–symmetric correspondence, we obtain a derived geometric Satake equivalence for the symmetric variety 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
. As an application, we compute the stalks of the IC-complexes for spherical orbit closures in the quaternionic affine Grassmannian and the loop space of
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
. As an application, we compute the stalks of the IC-complexes for spherical orbit closures in the quaternionic affine Grassmannian and the loop space of 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
. We show that the stalks are given by the Kostka–Foulkes polynomials for
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
. We show that the stalks are given by the Kostka–Foulkes polynomials for 
 ${\textrm{GL}}_n$
 but with all degrees doubled.
${\textrm{GL}}_n$
 but with all degrees doubled.
We explain how these equivalences fit into the general framework of a geometric Langlands correspondence for real groups, due to Ben-Zvi and the third author, and of the relative Langlands duality conjecture, due to Ben-Zvi, Sakellaridis, and Venkatesh.
 From the point of view of real groups, the quaternionic group 
 ${\textrm{GL}}_n({\mathbb H})$
 offers in some sense the simplest possible geometry: just as complex Grassmannians are simpler than real Grassmannians (Schubert cells are
${\textrm{GL}}_n({\mathbb H})$
 offers in some sense the simplest possible geometry: just as complex Grassmannians are simpler than real Grassmannians (Schubert cells are 
 $2d$
 versus
$2d$
 versus 
 $d$
 real-dimensional), quaternionic Grassmannians are simpler still than complex Grassmannians (Schubert cells are
$d$
 real-dimensional), quaternionic Grassmannians are simpler still than complex Grassmannians (Schubert cells are 
 $4d$
 versus
$4d$
 versus 
 $2d$
 real-dimensional). On the other hand, the geometry of the symmetric variety
$2d$
 real-dimensional). On the other hand, the geometry of the symmetric variety 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 is more complicated than that of
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 is more complicated than that of 
 ${\textrm{GL}}_{2n}$
. The real–symmetric correspondence allows us to use the simpler quaternionic geometry of
${\textrm{GL}}_{2n}$
. The real–symmetric correspondence allows us to use the simpler quaternionic geometry of 
 ${\textrm{GL}}_n({\mathbb H})$
 to answer questions about the more complicated geometry of
${\textrm{GL}}_n({\mathbb H})$
 to answer questions about the more complicated geometry of 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
.
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
.
We now describe the paper in more detail. We work throughout over the complex numbers, except where we specifically consider real forms. All topological sheaves are with respect to the classical topology with complex coefficients.
1.2 Reminder on derived Satake for 
 ${\textrm{GL}}_{2n}$
${\textrm{GL}}_{2n}$
 Let 
 $\mathfrak{L}{\textrm{GL}}_{2n}$
 and
$\mathfrak{L}{\textrm{GL}}_{2n}$
 and 
 $\mathfrak{L}^+{\textrm{GL}}_{2n}$
 be the Laurent loop group and Taylor arc group of
$\mathfrak{L}^+{\textrm{GL}}_{2n}$
 be the Laurent loop group and Taylor arc group of 
 ${\textrm{GL}}_{2n}$
. The affine Grassmannian
${\textrm{GL}}_{2n}$
. The affine Grassmannian 
 ${\textrm{Gr}}_{2n}=\mathfrak{L}{\textrm{GL}}_{2n}/\mathfrak{L}^+{\textrm{GL}}_{2n}$
 for
${\textrm{Gr}}_{2n}=\mathfrak{L}{\textrm{GL}}_{2n}/\mathfrak{L}^+{\textrm{GL}}_{2n}$
 for 
 ${\textrm{GL}}_{2n}$
 is the ind-variety classifying
${\textrm{GL}}_{2n}$
 is the ind-variety classifying 
 ${\mathbb C}[[t]]$
-lattices in
${\mathbb C}[[t]]$
-lattices in 
 ${\mathbb C}((t))^{n}$
. The arc group
${\mathbb C}((t))^{n}$
. The arc group 
 $\mathfrak{L}^+{\textrm{GL}}_{2n}$
 acts naturally on
$\mathfrak{L}^+{\textrm{GL}}_{2n}$
 acts naturally on 
 ${\textrm{Gr}}_{2n}$
, and we denote by
${\textrm{Gr}}_{2n}$
, and we denote by 
 $D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})$
 the monoidal dg-category of
$D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})$
 the monoidal dg-category of 
 $\mathfrak{L}^+{\textrm{GL}}_{2n}$
-equivariant constructible complexes on
$\mathfrak{L}^+{\textrm{GL}}_{2n}$
-equivariant constructible complexes on 
 ${\textrm{Gr}}_{2n}$
 with monoidal structure given by convolution.
${\textrm{Gr}}_{2n}$
 with monoidal structure given by convolution.
 Let 
 $\mathfrak{gl}_{2n}$
 be the Lie algebra of
$\mathfrak{gl}_{2n}$
 be the Lie algebra of 
 ${\textrm{GL}}_{2n}$
. Write
${\textrm{GL}}_{2n}$
. Write 
 $\textrm{Sym}(\mathfrak{gl}_{2n}[-2])$
 for the symmetric algebra of
$\textrm{Sym}(\mathfrak{gl}_{2n}[-2])$
 for the symmetric algebra of 
 $\mathfrak{gl}_{2n}[-2]$
 viewed as a dg-algebra with trivial differential. The group
$\mathfrak{gl}_{2n}[-2]$
 viewed as a dg-algebra with trivial differential. The group 
 ${\textrm{GL}}_{2n}$
 acts on
${\textrm{GL}}_{2n}$
 acts on 
 $\textrm{Sym}(\mathfrak{gl}_{2n}[-2])$
 via the adjoint action, and we denote by
$\textrm{Sym}(\mathfrak{gl}_{2n}[-2])$
 via the adjoint action, and we denote by 
 $D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
 the monoidal dg-category of perfect
$D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
 the monoidal dg-category of perfect 
 ${\textrm{GL}}_{2n}$
-equivariant dg-modules over
${\textrm{GL}}_{2n}$
-equivariant dg-modules over 
 $\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
 with monoidal structure given by the (derived) tensor product of dg-modules.
$\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
 with monoidal structure given by the (derived) tensor product of dg-modules.
One of the versions of the derived Satake equivalence in [Reference Bezrukavnikov and FinkelbergBF08] says that there is an equivalence of monoidal dg-categories
 \begin{equation*}\Psi :D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})\simeq D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))\end{equation*}
\begin{equation*}\Psi :D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})\simeq D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))\end{equation*}
extending the geometric Satake equivalence 
 $\textrm{Perv}({\textrm{Gr}}_{2n})\simeq \textrm{Rep}({\textrm{GL}}_{2n})$
, where
$\textrm{Perv}({\textrm{Gr}}_{2n})\simeq \textrm{Rep}({\textrm{GL}}_{2n})$
, where 
 $\textrm{Perv}({\textrm{Gr}}_{2n})\subset D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})$
 is the subcategory of
$\textrm{Perv}({\textrm{Gr}}_{2n})\subset D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})$
 is the subcategory of 
 $\mathfrak{L}^+{\textrm{GL}}_{2n}$
-equivariant perverse sheaves on
$\mathfrak{L}^+{\textrm{GL}}_{2n}$
-equivariant perverse sheaves on 
 ${\textrm{Gr}}_{2n}$
 and
${\textrm{Gr}}_{2n}$
 and 
 $\textrm{Rep}({\textrm{GL}}_{2n})\subset D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
 is the subcategory of representations of
$\textrm{Rep}({\textrm{GL}}_{2n})\subset D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
 is the subcategory of representations of 
 ${\textrm{GL}}_{2n}$
.Footnote 
1
${\textrm{GL}}_{2n}$
.Footnote 
1
1.3 Derived Satake for the quaternionic group 
 ${\textrm{GL}}_n({\mathbb H})$
${\textrm{GL}}_n({\mathbb H})$
 Let 
 ${\textrm{GL}}_{n}({\mathbb H})\subset {\textrm{GL}}_{2n}$
 be the real form given by the rank
${\textrm{GL}}_{n}({\mathbb H})\subset {\textrm{GL}}_{2n}$
 be the real form given by the rank 
 $n$
 quaternionic group. Let
$n$
 quaternionic group. Let 
 $\mathfrak{L}{\textrm{GL}}_n({\mathbb H})$
 and
$\mathfrak{L}{\textrm{GL}}_n({\mathbb H})$
 and 
 $\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 be the real Laurent loop group and Taylor arc group for
$\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 be the real Laurent loop group and Taylor arc group for 
 ${\textrm{GL}}_n({\mathbb H})$
. By the real affine Grassmannian for the quaternionic group
${\textrm{GL}}_n({\mathbb H})$
. By the real affine Grassmannian for the quaternionic group 
 ${\textrm{GL}}_n({\mathbb H})$
, we will mean the ind semi-analytic variety
${\textrm{GL}}_n({\mathbb H})$
, we will mean the ind semi-analytic variety 
 ${\textrm{Gr}}_{n,{\mathbb H}}=\mathfrak{L}{\textrm{GL}}_n({\mathbb H})/\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 classifying
${\textrm{Gr}}_{n,{\mathbb H}}=\mathfrak{L}{\textrm{GL}}_n({\mathbb H})/\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 classifying 
 ${\mathbb H}[[t]]$
-lattices in
${\mathbb H}[[t]]$
-lattices in 
 ${\mathbb H}((t))^{n}$
.Footnote 
2
${\mathbb H}((t))^{n}$
.Footnote 
2
 The real arc group 
 $\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 acts naturally on
$\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 acts naturally on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
, and we denote by
${\textrm{Gr}}_{n,{\mathbb H}}$
, and we denote by 
 $D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 the monoidal dg-category of
$D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 the monoidal dg-category of 
 $\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})$
-equivariant constructible complexes on
$\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})$
-equivariant constructible complexes on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 with monoidal structure given by convolution. The
${\textrm{Gr}}_{n,{\mathbb H}}$
 with monoidal structure given by convolution. The 
 $\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
-orbits on
$\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
-orbits on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 are all even real-dimensional (in fact,
${\textrm{Gr}}_{n,{\mathbb H}}$
 are all even real-dimensional (in fact, 
 $4d$
 real-dimensional; see § 4.4), and hence middle perversity makes sense. We denote by
$4d$
 real-dimensional; see § 4.4), and hence middle perversity makes sense. We denote by 
 $\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 the category of
$\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 the category of 
 $\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})$
-equivariant perverse sheaves on
$\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})$
-equivariant perverse sheaves on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
. In [Reference NadlerNad05], the third author established a real geometric Satake equivalence, giving an equivalence of monoidal abelian categories
${\textrm{Gr}}_{n,{\mathbb H}}$
. In [Reference NadlerNad05], the third author established a real geometric Satake equivalence, giving an equivalence of monoidal abelian categories 
 $\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Rep}({\textrm{GL}}_n)$
 in the case at hand.
$\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Rep}({\textrm{GL}}_n)$
 in the case at hand.
 The first main result of this paper is the following equivalence of monoidal dg derived categories, to be called derived Satake for 
 ${\textrm{GL}}_{n}({\mathbb H})$
.
${\textrm{GL}}_{n}({\mathbb H})$
.
Theorem 1.1 (See Theorem5.5). There is an equivalence of monoidal dg-categories
 \begin{equation*}\Psi _{\mathbb H}:D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^{{\textrm{GL}}_n}_{\textrm{perf}}(\textrm{Sym}(\mathfrak {gl}_n[-4]))\end{equation*}
\begin{equation*}\Psi _{\mathbb H}:D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^{{\textrm{GL}}_n}_{\textrm{perf}}(\textrm{Sym}(\mathfrak {gl}_n[-4]))\end{equation*}
extending the real geometric Satake equivalence 
 $\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Rep}({\textrm{GL}}_n)$
.
$\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Rep}({\textrm{GL}}_n)$
.
A key ingredient in the proof of Theorem1.1 (as in the proof of the abelian quaternionic geometric Satake) is a nearby cycles functor
 \begin{equation} \mathrm R:D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})\to D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \end{equation}
\begin{equation} \mathrm R:D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})\to D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \end{equation}
associated to a real form of the Beilinson–Drinfeld Grassmannian with generic fibers isomorphic to the complex affine Grassimannian 
 ${\textrm{Gr}}_{2n}$
 and special fiber isomorphic to the quaternionic affine Grassimannian
${\textrm{Gr}}_{2n}$
 and special fiber isomorphic to the quaternionic affine Grassimannian 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 (see § 4.5). Note that, unlike in the complex algebraic setting, the nearby cycles functor
${\textrm{Gr}}_{n,{\mathbb H}}$
 (see § 4.5). Note that, unlike in the complex algebraic setting, the nearby cycles functor 
 $\mathrm R$
 is not
$\mathrm R$
 is not 
 $t$
-exact: it maps perverse sheaves to direct sums of shifts of perverse sheaves (see Proposition 4.5). As a corollary of the proof, we obtain the following spectral description of the nearby cycles functor.
$t$
-exact: it maps perverse sheaves to direct sums of shifts of perverse sheaves (see Proposition 4.5). As a corollary of the proof, we obtain the following spectral description of the nearby cycles functor.
Consider the graded scheme
 \begin{equation*}\widetilde {\mathfrak{gl}}_{2n}=\bigg\{\bigg(\begin{array}{c@{\quad}c} A[0]& B[-2]\\[2pt] C[2]& D[0] \end{array}\bigg) \bigg| A,B,C,D\in \mathfrak{gl}_n\bigg\}.\end{equation*}
\begin{equation*}\widetilde {\mathfrak{gl}}_{2n}=\bigg\{\bigg(\begin{array}{c@{\quad}c} A[0]& B[-2]\\[2pt] C[2]& D[0] \end{array}\bigg) \bigg| A,B,C,D\in \mathfrak{gl}_n\bigg\}.\end{equation*}
We have the natural embedding of (even graded) schemes
 \begin{equation} \tau :\mathfrak{gl}_n[4]\to \widetilde {\mathfrak{gl}}_{2n}[2], \quad \tau (C[4]) =\bigg(\begin{array}{c@{\quad}c} 0& \textrm{Id}_n\\ C[4]& 0 \end{array}\bigg), \end{equation}
\begin{equation} \tau :\mathfrak{gl}_n[4]\to \widetilde {\mathfrak{gl}}_{2n}[2], \quad \tau (C[4]) =\bigg(\begin{array}{c@{\quad}c} 0& \textrm{Id}_n\\ C[4]& 0 \end{array}\bigg), \end{equation}
where 
 $\textrm{Id}_n$
 is the rank
$\textrm{Id}_n$
 is the rank 
 $n$
 identity matrix. Note that the map
$n$
 identity matrix. Note that the map 
 $\tau$
 is
$\tau$
 is 
 ${\textrm{GL}}_n$
 adjoint-equivariant, where
${\textrm{GL}}_n$
 adjoint-equivariant, where 
 ${\textrm{GL}}_n$
 acts on
${\textrm{GL}}_n$
 acts on 
 $\widetilde {\mathfrak{gl}}_{2n}[2]$
 via the diagonal embedding
$\widetilde {\mathfrak{gl}}_{2n}[2]$
 via the diagonal embedding 
 ${\textrm{GL}}_n\to {\textrm{GL}}_{2n}$
. Hence pullback along
${\textrm{GL}}_n\to {\textrm{GL}}_{2n}$
. Hence pullback along 
 $\tau$
 provides a functor
$\tau$
 provides a functor
 \begin{equation*}\tau ^*:D^{{\textrm{GL}}_{n}}_{\textrm{perf}}\big(\textrm{Sym}(\widetilde {\mathfrak{gl}}_{2n}[-2])\big)\to D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n} [-4])).\end{equation*}
\begin{equation*}\tau ^*:D^{{\textrm{GL}}_{n}}_{\textrm{perf}}\big(\textrm{Sym}(\widetilde {\mathfrak{gl}}_{2n}[-2])\big)\to D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n} [-4])).\end{equation*}
Here we view the rings of functions on 
 $\mathfrak{gl}_n[4]$
 and
$\mathfrak{gl}_n[4]$
 and 
 $\widetilde {\mathfrak{gl}}_{2n}[2]$
 as the dg symmetric algebras
$\widetilde {\mathfrak{gl}}_{2n}[2]$
 as the dg symmetric algebras 
 $\textrm{Sym}(\mathfrak{gl}_{2n}[-4])$
 and
$\textrm{Sym}(\mathfrak{gl}_{2n}[-4])$
 and 
 $\textrm{Sym}\big(\widetilde {\mathfrak{gl}}_{2n}[-2]\big)$
 with trivial differential. Introduce the functor
$\textrm{Sym}\big(\widetilde {\mathfrak{gl}}_{2n}[-2]\big)$
 with trivial differential. Introduce the functor
 \begin{equation*}\Phi :D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))\stackrel {}\longrightarrow D^{{\textrm{GL}}_{n}}_{\textrm{perf}}\big(\textrm{Sym}(\widetilde {\mathfrak{gl}}_{2n}[-2])\big)\stackrel {\tau ^*}\longrightarrow D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4]))\end{equation*}
\begin{equation*}\Phi :D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))\stackrel {}\longrightarrow D^{{\textrm{GL}}_{n}}_{\textrm{perf}}\big(\textrm{Sym}(\widetilde {\mathfrak{gl}}_{2n}[-2])\big)\stackrel {\tau ^*}\longrightarrow D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4]))\end{equation*}
where the first functor is the sheared forgetful functor associated to the 
 ${\mathbb G}_m$
-action on
${\mathbb G}_m$
-action on 
 $\mathfrak{gl}_{2n}[-2]$
 via the co-character
$\mathfrak{gl}_{2n}[-2]$
 via the co-character 
 $2\rho _L:{\mathbb G}_m\to {\textrm{GL}}_{2n}$
 (see (5.17)). Here
$2\rho _L:{\mathbb G}_m\to {\textrm{GL}}_{2n}$
 (see (5.17)). Here 
 $L$
 is the complexification of the Levi subgroup of the minimal parabolic subgroup of
$L$
 is the complexification of the Levi subgroup of the minimal parabolic subgroup of 
 ${\textrm{GL}}_{n}({\mathbb H})$
.
${\textrm{GL}}_{n}({\mathbb H})$
.
Theorem 1.2 (See Theorem5.7). The following square is naturally commutative

 
where 
 $\Psi$
 and
$\Psi$
 and 
 $\Psi _{\mathbb H}$
 are the complex and quaternionic derived Satake equivalences, respectively.
$\Psi _{\mathbb H}$
 are the complex and quaternionic derived Satake equivalences, respectively.
Later, in § 1.6, we will discuss how Theorem1.2 fits into the general framework of duality for Hamiltonian spaces.
1.4 Derived Satake for the symmetric variety 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 Let 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
 be the Laurent loop group of the symmetric subgroup
$\mathfrak{L}\;\textrm{Sp}_{2n}$
 be the Laurent loop group of the symmetric subgroup 
 $\textrm{Sp}_{2n}\subset {\textrm{GL}}_{2n}$
. There is a natural action of
$\textrm{Sp}_{2n}\subset {\textrm{GL}}_{2n}$
. There is a natural action of 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
 on
$\mathfrak{L}\;\textrm{Sp}_{2n}$
 on 
 ${\textrm{Gr}}_{2n}$
, and we denote by
${\textrm{Gr}}_{2n}$
, and we denote by 
 $D^b(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n})$
 the dg-category of
$D^b(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n})$
 the dg-category of 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
-equivariant constructible complexes on
$\mathfrak{L}\;\textrm{Sp}_{2n}$
-equivariant constructible complexes on 
 ${\textrm{Gr}}_{2n}$
.
${\textrm{Gr}}_{2n}$
.
In [Reference Chen and NadlerCN18, Theorem 8.1] it is shown that there is an equivalence of dg-categories
 \begin{equation} D^b\big(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\big)\simeq D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}}). \end{equation}
\begin{equation} D^b\big(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\big)\simeq D^b(\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})\backslash {\textrm{Gr}}_{n,{\mathbb H}}). \end{equation}
In [Reference Chen and NadlerCN18, Theorem 9.1] it is shown that the equivalence is compatible with the natural monoidal actions of 
 $D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})$
, where the action on the right-hand side is through the nearby cycles functor (1.1). One can view the above equivalence as an example of the real–symmetric correspondence for the affine Grassmannian
$D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})$
, where the action on the right-hand side is through the nearby cycles functor (1.1). One can view the above equivalence as an example of the real–symmetric correspondence for the affine Grassmannian 
 ${\textrm{Gr}}_{2n}$
. Combining this with Theorem1.1, we obtain a derived Satake equivalence for
${\textrm{Gr}}_{2n}$
. Combining this with Theorem1.1, we obtain a derived Satake equivalence for 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
.
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
.
Theorem 1.3. There is an equivalence of dg-categories
 \begin{equation*}\Psi _X:D^b\big(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\big)\simeq D^{{\textrm{GL}}_n}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4]))\end{equation*}
\begin{equation*}\Psi _X:D^b\big(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\big)\simeq D^{{\textrm{GL}}_n}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4]))\end{equation*}
compatible with the monoidal actions of 
 $D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})\simeq D^{{\textrm{GL}}_n}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
.
$D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n})\simeq D^{{\textrm{GL}}_n}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2]))$
.
Remark 1.4. In general, the 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits on
$\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits on 
 ${\textrm{Gr}}_{2n}$
 are neither finite-dimensional nor finite-codimensional. Thus there is not a naive approach to sheaves on
${\textrm{Gr}}_{2n}$
 are neither finite-dimensional nor finite-codimensional. Thus there is not a naive approach to sheaves on 
 $\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}$
 with traditional methods. To overcome this, we use the observation in [Reference Chen and NadlerCN18] that the based loop group
$\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}$
 with traditional methods. To overcome this, we use the observation in [Reference Chen and NadlerCN18] that the based loop group 
 $\Omega \textrm{Sp}(n)$
 of the compact real form
$\Omega \textrm{Sp}(n)$
 of the compact real form 
 $\textrm{Sp}(n)$
 of
$\textrm{Sp}(n)$
 of 
 $\textrm{Sp}_{2n}$
 acts freely on
$\textrm{Sp}_{2n}$
 acts freely on 
 ${\textrm{Gr}}_{2n}$
 and the quotient
${\textrm{Gr}}_{2n}$
 and the quotient 
 $\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
 is a semi-analytic space of ind-finite type, i.e. an inductive limit of real analytic schemes of finite type. We define
$\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
 is a semi-analytic space of ind-finite type, i.e. an inductive limit of real analytic schemes of finite type. We define 
 $D^b(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n})$
 to be the category of sheaves on
$D^b(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n})$
 to be the category of sheaves on 
 $\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
 constructible with respect to the stratification coming from the descent of the
$\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
 constructible with respect to the stratification coming from the descent of the 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits stratification on
$\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits stratification on 
 ${\textrm{Gr}}_{2n}$
; see [Reference Chen and NadlerCN18] Definition 1.3] and also Remark 1.10.
${\textrm{Gr}}_{2n}$
; see [Reference Chen and NadlerCN18] Definition 1.3] and also Remark 1.10.
1.5 Geometric Langlands correspondence for real groups
 We discuss here how one might interpret our results in terms of geometric Langlands for real groups [Reference Ben-Zvi and NadlerBZN15]. Our results specifically relate to the curve 
 $\mathbb P^1$
 with its standard real structure with real points
$\mathbb P^1$
 with its standard real structure with real points 
 $\mathbb R\mathbb P^1$
 (whereas connections to Langlands parameters have been explored [Reference Ben-Zvi and NadlerBZN13] for
$\mathbb R\mathbb P^1$
 (whereas connections to Langlands parameters have been explored [Reference Ben-Zvi and NadlerBZN13] for 
 $\mathbb P^1$
 with its antipodal real structure with no real points).
$\mathbb P^1$
 with its antipodal real structure with no real points).
 For complex reductive groups, it is known that the derived Satake equivalence implies the geometric Langlands correspondence over the projective line 
 $\mathbb P^1$
 via a Radon transform. To state a version of this in the setting at hand, let
$\mathbb P^1$
 via a Radon transform. To state a version of this in the setting at hand, let 
 $\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1)$
 be the moduli stack of
$\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1)$
 be the moduli stack of 
 ${\textrm{GL}}_{2n}$
-bundles over
${\textrm{GL}}_{2n}$
-bundles over 
 $\mathbb P^1$
, and let
$\mathbb P^1$
, and let 
 $\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2)$
 be the moduli stack of Betti
$\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2)$
 be the moduli stack of Betti 
 ${\textrm{GL}}_{2n}$
-local systems on the 2-sphere
${\textrm{GL}}_{2n}$
-local systems on the 2-sphere 
 $\textrm{S}^2$
. Let
$\textrm{S}^2$
. Let 
 $D_!(\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1))$
 be the dg-category of constructible complexes on
$D_!(\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1))$
 be the dg-category of constructible complexes on 
 $\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1)$
 that are extensions by zero off of a finite-type substack, and let
$\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1)$
 that are extensions by zero off of a finite-type substack, and let 
 $\textrm{Coh}(\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2))$
 be the dg-category of coherent complexes on
$\textrm{Coh}(\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2))$
 be the dg-category of coherent complexes on 
 $\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2)$
 with bounded cohomology.
$\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2)$
 with bounded cohomology.
 In this setting, the geometric Langlands correspondence for 
 $\mathbb P^1$
 constructed in [Reference LafforgueLaf09] takes the form of an equivalence
$\mathbb P^1$
 constructed in [Reference LafforgueLaf09] takes the form of an equivalence

Moreover, it fits into a commutative diagram of equivalences

where the left vertical equivalence
 \begin{equation} D_!\big(\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1)\big)\simeq D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n}) \end{equation}
\begin{equation} D_!\big(\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb P^1)\big)\simeq D^b(\mathfrak{L}^+{\textrm{GL}}_{2n}\backslash {\textrm{Gr}}_{2n}) \end{equation}
is given by the Radon transform (see [Reference LafforgueLaf09, Proposition 2.1]) and the right vertical equivalence is given by the the Koszul duality equivalence
 \begin{equation} \textrm{Coh}(\mathfrak{gl}_{2n}[-1]/{\textrm{GL}}_{2n})\stackrel {}\simeq D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2])) \end{equation}
\begin{equation} \textrm{Coh}(\mathfrak{gl}_{2n}[-1]/{\textrm{GL}}_{2n})\stackrel {}\simeq D^{{\textrm{GL}}_{2n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{2n}[-2])) \end{equation}
 under the isomorphisms 
 $\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2)\simeq \textrm{pt}/{\textrm{GL}}_{2n}\times _{\mathfrak{gl}_{2n}/{\textrm{GL}}_{2n}}\textrm{pt}/{\textrm{GL}}_{2n}\simeq \mathfrak{gl}_{2n}[-1]/{\textrm{GL}}_{2n}$
.
$\textrm{Loc}_{{\textrm{GL}}_{2n}}(\textrm{S}^2)\simeq \textrm{pt}/{\textrm{GL}}_{2n}\times _{\mathfrak{gl}_{2n}/{\textrm{GL}}_{2n}}\textrm{pt}/{\textrm{GL}}_{2n}\simeq \mathfrak{gl}_{2n}[-1]/{\textrm{GL}}_{2n}$
.
As a special case of the affine Matsuki correspondence established in [Reference Chen and NadlerCN18], we have a real group version of the equivalence (1.6) taking the form
 \begin{equation} D_!\big(\textrm{Bun}_{{\textrm{GL}}_{n}({\mathbb H})}(\mathbb {RP}^1)\big)\simeq D^b\big(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\big). \end{equation}
\begin{equation} D_!\big(\textrm{Bun}_{{\textrm{GL}}_{n}({\mathbb H})}(\mathbb {RP}^1)\big)\simeq D^b\big(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\big). \end{equation}
 Here 
 $\textrm{Bun}_{{\textrm{GL}}_{n}({\mathbb H})}(\mathbb {RP}^1)$
 is the real form of
$\textrm{Bun}_{{\textrm{GL}}_{n}({\mathbb H})}(\mathbb {RP}^1)$
 is the real form of 
 $\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb {P}^1)$
 classifying
$\textrm{Bun}_{{\textrm{GL}}_{2n}}(\mathbb {P}^1)$
 classifying 
 ${\textrm{GL}}_{n}({\mathbb H})$
-bundles on the real projective line
${\textrm{GL}}_{n}({\mathbb H})$
-bundles on the real projective line 
 $\mathbb {RP}^1$
. Combining this with the derived Satake equivalence for
$\mathbb {RP}^1$
. Combining this with the derived Satake equivalence for 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 in Theorem1.3, we obtain the following geometric Langlands correspondence for
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 in Theorem1.3, we obtain the following geometric Langlands correspondence for 
 ${\textrm{GL}}_n({\mathbb H})$
.
${\textrm{GL}}_n({\mathbb H})$
.
 Let 
 $\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)$
 be the moduli stack of Betti
$\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)$
 be the moduli stack of Betti 
 ${\textrm{GL}}_n$
-local systems on the
${\textrm{GL}}_n$
-local systems on the 
 $4$
-sphere
$4$
-sphere 
 $\textrm{S}^4$
. Note that the presentation
$\textrm{S}^4$
. Note that the presentation 
 $\textrm{S}^4=\textrm{D}^4\cup _{\textrm{S}^3}\textrm{D}^4$
 (where
$\textrm{S}^4=\textrm{D}^4\cup _{\textrm{S}^3}\textrm{D}^4$
 (where 
 $\textrm{D}^4$
 is the four-dimensional disk in
$\textrm{D}^4$
 is the four-dimensional disk in 
 $\mathbb R^4$
) gives an isomorphism of dg-stacks:
$\mathbb R^4$
) gives an isomorphism of dg-stacks:
 \begin{equation*}\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)\simeq \textrm{pt}/{\textrm{GL}}_{n}\times _{\mathfrak{gl}_{n}[-2]/{\textrm{GL}}_{n}}\textrm{pt}/{\textrm{GL}}_{n}\simeq \mathfrak{gl}_{n}[-3]/{\textrm{GL}}_{n}.\end{equation*}
\begin{equation*}\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)\simeq \textrm{pt}/{\textrm{GL}}_{n}\times _{\mathfrak{gl}_{n}[-2]/{\textrm{GL}}_{n}}\textrm{pt}/{\textrm{GL}}_{n}\simeq \mathfrak{gl}_{n}[-3]/{\textrm{GL}}_{n}.\end{equation*}
 From the Koszul duality 
 $\textrm{Coh}(\mathfrak{gl}_{n}[-3]/{\textrm{GL}}_{n})\stackrel {}\simeq D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4]))$
, we obtain
$\textrm{Coh}(\mathfrak{gl}_{n}[-3]/{\textrm{GL}}_{n})\stackrel {}\simeq D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4]))$
, we obtain
 \begin{equation} \textrm{Coh}\big(\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)\big)\simeq \textrm{Coh}(\mathfrak{gl}_{n}[-3]/{\textrm{GL}}_{n})\simeq D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4])).\end{equation}
\begin{equation} \textrm{Coh}\big(\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)\big)\simeq \textrm{Coh}(\mathfrak{gl}_{n}[-3]/{\textrm{GL}}_{n})\simeq D^{{\textrm{GL}}_{n}}_{\textrm{perf}}(\textrm{Sym}(\mathfrak{gl}_{n}[-4])).\end{equation}
Theorem 1.5. There is an equivalence
 \begin{equation*} D_!\big(\textrm{Bun}_{{\textrm{GL}}_{n}({\mathbb H})}(\mathbb {RP}^1)\big)\simeq \textrm{Coh}\big(\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)\big) \end{equation*}
\begin{equation*} D_!\big(\textrm{Bun}_{{\textrm{GL}}_{n}({\mathbb H})}(\mathbb {RP}^1)\big)\simeq \textrm{Coh}\big(\textrm{Loc}_{{\textrm{GL}}_n}(\textrm{S}^4)\big) \end{equation*}
that fits into a commutative diagram of equivalences as follows.

 
Here the left and right vertical equivalence are the affine Matsuki correspondence (
1.8
) and Koszul duality (
1.9
), respectively, and the bottom equivalence is the derived Satake equivalence for 
 ${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
.
${\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
.
Remark 1.6. On the one hand, the appearance of the 
 $4$
-sphere
$4$
-sphere 
 $S^4$
 in the above version of geometric Langlands for
$S^4$
 in the above version of geometric Langlands for 
 ${\textrm{GL}}_n({\mathbb H})$
 is perhaps not so surprising owing to the identification
${\textrm{GL}}_n({\mathbb H})$
 is perhaps not so surprising owing to the identification 
 $ {\mathbb H} \mathbb P^1 \simeq S^4$
. Moreover, the twistor fibration
$ {\mathbb H} \mathbb P^1 \simeq S^4$
. Moreover, the twistor fibration 
 $\mathbb C \mathbb P^3\to {\mathbb H} \mathbb P^1 \simeq S^4$
 arises naturally in the proof of Theorem 1.1 (see § 4.1). On the other hand, the appearance of connections on
$\mathbb C \mathbb P^3\to {\mathbb H} \mathbb P^1 \simeq S^4$
 arises naturally in the proof of Theorem 1.1 (see § 4.1). On the other hand, the appearance of connections on 
 $S^4$
 is quite mysterious (at least to the authors). From the perspective of geometric Langlands for real groups, we expect the spectral side to be expressible in terms of
$S^4$
 is quite mysterious (at least to the authors). From the perspective of geometric Langlands for real groups, we expect the spectral side to be expressible in terms of 
 ${\textrm{GL}}_{2n}$
-connections on a disk with a partial oper structure along the boundary. This should reflect the usual Satake
${\textrm{GL}}_{2n}$
-connections on a disk with a partial oper structure along the boundary. This should reflect the usual Satake 
 ${\textrm{GL}}_{2n}$
-Hecke operators in the bulk and the real Satake
${\textrm{GL}}_{2n}$
-Hecke operators in the bulk and the real Satake 
 ${\textrm{GL}}_{n}$
-Hecke operators along the boundary.
${\textrm{GL}}_{n}$
-Hecke operators along the boundary.
Remark 1.7. More generally, the real–symmetric correspondence (1.3) and affine Matsuki correspondence (1.8) hold for any real group 
 $G_{\mathbb R}$
. It follows that a derived Satake equivalence for real groups or symmetric varieties will imply a version of geometric Langlands correspondence over
$G_{\mathbb R}$
. It follows that a derived Satake equivalence for real groups or symmetric varieties will imply a version of geometric Langlands correspondence over 
 ${\mathbb R} {\mathbb P}^1$
 for real groups and vice versa.
${\mathbb R} {\mathbb P}^1$
 for real groups and vice versa.
1.6 Relative Langlands duality conjectures
 A far-reaching program of Ben-Zvi, Sakellaridis, and Venkatesh proposes relative Langlands duality conjectures between periods and L-functions (see e.g. [Reference Ben-Zvi, Sakellaridis and VenkateshBZSV24]). A fundamental conjecture in the program predicts that given a complex reductive group 
 $G$
, with dual group
$G$
, with dual group 
 $\check G$
, and a homogeneous spherical
$\check G$
, and a homogeneous spherical 
 $G$
-variety
$G$
-variety 
 $X$
, there exists a (graded) Hamiltonian
$X$
, there exists a (graded) Hamiltonian 
 $\check G$
-variety
$\check G$
-variety 
 $\check M$
 together with a moment map
$\check M$
 together with a moment map 
 $\mu :\check M\to \check {\mathfrak g}^*$
 equipped with a commuting
$\mu :\check M\to \check {\mathfrak g}^*$
 equipped with a commuting 
 ${\mathbb G}_m$
-action of weight
${\mathbb G}_m$
-action of weight 
 $2$
, and an equivalence
$2$
, and an equivalence
 \begin{equation} D^b(\mathfrak{L}X/\mathfrak{L}^+G)\simeq \textrm{Coh}(\check M/\check G) \end{equation}
\begin{equation} D^b(\mathfrak{L}X/\mathfrak{L}^+G)\simeq \textrm{Coh}(\check M/\check G) \end{equation}
where 
 $\textrm{Coh}(\check M/\check G)$
 is the dg-category of
$\textrm{Coh}(\check M/\check G)$
 is the dg-category of 
 $\check G$
-equivariant perfect dg-modules over the ring of functions on
$\check G$
-equivariant perfect dg-modules over the ring of functions on 
 $\check M$
 viewed as a dg-algebra with trivial differential and grading given by the above
$\check M$
 viewed as a dg-algebra with trivial differential and grading given by the above 
 ${\mathbb G}_m$
-action. Moreover, this equivalence should be compatible with the derived Satake equivalence
${\mathbb G}_m$
-action. Moreover, this equivalence should be compatible with the derived Satake equivalence 
 $D^b(\mathfrak{L}^+G\backslash {\textrm{Gr}}_G)\simeq D^{\check G}_{\textrm{perf}}(\textrm{Sym}(\check {\mathfrak g}[-2]))\simeq \textrm{Coh}(\check {\mathfrak g}^*[2]/\check G)$
, in the sense that the right convolution action of
$D^b(\mathfrak{L}^+G\backslash {\textrm{Gr}}_G)\simeq D^{\check G}_{\textrm{perf}}(\textrm{Sym}(\check {\mathfrak g}[-2]))\simeq \textrm{Coh}(\check {\mathfrak g}^*[2]/\check G)$
, in the sense that the right convolution action of 
 $D^b(\mathfrak{L}^+G\backslash {\textrm{Gr}}_G)$
 on
$D^b(\mathfrak{L}^+G\backslash {\textrm{Gr}}_G)$
 on 
 $D^b(\mathfrak{L}X/\mathfrak{L}^+G)$
 should correspond to the tensor product action of
$D^b(\mathfrak{L}X/\mathfrak{L}^+G)$
 should correspond to the tensor product action of 
 $\textrm{Coh}(\check {\mathfrak g}^*[2]/\check G)$
 on
$\textrm{Coh}(\check {\mathfrak g}^*[2]/\check G)$
 on 
 $\textrm{Coh}(\check M/\check G)$
 through the moment map
$\textrm{Coh}(\check M/\check G)$
 through the moment map 
 $\mu$
.
$\mu$
.
 We now explain how the derived Satake equivalence for the symmetric variety 
 $X={\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 fits into the general setting of relative Langlands duality. On the one hand, by [Reference Chen and NadlerCN24, Proposition 8.1], there are an isomorphism of stacks
$X={\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 fits into the general setting of relative Langlands duality. On the one hand, by [Reference Chen and NadlerCN24, Proposition 8.1], there are an isomorphism of stacks
 \begin{equation*}\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n}\cong \mathfrak{L}\;\textrm{Sp}_{2n}\backslash \mathfrak{L}{\textrm{GL}}_{2n}/\mathfrak{L}^+{\textrm{GL}}_{2n}\cong \mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\end{equation*}
\begin{equation*}\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n}\cong \mathfrak{L}\;\textrm{Sp}_{2n}\backslash \mathfrak{L}{\textrm{GL}}_{2n}/\mathfrak{L}^+{\textrm{GL}}_{2n}\cong \mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}\end{equation*}
and hence equivalences of categories
 \begin{equation} D^b(\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n})\simeq D^b(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}) \end{equation}
\begin{equation} D^b(\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n})\simeq D^b(\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n}) \end{equation}
where 
 $D^b(\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n})$
 is the dg-category of
$D^b(\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n})$
 is the dg-category of 
 $\mathfrak{L}^+{\textrm{GL}}_{2n}$
-equivariant constuctible complexes on the loop space
$\mathfrak{L}^+{\textrm{GL}}_{2n}$
-equivariant constuctible complexes on the loop space 
 $\mathfrak{L}X$
 of
$\mathfrak{L}X$
 of 
 $X$
.
$X$
.
 On the other hand, it is expected that the Hamiltonian 
 $\check G$
-space
$\check G$
-space 
 $\check M$
 associated to the symmetric variety
$\check M$
 associated to the symmetric variety 
 $X={\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 (note that symmetric varieties are spherical) is given by
$X={\textrm{GL}}_{2n}/\textrm{Sp}_{2n}$
 (note that symmetric varieties are spherical) is given by 
 $\check M=T^*({\textrm{GL}}_{2n}/{\textrm{GL}}_n^{\textrm{}}\ltimes U,\psi )$
, the partial Whittaker reduction of
$\check M=T^*({\textrm{GL}}_{2n}/{\textrm{GL}}_n^{\textrm{}}\ltimes U,\psi )$
, the partial Whittaker reduction of 
 $T^*{\textrm{GL}}_{2n}$
 with respect to the generic homomorphism
$T^*{\textrm{GL}}_{2n}$
 with respect to the generic homomorphism 
 $\psi$
 of the Shalika subgroup
$\psi$
 of the Shalika subgroup 
 ${\textrm{GL}}^{\textrm{}}_n\ltimes U$
 of
${\textrm{GL}}^{\textrm{}}_n\ltimes U$
 of 
 ${\textrm{GL}}_{2n}$
:
${\textrm{GL}}_{2n}$
:
 \begin{equation} {\textrm{GL}}^{\textrm{}}_n\ltimes U=\bigg\{\!\begin{pmatrix} A&\quad\! 0\\ 0&\quad\! A \end{pmatrix}\begin{pmatrix} \textrm{Id}_n&\quad\! 0\\ C&\quad\! \textrm{Id}_n \end{pmatrix}\!\bigg|\! \ A\in {\textrm{GL}}_n,\, C\in \mathfrak{gl}_n\bigg\},\quad\! \psi \left(\!\begin{pmatrix} A&\quad\! 0\\ 0&\quad\! A \end{pmatrix}\begin{pmatrix} \textrm{Id}_n&\quad\! 0\\ C&\quad\! \textrm{Id}_n \end{pmatrix}\!\right)=-\textrm{tr}(C) \end{equation}
\begin{equation} {\textrm{GL}}^{\textrm{}}_n\ltimes U=\bigg\{\!\begin{pmatrix} A&\quad\! 0\\ 0&\quad\! A \end{pmatrix}\begin{pmatrix} \textrm{Id}_n&\quad\! 0\\ C&\quad\! \textrm{Id}_n \end{pmatrix}\!\bigg|\! \ A\in {\textrm{GL}}_n,\, C\in \mathfrak{gl}_n\bigg\},\quad\! \psi \left(\!\begin{pmatrix} A&\quad\! 0\\ 0&\quad\! A \end{pmatrix}\begin{pmatrix} \textrm{Id}_n&\quad\! 0\\ C&\quad\! \textrm{Id}_n \end{pmatrix}\!\right)=-\textrm{tr}(C) \end{equation}
(see the list of examples of relative duality in [Reference WangWan]).
 By Lemma 3.2, there is an isomorphism 
 $\check M\simeq {\textrm{GL}}_{2n}\times ^{{\textrm{GL}}_n^{\textrm{}}}\mathfrak{gl}_n$
 such that the induced isomorphism
$\check M\simeq {\textrm{GL}}_{2n}\times ^{{\textrm{GL}}_n^{\textrm{}}}\mathfrak{gl}_n$
 such that the induced isomorphism 
 $\check M/{\textrm{GL}}_{2n}\simeq \mathfrak{gl}_n/{\textrm{GL}}_n$
 fits into a commutative diagram as follows.
$\check M/{\textrm{GL}}_{2n}\simeq \mathfrak{gl}_n/{\textrm{GL}}_n$
 fits into a commutative diagram as follows.

 Here 
 $\mu$
 is the moment map,
$\mu$
 is the moment map, 
 $\tau$
 is the embedding in (1.2) (disregarding the cohomological grading), and we identify
$\tau$
 is the embedding in (1.2) (disregarding the cohomological grading), and we identify 
 $\mathfrak{gl}^*_{2n}/{\textrm{GL}}_{2n}\simeq \mathfrak{gl}_{2n}/{\textrm{GL}}_{2n}$
 via the trace form. Thus, in view of (1.12), the equivalence of Theorem1.3 gives an instance of (1.11) of the formFootnote 
3
$\mathfrak{gl}^*_{2n}/{\textrm{GL}}_{2n}\simeq \mathfrak{gl}_{2n}/{\textrm{GL}}_{2n}$
 via the trace form. Thus, in view of (1.12), the equivalence of Theorem1.3 gives an instance of (1.11) of the formFootnote 
3
 \begin{equation*}D^b(\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n})\simeq \textrm{Coh}(\check M/{\textrm{GL}}_{2n}).\end{equation*}
\begin{equation*}D^b(\mathfrak{L}X/\mathfrak{L}^+{\textrm{GL}}_{2n})\simeq \textrm{Coh}(\check M/{\textrm{GL}}_{2n}).\end{equation*}
Remark 1.8. Our work suggests an interesting relationship between real groups and periods of automorphic forms associated to the corresponding symmetric varieties, and we plan to investigate this relationship in more detail. The case of the quaternionic group 
 $\textrm{GL}_n(\mathbb H)$
 is related to the so-called symplectic periods and Jacquet–Shalika periods [Reference Jacquet and RallisJR92, Reference Jacquet and ShalikaJS90].
$\textrm{GL}_n(\mathbb H)$
 is related to the so-called symplectic periods and Jacquet–Shalika periods [Reference Jacquet and RallisJR92, Reference Jacquet and ShalikaJS90].
1.7 IC-stalks and Kostka–Foulkes polynomials
 As an application of the proof of Theorem1.1, we determine the stalk cohomology of the IC-complexes for the 
 $\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
-orbit closures in the quaternionic affine Grassmannian
$\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
-orbit closures in the quaternionic affine Grassmannian 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 and the
${\textrm{Gr}}_{n,{\mathbb H}}$
 and the 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbit closures in the complex affine Grassmannian
$\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbit closures in the complex affine Grassmannian 
 ${\textrm{Gr}}_{2n}$
.
${\textrm{Gr}}_{2n}$
.
 The 
 $\mathfrak{L}^+{\textrm{GL}}_n$
-orbits (respectively,
$\mathfrak{L}^+{\textrm{GL}}_n$
-orbits (respectively, 
 $\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 and
$\mathfrak{L}^+{\textrm{GL}}_n({\mathbb H})$
 and 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits) on
$\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits) on 
 ${\textrm{Gr}}_n$
 (respectively,
${\textrm{Gr}}_n$
 (respectively, 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 and
${\textrm{Gr}}_{n,{\mathbb H}}$
 and 
 ${\textrm{Gr}}_{2n}$
) are in bijection with the set of dominant coweights
${\textrm{Gr}}_{2n}$
) are in bijection with the set of dominant coweights 
 $\Lambda ^+_n$
 of
$\Lambda ^+_n$
 of 
 ${\textrm{GL}}_n$
; see § 4.4. For any
${\textrm{GL}}_n$
; see § 4.4. For any 
 $\lambda \in \Lambda ^+_n$
 we denote by
$\lambda \in \Lambda ^+_n$
 we denote by 
 ${\textrm{Gr}}_n^\lambda$
 (respectively,
${\textrm{Gr}}_n^\lambda$
 (respectively, 
 ${\textrm{Gr}}_{n,{\mathbb H}}^\lambda$
 and
${\textrm{Gr}}_{n,{\mathbb H}}^\lambda$
 and 
 ${\textrm{Gr}}^\lambda _{2n,X}$
) the corresponding orbit and by
${\textrm{Gr}}^\lambda _{2n,X}$
) the corresponding orbit and by 
 $\textrm{IC}(\overline {{\textrm{Gr}}_n^\lambda })$
 (respectively,
$\textrm{IC}(\overline {{\textrm{Gr}}_n^\lambda })$
 (respectively, 
 $\textrm{IC}(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda })$
 and
$\textrm{IC}(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda })$
 and 
 $\textrm{IC}(\overline {{\textrm{Gr}}^\lambda _{2n,X}})$
) the intersection cohomology complex on the orbit closure. We will write
$\textrm{IC}(\overline {{\textrm{Gr}}^\lambda _{2n,X}})$
) the intersection cohomology complex on the orbit closure. We will write 
 $\mathscr{H}^{\;\;i}(\mathcal {F})$
 for the
$\mathscr{H}^{\;\;i}(\mathcal {F})$
 for the 
 $i$
th cohomology sheaf of a complex
$i$
th cohomology sheaf of a complex 
 $\mathcal {F}$
 and
$\mathcal {F}$
 and 
 $\mathscr H_x^{\;\,i}(\mathcal {F})$
 for its stalk at a point
$\mathscr H_x^{\;\,i}(\mathcal {F})$
 for its stalk at a point 
 $x$
.
$x$
.
 For any pair of dominant coweights 
 $\lambda ,\mu \in \Lambda ^+_n$
, we denote by
$\lambda ,\mu \in \Lambda ^+_n$
, we denote by 
 $K_{\lambda ,\mu }(q)$
 the associated Kostka–Foulkes polynomial with variable
$K_{\lambda ,\mu }(q)$
 the associated Kostka–Foulkes polynomial with variable 
 $q$
. Denote by
$q$
. Denote by 
 $\rho _n$
 the half-sum of positive roots of
$\rho _n$
 the half-sum of positive roots of 
 ${\textrm{GL}}_n$
. A well-known result of Lusztig [Reference LusztigLus81] says that we have
${\textrm{GL}}_n$
. A well-known result of Lusztig [Reference LusztigLus81] says that we have 
 $\mathscr H^{\,\,i-2\langle \lambda ,\rho _n\rangle }(\textrm{IC}(\overline {{\textrm{Gr}}_n^\lambda }))=0$
 for
$\mathscr H^{\,\,i-2\langle \lambda ,\rho _n\rangle }(\textrm{IC}(\overline {{\textrm{Gr}}_n^\lambda }))=0$
 for 
 $2\nmid i$
 and
$2\nmid i$
 and
 \begin{equation*}\sum _i\textrm{dim}\;\mathscr H^{\,\,2i-2\langle \lambda ,\rho _n\rangle }_x\big(\textrm{IC}\big(\overline {{\textrm{Gr}}_n^\lambda }\big)\big)q^i =q^{\langle \lambda -\mu ,\rho _n\rangle }K_{\lambda ,\mu }(q^{-1}) \quad \text {for } x\in {\textrm{Gr}}_{n}^\mu .\end{equation*}
\begin{equation*}\sum _i\textrm{dim}\;\mathscr H^{\,\,2i-2\langle \lambda ,\rho _n\rangle }_x\big(\textrm{IC}\big(\overline {{\textrm{Gr}}_n^\lambda }\big)\big)q^i =q^{\langle \lambda -\mu ,\rho _n\rangle }K_{\lambda ,\mu }(q^{-1}) \quad \text {for } x\in {\textrm{Gr}}_{n}^\mu .\end{equation*}
We have the following real and symmetric analogue.
Theorem 1.9 (See Corollary 4.14 and Theorem 4.22). Let 
 $\lambda ,\mu \in \Lambda _n^+$
. For any
$\lambda ,\mu \in \Lambda _n^+$
. For any 
 $x\in {\textrm{Gr}}_{n,{\mathbb H}}^\mu$
 and
$x\in {\textrm{Gr}}_{n,{\mathbb H}}^\mu$
 and 
 $y\in {\textrm{Gr}}_{2n,X}^\mu$
, the following hold:
$y\in {\textrm{Gr}}_{2n,X}^\mu$
, the following hold:
- 
(1)  $\mathscr H^{\,\,i-4\langle \lambda ,\rho _n\rangle }\big(\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\big)\big)= \mathscr H^{\,\,i-4\langle \lambda ,\rho _n\rangle }\big(\textrm{IC}\big(\overline {{\textrm{Gr}}_{2n,X}^\lambda }\big)\big)= 0$
 for $\mathscr H^{\,\,i-4\langle \lambda ,\rho _n\rangle }\big(\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\big)\big)= \mathscr H^{\,\,i-4\langle \lambda ,\rho _n\rangle }\big(\textrm{IC}\big(\overline {{\textrm{Gr}}_{2n,X}^\lambda }\big)\big)= 0$
 for $4\nmid i$
; $4\nmid i$
;
- 
(2)  $\underset {i}\sum \textrm{dim}\;\mathscr H^{\,\,4i-4\langle \lambda ,\rho _n\rangle }_x\big(\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\big)\big)q^i= \underset {i}\sum \textrm{dim}\;\mathscr H^{\,\,4i-4\langle \lambda ,\rho _n\rangle }_y\big(\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{2n,X}}\big)\big)q^i=q^{\langle \lambda -\mu ,\rho _n\rangle }K_{\lambda ,\mu }(q^{-1})$
. $\underset {i}\sum \textrm{dim}\;\mathscr H^{\,\,4i-4\langle \lambda ,\rho _n\rangle }_x\big(\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\big)\big)q^i= \underset {i}\sum \textrm{dim}\;\mathscr H^{\,\,4i-4\langle \lambda ,\rho _n\rangle }_y\big(\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{2n,X}}\big)\big)q^i=q^{\langle \lambda -\mu ,\rho _n\rangle }K_{\lambda ,\mu }(q^{-1})$
.
 In other words, the theorem says that the IC-complex for the 
 $\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})$
 and
$\mathfrak{L}^+{\textrm{GL}}_{n}({\mathbb H})$
 and 
 $\mathfrak{L}K$
-orbit closures
$\mathfrak{L}K$
-orbit closures 
 $\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}$
 and
$\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}$
 and 
 $\overline {{\textrm{Gr}}^\lambda _{2n,X}}$
 have the same stalk cohomology as the
$\overline {{\textrm{Gr}}^\lambda _{2n,X}}$
 have the same stalk cohomology as the 
 $\overline {{\textrm{Gr}}_n^\lambda }$
 ones for
$\overline {{\textrm{Gr}}_n^\lambda }$
 ones for 
 ${\textrm{GL}}_n$
, but with all degrees doubled.
${\textrm{GL}}_n$
, but with all degrees doubled.
Remark 1.10. To define the IC-stalk 
 $\mathscr H_y^{\ i}\textrm{IC}\big(\overline {{\textrm{Gr}}_{2n,X}^\lambda }\big)$
 at
$\mathscr H_y^{\ i}\textrm{IC}\big(\overline {{\textrm{Gr}}_{2n,X}^\lambda }\big)$
 at 
 $y\in {\textrm{Gr}}_{2n,X}^\mu$
, we use the observation that
$y\in {\textrm{Gr}}_{2n,X}^\mu$
, we use the observation that 
 ${\textrm{Gr}}_{2n,X}^\mu$
 has finite codimension in
${\textrm{Gr}}_{2n,X}^\mu$
 has finite codimension in 
 $\overline {{\textrm{Gr}}_{2n,X}^\lambda }$
 and hence the IC-stalk makes sense. This can be made precise using the observation in [Reference Chen and NadlerCN18] that the image
$\overline {{\textrm{Gr}}_{2n,X}^\lambda }$
 and hence the IC-stalk makes sense. This can be made precise using the observation in [Reference Chen and NadlerCN18] that the image 
 $\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n,X}^\lambda$
 of the
$\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n,X}^\lambda$
 of the 
 $\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits
$\mathfrak{L}\;\textrm{Sp}_{2n}$
-orbits 
 ${\textrm{Gr}}_{2n,X}^\lambda$
 in the quotient
${\textrm{Gr}}_{2n,X}^\lambda$
 in the quotient 
 $\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
 is finite-dimensional with even real dimension and the collection
$\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
 is finite-dimensional with even real dimension and the collection 
 $\big\{\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n,X}^\lambda \big\}_{\lambda \in \Lambda _n^+}$
 forms a nice stratification of
$\big\{\mathfrak{L}\;\textrm{Sp}_{2n}\backslash {\textrm{Gr}}_{2n,X}^\lambda \big\}_{\lambda \in \Lambda _n^+}$
 forms a nice stratification of 
 $\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
. This allows us to define the IC-complex
$\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
. This allows us to define the IC-complex 
 $\textrm{IC}\big(\overline {{\textrm{Gr}}_{2n,X}^\lambda }\big)$
 of
$\textrm{IC}\big(\overline {{\textrm{Gr}}_{2n,X}^\lambda }\big)$
 of 
 $\overline {{\textrm{Gr}}_{2n,X}^\lambda }$
 (and hence the IC-stalks) as the IC-complex for the orbit closure
$\overline {{\textrm{Gr}}_{2n,X}^\lambda }$
 (and hence the IC-stalks) as the IC-complex for the orbit closure 
 $\Omega \textrm{Sp}(n)\backslash \overline {{\textrm{Gr}}_{2n,X}^\lambda }$
 of
$\Omega \textrm{Sp}(n)\backslash \overline {{\textrm{Gr}}_{2n,X}^\lambda }$
 of 
 $\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n,X}^\lambda$
 inside
$\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n,X}^\lambda$
 inside 
 $\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
.
$\Omega \textrm{Sp}(n)\backslash {\textrm{Gr}}_{2n}$
.
Remark 1.11. We first prove Theorem 1.9 in the real case using the nice geometry of the quaternionic Grassmannian 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 and then deduce the symmetric case via the real–symmetric correspondence. At the moment, we do not know a direct argument in the symmetric case.
${\textrm{Gr}}_{n,{\mathbb H}}$
 and then deduce the symmetric case via the real–symmetric correspondence. At the moment, we do not know a direct argument in the symmetric case.
1.8 Outline of the proof
 We briefly explain the proof of Theorem1.1. Similar to the proof of the derived Satake for complex reductive groups [Reference Bezrukavnikov and FinkelbergBF08], the desired equivalence follows from the following two statements: (1) the de-equivariantized extension algebra 
 $\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
 is isomorphic to the dg symmetric algebra
$\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
 is isomorphic to the dg symmetric algebra 
 $\textrm{Sym}(\mathfrak{gl}_n[-4])$
 (see Proposition 4.21) and (2) the dg algebra
$\textrm{Sym}(\mathfrak{gl}_n[-4])$
 (see Proposition 4.21) and (2) the dg algebra 
 $\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
 is formal (see Proposition 5.3).
$\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
 is formal (see Proposition 5.3).
We deduce (1) from a fully-faithfulness property of the equivariant cohomology functor. In [Reference Bezrukavnikov and FinkelbergBF08], this was proved using a general result of Ginzburg [Reference GinzburgGin91], whose proof uses Hodge theory and hence does not apply directly to the real analytic setting. We observe that in the situation of the quaternionic affine Grassmannian, the stalks of the IC-complexes satisfy a parity vanishing property and, as observed in [Reference Achar and RiderAR15], one can use parity considerations in place of Hodge theory. To prove the parity vanishing result of the IC-stalks, we show that fibers of certain convolution Grassmannians (which are basically quaternionic Springer fibers) admit pavings by quaternionic affine spaces.
 The proof of (2) in [Reference Bezrukavnikov and FinkelbergBF08] also relies on Hodge theory (or theory of weights) and hence must be modified in the real setting. We observe that the nearby cycles functor (1.1) induces a surjective homomorphism from the dg-algebra 
 $\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_{2n}))$
 associated to the complex affine Grassimannian
$\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_{2n}))$
 associated to the complex affine Grassimannian 
 ${\textrm{Gr}}_{2n}$
 to the dg-algebra
${\textrm{Gr}}_{2n}$
 to the dg-algebra 
 $\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
. Since the former dg-algebra is formal, thanks to [Reference Bezrukavnikov and FinkelbergBF08], the desired claim follows from a general criterion of formality; see Lemma 5.4.
$\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
. Since the former dg-algebra is formal, thanks to [Reference Bezrukavnikov and FinkelbergBF08], the desired claim follows from a general criterion of formality; see Lemma 5.4.
Remark 1.12. We expect that the proof strategy outlined above is applicable to general real groups: the parity vanishing, fully-faithfulness, and formality results should hold in general.
1.9 Organization
We briefly summarize the main goals of each section. In § 2, we recall some notation and results on constructible sheaves on a semi-analytic stack. In § 3, we study the spectral side of the quaternionic Satake equivalence, including results on quaternionic groups, symplectic groups, regular centralizers group schemes, and Whittaker reduction. In § 4, we study the constructible side of the equivalence, including the study of nearby cycles functors, parity vanishing results, fully-faithfulness of the equivariant cohomology functor, and the computation of the IC-stalks and the de-equivariantized extension algebra. Finally, in § 5, we prove the formality result and deduce the derived Satake equivalence for quaternionic groups, including a version involving nilpotent singular supports (Theorem5.5), and also the spectral description of the nearby cycles functor (Theorem5.7).
2. Constructible sheaves on a semi-analytic stack
 We will work with 
 $\mathbb C$
-linear dg-categories (see e.g. [Reference Gaitsgory and RozenblyumGR17, Chapter 1, § 10] for a concise summary of the theory of dg-categories). Unless specified otherwise, all dg-categories will be assumed to be cocomplete, i.e. containing all small colimits, and all functors between dg-categories will be assumed to be continuous, i.e. preserving all small colimits.
$\mathbb C$
-linear dg-categories (see e.g. [Reference Gaitsgory and RozenblyumGR17, Chapter 1, § 10] for a concise summary of the theory of dg-categories). Unless specified otherwise, all dg-categories will be assumed to be cocomplete, i.e. containing all small colimits, and all functors between dg-categories will be assumed to be continuous, i.e. preserving all small colimits.
 We collect some facts about constructible sheaves on a semi-analytic stack. Recall that a subset 
 $Y$
 of a real analytic manifold
$Y$
 of a real analytic manifold 
 $M$
 is said to be semi-analytic if any point
$M$
 is said to be semi-analytic if any point 
 $y\in Y$
 has a open neighbourhood
$y\in Y$
 has a open neighbourhood 
 $U$
 such that the intersection
$U$
 such that the intersection 
 $Y\cap U$
 is a finite union of sets of the form
$Y\cap U$
 is a finite union of sets of the form
 \begin{equation*}\{y\in U\mid f_1(y)=\cdot \cdot \cdot =f_r(y)=0, g_1(y)\gt 0,\ldots , g_l(y)\gt 0\},\end{equation*}
\begin{equation*}\{y\in U\mid f_1(y)=\cdot \cdot \cdot =f_r(y)=0, g_1(y)\gt 0,\ldots , g_l(y)\gt 0\},\end{equation*}
where the 
 $f_i$
 and
$f_i$
 and 
 $g_j$
 are real analytic functions on
$g_j$
 are real analytic functions on 
 $U$
. A map
$U$
. A map 
 $f:Y\to Y'$
 between two semi-analytic sets is said to be semi-analytic if it is continuous and its graph is a semi-analytic set.
$f:Y\to Y'$
 between two semi-analytic sets is said to be semi-analytic if it is continuous and its graph is a semi-analytic set.
 Let 
 $\textrm{Grpd}$
 be the
$\textrm{Grpd}$
 be the 
 $\infty$
-category of spaces or, equivalently,
$\infty$
-category of spaces or, equivalently, 
 $\infty$
-groupoids. Let
$\infty$
-groupoids. Let 
 $\textrm{RSp}$
 be the site of semi-analytic sets where the coverings are étale (i.e. locally bi-analytic) maps
$\textrm{RSp}$
 be the site of semi-analytic sets where the coverings are étale (i.e. locally bi-analytic) maps 
 $\{S_i\to S\}_{i\in I}$
 such that the map
$\{S_i\to S\}_{i\in I}$
 such that the map 
 $\bigsqcup S_i\to S$
 is surjective. A semi-analytic pre-stack is a functor
$\bigsqcup S_i\to S$
 is surjective. A semi-analytic pre-stack is a functor 
 ${\mathcal Y}:\textrm{RSp}\to \textrm{Grpd}$
, and a semi-analytic stack is a pre-stack that is a sheaf. We will view any semi-analytic set as a semi-analytic stack via the Yoneda embedding.
${\mathcal Y}:\textrm{RSp}\to \textrm{Grpd}$
, and a semi-analytic stack is a pre-stack that is a sheaf. We will view any semi-analytic set as a semi-analytic stack via the Yoneda embedding.
 For any semi-analytic set 
 $Y$
, we define
$Y$
, we define 
 $D(Y)=\textrm{Ind}(D^b(Y))$
 to be the ind-completion of the bounded dg-category
$D(Y)=\textrm{Ind}(D^b(Y))$
 to be the ind-completion of the bounded dg-category 
 $D^b(Y)$
 of
$D^b(Y)$
 of 
 $\mathbb C$
-constructible sheaves on
$\mathbb C$
-constructible sheaves on 
 $Y$
. For any semi-analytic stack
$Y$
. For any semi-analytic stack 
 $\mathcal Y$
 we define
$\mathcal Y$
 we define 
 $D({\mathcal Y}):={\textrm{lim}}_I D(Y)$
, where the index category is that of semi-analytic sets equipped with a semi-analytic map to
$D({\mathcal Y}):={\textrm{lim}}_I D(Y)$
, where the index category is that of semi-analytic sets equipped with a semi-analytic map to 
 $\mathcal Y$
 and the transition functors are given by
$\mathcal Y$
 and the transition functors are given by 
 $!$
-pullback. Since we are in the constructible context,
$!$
-pullback. Since we are in the constructible context, 
 $!$
-pullback admits a left adjoint, given by
$!$
-pullback admits a left adjoint, given by 
 $!$
-pushforward, and it follows that
$!$
-pushforward, and it follows that 
 $D({\mathcal Y})={\textrm{colim}}_I D(Y)$
. In particular,
$D({\mathcal Y})={\textrm{colim}}_I D(Y)$
. In particular, 
 $D({\mathcal Y})$
 is compactly generated. We let
$D({\mathcal Y})$
 is compactly generated. We let 
 $D({\mathcal Y})^c$
 be the full subcategory of compact objects and
$D({\mathcal Y})^c$
 be the full subcategory of compact objects and 
 $D^b({\mathcal Y})\subset D({\mathcal Y})$
 the full subcategory of objects that pull back to an object of
$D^b({\mathcal Y})\subset D({\mathcal Y})$
 the full subcategory of objects that pull back to an object of 
 $D^b(Y)$
 for any
$D^b(Y)$
 for any 
 $Y$
 mapping to
$Y$
 mapping to 
 $\mathcal Y$
. Note that we have the natural inclusion
$\mathcal Y$
. Note that we have the natural inclusion 
 $D({\mathcal Y})^c\subset D^b({\mathcal Y})$
, but the inclusion is in general not an equality. For example, the constant sheaf
$D({\mathcal Y})^c\subset D^b({\mathcal Y})$
, but the inclusion is in general not an equality. For example, the constant sheaf 
 ${\mathbb C}_{{\mathcal Y}}\in D^b({\mathcal Y})$
 for the classifying stack
${\mathbb C}_{{\mathcal Y}}\in D^b({\mathcal Y})$
 for the classifying stack 
 ${\mathcal Y}=B({\textrm{GL}}_1({\mathbb C}))$
 is not compact.
${\mathcal Y}=B({\textrm{GL}}_1({\mathbb C}))$
 is not compact.
 Let 
 $f:{\mathcal Y}\to {\mathcal Y}'$
 be a map between semi-analytic stacks. We have the usual six-functor formalism
$f:{\mathcal Y}\to {\mathcal Y}'$
 be a map between semi-analytic stacks. We have the usual six-functor formalism 
 $f_*$
,
$f_*$
, 
 $f^!$
,
$f^!$
, 
 $f_*$
,
$f_*$
, 
 $f_!$
,
$f_!$
, 
 $\otimes$
,
$\otimes$
, 
 $\underline {\textrm{Hom}}$
.
$\underline {\textrm{Hom}}$
.
 For a semi-analytic stack 
 $\mathcal Y$
, with projection map
$\mathcal Y$
, with projection map 
 $p:{\mathcal Y}\to \textrm{pt}$
, and
$p:{\mathcal Y}\to \textrm{pt}$
, and 
 $\mathcal {F}\in D(Y)$
, we will write
$\mathcal {F}\in D(Y)$
, we will write 
 $H^*(Y,\mathcal {F}):=p_*(\mathcal {F})$
 for the cohomology of
$H^*(Y,\mathcal {F}):=p_*(\mathcal {F})$
 for the cohomology of 
 $\mathcal {F}$
. If
$\mathcal {F}$
. If 
 $\mathcal Y$
 is isomorphic to the quotient stack
$\mathcal Y$
 is isomorphic to the quotient stack 
 ${\mathcal Y}\simeq G\backslash Y$
, where
${\mathcal Y}\simeq G\backslash Y$
, where 
 $Y$
 is a semi-analytic set acted on real analytically by a Lie group
$Y$
 is a semi-analytic set acted on real analytically by a Lie group 
 $G$
, we will write
$G$
, we will write 
 $H^*_G(Y,\mathcal {F}):=(p_{BG})_*(\mathcal {F})$
 for the equivariant cohomology of
$H^*_G(Y,\mathcal {F}):=(p_{BG})_*(\mathcal {F})$
 for the equivariant cohomology of 
 $\mathcal {F}$
, where
$\mathcal {F}$
, where 
 $p_{BG}:{\mathcal Y}=G\backslash Y\to BG$
 is the projection map. When it is clear from the context we will abbreviate
$p_{BG}:{\mathcal Y}=G\backslash Y\to BG$
 is the projection map. When it is clear from the context we will abbreviate 
 $H^*(Y,\mathcal {F})$
 and
$H^*(Y,\mathcal {F})$
 and 
 $H^*_G(Y,\mathcal {F})$
 by
$H^*_G(Y,\mathcal {F})$
 by 
 $H^*(\mathcal {F})$
 and
$H^*(\mathcal {F})$
 and 
 $H^*_G(\mathcal {F})$
.
$H^*_G(\mathcal {F})$
.
 For an ind semi-analytic stack 
 ${\mathcal Y}={\textrm{colim}}_I {\mathcal Y}_i$
 we define
${\mathcal Y}={\textrm{colim}}_I {\mathcal Y}_i$
 we define 
 $D({\mathcal Y})={\textrm{lim}}_I D({\mathcal Y}_i)$
, where the limit is taken with respect to the
$D({\mathcal Y})={\textrm{lim}}_I D({\mathcal Y}_i)$
, where the limit is taken with respect to the 
 $!$
-pull-back along the closed embedding
$!$
-pull-back along the closed embedding 
 $\iota _{i,i'}:{\mathcal Y}_i\to {\mathcal Y}_{i'}, i,i'\in I$
.
$\iota _{i,i'}:{\mathcal Y}_i\to {\mathcal Y}_{i'}, i,i'\in I$
.
3. Spectral side
3.1 Quaternionic group
 For any positive integer 
 $n$
, we denote by
$n$
, we denote by 
 $G_n={\textrm{GL}}_n({\mathbb C})$
 the complex Lie group of
$G_n={\textrm{GL}}_n({\mathbb C})$
 the complex Lie group of 
 $n\times n$
 invertible matrices and
$n\times n$
 invertible matrices and 
 ${\mathfrak g}_n=\mathfrak{gl}_n({\mathbb C})$
 its Lie algebra of
${\mathfrak g}_n=\mathfrak{gl}_n({\mathbb C})$
 its Lie algebra of 
 $n\times n$
 matrices. We write
$n\times n$
 matrices. We write 
 $B_n$
,
$B_n$
, 
 $N_n$
, and
$N_n$
, and 
 $T_n$
 for the subgroups of
$T_n$
 for the subgroups of 
 $G_n$
 consisting of upper triangular matrices, upper triangular unipotent matrices, and diagonal matrices, respectively, and write
$G_n$
 consisting of upper triangular matrices, upper triangular unipotent matrices, and diagonal matrices, respectively, and write 
 ${\mathfrak b}_n$
,
${\mathfrak b}_n$
, 
 $\mathfrak{n}_n$
, and
$\mathfrak{n}_n$
, and 
 ${\mathfrak t}_n$
 for their Lie algebras. We denote by
${\mathfrak t}_n$
 for their Lie algebras. We denote by 
 ${\mathrm {W}}_n$
 the Weyl group of
${\mathrm {W}}_n$
 the Weyl group of 
 $G_n$
 acting on
$G_n$
 acting on 
 ${\mathfrak t}_n$
 by the permutation action. We let
${\mathfrak t}_n$
 by the permutation action. We let 
 $\mathfrak{c}_n={\mathfrak t}_n//{\mathrm {W}}_n$
. We will identify
$\mathfrak{c}_n={\mathfrak t}_n//{\mathrm {W}}_n$
. We will identify 
 $\mathfrak{c}_n$
 with the space of degree
$\mathfrak{c}_n$
 with the space of degree 
 $n$
 monic polynomials in such a way that under the above identification, the Chevalley map
$n$
 monic polynomials in such a way that under the above identification, the Chevalley map 
 $\chi _n:{\mathfrak g}_n\to {\mathfrak g}_n//G_n\simeq {\mathfrak c}_n$
 becomes the map sending a matrix to its characteristic polynomial. We will fix the coordinates
$\chi _n:{\mathfrak g}_n\to {\mathfrak g}_n//G_n\simeq {\mathfrak c}_n$
 becomes the map sending a matrix to its characteristic polynomial. We will fix the coordinates 
 $(c_1, \ldots , c_{n})$
 on
$(c_1, \ldots , c_{n})$
 on 
 ${\mathfrak c}_n$
 given by the coefficients of a degree
${\mathfrak c}_n$
 given by the coefficients of a degree 
 $n$
 monic polynomial listed in increasing degree. We will identify
$n$
 monic polynomial listed in increasing degree. We will identify 
 ${\mathfrak g}_n\simeq {\mathfrak g}_n^*$
 using the trace paring
${\mathfrak g}_n\simeq {\mathfrak g}_n^*$
 using the trace paring 
 ${\mathfrak g}_n\times {\mathfrak g}_n\to {\mathbb C}$
,
${\mathfrak g}_n\times {\mathfrak g}_n\to {\mathbb C}$
, 
 $(A,B)\mapsto \textrm{tr}(AB)$
.
$(A,B)\mapsto \textrm{tr}(AB)$
.
 Let 
 $\mathbb H= \{a+ib+jc+kd\}$
 denote the quaternions. Consider the quaternionic vector space
$\mathbb H= \{a+ib+jc+kd\}$
 denote the quaternions. Consider the quaternionic vector space 
 ${\mathbb H}^n$
 where
${\mathbb H}^n$
 where 
 $\mathbb H$
 acts via right multiplication. Let
$\mathbb H$
 acts via right multiplication. Let 
 $G_{n,{\mathbb H}}$
 be the Lie group of
$G_{n,{\mathbb H}}$
 be the Lie group of 
 $\mathbb H$
-linear invertible endomorphisms of
$\mathbb H$
-linear invertible endomorphisms of 
 ${\mathbb H}^n$
, which can be identified with the space
${\mathbb H}^n$
, which can be identified with the space 
 ${\textrm{GL}}_n({\mathbb H})$
 of
${\textrm{GL}}_n({\mathbb H})$
 of 
 $n\times n$
 invertible quaternionic matrices, and let
$n\times n$
 invertible quaternionic matrices, and let 
 $\mathfrak{g}_{n,{\mathbb H}}$
 be the Lie algebra of
$\mathfrak{g}_{n,{\mathbb H}}$
 be the Lie algebra of 
 $\mathbb H$
-linear endomorphisms of
$\mathbb H$
-linear endomorphisms of 
 ${\mathbb H}^n$
, which can be identified with the space
${\mathbb H}^n$
, which can be identified with the space 
 $\mathfrak{gl}_n({\mathbb H})$
 of
$\mathfrak{gl}_n({\mathbb H})$
 of 
 $n\times n$
 quaternionic matrices.
$n\times n$
 quaternionic matrices.
 Using the identification 
 ${\mathbb C}^{2n}\simeq {\mathbb H}^n$
 sending
${\mathbb C}^{2n}\simeq {\mathbb H}^n$
 sending 
 $(z,w)\to q=z+jw$
 for
$(z,w)\to q=z+jw$
 for 
 $z,w\in {\mathbb C}^n$
, one can realize
$z,w\in {\mathbb C}^n$
, one can realize 
 $G_{n,{\mathbb H}}$
 as a real form of
$G_{n,{\mathbb H}}$
 as a real form of 
 $G_n$
. Specifically, the endomorphism of
$G_n$
. Specifically, the endomorphism of 
 ${\mathbb H}^n$
 sending
${\mathbb H}^n$
 sending 
 $q\to qj$
 corresponds to the endomorphism of
$q\to qj$
 corresponds to the endomorphism of 
 ${\mathbb C}^{2n}$
 sending
${\mathbb C}^{2n}$
 sending
 \begin{equation} (z,w)\to (-\bar w,\bar z), \end{equation}
\begin{equation} (z,w)\to (-\bar w,\bar z), \end{equation}
and we can identify 
 ${\mathfrak g}_{n,{\mathbb H}}$
 and
${\mathfrak g}_{n,{\mathbb H}}$
 and 
 $G_{n,{\mathbb H}}$
 as the subsets of
$G_{n,{\mathbb H}}$
 as the subsets of 
 ${\mathfrak g}_n$
 and
${\mathfrak g}_n$
 and 
 $G_n$
 consisting of
$G_n$
 consisting of 
 $\mathbb C$
-linear endomorphisms of
$\mathbb C$
-linear endomorphisms of 
 ${\mathbb C}^{2n}$
 that commute with the map (3.1). Equivalently, consider the
${\mathbb C}^{2n}$
 that commute with the map (3.1). Equivalently, consider the 
 $2n\times 2n$
 matrix
$2n\times 2n$
 matrix
 \begin{equation*}S_n=\begin{pmatrix} 0 & \quad -\textrm{Id}_n \\ \textrm{Id}_n &\quad 0 \end{pmatrix}\end{equation*}
\begin{equation*}S_n=\begin{pmatrix} 0 & \quad -\textrm{Id}_n \\ \textrm{Id}_n &\quad 0 \end{pmatrix}\end{equation*}
where 
 $\textrm{Id}_n$
 is the
$\textrm{Id}_n$
 is the 
 $n\times n$
 identity matrix. Then the endomorphism
$n\times n$
 identity matrix. Then the endomorphism 
 $\eta$
 of
$\eta$
 of 
 ${\mathfrak g}_{2n}$
 (respectively,
${\mathfrak g}_{2n}$
 (respectively, 
 $G_{2n}$
) sending
$G_{2n}$
) sending 
 $X\in {\mathfrak g}_{2n}$
 (respectively,
$X\in {\mathfrak g}_{2n}$
 (respectively, 
 $X\in G_{2n}$
) to
$X\in G_{2n}$
) to
 \begin{equation*}\eta (X)=S_n\overline XS_n^{-1}\end{equation*}
\begin{equation*}\eta (X)=S_n\overline XS_n^{-1}\end{equation*}
defines a real form of 
 ${\mathfrak g}_{2n}$
 (respectively,
${\mathfrak g}_{2n}$
 (respectively, 
 $G_{2n}$
), that is, an anti-holomorphic conjugation on
$G_{2n}$
), that is, an anti-holomorphic conjugation on 
 ${\mathfrak g}_{2n}$
 (respectively,
${\mathfrak g}_{2n}$
 (respectively, 
 $G_{2n}$
), and
$G_{2n}$
), and 
 ${\mathfrak g}_{n,{\mathbb H}}$
 and
${\mathfrak g}_{n,{\mathbb H}}$
 and 
 $G_{n,{\mathbb H}}$
 are the
$G_{n,{\mathbb H}}$
 are the 
 $\eta$
-fixed points in
$\eta$
-fixed points in 
 ${\mathfrak g}_{2n}$
 and
${\mathfrak g}_{2n}$
 and 
 $G_{2n}$
. Concretely,
$G_{2n}$
. Concretely, 
 ${\mathfrak g}_{n,{\mathbb H}}$
 (respectively,
${\mathfrak g}_{n,{\mathbb H}}$
 (respectively, 
 $G_{n,{\mathbb H}}$
) consists of
$G_{n,{\mathbb H}}$
) consists of 
 $2n\times 2n$
 matrices (respectively, invertible matrices) of the form
$2n\times 2n$
 matrices (respectively, invertible matrices) of the form
 \begin{equation*}\begin{pmatrix} A & \quad B \\ -\overline B &\quad \overline A \end{pmatrix}\end{equation*}
\begin{equation*}\begin{pmatrix} A & \quad B \\ -\overline B &\quad \overline A \end{pmatrix}\end{equation*}
where 
 $A,B\in {\mathfrak g}_n$
.
$A,B\in {\mathfrak g}_n$
.
 We denote by 
 ${\mathfrak t}_{n,{\mathbb H}}\subset {\mathfrak g}_{n,{\mathbb H}}$
 (respectively,
${\mathfrak t}_{n,{\mathbb H}}\subset {\mathfrak g}_{n,{\mathbb H}}$
 (respectively, 
 $T_{n,{\mathbb H}}\subset G_{n,{\mathbb H}}$
) the Cartan subalgebra (respectively, Cartan subgroup) consisting of matrices (respectively, invertible matrices)
$T_{n,{\mathbb H}}\subset G_{n,{\mathbb H}}$
) the Cartan subalgebra (respectively, Cartan subgroup) consisting of matrices (respectively, invertible matrices)
 \begin{equation*}\begin{pmatrix} A & \quad 0 \\ 0 &\quad \overline A \end{pmatrix}\end{equation*}
\begin{equation*}\begin{pmatrix} A & \quad 0 \\ 0 &\quad \overline A \end{pmatrix}\end{equation*}
where 
 $A\in {\mathfrak t}_n$
.
$A\in {\mathfrak t}_n$
.
 We denote by 
 $P_{n,{\mathbb H}}=M_{n,{\mathbb H}} A_{n,{\mathbb H}} N_{n,{\mathbb H}}$
 the standard minimal parabolic subgroup of
$P_{n,{\mathbb H}}=M_{n,{\mathbb H}} A_{n,{\mathbb H}} N_{n,{\mathbb H}}$
 the standard minimal parabolic subgroup of 
 $G_{n,{\mathbb H}}$
 consisting of invertible upper triangular quaternionic matrices and
$G_{n,{\mathbb H}}$
 consisting of invertible upper triangular quaternionic matrices and 
 ${\mathfrak p}_{n,{\mathbb H}}=\mathfrak{m}_{n,{\mathbb H}}\oplus {\mathfrak a}_{n,{\mathbb H}}\oplus \mathfrak{n}_{n,{\mathbb H}}$
 its Lie algebra.
${\mathfrak p}_{n,{\mathbb H}}=\mathfrak{m}_{n,{\mathbb H}}\oplus {\mathfrak a}_{n,{\mathbb H}}\oplus \mathfrak{n}_{n,{\mathbb H}}$
 its Lie algebra.
3.2 Symplectic group
 According to the Cartan classification of real forms, the conjugation 
 $\eta$
 corresponds to a holomorphic involution
$\eta$
 corresponds to a holomorphic involution 
 $\theta$
 on
$\theta$
 on 
 $G_{2n}$
 (respectively,
$G_{2n}$
 (respectively, 
 $\mathbb C$
-linear involution of
$\mathbb C$
-linear involution of 
 ${\mathfrak g}_{2n}$
) characterized by the property that
${\mathfrak g}_{2n}$
) characterized by the property that 
 $\eta \circ \theta =\theta \circ \eta$
 is a compact real form, that is, the fixed-point subgroup (respectively, subalgebra) of
$\eta \circ \theta =\theta \circ \eta$
 is a compact real form, that is, the fixed-point subgroup (respectively, subalgebra) of 
 $\eta _c:=\eta \circ \theta$
,
$\eta _c:=\eta \circ \theta$
,
 \begin{equation*}G_c=(G_{2n})^{\eta _c}\quad \big(\mbox{respectively,}\ {\mathfrak g}_c=({\mathfrak g}_{2n})^{\eta _c}\big),\end{equation*}
\begin{equation*}G_c=(G_{2n})^{\eta _c}\quad \big(\mbox{respectively,}\ {\mathfrak g}_c=({\mathfrak g}_{2n})^{\eta _c}\big),\end{equation*}
is compact. In our case, we will take 
 $\theta$
 to be
$\theta$
 to be
 \begin{equation*}\theta (X)=S_n(X^t)^{-1}S_n^{-1}\quad \big(\mbox{respectively,}\ \theta (X)=-S_n(X^t)S_n^{-1}\big),\end{equation*}
\begin{equation*}\theta (X)=S_n(X^t)^{-1}S_n^{-1}\quad \big(\mbox{respectively,}\ \theta (X)=-S_n(X^t)S_n^{-1}\big),\end{equation*}
where 
 $X\in G_{2n}$
 (respectively,
$X\in G_{2n}$
 (respectively, 
 $X\in {\mathfrak g}_{2n}$
), and we have
$X\in {\mathfrak g}_{2n}$
), and we have
 \begin{equation*}\eta _c(X)=(\overline X^t)^{-1}\quad \big(\mbox{respectively,}\ \theta (X)=-\overline X^t\big), \end{equation*}
\begin{equation*}\eta _c(X)=(\overline X^t)^{-1}\quad \big(\mbox{respectively,}\ \theta (X)=-\overline X^t\big), \end{equation*}
and the corresponding compact subgroup 
 $G_c=(G_{2n})^{\eta _c}$
 is the group of
$G_c=(G_{2n})^{\eta _c}$
 is the group of 
 $2n\times 2n$
 unitary matrices.
$2n\times 2n$
 unitary matrices.
 The 
 $\theta$
-fixed-point subgroup
$\theta$
-fixed-point subgroup 
 $K=(G_{2n})^\theta =\textrm{Sp}_{2n}$
 is the symplectic group of rank
$K=(G_{2n})^\theta =\textrm{Sp}_{2n}$
 is the symplectic group of rank 
 $n$
, and the intersection
$n$
, and the intersection
 \begin{equation*}K_c:= \textrm{Sp}_{2n}\cap G_c=\textrm{Sp}(n)\end{equation*}
\begin{equation*}K_c:= \textrm{Sp}_{2n}\cap G_c=\textrm{Sp}(n)\end{equation*}
is the compact symplectic group. The Lie algebras 
 $\mathfrak{k}_{}$
 and
$\mathfrak{k}_{}$
 and 
 $\mathfrak{k}_c$
 consist of matrices
$\mathfrak{k}_c$
 consist of matrices
 \begin{equation*}\begin{pmatrix} A& \quad B\\ C& \quad -A^t \end{pmatrix},\end{equation*}
\begin{equation*}\begin{pmatrix} A& \quad B\\ C& \quad -A^t \end{pmatrix},\end{equation*}
where 
 $A,B, C\in \mathfrak{g}_n$
 satisfy
$A,B, C\in \mathfrak{g}_n$
 satisfy 
 $B=B^t$
 and
$B=B^t$
 and 
 $C=C^t$
 for
$C=C^t$
 for 
 $\mathfrak{k}_{}$
 and the additional conditions
$\mathfrak{k}_{}$
 and the additional conditions 
 $A=-\bar A^t$
 and
$A=-\bar A^t$
 and 
 $C=-\overline B$
 for
$C=-\overline B$
 for 
 $\mathfrak{k}_c$
.
$\mathfrak{k}_c$
.
 Recall the Cartan decompostion of the Lie algebra 
 ${\mathfrak g}_{2n}=\mathfrak{k}\oplus \mathfrak{p}$
 where
${\mathfrak g}_{2n}=\mathfrak{k}\oplus \mathfrak{p}$
 where 
 $\mathfrak{p}$
 is the
$\mathfrak{p}$
 is the 
 $(-1)$
-eigenspace of
$(-1)$
-eigenspace of 
 $\theta$
. The Cartan decompostion induces a decomposition
$\theta$
. The Cartan decompostion induces a decomposition 
 ${\mathfrak t}_{2n}={\mathfrak t}\oplus \mathfrak{s}$
 where
${\mathfrak t}_{2n}={\mathfrak t}\oplus \mathfrak{s}$
 where 
 ${\mathfrak t}={\mathfrak t}_{2n}\cap \mathfrak{k}$
 is a Cartan subalgebra of
${\mathfrak t}={\mathfrak t}_{2n}\cap \mathfrak{k}$
 is a Cartan subalgebra of 
 $\mathfrak{k}$
 consisting of diagonal matrices of the form
$\mathfrak{k}$
 consisting of diagonal matrices of the form
 \begin{equation*}\textrm{diag}(h_1,\ldots ,h_n,-h_1,\ldots ,-h_n)\end{equation*}
\begin{equation*}\textrm{diag}(h_1,\ldots ,h_n,-h_1,\ldots ,-h_n)\end{equation*}
and 
 $\mathfrak{s}=\mathfrak{t}_{2n}\cap {\mathfrak p}\subset {\mathfrak p}$
 is a maximal abelian subspace of
$\mathfrak{s}=\mathfrak{t}_{2n}\cap {\mathfrak p}\subset {\mathfrak p}$
 is a maximal abelian subspace of 
 $\mathfrak p$
 consisting of diagonal matrices of the form
$\mathfrak p$
 consisting of diagonal matrices of the form
 \begin{equation*}\textrm{diag}(s_1,\ldots ,s_n,s_1,\ldots ,s_n).\end{equation*}
\begin{equation*}\textrm{diag}(s_1,\ldots ,s_n,s_1,\ldots ,s_n).\end{equation*}
We denote by 
 $\mathrm {W}$
 the Weyl group of
$\mathrm {W}$
 the Weyl group of 
 $K$
 and
$K$
 and 
 ${\mathrm {W}}_{\mathfrak{s}}=N_K(\mathfrak{s})/Z_K(\mathfrak{s})$
 the little Weyl group. We have
${\mathrm {W}}_{\mathfrak{s}}=N_K(\mathfrak{s})/Z_K(\mathfrak{s})$
 the little Weyl group. We have 
 ${\mathrm {W}}\simeq {\mathrm {W}}_n\ltimes \{\pm 1\}^n$
 and
${\mathrm {W}}\simeq {\mathrm {W}}_n\ltimes \{\pm 1\}^n$
 and 
 ${\mathrm {W}}_{\mathfrak{s}}\simeq {\mathrm {W}}_n$
. We let
${\mathrm {W}}_{\mathfrak{s}}\simeq {\mathrm {W}}_n$
. We let 
 ${\mathfrak c}=\mathfrak{t}//{\mathrm {W}}$
. Then the natural inclusion
${\mathfrak c}=\mathfrak{t}//{\mathrm {W}}$
. Then the natural inclusion 
 ${\mathfrak t}\to {\mathfrak t}_{2n}$
 gives rise to an embedding
${\mathfrak t}\to {\mathfrak t}_{2n}$
 gives rise to an embedding
 \begin{equation} \mathfrak{c}={\mathfrak t}//{\mathrm {W}}\longrightarrow {\mathfrak t}_{2n}//{\mathrm {W}}_{2n}=\mathfrak{c}_{2n} \end{equation}
\begin{equation} \mathfrak{c}={\mathfrak t}//{\mathrm {W}}\longrightarrow {\mathfrak t}_{2n}//{\mathrm {W}}_{2n}=\mathfrak{c}_{2n} \end{equation}
whose image consists of monic polynomials of degree 
 $2n$
 with vanishing odd terms.
$2n$
 with vanishing odd terms.
 Finally, we denote by 
 $X_{}=G_{2n}/K_{}$
 and
$X_{}=G_{2n}/K_{}$
 and 
 $X_c=G_c/K_c$
 the symmetric space and compact symmetric space associated to
$X_c=G_c/K_c$
 the symmetric space and compact symmetric space associated to 
 $K_{}$
 and
$K_{}$
 and 
 $K_c$
.
$K_c$
.
3.3 Notation related to root structure
 Let 
 $\Lambda _{n}={\textrm{Hom}}({\mathbb C}^\times , T_n)$
 be the coweight lattice of
$\Lambda _{n}={\textrm{Hom}}({\mathbb C}^\times , T_n)$
 be the coweight lattice of 
 $T_n$
 and let
$T_n$
 and let 
 $\Lambda _n^+$
 be the set of dominant coweights with respect to
$\Lambda _n^+$
 be the set of dominant coweights with respect to 
 $B_n$
. Let
$B_n$
. Let 
 $2\rho _n$
 be the sum of the positive roots of
$2\rho _n$
 be the sum of the positive roots of 
 $G_n$
 and let
$G_n$
 and let 
 $\langle \lambda ,2\rho _n\rangle \in {\mathbb Z}$
 be the natural paring for a coweight
$\langle \lambda ,2\rho _n\rangle \in {\mathbb Z}$
 be the natural paring for a coweight 
 $\lambda \in \Lambda _n$
.
$\lambda \in \Lambda _n$
.
 Let 
 $S\subset T$
 be the maximal split torus corresponding to the maximal abelian subspace
$S\subset T$
 be the maximal split torus corresponding to the maximal abelian subspace 
 $\mathfrak{s}\subset {\mathfrak p}$
, let
$\mathfrak{s}\subset {\mathfrak p}$
, let 
 $\Lambda _S={\textrm{Hom}}({\mathbb C}^\times ,S)$
 be the set of real coweights, and let
$\Lambda _S={\textrm{Hom}}({\mathbb C}^\times ,S)$
 be the set of real coweights, and let 
 $\Lambda _S^+=\Lambda _S\cap \Lambda _{2n}^+$
 be the set of dominant real coweights. There is a natural identification
$\Lambda _S^+=\Lambda _S\cap \Lambda _{2n}^+$
 be the set of dominant real coweights. There is a natural identification 
 $S\simeq T_n$
 sending
$S\simeq T_n$
 sending 
 $\textrm{diag}(s_1,\ldots s_n,s_1,\ldots ,s_n)$
 to
$\textrm{diag}(s_1,\ldots s_n,s_1,\ldots ,s_n)$
 to 
 $\textrm{diag}(s_1,\ldots ,s_n)$
 and hence natural identifications
$\textrm{diag}(s_1,\ldots ,s_n)$
 and hence natural identifications 
 $\Lambda _S\simeq \Lambda _n$
 and
$\Lambda _S\simeq \Lambda _n$
 and 
 $\Lambda _S^+\simeq \Lambda ^+_n$
.
$\Lambda _S^+\simeq \Lambda ^+_n$
.
3.4 Regular centralizers
3.4.1
 Recall the group scheme of centralizers 
 $I_{n}\to {\mathfrak g}_n$
 whose fiber over
$I_{n}\to {\mathfrak g}_n$
 whose fiber over 
 $x\in {\mathfrak g}_n$
 is the centralizer
$x\in {\mathfrak g}_n$
 is the centralizer 
 $(G_n)^x=\{g\in G_n\mid\textrm{Ad}_g(x)=x\}$
 of
$(G_n)^x=\{g\in G_n\mid\textrm{Ad}_g(x)=x\}$
 of 
 $x$
 in
$x$
 in 
 $G_n$
. Let
$G_n$
. Let 
 ${\mathfrak g}_n^{\rm reg}\subset {\mathfrak g}_n$
 be the open subset of regular elements in
${\mathfrak g}_n^{\rm reg}\subset {\mathfrak g}_n$
 be the open subset of regular elements in 
 ${\mathfrak g}_n$
. It is shown in [Reference NgôNgô06, § 3] that the base-change
${\mathfrak g}_n$
. It is shown in [Reference NgôNgô06, § 3] that the base-change 
 $I_n|_{{\mathfrak g}_n^{\rm reg}}\to {\mathfrak g}_n^{\rm reg}$
 is a smooth group scheme over
$I_n|_{{\mathfrak g}_n^{\rm reg}}\to {\mathfrak g}_n^{\rm reg}$
 is a smooth group scheme over 
 ${\mathfrak g}_n^{\rm reg}$
 which descends to a smooth group scheme
${\mathfrak g}_n^{\rm reg}$
 which descends to a smooth group scheme 
 $J_n\to {\mathfrak c}_n$
 over
$J_n\to {\mathfrak c}_n$
 over 
 ${\mathfrak c}_n$
, known as the group scheme of regular centralizers.
${\mathfrak c}_n$
, known as the group scheme of regular centralizers.
3.4.2
Consider the embedding
 \begin{equation} \tau :{\mathfrak g}_n\to {\mathfrak g}_{2n},\quad \tau (C)=\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}. \end{equation}
\begin{equation} \tau :{\mathfrak g}_n\to {\mathfrak g}_{2n},\quad \tau (C)=\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}. \end{equation}
Note that the map 
 $\tau$
 is
$\tau$
 is 
 $G_n$
-equivariant, where
$G_n$
-equivariant, where 
 $G_n$
 acts on
$G_n$
 acts on 
 ${\mathfrak g}_{2n}$
 via diagonal embedding
${\mathfrak g}_{2n}$
 via diagonal embedding 
 $\delta :G_n\to G_{2n}$
. Thus it induces an embedding on the invariant quotients (denoted again by
$\delta :G_n\to G_{2n}$
. Thus it induces an embedding on the invariant quotients (denoted again by 
 $\tau$
),
$\tau$
),
 \begin{equation} \tau :{\mathfrak c}_n={\mathfrak g}_n//G_n\to {\mathfrak g}_{2n}//G_{2n}\simeq {\mathfrak c}_{2n},\quad \tau (c_1,\ldots ,c_n)=(0,c_1,0,c_2,\ldots ,0,c_{n}), \end{equation}
\begin{equation} \tau :{\mathfrak c}_n={\mathfrak g}_n//G_n\to {\mathfrak g}_{2n}//G_{2n}\simeq {\mathfrak c}_{2n},\quad \tau (c_1,\ldots ,c_n)=(0,c_1,0,c_2,\ldots ,0,c_{n}), \end{equation}
whose image consists of monic polynomials of degree 
 $2n$
 with vanishing odd terms. Note that the image of
$2n$
 with vanishing odd terms. Note that the image of 
 $\tau$
 is equal to the image of the map
$\tau$
 is equal to the image of the map 
 ${\mathfrak c}={\mathfrak t}//{\mathrm {W}}\to {\mathfrak c}_{2n}={\mathfrak t}_{2n}//{\mathrm {W}}_{2n}$
 in (3.2), and hence there is a natural identification
${\mathfrak c}={\mathfrak t}//{\mathrm {W}}\to {\mathfrak c}_{2n}={\mathfrak t}_{2n}//{\mathrm {W}}_{2n}$
 in (3.2), and hence there is a natural identification
 \begin{equation} {\mathfrak c}\simeq {\mathfrak c}_n \end{equation}
\begin{equation} {\mathfrak c}\simeq {\mathfrak c}_n \end{equation}
such that 
 $\tau :{\mathfrak c}_n\simeq {\mathfrak c}\to {\mathfrak c}_{2n}$
.
$\tau :{\mathfrak c}_n\simeq {\mathfrak c}\to {\mathfrak c}_{2n}$
.
 Recall the group scheme of centralizers 
 $I_{n}$
 (respectively,
$I_{n}$
 (respectively, 
 $I_{2n}$
) over
$I_{2n}$
) over 
 ${\mathfrak g}_{n}$
 (respectively,
${\mathfrak g}_{n}$
 (respectively, 
 ${\mathfrak g}_{2n}$
) and the group scheme of regular centralizers
${\mathfrak g}_{2n}$
) and the group scheme of regular centralizers 
 $J_n$
 (respectively,
$J_n$
 (respectively, 
 $J_{2n}$
) over
$J_{2n}$
) over 
 ${\mathfrak c}_n$
 (respectively,
${\mathfrak c}_n$
 (respectively, 
 ${\mathfrak c}_{2n}$
).
${\mathfrak c}_{2n}$
).
Lemma 3.1. 
There is a natural closed embedding of group schemes, 
 $J_n\to J_{2n}$
, that fits into a commutative diagram
$J_n\to J_{2n}$
, that fits into a commutative diagram

where the bottom arrow is the map in ( 3.4 ).
Proof.
 We first claim that 
 $\tau \big({\mathfrak g}_n^{\rm reg}\big)={\mathfrak g}_{2n}^{\rm reg}\cap \tau ({\mathfrak g}_n)$
. Let
$\tau \big({\mathfrak g}_n^{\rm reg}\big)={\mathfrak g}_{2n}^{\rm reg}\cap \tau ({\mathfrak g}_n)$
. Let 
 $x=\tau (C)=\small\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}$
. If
$x=\tau (C)=\small\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}$
. If 
 $x$
 is in
$x$
 is in 
 ${\mathfrak g}_{2n}^{\rm reg}$
, then the centralizer
${\mathfrak g}_{2n}^{\rm reg}$
, then the centralizer 
 $(G_{2n})^x$
 of
$(G_{2n})^x$
 of 
 $x$
 in
$x$
 in 
 $G_{2n}$
 is abelian. By direct calculation, the inclusion
$G_{2n}$
 is abelian. By direct calculation, the inclusion 
 $G_n \to G_{2n}$
 takes the centralizer
$G_n \to G_{2n}$
 takes the centralizer 
 $(G_n)^C$
 of
$(G_n)^C$
 of 
 $C$
 in
$C$
 in 
 $G_n$
 into the centralizer
$G_n$
 into the centralizer 
 $(G_{2n})^x$
, and hence
$(G_{2n})^x$
, and hence 
 $(G_n)^C$
 is also abelian. Hence the characterization of regular elements for
$(G_n)^C$
 is also abelian. Hence the characterization of regular elements for 
 ${\mathfrak g}_n$
 implies that
${\mathfrak g}_n$
 implies that 
 $C\in {\mathfrak g}_{n}^{\textrm{reg}}$
. On the other hand, if
$C\in {\mathfrak g}_{n}^{\textrm{reg}}$
. On the other hand, if 
 $C\in {\mathfrak g}_n^{\rm reg}$
, then without loss of generality we can assume that
$C\in {\mathfrak g}_n^{\rm reg}$
, then without loss of generality we can assume that 
 $C$
 is a companion matrix, and an easy computation shows that
$C$
 is a companion matrix, and an easy computation shows that 
 $x$
 is in
$x$
 is in 
 ${\mathfrak g}_{2n}^{\rm reg}$
 (see (3.8)). The claim follows.
${\mathfrak g}_{2n}^{\rm reg}$
 (see (3.8)). The claim follows.
 Let 
 $I_n^{\rm reg}=I_n|_{{\mathfrak g}_n^{\rm reg}}$
 and
$I_n^{\rm reg}=I_n|_{{\mathfrak g}_n^{\rm reg}}$
 and 
 $I_{2n}^{\rm reg}=I_{2n}|_{{\mathfrak g}_{2n}^{\rm reg}}$
. Then the claim implies that we have a commutative diagram as follows.
$I_{2n}^{\rm reg}=I_{2n}|_{{\mathfrak g}_{2n}^{\rm reg}}$
. Then the claim implies that we have a commutative diagram as follows.

 Since 
 $J_{2n}\simeq I_{2n}^{\rm reg}//G_{2n}$
 is the descent of
$J_{2n}\simeq I_{2n}^{\rm reg}//G_{2n}$
 is the descent of 
 $I_{2n}^{\rm reg}$
 along the map
$I_{2n}^{\rm reg}$
 along the map 
 ${\mathfrak g}_{2n}^{\rm reg}\to {\mathfrak c}_{2n}$
, the restriction
${\mathfrak g}_{2n}^{\rm reg}\to {\mathfrak c}_{2n}$
, the restriction 
 $J_{2n}|_{{\mathfrak c}_n}$
 is the descent of
$J_{2n}|_{{\mathfrak c}_n}$
 is the descent of 
 $I_{2n}^{\rm reg}|_{{\mathfrak g}_n^{\rm reg}}$
 along the map
$I_{2n}^{\rm reg}|_{{\mathfrak g}_n^{\rm reg}}$
 along the map 
 ${\mathfrak g}_{n}^{\rm reg}\to {\mathfrak c}_n$
:
${\mathfrak g}_{n}^{\rm reg}\to {\mathfrak c}_n$
:
 \begin{equation*}J_{2n}|_{{\mathfrak c}_n}\simeq \big(I_{2n}^{\rm reg}//G_{2n}\big)\big|_{{\mathfrak c}_n}\simeq I_{2n}^{\rm reg}\big|_{{\mathfrak g}_n^{\rm reg}}// G_n.\end{equation*}
\begin{equation*}J_{2n}|_{{\mathfrak c}_n}\simeq \big(I_{2n}^{\rm reg}//G_{2n}\big)\big|_{{\mathfrak c}_n}\simeq I_{2n}^{\rm reg}\big|_{{\mathfrak g}_n^{\rm reg}}// G_n.\end{equation*}
Since the maps in (3.7) are compatible with the natural 
 $G_n$
-action and the desired map is the map on the GIT quotients,
$G_n$
-action and the desired map is the map on the GIT quotients,
 \begin{equation*}J_n\simeq I_n^{\rm reg}//G_n\longrightarrow I_{2n}^{\rm reg}|_{{\mathfrak g}_n^{\rm reg}}// G_n\simeq J_{2n}|_{{\mathfrak c}_n}.\end{equation*}
\begin{equation*}J_n\simeq I_n^{\rm reg}//G_n\longrightarrow I_{2n}^{\rm reg}|_{{\mathfrak g}_n^{\rm reg}}// G_n\simeq J_{2n}|_{{\mathfrak c}_n}.\end{equation*}
3.4.3 Kostant sections
 We give an alternative construction of the map 
 $J_n\to J_{2n}$
 in (3.6) using Kostant sections.
$J_n\to J_{2n}$
 in (3.6) using Kostant sections.
 Consider the following two ordered bases of 
 ${\mathbb C}^{2n}$
: the standard basis
${\mathbb C}^{2n}$
: the standard basis 
 $\{e_1=(1,0,\ldots,0),\ldots , e_{2n}=(0,\ldots ,0,1)\}$
 and the basis
$\{e_1=(1,0,\ldots,0),\ldots , e_{2n}=(0,\ldots ,0,1)\}$
 and the basis 
 $\{w_1=e_1, w_2=e_{3},\ldots ,w_n=e_{2n-1}, w_{n+1}=e_2,w_{n+2}=e_4,\ldots ,w_{2n}=e_{2n}\}$
. Let
$\{w_1=e_1, w_2=e_{3},\ldots ,w_n=e_{2n-1}, w_{n+1}=e_2,w_{n+2}=e_4,\ldots ,w_{2n}=e_{2n}\}$
. Let 
 $P\in G_{2n}$
 be the matrix associated to the linear map
$P\in G_{2n}$
 be the matrix associated to the linear map 
 $w_i \mapsto e_i$
 in the basis
$w_i \mapsto e_i$
 in the basis 
 $w_1,\ldots ,w_{2n}$
.
$w_1,\ldots ,w_{2n}$
.
 For any positive integer 
 $s$
, consider the Kostant section
$s$
, consider the Kostant section 
 $\kappa _s:{\mathfrak c}_s\to {\mathfrak g}_s$
 for
$\kappa _s:{\mathfrak c}_s\to {\mathfrak g}_s$
 for 
 $G_s$
 given by
$G_s$
 given by
 \begin{equation*} \kappa _s(c)=\begin{pmatrix} 0 & \quad 1 & \\ \vdots & \quad 0 & \quad \ddots & \\ \vdots & & \quad \ddots & 1 \\ -c_{s}& \quad -c_{s-1}& \quad \ldots & \quad -c_1 \end{pmatrix},\quad c=(c_1,\ldots ,c_s)\in {\mathfrak c}_s. \end{equation*}
\begin{equation*} \kappa _s(c)=\begin{pmatrix} 0 & \quad 1 & \\ \vdots & \quad 0 & \quad \ddots & \\ \vdots & & \quad \ddots & 1 \\ -c_{s}& \quad -c_{s-1}& \quad \ldots & \quad -c_1 \end{pmatrix},\quad c=(c_1,\ldots ,c_s)\in {\mathfrak c}_s. \end{equation*}
A direct computation shows that
 \begin{equation*} \begin{pmatrix} 0 & \quad 1 & \\ \vdots & \quad 0 & \quad \ddots & \\ \vdots & & \quad \ddots & \quad 1 \\ -c_{2n}& \quad -c_{2n-1}& \quad \ldots & \quad -c_1 \end{pmatrix}= P\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad D\end{pmatrix}P^{-1} \end{equation*}
\begin{equation*} \begin{pmatrix} 0 & \quad 1 & \\ \vdots & \quad 0 & \quad \ddots & \\ \vdots & & \quad \ddots & \quad 1 \\ -c_{2n}& \quad -c_{2n-1}& \quad \ldots & \quad -c_1 \end{pmatrix}= P\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad D\end{pmatrix}P^{-1} \end{equation*}
where
 \begin{equation*}C=\begin{pmatrix} 0 & \quad 1 & \\ \vdots & \quad 0 & \quad \ddots & \\ \vdots & & \quad \ddots & 1 \\ -c_{2n}& \quad -c_{2n-2}& \quad \ldots & \quad -c_2 \end{pmatrix},\quad D=\begin{pmatrix} 0 & \quad 0 & \quad \ldots & \quad 0 \\ \vdots & \quad \vdots & \quad \ldots & \quad 0 \\ 0 & \quad 0 & \quad \ldots & \quad 0 \\ -c_{2n-1}& \quad -c_{2n-3}& \quad \ldots & \quad -c_1\end{pmatrix}.\end{equation*}
\begin{equation*}C=\begin{pmatrix} 0 & \quad 1 & \\ \vdots & \quad 0 & \quad \ddots & \\ \vdots & & \quad \ddots & 1 \\ -c_{2n}& \quad -c_{2n-2}& \quad \ldots & \quad -c_2 \end{pmatrix},\quad D=\begin{pmatrix} 0 & \quad 0 & \quad \ldots & \quad 0 \\ \vdots & \quad \vdots & \quad \ldots & \quad 0 \\ 0 & \quad 0 & \quad \ldots & \quad 0 \\ -c_{2n-1}& \quad -c_{2n-3}& \quad \ldots & \quad -c_1\end{pmatrix}.\end{equation*}
It follows that for any 
 ${\mathfrak c}=(c_1,\ldots ,c_n)\in {\mathfrak c}_n$
 with
${\mathfrak c}=(c_1,\ldots ,c_n)\in {\mathfrak c}_n$
 with 
 $\tau (c)=(0,c_1,0,c_2,\ldots ,0,c_{n}) \in {\mathfrak c}_{2n}$
, we have
$\tau (c)=(0,c_1,0,c_2,\ldots ,0,c_{n}) \in {\mathfrak c}_{2n}$
, we have
 \begin{equation} \begin{pmatrix} 0 & \quad 1 & \quad \ldots & \quad \ldots & \quad \ldots & \quad \ldots & \quad 0\\ \vdots & \quad 0 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad \ddots \\ \vdots & \quad & \quad \ddots & \quad \ddots & \quad & \quad \\ \vdots & \quad & \quad \ddots & \quad \ddots & \quad \ddots & \quad & \quad 1\\ -c_{n}& \quad 0& \quad -c_{n-1} & \quad 0& \quad \ldots & \quad 0& \quad -c_1 \end{pmatrix}=P\begin{pmatrix}0& \quad \textrm{Id}_n\\ \kappa _n(c)& \quad 0\end{pmatrix}P^{-1}. \end{equation}
\begin{equation} \begin{pmatrix} 0 & \quad 1 & \quad \ldots & \quad \ldots & \quad \ldots & \quad \ldots & \quad 0\\ \vdots & \quad 0 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad \ddots \\ \vdots & \quad & \quad \ddots & \quad \ddots & \quad & \quad \\ \vdots & \quad & \quad \ddots & \quad \ddots & \quad \ddots & \quad & \quad 1\\ -c_{n}& \quad 0& \quad -c_{n-1} & \quad 0& \quad \ldots & \quad 0& \quad -c_1 \end{pmatrix}=P\begin{pmatrix}0& \quad \textrm{Id}_n\\ \kappa _n(c)& \quad 0\end{pmatrix}P^{-1}. \end{equation}
Thus there is the following commutative diagram.

In particular, we have
 \begin{equation*}\tau \circ \kappa _n:{\mathfrak c}_n\to {\mathfrak g}_{2n}^{\rm reg}.\end{equation*}
\begin{equation*}\tau \circ \kappa _n:{\mathfrak c}_n\to {\mathfrak g}_{2n}^{\rm reg}.\end{equation*}
The pullback of the group scheme 
 $I_{2n}^{\rm reg}$
 along the map above
$I_{2n}^{\rm reg}$
 along the map above 
 $\tau \circ \kappa _n$
 is isomorphic to
$\tau \circ \kappa _n$
 is isomorphic to
 \begin{equation*}(\tau \circ \kappa _n)^*\big(I_{2n}^{\rm reg}\big)\simeq \big((\kappa _{2n})^*\textrm{Ad}_{P^{-1}}^*\big(I_{2n}^{\rm reg}\big)\big)\big|_{{\mathfrak c}_n}\simeq (\kappa _{2n})^*\big(I_{2n}^{\rm reg}\big)\big|_{{\mathfrak c}_n}\simeq J_{2n}|_{{\mathfrak c}_n}, \end{equation*}
\begin{equation*}(\tau \circ \kappa _n)^*\big(I_{2n}^{\rm reg}\big)\simeq \big((\kappa _{2n})^*\textrm{Ad}_{P^{-1}}^*\big(I_{2n}^{\rm reg}\big)\big)\big|_{{\mathfrak c}_n}\simeq (\kappa _{2n})^*\big(I_{2n}^{\rm reg}\big)\big|_{{\mathfrak c}_n}\simeq J_{2n}|_{{\mathfrak c}_n}, \end{equation*}
and the desired map is given by pullback of (3.7) along the map 
 $\tau \circ \kappa _n$
:
$\tau \circ \kappa _n$
:
 \begin{equation} J_n\simeq \kappa _n^*\big(I_n^{\rm reg}\big)\longrightarrow \kappa _n^{*}\big(I^{\rm reg}_{2n}\big|_{{\mathfrak g}_n^{\rm reg}}\big)\simeq (\tau \circ \kappa _n)^*\big(I_{2n}^{\rm reg}\big)\simeq J_{2n}|_{{\mathfrak c}_n} . \end{equation}
\begin{equation} J_n\simeq \kappa _n^*\big(I_n^{\rm reg}\big)\longrightarrow \kappa _n^{*}\big(I^{\rm reg}_{2n}\big|_{{\mathfrak g}_n^{\rm reg}}\big)\simeq (\tau \circ \kappa _n)^*\big(I_{2n}^{\rm reg}\big)\simeq J_{2n}|_{{\mathfrak c}_n} . \end{equation}
3.4.4
 The identification 
 ${\mathfrak c}\simeq {\mathfrak c}_n$
 in (3.5) gives rise to a map
${\mathfrak c}\simeq {\mathfrak c}_n$
 in (3.5) gives rise to a map
 \begin{equation*}{\mathfrak t}\to {\mathfrak c}\simeq {\mathfrak c}_n\end{equation*}
\begin{equation*}{\mathfrak t}\to {\mathfrak c}\simeq {\mathfrak c}_n\end{equation*}
sending 
 $(t_1,\ldots ,t_n,-t_1,\ldots ,-t_n)\in {\mathfrak t}$
 to the coefficients
$(t_1,\ldots ,t_n,-t_1,\ldots ,-t_n)\in {\mathfrak t}$
 to the coefficients 
 $(c_1,\ldots ,c_n)\in {\mathfrak c}_n$
 of the monic polynomial
$(c_1,\ldots ,c_n)\in {\mathfrak c}_n$
 of the monic polynomial 
 $f(x)=\prod _{i=1}^n\big(x-t_i^2\big)$
 of degree
$f(x)=\prod _{i=1}^n\big(x-t_i^2\big)$
 of degree 
 $n$
. We shall give a description of the pullback
$n$
. We shall give a description of the pullback
 \begin{equation} J_n\times _{{\mathfrak c}_n}{\mathfrak t}\to J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t} \end{equation}
\begin{equation} J_n\times _{{\mathfrak c}_n}{\mathfrak t}\to J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t} \end{equation}
of (3.6) along 
 ${\mathfrak t}\to {\mathfrak c}_n$
. Consider the map
${\mathfrak t}\to {\mathfrak c}_n$
. Consider the map
 \begin{equation} e^{T_{2n}}:{\mathfrak t}_{2n}\to {\mathfrak g}_{2n},\quad e^{T_{2n}}(t)= \begin{pmatrix} t_1 & \quad 1 & \quad \\ 0& \quad t_2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad \ldots & \quad 0 & \quad t_{2n} \end{pmatrix},\quad t=\textrm{diag}(t_1,\ldots ,t_{2n}). \end{equation}
\begin{equation} e^{T_{2n}}:{\mathfrak t}_{2n}\to {\mathfrak g}_{2n},\quad e^{T_{2n}}(t)= \begin{pmatrix} t_1 & \quad 1 & \quad \\ 0& \quad t_2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad \ldots & \quad 0 & \quad t_{2n} \end{pmatrix},\quad t=\textrm{diag}(t_1,\ldots ,t_{2n}). \end{equation}
Note that the image of 
 $e^{T_{2n}}$
 consists of regular elements. We have a commutative diagram
$e^{T_{2n}}$
 consists of regular elements. We have a commutative diagram

where the vertical arrows are the natural adjoint quotient maps. If follows that there is a canonical isomorphism
 \begin{equation*}J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}_{2n}\simeq (e^T)^*I_{2n}=(G_{2n}\times {\mathfrak t}_{2n})^{e^{T_{2n}}}\end{equation*}
\begin{equation*}J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}_{2n}\simeq (e^T)^*I_{2n}=(G_{2n}\times {\mathfrak t}_{2n})^{e^{T_{2n}}}\end{equation*}
where 
 $(G_{2n}\times {\mathfrak t}_{2n})^{e^{T_{2n}}} = \{( g, t) \in G_{2n}\times {\mathfrak t}_{2n} \, |\, \textrm{Ad}_g( e^{T_{2n}}(t)) = e^{T_{2n}}(t)\}$
 denotes the subgroup scheme of the constant group scheme
$(G_{2n}\times {\mathfrak t}_{2n})^{e^{T_{2n}}} = \{( g, t) \in G_{2n}\times {\mathfrak t}_{2n} \, |\, \textrm{Ad}_g( e^{T_{2n}}(t)) = e^{T_{2n}}(t)\}$
 denotes the subgroup scheme of the constant group scheme 
 $G_{2n}\times {\mathfrak t}_{2n}$
 over
$G_{2n}\times {\mathfrak t}_{2n}$
 over 
 ${\mathfrak t}_{2n}$
 of centralizers of the section
${\mathfrak t}_{2n}$
 of centralizers of the section 
 $e^{T_{2n}}$
.
$e^{T_{2n}}$
.
 Consider the restriction 
 $e^T=e^{T_{2n}}|_{{\mathfrak t}}:{\mathfrak t}\to {\mathfrak g}_{2n}$
. Concretely, we have
$e^T=e^{T_{2n}}|_{{\mathfrak t}}:{\mathfrak t}\to {\mathfrak g}_{2n}$
. Concretely, we have
 \begin{equation} e^{T}:{\mathfrak t}_{}\to {\mathfrak g}_{2n},\quad e^{T_{}}(t)= \begin{pmatrix} t_1 & \quad 1 & \quad 0& \quad \ldots & \quad \ldots & \quad 0 \\ 0& \quad \ddots & \quad \ddots & \quad & \quad & \quad \vdots \\ \vdots & \quad \ddots & \quad t_n & \quad \ddots & \quad & \quad \vdots \\ \vdots & \quad & \quad \ddots & \quad -t_1& \quad \ddots & \quad 0\\ \vdots & \quad & \quad & \quad \ddots & \quad \ddots & \quad 1\\ 0& \quad \ldots & \quad \ldots & \quad \ldots & \quad 0 & \quad -t_n \end{pmatrix},\quad t=\textrm{diag}(t_1,\ldots ,t_n,-t_1,\ldots ,-t_{2n}). \end{equation}
\begin{equation} e^{T}:{\mathfrak t}_{}\to {\mathfrak g}_{2n},\quad e^{T_{}}(t)= \begin{pmatrix} t_1 & \quad 1 & \quad 0& \quad \ldots & \quad \ldots & \quad 0 \\ 0& \quad \ddots & \quad \ddots & \quad & \quad & \quad \vdots \\ \vdots & \quad \ddots & \quad t_n & \quad \ddots & \quad & \quad \vdots \\ \vdots & \quad & \quad \ddots & \quad -t_1& \quad \ddots & \quad 0\\ \vdots & \quad & \quad & \quad \ddots & \quad \ddots & \quad 1\\ 0& \quad \ldots & \quad \ldots & \quad \ldots & \quad 0 & \quad -t_n \end{pmatrix},\quad t=\textrm{diag}(t_1,\ldots ,t_n,-t_1,\ldots ,-t_{2n}). \end{equation}
It is clear that
 \begin{equation*}J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}\simeq (e^T)^*I_{2n}=(G_{2n}\times {\mathfrak t}_{})^{e^{T_{}}}.\end{equation*}
\begin{equation*}J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}\simeq (e^T)^*I_{2n}=(G_{2n}\times {\mathfrak t}_{})^{e^{T_{}}}.\end{equation*}
Consider the map
 \begin{equation} e^T_X:{\mathfrak t}\to {\mathfrak g}_n,\quad e^T_X(t)=\begin{pmatrix} t_1^2 & \quad 1 & \quad \\ \vdots & \quad t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad t_n^2 \end{pmatrix}. \end{equation}
\begin{equation} e^T_X:{\mathfrak t}\to {\mathfrak g}_n,\quad e^T_X(t)=\begin{pmatrix} t_1^2 & \quad 1 & \quad \\ \vdots & \quad t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad t_n^2 \end{pmatrix}. \end{equation}
The image of 
 $e_X^T$
 consists of regular elements, and we have the following commutative diagram.
$e_X^T$
 consists of regular elements, and we have the following commutative diagram.

It follows that we have a canonical isomorphism
 \begin{equation*}J_n\times _{{\mathfrak c}_n}{\mathfrak t}\simeq \big(e^T_X\big)^*I_n=(G_n\times {\mathfrak t})^{e^T_X}\end{equation*}
\begin{equation*}J_n\times _{{\mathfrak c}_n}{\mathfrak t}\simeq \big(e^T_X\big)^*I_n=(G_n\times {\mathfrak t})^{e^T_X}\end{equation*}
of group schemes over 
 $\mathfrak t$
. For any
$\mathfrak t$
. For any 
 $t\in {\mathfrak t}$
, we have
$t\in {\mathfrak t}$
, we have
 \begin{equation} \tau \circ e^T_X(t)= \begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}, \quad C=\begin{pmatrix} t_1^2 & \quad 1 & \quad \\ \vdots & \quad t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad t_n^2 \end{pmatrix}. \end{equation}
\begin{equation} \tau \circ e^T_X(t)= \begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}, \quad C=\begin{pmatrix} t_1^2 & \quad 1 & \quad \\ \vdots & \quad t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad t_n^2 \end{pmatrix}. \end{equation}
Note that the elements 
 $\tau \circ e^T_X(t)$
 and
$\tau \circ e^T_X(t)$
 and 
 $e^T(t)$
 are regular and have the same characteristic polynomial and hence lie in the same
$e^T(t)$
 are regular and have the same characteristic polynomial and hence lie in the same 
 $G_{2n}$
-orbit. Pick an element
$G_{2n}$
-orbit. Pick an element 
 $g_t\in G_{2n}$
 such that
$g_t\in G_{2n}$
 such that
 \begin{equation*}e^T(t)=g_t\big(\tau \circ e^T_X(t)\big) g_t^{-1}.\end{equation*}
\begin{equation*}e^T(t)=g_t\big(\tau \circ e^T_X(t)\big) g_t^{-1}.\end{equation*}
Then the conjugation map 
 $\textrm{Ad}_{g_t}:G_{2n}\to G_{2n}, g\to g_t g g_t^{-1}$
 restricts to a map between the centralizers,
$\textrm{Ad}_{g_t}:G_{2n}\to G_{2n}, g\to g_t g g_t^{-1}$
 restricts to a map between the centralizers,
 \begin{equation*}(G_n)^{e^T_X(t)}\stackrel {\delta }\to (G_{2n})^{e^T(t)}\stackrel {\textrm{Ad}_{g_t}}\to (G_{2n})^{e^T(t)}.\end{equation*}
\begin{equation*}(G_n)^{e^T_X(t)}\stackrel {\delta }\to (G_{2n})^{e^T(t)}\stackrel {\textrm{Ad}_{g_t}}\to (G_{2n})^{e^T(t)}.\end{equation*}
Since centralizers of a regular element form a commutative group, the map above is independent of the choice of the element 
 $g_t$
 and hence is canonical. Then, as
$g_t$
 and hence is canonical. Then, as 
 $t$
 varies over
$t$
 varies over 
 $\mathfrak t$
, we obtain a map between the corresponding centralizer group schemes,
$\mathfrak t$
, we obtain a map between the corresponding centralizer group schemes,
 \begin{equation} J_n\times _{{\mathfrak c}_n}{\mathfrak t}\simeq (G_n\times {\mathfrak t})^{e^T_X}\stackrel {}\to (G_{2n}\times {\mathfrak t})^{e^T}\simeq J_{2n}\times _{{\mathfrak c}_X}{\mathfrak t}, \end{equation}
\begin{equation} J_n\times _{{\mathfrak c}_n}{\mathfrak t}\simeq (G_n\times {\mathfrak t})^{e^T_X}\stackrel {}\to (G_{2n}\times {\mathfrak t})^{e^T}\simeq J_{2n}\times _{{\mathfrak c}_X}{\mathfrak t}, \end{equation}
which is the map in (3.11).
 Alternatively, the assignment 
 $t\to g_t$
 gives rise to an element
$t\to g_t$
 gives rise to an element
 \begin{equation*}\Phi \in G_{2n}\otimes R_T,\quad \Phi (t)=g_t,\end{equation*}
\begin{equation*}\Phi \in G_{2n}\otimes R_T,\quad \Phi (t)=g_t,\end{equation*}
where we set 
 $R_T = \mathcal {O}({\mathfrak t})$
. If we regard the maps
$R_T = \mathcal {O}({\mathfrak t})$
. If we regard the maps 
 $e^T$
 and
$e^T$
 and 
 $\tau \circ e^T_X$
 as elements in
$\tau \circ e^T_X$
 as elements in 
 ${\mathfrak g}_{2n}\otimes R_T$
, we have
${\mathfrak g}_{2n}\otimes R_T$
, we have
 \begin{equation} e^T=\Phi \big(\tau \circ e^T_X\big)\Phi ^{-1}\in {\mathfrak g}_{2n}\otimes R_T \end{equation}
\begin{equation} e^T=\Phi \big(\tau \circ e^T_X\big)\Phi ^{-1}\in {\mathfrak g}_{2n}\otimes R_T \end{equation}
(we will give a canonical construction of the element 
 $\Phi$
; see Remark 5.2). Then the composition
$\Phi$
; see Remark 5.2). Then the composition
 \begin{equation*}\textrm{Ad}_\Phi \circ \delta : G_{n}\times {\mathfrak t}\to G_{2n}\times {\mathfrak t},\quad (g,t)\to \Phi (t)(\delta (g))\Phi ^{-1}(t)\end{equation*}
\begin{equation*}\textrm{Ad}_\Phi \circ \delta : G_{n}\times {\mathfrak t}\to G_{2n}\times {\mathfrak t},\quad (g,t)\to \Phi (t)(\delta (g))\Phi ^{-1}(t)\end{equation*}
restricts to the map (3.16) between the corresponding centralizer group schemes.
3.5 Dual group
 In [Reference NadlerNad05], the author associated to each real form 
 $G_{\mathbb R}$
 of a complex reductive group
$G_{\mathbb R}$
 of a complex reductive group 
 $G$
, or equivalently a symmetric space
$G$
, or equivalently a symmetric space 
 $X$
 of
$X$
 of 
 $G$
, a complex reductive group
$G$
, a complex reductive group 
 $\check G_X$
 together with a homomorphism
$\check G_X$
 together with a homomorphism 
 $\delta : \check G_X\to \check G$
. In the case of
$\delta : \check G_X\to \check G$
. In the case of 
 $G=G_{2n}$
 and
$G=G_{2n}$
 and 
 $G_{\mathbb R}=G_{n,{\mathbb H}}$
, or equivalently
$G_{\mathbb R}=G_{n,{\mathbb H}}$
, or equivalently 
 $X=G_{2n}/\textrm{Sp}_{2n}$
, we have
$X=G_{2n}/\textrm{Sp}_{2n}$
, we have 
 $\check G=G_{2n}$
 and
$\check G=G_{2n}$
 and 
 $\check G_X = G_n$
, and the homomorphism is the diagonal embedding
$\check G_X = G_n$
, and the homomorphism is the diagonal embedding
 \begin{equation} \delta :G_n\to G_{2n},\quad \delta (A)=\begin{pmatrix} A& \quad 0\\ 0& \quad A \end{pmatrix}. \end{equation}
\begin{equation} \delta :G_n\to G_{2n},\quad \delta (A)=\begin{pmatrix} A& \quad 0\\ 0& \quad A \end{pmatrix}. \end{equation}
Let 
 $P=LN$
 be the complexification of the minimal parabolic
$P=LN$
 be the complexification of the minimal parabolic 
 $P_{n,{\mathbb H}}$
. The Levi subgroup
$P_{n,{\mathbb H}}$
. The Levi subgroup 
 $L$
 consists of matrices of the form
$L$
 consists of matrices of the form
 \begin{equation*}L=\biggl\{\begin{pmatrix} A& \quad B\\ C& \quad D \end{pmatrix}\in G_{2n}\ \bigg|\ A,B,C,\text{and }D\text { are diagonal matrices}\biggr\}. \end{equation*}
\begin{equation*}L=\biggl\{\begin{pmatrix} A& \quad B\\ C& \quad D \end{pmatrix}\in G_{2n}\ \bigg|\ A,B,C,\text{and }D\text { are diagonal matrices}\biggr\}. \end{equation*}
Consider the principal 
 $\textrm{SL}_2$
 of
$\textrm{SL}_2$
 of 
 $L$
 given by
$L$
 given by
 \begin{equation*}\psi :\textrm{SL}_2\to L,\quad \psi \begin{pmatrix} a& \quad b\\ c& \quad d \end{pmatrix}=\begin{pmatrix} A& \quad B\\ C& \quad D \end{pmatrix},\end{equation*}
\begin{equation*}\psi :\textrm{SL}_2\to L,\quad \psi \begin{pmatrix} a& \quad b\\ c& \quad d \end{pmatrix}=\begin{pmatrix} A& \quad B\\ C& \quad D \end{pmatrix},\end{equation*}
where 
 $A=a\textrm{Id}_n$
,
$A=a\textrm{Id}_n$
, 
 $B=b\textrm{Id}_n$
, etc. The restriction of
$B=b\textrm{Id}_n$
, etc. The restriction of 
 $\psi$
 to the torus
$\psi$
 to the torus 
 ${\mathbb G}_m\subset \textrm{SL}_2$
 is the co-character
${\mathbb G}_m\subset \textrm{SL}_2$
 is the co-character
 \begin{align*} 2\rho _{L}:{\mathbb G}_m\to L, \quad 2\rho _{L}(h)=\textrm{diag}(h,\ldots ,h,h^{-1},\ldots ,h^{-1}) \end{align*}
\begin{align*} 2\rho _{L}:{\mathbb G}_m\to L, \quad 2\rho _{L}(h)=\textrm{diag}(h,\ldots ,h,h^{-1},\ldots ,h^{-1}) \end{align*}
corresponding to the sum of the positive roots of the Levi factor 
 $L$
. A direct computation shows that the image
$L$
. A direct computation shows that the image 
 $\psi (\textrm{SL}_2)\subset G_{2n}$
 centralizes the subgroup
$\psi (\textrm{SL}_2)\subset G_{2n}$
 centralizes the subgroup 
 $\delta (G_n)\subset G_{2n}$
, and hence we obtain a homomorphism
$\delta (G_n)\subset G_{2n}$
, and hence we obtain a homomorphism
 \begin{equation} \psi _X:\check G_X\times \textrm{SL}_2\to G_{2n},\quad \psi _X(g,y)\to \delta (g)\psi (y). \end{equation}
\begin{equation} \psi _X:\check G_X\times \textrm{SL}_2\to G_{2n},\quad \psi _X(g,y)\to \delta (g)\psi (y). \end{equation}
3.6 The partial Whittaker reduction
 Consider the identification 
 $T^*G_{2n}\simeq G_{2n}\times {\mathfrak g}_{2n}^*$
 by considering
$T^*G_{2n}\simeq G_{2n}\times {\mathfrak g}_{2n}^*$
 by considering 
 ${\mathfrak g}_{2n}^*$
 as left-invariant differential forms on
${\mathfrak g}_{2n}^*$
 as left-invariant differential forms on 
 $G_{2n}$
. The group
$G_{2n}$
. The group 
 $G_{2n}\times G_{2n}$
 acts on
$G_{2n}\times G_{2n}$
 acts on 
 $G_{2n}$
 via left and right multiplication, and the induced action on
$G_{2n}$
 via left and right multiplication, and the induced action on 
 $T^*G_{2n}\simeq G_{2n}\times {\mathfrak g}_{2n}^*$
 is given by
$T^*G_{2n}\simeq G_{2n}\times {\mathfrak g}_{2n}^*$
 is given by 
 $(g,h)(x,v)=(gxh^{-1},\textrm{Ad}_hv)$
. The moment map
$(g,h)(x,v)=(gxh^{-1},\textrm{Ad}_hv)$
. The moment map 
 $(\mu _l,\mu _r):T^*G_{2n}\to {\mathfrak g}_{2n}^*\times {\mathfrak g}_{2n}^*$
 with respect to the
$(\mu _l,\mu _r):T^*G_{2n}\to {\mathfrak g}_{2n}^*\times {\mathfrak g}_{2n}^*$
 with respect to the 
 $G_{2n}\times G_{2n}$
-action is given by
$G_{2n}\times G_{2n}$
-action is given by 
 $(\mu _l,\mu _r)(x,v)=(\textrm{Ad}_xv,-v)$
.
$(\mu _l,\mu _r)(x,v)=(\textrm{Ad}_xv,-v)$
.
 Consider the Shalika subgroup 
 $G_n\ltimes U$
 and the generic morphism
$G_n\ltimes U$
 and the generic morphism 
 $\psi$
 in (1.13). Let
$\psi$
 in (1.13). Let 
 ${\mathfrak g}_n\times {\mathfrak u}$
 be the Lie algebra of
${\mathfrak g}_n\times {\mathfrak u}$
 be the Lie algebra of 
 $G_n\ltimes U$
. Then one can view
$G_n\ltimes U$
. Then one can view 
 $\psi$
 as an element
$\psi$
 as an element 
 $\psi =(0,-\textrm{tr})$
 in
$\psi =(0,-\textrm{tr})$
 in 
 ${\mathfrak g}_n^*\times {\mathfrak u}^*$
:
${\mathfrak g}_n^*\times {\mathfrak u}^*$
:
 \begin{equation*}\psi \left(\begin{pmatrix} A& \quad 0\\ 0& \quad A \end{pmatrix},\begin{pmatrix} 0& \quad 0\\ C& \quad 0 \end{pmatrix}\right)=-\textrm{tr}(C).\end{equation*}
\begin{equation*}\psi \left(\begin{pmatrix} A& \quad 0\\ 0& \quad A \end{pmatrix},\begin{pmatrix} 0& \quad 0\\ C& \quad 0 \end{pmatrix}\right)=-\textrm{tr}(C).\end{equation*}
The moment map for the right 
 $G_n\ltimes U$
-action on
$G_n\ltimes U$
-action on 
 $T^*G_{2n}$
 is given by
$T^*G_{2n}$
 is given by
 \begin{equation*}\mu :T^*G_{2n}\stackrel {\mu _r}\to {\mathfrak g}_{2n}^*\to {\mathfrak g}_n^*\times {\mathfrak u}^*\end{equation*}
\begin{equation*}\mu :T^*G_{2n}\stackrel {\mu _r}\to {\mathfrak g}_{2n}^*\to {\mathfrak g}_n^*\times {\mathfrak u}^*\end{equation*}
where 
 $\mu _r$
 is the right moment map above and the second map is the natural restriction map. The partial Whittaker reduction
$\mu _r$
 is the right moment map above and the second map is the natural restriction map. The partial Whittaker reduction 
 $\check M$
 of
$\check M$
 of 
 $T^*G_{2n}$
 with respect to the right
$T^*G_{2n}$
 with respect to the right 
 $G_n\ltimes U$
-action is given by
$G_n\ltimes U$
-action is given by
 \begin{equation*}\check M=T^*(G_{2n}/G_n\ltimes U,\psi ):=\mu ^{-1}(\psi )/G_n\ltimes U.\end{equation*}
\begin{equation*}\check M=T^*(G_{2n}/G_n\ltimes U,\psi ):=\mu ^{-1}(\psi )/G_n\ltimes U.\end{equation*}
Lemma 3.2. 
There is an isomorphism 
 $\check M\simeq G_{2n}\times ^{G_n}{\mathfrak g}_n$
 fitting into a commutative diagram
$\check M\simeq G_{2n}\times ^{G_n}{\mathfrak g}_n$
 fitting into a commutative diagram

 
where the left vertical arrow is the left moment map 
 $\mu _l$
, the bottom arrow is induced by the trace pairing
$\mu _l$
, the bottom arrow is induced by the trace pairing 
 $(A,B)\to \textrm{tr}(AB)$
, and the right vertical map is given by
$(A,B)\to \textrm{tr}(AB)$
, and the right vertical map is given by
 \begin{equation*}(x,C)\mapsto \textrm{Ad}_x\begin{pmatrix} 0& \quad \textrm{Id}_n\\ C& \quad 0 \end{pmatrix}.\end{equation*}
\begin{equation*}(x,C)\mapsto \textrm{Ad}_x\begin{pmatrix} 0& \quad \textrm{Id}_n\\ C& \quad 0 \end{pmatrix}.\end{equation*}
Proof.
 We will identify 
 ${\mathfrak g}_{2n}^*$
 with
${\mathfrak g}_{2n}^*$
 with 
 ${\mathfrak g}_{2n}$
 via the trace pairing. The pre-image of
${\mathfrak g}_{2n}$
 via the trace pairing. The pre-image of 
 $\psi =(0,-\textrm{tr})\in {\mathfrak g}_n^*\times {\mathfrak u}^*$
 in
$\psi =(0,-\textrm{tr})\in {\mathfrak g}_n^*\times {\mathfrak u}^*$
 in 
 ${\mathfrak g}_{2n}^*\simeq {\mathfrak g}_{2n}$
 is given by
${\mathfrak g}_{2n}^*\simeq {\mathfrak g}_{2n}$
 is given by
 \begin{equation*}{\mathfrak g}_{2n,\psi }^*:=\bigg\{\bigg(\begin{array}{c@{\quad}c} A& -\textrm{Id}_n\\ C& -A \end{array}\bigg)\bigg| A,C\in {\mathfrak g}_n\bigg\}, \end{equation*}
\begin{equation*}{\mathfrak g}_{2n,\psi }^*:=\bigg\{\bigg(\begin{array}{c@{\quad}c} A& -\textrm{Id}_n\\ C& -A \end{array}\bigg)\bigg| A,C\in {\mathfrak g}_n\bigg\}, \end{equation*}
and it follows that
 \begin{equation*}\check M\simeq \mu ^{-1}(\psi )/G_n\ltimes U\simeq \mu _r^{-1}\big({\mathfrak g}_{2n,\psi }^*\big)\big/G_n\ltimes U\simeq G_{2n}\times ^{G_n\ltimes U} \big(-{\mathfrak g}_{2n,\psi }^*\big)\end{equation*}
\begin{equation*}\check M\simeq \mu ^{-1}(\psi )/G_n\ltimes U\simeq \mu _r^{-1}\big({\mathfrak g}_{2n,\psi }^*\big)\big/G_n\ltimes U\simeq G_{2n}\times ^{G_n\ltimes U} \big(-{\mathfrak g}_{2n,\psi }^*\big)\end{equation*}
(recall that 
 $\mu _r(x,v)=-v$
). On the other hand, a direct computation shows that the action of
$\mu _r(x,v)=-v$
). On the other hand, a direct computation shows that the action of 
 $U$
 on
$U$
 on 
 $-{\mathfrak g}_{2n,\psi }^*$
 is free and any
$-{\mathfrak g}_{2n,\psi }^*$
 is free and any 
 $U$
-orbit on
$U$
-orbit on 
 $-{\mathfrak g}_{2n,\psi }^*$
 contains a unique element of the form
$-{\mathfrak g}_{2n,\psi }^*$
 contains a unique element of the form 
 $\small\begin{pmatrix} 0& \quad \textrm{Id}_n\\ C& \quad 0 \end{pmatrix}$
 with
$\small\begin{pmatrix} 0& \quad \textrm{Id}_n\\ C& \quad 0 \end{pmatrix}$
 with 
 $C\in {\mathfrak g}_n$
.Footnote 
4
 Thus there is an isomorphism
$C\in {\mathfrak g}_n$
.Footnote 
4
 Thus there is an isomorphism
 \begin{equation*}\check M\simeq G_{2n}\times ^{G_n\ltimes U} \big({-}{\mathfrak g}_{2n,\psi }^*\big)\simeq G_{2n}\times ^{G_n}{\mathfrak g}_n\end{equation*}
\begin{equation*}\check M\simeq G_{2n}\times ^{G_n\ltimes U} \big({-}{\mathfrak g}_{2n,\psi }^*\big)\simeq G_{2n}\times ^{G_n}{\mathfrak g}_n\end{equation*}
such that the left moment map is given by 
 $\mu _l(x,C)=\textrm{Ad}_x\small\begin{pmatrix} 0& \quad \textrm{Id}_n\\ C& \quad 0 \end{pmatrix}$
. The lemma follows.
$\mu _l(x,C)=\textrm{Ad}_x\small\begin{pmatrix} 0& \quad \textrm{Id}_n\\ C& \quad 0 \end{pmatrix}$
. The lemma follows.
4. Constructible side
4.1 Twistor fibration
 Consider the complex projective space 
 $\mathbb P^{2n-1}$
 and the quaternionic projective space
$\mathbb P^{2n-1}$
 and the quaternionic projective space 
 $\mathbb {HP}^{n-1}$
. Recall the identification
$\mathbb {HP}^{n-1}$
. Recall the identification 
 ${\mathbb C}^{2n}\simeq \mathbb H^n$
 sending
${\mathbb C}^{2n}\simeq \mathbb H^n$
 sending
 \begin{equation*}(z,w)=(z_1,\ldots ,z_n,w_1,\ldots ,w_n)\rightarrow z+jw=(q_1=z_1+jw_1,\ldots ,q_n=z_n+jw_n).\end{equation*}
\begin{equation*}(z,w)=(z_1,\ldots ,z_n,w_1,\ldots ,w_n)\rightarrow z+jw=(q_1=z_1+jw_1,\ldots ,q_n=z_n+jw_n).\end{equation*}
If to each complex line in 
 ${\mathbb C}^{2n}\simeq \mathbb H^n$
 we associate the quaternionic line it generates, we get a map
${\mathbb C}^{2n}\simeq \mathbb H^n$
 we associate the quaternionic line it generates, we get a map
 \begin{equation} f:\mathbb P^{2n-1}\to \mathbb {HP}^{n-1},\quad [z,w]\to [q_1,\ldots ,q_n] \end{equation}
\begin{equation} f:\mathbb P^{2n-1}\to \mathbb {HP}^{n-1},\quad [z,w]\to [q_1,\ldots ,q_n] \end{equation}
between the corresponding complex and quaternionic projective spaces, called the twistor fibration for 
 $\mathbb {HP}^{n-1}$
. The fiber of
$\mathbb {HP}^{n-1}$
. The fiber of 
 $f$
 over a quaternionic line (a copy of
$f$
 over a quaternionic line (a copy of 
 ${\mathbb H}\simeq {\mathbb C}^2$
) consists of all complex lines generating that quaternionic line, which is a copy of
${\mathbb H}\simeq {\mathbb C}^2$
) consists of all complex lines generating that quaternionic line, which is a copy of 
 $\mathbb P^1\simeq S^2$
. Thus the twistor fibration
$\mathbb P^1\simeq S^2$
. Thus the twistor fibration 
 $f$
 is a fiber bundle with fiber
$f$
 is a fiber bundle with fiber 
 $\mathbb P^1$
. In the
$\mathbb P^1$
. In the 
 $n=2$
 case, we have
$n=2$
 case, we have 
 $\mathbb {HP}^{n-1}=\mathbb {HP}^{1}\simeq S^4$
 and the map (4.1) is the well-known twistor fibration
$\mathbb {HP}^{n-1}=\mathbb {HP}^{1}\simeq S^4$
 and the map (4.1) is the well-known twistor fibration
 \begin{equation*}f:\mathbb P^3\to S^4\end{equation*}
\begin{equation*}f:\mathbb P^3\to S^4\end{equation*}
for 
 $S^4$
.
$S^4$
.
 Consider the standard action of the complex torus 
 $T_{2n}$
 (respectively,
$T_{2n}$
 (respectively, 
 $T_n$
) on
$T_n$
) on 
 $\mathbb P^{2n-1}$
 (respectively,
$\mathbb P^{2n-1}$
 (respectively, 
 $\mathbb {HP}^{n-1}$
):
$\mathbb {HP}^{n-1}$
):
 \begin{align*}x\cdot [z_1,\ldots ,z_{2n}]&=[x_1z_1,\ldots ,x_{2n}z_{2n}],\quad x=(x_1,\ldots ,x_{2n})\in T_{2n}\\[6pt](\textrm{respectively,}\quad x\cdot [q_1,\ldots ,q_{n}]&=[x_1q_1,\ldots ,x_{n}q_{n}],\quad\quad \!x=(x_1,\ldots ,x_n)\in T_n).\end{align*}
\begin{align*}x\cdot [z_1,\ldots ,z_{2n}]&=[x_1z_1,\ldots ,x_{2n}z_{2n}],\quad x=(x_1,\ldots ,x_{2n})\in T_{2n}\\[6pt](\textrm{respectively,}\quad x\cdot [q_1,\ldots ,q_{n}]&=[x_1q_1,\ldots ,x_{n}q_{n}],\quad\quad \!x=(x_1,\ldots ,x_n)\in T_n).\end{align*}
Then the twistor map 
 $f:\mathbb P^{2n-1}\to \mathbb {HP}^{n-1}$
 is
$f:\mathbb P^{2n-1}\to \mathbb {HP}^{n-1}$
 is 
 $T_n$
-equivariant, where
$T_n$
-equivariant, where 
 $T_n$
 acts on
$T_n$
 acts on 
 $\mathbb P^{2n-1}$
 through the embedding
$\mathbb P^{2n-1}$
 through the embedding
 \begin{equation*}T_n\stackrel {\sim }\longrightarrow T_{n,{\mathbb H}}\subset T_{2n},\quad (x_1,\ldots ,x_n)\to (x_1,\ldots ,x_n,\bar x_1,\ldots ,\bar x_n)\end{equation*}
\begin{equation*}T_n\stackrel {\sim }\longrightarrow T_{n,{\mathbb H}}\subset T_{2n},\quad (x_1,\ldots ,x_n)\to (x_1,\ldots ,x_n,\bar x_1,\ldots ,\bar x_n)\end{equation*}
(recall that 
 $T_{n,{\mathbb H}}$
 is the Cartan subgroup of
$T_{n,{\mathbb H}}$
 is the Cartan subgroup of 
 $G_{n,{\mathbb H}}$
). Indeed, for any
$G_{n,{\mathbb H}}$
). Indeed, for any 
 $x=(x_1,\ldots ,x_n)\in T_n$
, we have
$x=(x_1,\ldots ,x_n)\in T_n$
, we have
 \begin{align*}f(x\cdot [z,w])&=f([x_1z_1,\ldots ,x_{n}z_{n},\bar x_1w_1,\ldots ,\bar x_nw_n])= [x_1z_1+j\bar x_1w_1,\ldots ,x_nz_n+j\bar x_nw_n]\\[4pt] &=[x_1z_1+x_1jw_1,\ldots ,x_nz_n+x_njw_n]=x\cdot [q_1,\ldots ,q_n]. \end{align*}
\begin{align*}f(x\cdot [z,w])&=f([x_1z_1,\ldots ,x_{n}z_{n},\bar x_1w_1,\ldots ,\bar x_nw_n])= [x_1z_1+j\bar x_1w_1,\ldots ,x_nz_n+j\bar x_nw_n]\\[4pt] &=[x_1z_1+x_1jw_1,\ldots ,x_nz_n+x_njw_n]=x\cdot [q_1,\ldots ,q_n]. \end{align*}
4.2 Equivariant cohomology of quaternionic projective spaces
 Consider the inverse action of 
 $T_{2n}$
 on
$T_{2n}$
 on 
 $\mathbb P^{2n-1}$
.Footnote 
5
 Recall the following well known description of the
$\mathbb P^{2n-1}$
.Footnote 
5
 Recall the following well known description of the 
 $T_{2n}$
-equivariant cohomology of
$T_{2n}$
-equivariant cohomology of 
 $\mathbb {P}^{2n-1}$
:
$\mathbb {P}^{2n-1}$
:
 \begin{equation} H^*_{T_{2n}}(\mathbb P^{2n-1})\simeq {\mathbb C}[t_1,\ldots ,t_{2n}][\xi ]/\prod _{i=1}^{2n}(\xi -t_i) \end{equation}
\begin{equation} H^*_{T_{2n}}(\mathbb P^{2n-1})\simeq {\mathbb C}[t_1,\ldots ,t_{2n}][\xi ]/\prod _{i=1}^{2n}(\xi -t_i) \end{equation}
where
 \begin{equation*}\xi =c_1^{T_{2n}}(\mathcal {O}(1))\in H^*_{T_{2n}}(\mathbb P^{2n-1})\end{equation*}
\begin{equation*}\xi =c_1^{T_{2n}}(\mathcal {O}(1))\in H^*_{T_{2n}}(\mathbb P^{2n-1})\end{equation*}
is the first equivariant Chern class of the line bundle 
 $\mathcal {O}(1)$
 over
$\mathcal {O}(1)$
 over 
 $\mathbb P^{2n-1}$
 and
$\mathbb P^{2n-1}$
 and 
 $H^*_{T_{2n}}(\textrm{pt})\simeq \mathcal {O}({\mathfrak t}_{2n})\simeq {\mathbb C}[t_1,\ldots ,t_{2n}]$
.
$H^*_{T_{2n}}(\textrm{pt})\simeq \mathcal {O}({\mathfrak t}_{2n})\simeq {\mathbb C}[t_1,\ldots ,t_{2n}]$
.
 The imbedding 
 $T_n\simeq T_{n,{\mathbb H}}\subset T_{2n}$
 gives rise to a map
$T_n\simeq T_{n,{\mathbb H}}\subset T_{2n}$
 gives rise to a map 
 $H^*_{T_{2n}}(\textrm{pt})\to H^*_{T_{n}}(\textrm{pt})$
, and a direct computation show that under the isomorphism
$H^*_{T_{2n}}(\textrm{pt})\to H^*_{T_{n}}(\textrm{pt})$
, and a direct computation show that under the isomorphism 
 ${\mathbb C}[t_1,\ldots ,t_{2n}]\simeq H^*_{T_{2n}}(\textrm{pt})$
 and
${\mathbb C}[t_1,\ldots ,t_{2n}]\simeq H^*_{T_{2n}}(\textrm{pt})$
 and 
 ${\mathbb C}[t_1,\ldots ,t_{n}]\simeq H^*_{T_{n}}(\textrm{pt})$
, the map is given by
${\mathbb C}[t_1,\ldots ,t_{n}]\simeq H^*_{T_{n}}(\textrm{pt})$
, the map is given by
 \begin{equation*}{\mathbb C}[t_1,\ldots ,t_{2n}]\to {\mathbb C}[t_1,\ldots ,t_{2n}]/(t_1+t_{n+1},t_2+t_{n+2},\ldots ,t_n+t_{2n})\simeq {\mathbb C}[t_1,\ldots ,t_n].\end{equation*}
\begin{equation*}{\mathbb C}[t_1,\ldots ,t_{2n}]\to {\mathbb C}[t_1,\ldots ,t_{2n}]/(t_1+t_{n+1},t_2+t_{n+2},\ldots ,t_n+t_{2n})\simeq {\mathbb C}[t_1,\ldots ,t_n].\end{equation*}
It follows that
 \begin{equation} H^*_{T_{n}}(\mathbb P^{2n-1})\simeq H^*_{T_{2n}}(\mathbb P^{2n-1})\otimes _{H^*_{T_{2n}}(\textrm{pt})}H^*_{T_{n}}(\textrm{pt})\simeq {\mathbb C}[t_1,\ldots ,t_n][\xi ]/\prod _{i=1}^{n}\big(\xi ^2-t^2_i\big). \end{equation}
\begin{equation} H^*_{T_{n}}(\mathbb P^{2n-1})\simeq H^*_{T_{2n}}(\mathbb P^{2n-1})\otimes _{H^*_{T_{2n}}(\textrm{pt})}H^*_{T_{n}}(\textrm{pt})\simeq {\mathbb C}[t_1,\ldots ,t_n][\xi ]/\prod _{i=1}^{n}\big(\xi ^2-t^2_i\big). \end{equation}
 Similarly, we consider the inverse 
 $T_n$
-action on
$T_n$
-action on 
 $\mathbb {HP}^{n-1}$
. Let
$\mathbb {HP}^{n-1}$
. Let 
 $\mathcal {O}_{\mathbb H}(-1)$
 be the tautological
$\mathcal {O}_{\mathbb H}(-1)$
 be the tautological 
 $\mathbb H$
-line bundle
$\mathbb H$
-line bundle 
 $\mathcal {O}_{\mathbb H}(-1)$
 over
$\mathcal {O}_{\mathbb H}(-1)$
 over 
 $\mathbb {HP}^{n-1}$
. It is canonically
$\mathbb {HP}^{n-1}$
. It is canonically 
 $T_n$
-equivariant, and we denote by
$T_n$
-equivariant, and we denote by
 \begin{equation*}\eta =-e^{T_n}(\mathcal {O}_{\mathbb H}(-1))\in H^4_{T_n}(\mathbb {HP}^{n-1})\end{equation*}
\begin{equation*}\eta =-e^{T_n}(\mathcal {O}_{\mathbb H}(-1))\in H^4_{T_n}(\mathbb {HP}^{n-1})\end{equation*}
the negative of the equivariant Euler class of 
 $\mathcal {O}_{\mathbb H}(-1)$
.
$\mathcal {O}_{\mathbb H}(-1)$
.
Lemma 4.1. There is an isomorphism
 \begin{equation*}H_{T_n}^*(\mathbb {HP}^{n-1})\simeq {\mathbb C}[t_1,\ldots t_n][\eta ]/\prod _{i=1}^n\big(\eta -t_i^2\big)\end{equation*}
\begin{equation*}H_{T_n}^*(\mathbb {HP}^{n-1})\simeq {\mathbb C}[t_1,\ldots t_n][\eta ]/\prod _{i=1}^n\big(\eta -t_i^2\big)\end{equation*}
making a diagram commute

 
where the bottom arrow is the natural 
 ${\mathbb C}[t_1,\ldots t_n]$
-linear embedding sending
${\mathbb C}[t_1,\ldots t_n]$
-linear embedding sending 
 $\eta$
 to
$\eta$
 to 
 $\xi ^2$
, that is, we have
$\xi ^2$
, that is, we have 
 $f^*(\eta )=\xi ^2$
.
$f^*(\eta )=\xi ^2$
.
Proof.
 The 
 $T_n$
-fixed points on
$T_n$
-fixed points on 
 $\mathbb {HP}^{n-1}$
 are
$\mathbb {HP}^{n-1}$
 are 
 $p_0=[1,0,\ldots ,0]_{\mathbb H}$
,
$p_0=[1,0,\ldots ,0]_{\mathbb H}$
, 
 $p_2=[0,1,0,\ldots ,0]_{\mathbb H}, \ldots, $
$p_2=[0,1,0,\ldots ,0]_{\mathbb H}, \ldots, $
 $p_n=[0,0,\ldots ,0,1]_{\mathbb H}$
. Write
$p_n=[0,0,\ldots ,0,1]_{\mathbb H}$
. Write 
 $s_i:\{p_i\}\to \mathbb {HP}^{n-1}$
 for the inclusion map. Then equivariant localization says that we have an injective map of rings
$s_i:\{p_i\}\to \mathbb {HP}^{n-1}$
 for the inclusion map. Then equivariant localization says that we have an injective map of rings
 \begin{equation*}\mathrm{Loc}=\bigoplus s^*_i:H_{T_n}^*(\mathbb {HP}^{n-1})\longrightarrow R_{T_n}^{\oplus n}.\end{equation*}
\begin{equation*}\mathrm{Loc}=\bigoplus s^*_i:H_{T_n}^*(\mathbb {HP}^{n-1})\longrightarrow R_{T_n}^{\oplus n}.\end{equation*}
The fiber of 
 $\mathcal {O}_{\mathbb H}(-1)|_{p_i}$
 over
$\mathcal {O}_{\mathbb H}(-1)|_{p_i}$
 over 
 $p_i$
 is the
$p_i$
 is the 
 $\mathbb H$
-line spanned by the
$\mathbb H$
-line spanned by the 
 $i$
th coordinate vector of
$i$
th coordinate vector of 
 $\mathbb H^n$
, and hence the action of
$\mathbb H^n$
, and hence the action of 
 $T_n$
 factors though the
$T_n$
 factors though the 
 $i$
th projection
$i$
th projection 
 $T_n\to {\mathbb G}_m$
,
$T_n\to {\mathbb G}_m$
, 
 $(x_1,\ldots ,x_n)\to x_i$
. It follows that, in terms of the coordinate
$(x_1,\ldots ,x_n)\to x_i$
. It follows that, in terms of the coordinate 
 ${\mathbb C}^2\simeq \mathcal {O}_{\mathbb H}(-1)|_{p_i}, (z_i,w_i)\to z_i+jw_i$
 (and hence a chosen orientation), the (inverse) action is given by
${\mathbb C}^2\simeq \mathcal {O}_{\mathbb H}(-1)|_{p_i}, (z_i,w_i)\to z_i+jw_i$
 (and hence a chosen orientation), the (inverse) action is given by 
 $x_i (z_i,w_i)=(x_i^{-1}z_i,\bar x_i^{-1}w_i)$
) and hence
$x_i (z_i,w_i)=(x_i^{-1}z_i,\bar x_i^{-1}w_i)$
) and hence
 \begin{equation*}s_i^*(\eta )= s_i^*\big({-}e^{T_n}(\mathcal {O}_{\mathbb H}(-1))\big)= -e^{T_n}(\mathcal {O}_{\mathbb H}(-1)|_{p_i})=t_i^2. \end{equation*}
\begin{equation*}s_i^*(\eta )= s_i^*\big({-}e^{T_n}(\mathcal {O}_{\mathbb H}(-1))\big)= -e^{T_n}(\mathcal {O}_{\mathbb H}(-1)|_{p_i})=t_i^2. \end{equation*}
Thus we have
 \begin{equation*}\mathrm{Loc}(\eta )=\mathrm{Loc}\big({-}e^{T_n}(\mathcal {O}_{\mathbb H}(-1))\big)=\big(t_1^2,\ldots ,t_n^2\big)\in R_{T_n}^{\oplus n},\end{equation*}
\begin{equation*}\mathrm{Loc}(\eta )=\mathrm{Loc}\big({-}e^{T_n}(\mathcal {O}_{\mathbb H}(-1))\big)=\big(t_1^2,\ldots ,t_n^2\big)\in R_{T_n}^{\oplus n},\end{equation*}
and it follows that 
 $\mathrm{Loc}\big(\prod _{i=1}^n\big(\eta -t_i^2\big)\big)=0$
; as
$\mathrm{Loc}\big(\prod _{i=1}^n\big(\eta -t_i^2\big)\big)=0$
; as 
 $\mathrm{Loc}$
 is injective, this implies
$\mathrm{Loc}$
 is injective, this implies 
 $\prod _{i=1}^n\big(\eta -t_i^2\big)=0$
.
$\prod _{i=1}^n\big(\eta -t_i^2\big)=0$
.
 To see that 
 $f^*(\eta )=\xi ^2$
, we observe that the pre-image
$f^*(\eta )=\xi ^2$
, we observe that the pre-image 
 $f^{-1}(p_i)$
 is isomorphic to the projection line
$f^{-1}(p_i)$
 is isomorphic to the projection line 
 $\mathbb P^1_i=[z_i,w_i]\subset \mathbb {P}^{2n-1}$
. The
$\mathbb P^1_i=[z_i,w_i]\subset \mathbb {P}^{2n-1}$
. The 
 $T_n$
-action preserves
$T_n$
-action preserves 
 $\mathbb P^1_i$
 and is given by
$\mathbb P^1_i$
 and is given by 
 $(x_1,\ldots ,x_n)[z_i,w_i]=\big[x_i^{-1}z_i,\bar x_i^{-1}w_i\big]$
. The localization map
$(x_1,\ldots ,x_n)[z_i,w_i]=\big[x_i^{-1}z_i,\bar x_i^{-1}w_i\big]$
. The localization map 
 $\mathrm{Loc}':H_{T_n}^*(\mathbb P^{2n-1})\to \bigoplus H_{T_n}^*\big(\mathbb P^{1}_i\big)={\mathbb C}[t_i][\xi _i]/\big(\xi _i^2-t_i^2\big)$
 is injective and we have
$\mathrm{Loc}':H_{T_n}^*(\mathbb P^{2n-1})\to \bigoplus H_{T_n}^*\big(\mathbb P^{1}_i\big)={\mathbb C}[t_i][\xi _i]/\big(\xi _i^2-t_i^2\big)$
 is injective and we have 
 $\mathrm{Loc}'(\xi ^2)=\big(\xi ^2_1,\ldots ,\xi ^2_n\big)$
. On the other hand, we have
$\mathrm{Loc}'(\xi ^2)=\big(\xi ^2_1,\ldots ,\xi ^2_n\big)$
. On the other hand, we have
 \begin{equation*}\mathrm{Loc}'(f^*\eta )=f^*(\mathrm{Loc}(\eta ))=f^*\big(\big(t_1^2,\ldots ,t^2_n\big)\big)= \big(t_1^2,\ldots ,t^2_n\big)= \big(\xi ^2_1,\ldots ,\xi ^2_n\big)\in \bigoplus H_{T_n}^*(\mathbb P^{1}_i)\end{equation*}
\begin{equation*}\mathrm{Loc}'(f^*\eta )=f^*(\mathrm{Loc}(\eta ))=f^*\big(\big(t_1^2,\ldots ,t^2_n\big)\big)= \big(t_1^2,\ldots ,t^2_n\big)= \big(\xi ^2_1,\ldots ,\xi ^2_n\big)\in \bigoplus H_{T_n}^*(\mathbb P^{1}_i)\end{equation*}
as 
 $\xi ^2_i=t_i^2$
 in
$\xi ^2_i=t_i^2$
 in 
 $H_{T_n}^*\big(\mathbb P^{1}_i\big)$
. We conclude that
$H_{T_n}^*\big(\mathbb P^{1}_i\big)$
. We conclude that 
 $\mathrm{Loc}'(\xi ^2)=\mathrm{Loc}'(f^*\eta )$
 and hence
$\mathrm{Loc}'(\xi ^2)=\mathrm{Loc}'(f^*\eta )$
 and hence 
 $\xi ^2=f^*\eta$
.
$\xi ^2=f^*\eta$
.
Remark 4.2. Here is an alternative argument. One can show that there is an isomorphism of 
 $T_n$
-equivariant complex vector bundles
$T_n$
-equivariant complex vector bundles
 \begin{equation*}f^*\mathcal {O}_{\mathbb H}(-1)\simeq \mathcal {O}(-1)\oplus \overline {\mathcal {O}(-1)}\end{equation*}
\begin{equation*}f^*\mathcal {O}_{\mathbb H}(-1)\simeq \mathcal {O}(-1)\oplus \overline {\mathcal {O}(-1)}\end{equation*}
over 
 $\mathbb P^{2n-1}$
. Here
$\mathbb P^{2n-1}$
. Here 
 $\overline {\mathcal {O}(-1)}$
 is the complex conjugate of
$\overline {\mathcal {O}(-1)}$
 is the complex conjugate of 
 $\mathcal {O}(-1)$
 (note that a choice of a hermitian metric on
$\mathcal {O}(-1)$
 (note that a choice of a hermitian metric on 
 $\mathcal {O}(-1)$
 induces an isomorphism
$\mathcal {O}(-1)$
 induces an isomorphism 
 $\overline {\mathcal {O}(-1)}\simeq \mathcal {O}(-1)^\vee \simeq \mathcal {O}(1)$
). Since
$\overline {\mathcal {O}(-1)}\simeq \mathcal {O}(-1)^\vee \simeq \mathcal {O}(1)$
). Since 
 $e^T(\overline {\mathcal {O}(-1)})=-e^T(\mathcal {O}(-1))=-\xi$
, it follows that
$e^T(\overline {\mathcal {O}(-1)})=-e^T(\mathcal {O}(-1))=-\xi$
, it follows that
 \begin{equation*}f^*(\eta )=-f^*(e^T(\mathcal {O}_\mathbb H(-1))= -e^T(\mathcal {O}(-1)\oplus \overline {\mathcal {O}(-1)})= e^T(\mathcal {O}(-1))^2=\xi ^2.\end{equation*}
\begin{equation*}f^*(\eta )=-f^*(e^T(\mathcal {O}_\mathbb H(-1))= -e^T(\mathcal {O}(-1)\oplus \overline {\mathcal {O}(-1)})= e^T(\mathcal {O}(-1))^2=\xi ^2.\end{equation*}
Now the lemma follows from the fact that 
 $f^*:H_{T_n}^*(\mathbb {HP}^{n-1})\to H_{T_n}^*(\mathbb {P}^{2n-1})$
 is injective and
$f^*:H_{T_n}^*(\mathbb {HP}^{n-1})\to H_{T_n}^*(\mathbb {P}^{2n-1})$
 is injective and 
 $f^*\big(\prod _{i=1}^n\big(\eta -t_i^2\big)\big)=\prod _{i=1}^n\big(f^*\eta -t_i^2\big)=\prod _{i=1}^n\big(\xi ^2-t_i^2\big)=0$
 in
$f^*\big(\prod _{i=1}^n\big(\eta -t_i^2\big)\big)=\prod _{i=1}^n\big(f^*\eta -t_i^2\big)=\prod _{i=1}^n\big(\xi ^2-t_i^2\big)=0$
 in 
 $H_{T_n}^*(\mathbb {P}^{2n-1})$
.
$H_{T_n}^*(\mathbb {P}^{2n-1})$
.
 Consider the pushforward functor 
 $f_*:D_{T_n}^b(\mathbb {P}^{2n-1})\to D_{T_n}^b(\mathbb {HP}^{n-1})$
.
$f_*:D_{T_n}^b(\mathbb {P}^{2n-1})\to D_{T_n}^b(\mathbb {HP}^{n-1})$
.
Lemma 4.3. 
We have that 
 $f_*\big({\mathbb C}_{\mathbb {P}^{2n-1}}\big)\simeq {\mathbb C}_{\mathbb {HP}^{n-1}}\oplus {\mathbb C}_{\mathbb {HP}^{n-1}}[-2]$
.
$f_*\big({\mathbb C}_{\mathbb {P}^{2n-1}}\big)\simeq {\mathbb C}_{\mathbb {HP}^{n-1}}\oplus {\mathbb C}_{\mathbb {HP}^{n-1}}[-2]$
.
Proof.
 Since 
 $f$
 is a
$f$
 is a 
 $\mathbb P^1$
-fibration, we have a distinguished triangle
$\mathbb P^1$
-fibration, we have a distinguished triangle
 \begin{equation*}{\mathbb C}_{\mathbb {HP}^{n-1}}\to f_*\big({\mathbb C}_{\mathbb {P}^{2n-1}}\big)\to {\mathbb C}_{\mathbb {HP}^{n-1}}[-2]\to {\mathbb C}_{\mathbb {HP}^{n-1}}[1]\end{equation*}
\begin{equation*}{\mathbb C}_{\mathbb {HP}^{n-1}}\to f_*\big({\mathbb C}_{\mathbb {P}^{2n-1}}\big)\to {\mathbb C}_{\mathbb {HP}^{n-1}}[-2]\to {\mathbb C}_{\mathbb {HP}^{n-1}}[1]\end{equation*}
and we need to show that it splits. But this follows from
 \begin{equation*}{\textrm{Hom}}\big({\mathbb C}_{\mathbb {HP}^{n-1}}[-2],{\mathbb C}_{\mathbb {HP}^{n-1}}[1]\big)\simeq \textrm{Ext}^3\big({\mathbb C}_{\mathbb {HP}^{n-1}},{\mathbb C}_{\mathbb {HP}^{n-1}}\big)\simeq H^3_{T_n}(\mathbb {HP}^{n-1})=0.\end{equation*}
\begin{equation*}{\textrm{Hom}}\big({\mathbb C}_{\mathbb {HP}^{n-1}}[-2],{\mathbb C}_{\mathbb {HP}^{n-1}}[1]\big)\simeq \textrm{Ext}^3\big({\mathbb C}_{\mathbb {HP}^{n-1}},{\mathbb C}_{\mathbb {HP}^{n-1}}\big)\simeq H^3_{T_n}(\mathbb {HP}^{n-1})=0.\end{equation*}
4.3 Two bases
 Consider the subvarieties 
 $\mathbb P^{i-1}=\{[z_1,\ldots ,z_i,0,\ldots ,0]\}\subset \mathbb P^{2n-1}$
 for
$\mathbb P^{i-1}=\{[z_1,\ldots ,z_i,0,\ldots ,0]\}\subset \mathbb P^{2n-1}$
 for 
 $i=1,\ldots ,2n$
. If we write
$i=1,\ldots ,2n$
. If we write 
 $[\mathbb P^{i-1}]\in H_{2i-2}^{T_{2n}}(\mathbb P^{2n-1})\simeq H^{4n-2i}_{T_{2n}}(\mathbb P^{2n-1})$
 for the corresponding fundamental class in the equi-variant Borel–Moore homology, then the collection
$[\mathbb P^{i-1}]\in H_{2i-2}^{T_{2n}}(\mathbb P^{2n-1})\simeq H^{4n-2i}_{T_{2n}}(\mathbb P^{2n-1})$
 for the corresponding fundamental class in the equi-variant Borel–Moore homology, then the collection 
 $\{[\mathbb P^{i-1}]\}_{i=1,\ldots ,2n}$
 forms a basis of the free
$\{[\mathbb P^{i-1}]\}_{i=1,\ldots ,2n}$
 forms a basis of the free 
 $R_{T_{2n}}$
-module
$R_{T_{2n}}$
-module 
 $H^{*}_{T_{2n}}(\mathbb P^{2n-1})$
. Moreover, one can check that the image of the fundamental class
$H^{*}_{T_{2n}}(\mathbb P^{2n-1})$
. Moreover, one can check that the image of the fundamental class 
 $[\mathbb P^{i-1}]$
 under the identification (4.3) is given by
$[\mathbb P^{i-1}]$
 under the identification (4.3) is given by
 \begin{equation*}\Upsilon :H^{*}_{T_{2n}}(\mathbb P^{2n-1})\simeq {\mathbb C}[t_1,\ldots ,t_{2n}][\xi ]/\prod _{i=1}^{2n}(\xi -t_i),\end{equation*}
\begin{equation*}\Upsilon :H^{*}_{T_{2n}}(\mathbb P^{2n-1})\simeq {\mathbb C}[t_1,\ldots ,t_{2n}][\xi ]/\prod _{i=1}^{2n}(\xi -t_i),\end{equation*}
 \begin{equation*}\Upsilon ([\mathbb P^{2n-1}])=1,\quad \Upsilon ([\mathbb P^{i-1}])=\prod _{s=i+1}^{2n}(\xi -t_s)\quad \mbox{for } i=1,\ldots ,2n-1.\end{equation*}
\begin{equation*}\Upsilon ([\mathbb P^{2n-1}])=1,\quad \Upsilon ([\mathbb P^{i-1}])=\prod _{s=i+1}^{2n}(\xi -t_s)\quad \mbox{for } i=1,\ldots ,2n-1.\end{equation*}
Consider the subvarieties 
 $\mathbb {HP}^{i-1}=\{[q_1,\ldots ,q_i,0,\ldots ,0]\}\subset \mathbb {HP}^{n-1}$
 for
$\mathbb {HP}^{i-1}=\{[q_1,\ldots ,q_i,0,\ldots ,0]\}\subset \mathbb {HP}^{n-1}$
 for 
 $i=1,\ldots ,n$
. If we write
$i=1,\ldots ,n$
. If we write 
 $[\mathbb {HP}^{i-1}]\in H_{4i-4,T_n}(\mathbb {HP}^{n-1})\simeq H^{4n-4i}_{T_n}(\mathbb P^{2n-1})$
 for the corresponding fundamental class in the equivariant Borel–Moore homology, then the collection
$[\mathbb {HP}^{i-1}]\in H_{4i-4,T_n}(\mathbb {HP}^{n-1})\simeq H^{4n-4i}_{T_n}(\mathbb P^{2n-1})$
 for the corresponding fundamental class in the equivariant Borel–Moore homology, then the collection 
 $\{[\mathbb {HP}^{i-1}]\}_{i=1,\ldots ,n}$
 forms a basis of the free
$\{[\mathbb {HP}^{i-1}]\}_{i=1,\ldots ,n}$
 forms a basis of the free 
 $R_{T_n}$
-module
$R_{T_n}$
-module 
 $H^{*}_{T_n}(\mathbb {HP}^{n-1})$
. Moreover, one can check that the image of the fundamental class
$H^{*}_{T_n}(\mathbb {HP}^{n-1})$
. Moreover, one can check that the image of the fundamental class 
 $[\mathbb {HP}^{i-1}]$
 under the identification in (4.1) is given by
$[\mathbb {HP}^{i-1}]$
 under the identification in (4.1) is given by
 \begin{equation*}\Upsilon _{\mathbb H}:H^{*}_{T_n}(\mathbb {HP}^{n-1})={\mathbb C}[t_1,\ldots ,t_n][\eta ]/\prod _{i=1}^n\big(\eta -t_i^2\big),\end{equation*}
\begin{equation*}\Upsilon _{\mathbb H}:H^{*}_{T_n}(\mathbb {HP}^{n-1})={\mathbb C}[t_1,\ldots ,t_n][\eta ]/\prod _{i=1}^n\big(\eta -t_i^2\big),\end{equation*}
 \begin{equation*}\Upsilon _{\mathbb H}([\mathbb {HP}^{n-1}])=1,\quad \Upsilon _{\mathbb H}([\mathbb {HP}^{i-1}])=\prod _{s=i+1}^{n}(\eta -t^2_s)\quad\mbox{for } i=1,\ldots ,n-1.\end{equation*}
\begin{equation*}\Upsilon _{\mathbb H}([\mathbb {HP}^{n-1}])=1,\quad \Upsilon _{\mathbb H}([\mathbb {HP}^{i-1}])=\prod _{s=i+1}^{n}(\eta -t^2_s)\quad\mbox{for } i=1,\ldots ,n-1.\end{equation*}
 The isomorphism 
 $f_*{\mathbb C}_{\mathbb P^{2n-1}}\simeq {\mathbb C}_{\mathbb {HP}^{n-1}}\oplus {\mathbb C}_{\mathbb {HP}^{n-1}}[-2]$
 gives rise to a decomposition
$f_*{\mathbb C}_{\mathbb P^{2n-1}}\simeq {\mathbb C}_{\mathbb {HP}^{n-1}}\oplus {\mathbb C}_{\mathbb {HP}^{n-1}}[-2]$
 gives rise to a decomposition
 \begin{equation*}\Upsilon ':H^*_{T_n}(\mathbb {HP}^{n-1})\oplus H^{*-2}_{T_n}(\mathbb {HP}^{n-1})\simeq H^*_{T_n}(\mathbb P^{2n-1})\simeq {\mathbb C}[t_1,\ldots ,t_n][\xi ]/\prod _{i=1}^n\big(\xi ^2-t_i^2\big),\end{equation*}
\begin{equation*}\Upsilon ':H^*_{T_n}(\mathbb {HP}^{n-1})\oplus H^{*-2}_{T_n}(\mathbb {HP}^{n-1})\simeq H^*_{T_n}(\mathbb P^{2n-1})\simeq {\mathbb C}[t_1,\ldots ,t_n][\xi ]/\prod _{i=1}^n\big(\xi ^2-t_i^2\big),\end{equation*}
and one can check that the image of the basis 
 $\{[\mathbb {HP}^{i-1}]\}\cup \{[\mathbb {HP}^{i-1}][2]\}$
 of
$\{[\mathbb {HP}^{i-1}]\}\cup \{[\mathbb {HP}^{i-1}][2]\}$
 of 
 $H^*_{T_n}(\mathbb {HP}^{n-1})\oplus H^{*-2}_{T_n}(\mathbb {HP}^{n-1})$
 under the map above is given by
$H^*_{T_n}(\mathbb {HP}^{n-1})\oplus H^{*-2}_{T_n}(\mathbb {HP}^{n-1})$
 under the map above is given by 
 \begin{align} \Upsilon '([\mathbb {HP}^{n-1}])=1,\quad \Upsilon '([\mathbb {HP}^{i-1}])=\prod _{s=i+1}^{n}\big(\xi ^2-t^2_s\big)\quad\mbox{for } i=1,\ldots ,n-1,\nonumber\\[-8pt]\\[-8pt]\Upsilon '([\mathbb {HP}^{n-1}[2]])=\xi ,\quad \Upsilon '([\mathbb {HP}^{i-1}[2]])=\xi \prod _{s=i+1}^{n}\big(\xi ^2-t^2_s\big)\quad\mbox{for } i=1,\ldots ,n-1 .\nonumber \end{align}
\begin{align} \Upsilon '([\mathbb {HP}^{n-1}])=1,\quad \Upsilon '([\mathbb {HP}^{i-1}])=\prod _{s=i+1}^{n}\big(\xi ^2-t^2_s\big)\quad\mbox{for } i=1,\ldots ,n-1,\nonumber\\[-8pt]\\[-8pt]\Upsilon '([\mathbb {HP}^{n-1}[2]])=\xi ,\quad \Upsilon '([\mathbb {HP}^{i-1}[2]])=\xi \prod _{s=i+1}^{n}\big(\xi ^2-t^2_s\big)\quad\mbox{for } i=1,\ldots ,n-1 .\nonumber \end{align}
Lemma 4.4.
- 
(1) In terms of the ordered basis  $\{[\mathbb P^{0}],[\mathbb P^{1}],\ldots ,[\mathbb P^{2n}]\}$
, the cup product action $\{[\mathbb P^{0}],[\mathbb P^{1}],\ldots ,[\mathbb P^{2n}]\}$
, the cup product action $c_1^{T_{2n}}(\mathcal {O}(1))\cup (-)\in {\textrm{End}}_{R_{T_n}} \big(H^*_{T_n}(\mathbb P^{2n-1})\big)$
 is given by the element $c_1^{T_{2n}}(\mathcal {O}(1))\cup (-)\in {\textrm{End}}_{R_{T_n}} \big(H^*_{T_n}(\mathbb P^{2n-1})\big)$
 is given by the element $e^{T_{2n}}$
 in (
3.12
): $e^{T_{2n}}$
 in (
3.12
): \begin{equation*}e^{T_{2n}}=\begin{pmatrix} t_1 & \quad 1 & \quad \\ 0& \quad t_2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad \ldots & \quad 0 & \quad t_{2n} \end{pmatrix}. \end{equation*} \begin{equation*}e^{T_{2n}}=\begin{pmatrix} t_1 & \quad 1 & \quad \\ 0& \quad t_2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad \ldots & \quad 0 & \quad t_{2n} \end{pmatrix}. \end{equation*}
- 
(2) In terms of the ordered basis  $\{[\mathbb {HP}^{0}][2],\ldots ,[\mathbb {HP}^{n-1}][2], [\mathbb {HP}^{0}],\ldots,[\mathbb {HP}^{n-1}]\}$
, the cup product action $\{[\mathbb {HP}^{0}][2],\ldots ,[\mathbb {HP}^{n-1}][2], [\mathbb {HP}^{0}],\ldots,[\mathbb {HP}^{n-1}]\}$
, the cup product action $c_1^{T_{n}}(\mathcal {O}(1))\cup (-)\in {\textrm{End}}_{R_{T_n}} \big(H^*_{T_n}(\mathbb P^{2n-1})\big)$
 is given by the element $c_1^{T_{n}}(\mathcal {O}(1))\cup (-)\in {\textrm{End}}_{R_{T_n}} \big(H^*_{T_n}(\mathbb P^{2n-1})\big)$
 is given by the element $\tau \circ e^T_X$
 in (
3.14
): $\tau \circ e^T_X$
 in (
3.14
): \begin{equation*} \tau \circ e^T_X=\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}, \quad C=\begin{pmatrix} t_1^2 & \quad 1 & \quad \\ \vdots & \quad t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad t_n^2 \end{pmatrix}. \end{equation*} \begin{equation*} \tau \circ e^T_X=\begin{pmatrix}0& \quad \textrm{Id}_n\\ C& \quad 0\end{pmatrix}, \quad C=\begin{pmatrix} t_1^2 & \quad 1 & \quad \\ \vdots & \quad t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad t_n^2 \end{pmatrix}. \end{equation*}
Proof.
 The cup product action is given by mutiplication by 
 $\xi$
 and the claim is a straightforward computation.
$\xi$
 and the claim is a straightforward computation.
4.4 Complex and quaternionic affine Grassmannians
 We denote by 
 ${\textrm{Gr}}_{2n}=\mathfrak{L}G_{2n}/\mathfrak{L}^+G_{2n}$
 the complex affine Grassmannian for
${\textrm{Gr}}_{2n}=\mathfrak{L}G_{2n}/\mathfrak{L}^+G_{2n}$
 the complex affine Grassmannian for 
 $G_{2n}$
, where
$G_{2n}$
, where 
 $\mathfrak{L}G_{2n}=G_{2n}({\mathbb C}((t)))$
 and
$\mathfrak{L}G_{2n}=G_{2n}({\mathbb C}((t)))$
 and 
 $\mathfrak{L}^+G_{2n}=G_{2n}({\mathbb C}[[t]])$
 are the Laurent loop group and Taylor arc group for
$\mathfrak{L}^+G_{2n}=G_{2n}({\mathbb C}[[t]])$
 are the Laurent loop group and Taylor arc group for 
 $G_{2n}$
, respectively. We denote by
$G_{2n}$
, respectively. We denote by 
 $D^b(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})$
 the dg-category of
$D^b(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})$
 the dg-category of 
 $\mathfrak{L}^+G_{2n}$
-equivariant constructible complexes on
$\mathfrak{L}^+G_{2n}$
-equivariant constructible complexes on 
 ${\textrm{Gr}}_{2n}$
 and by
${\textrm{Gr}}_{2n}$
 and by 
 $\textrm{Perv}({\textrm{Gr}}_{2n})$
 the abelian category of
$\textrm{Perv}({\textrm{Gr}}_{2n})$
 the abelian category of 
 $\mathfrak{L}^+G_{2n}$
-equivariant perverse sheaves on
$\mathfrak{L}^+G_{2n}$
-equivariant perverse sheaves on 
 ${\textrm{Gr}}_{2n}$
.
${\textrm{Gr}}_{2n}$
.
 We denote by 
 ${\textrm{Gr}}_{n,{\mathbb H}}=\mathfrak{L}G_{n,{\mathbb H}}/\mathfrak{L}^+G_{n,{\mathbb H}}$
 the real affine Grassmannian for the quaternionic group
${\textrm{Gr}}_{n,{\mathbb H}}=\mathfrak{L}G_{n,{\mathbb H}}/\mathfrak{L}^+G_{n,{\mathbb H}}$
 the real affine Grassmannian for the quaternionic group 
 $G_{n,{\mathbb H}}$
, where
$G_{n,{\mathbb H}}$
, where 
 $\mathfrak{L}G_{n,{\mathbb H}}=G_{n,{\mathbb H}}({\mathbb R}((t)))$
 and
$\mathfrak{L}G_{n,{\mathbb H}}=G_{n,{\mathbb H}}({\mathbb R}((t)))$
 and 
 $\mathfrak{L}^+G_{n,{\mathbb H}}=G_{n,{\mathbb H}}({\mathbb R}[[t]])$
 are the real Laurent loop group and real Taylor arc group for
$\mathfrak{L}^+G_{n,{\mathbb H}}=G_{n,{\mathbb H}}({\mathbb R}[[t]])$
 are the real Laurent loop group and real Taylor arc group for 
 $G_{n,{\mathbb H}}$
. The
$G_{n,{\mathbb H}}$
. The 
 $\mathfrak{L}^+G_{n,{\mathbb H}}$
-orbits on
$\mathfrak{L}^+G_{n,{\mathbb H}}$
-orbits on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 are of the form
${\textrm{Gr}}_{n,{\mathbb H}}$
 are of the form 
 ${\textrm{Gr}}^\lambda _{n,{\mathbb H}}=\mathfrak{L}^+G_{n,{\mathbb H}}\cdot t^\lambda$
 where
${\textrm{Gr}}^\lambda _{n,{\mathbb H}}=\mathfrak{L}^+G_{n,{\mathbb H}}\cdot t^\lambda$
 where 
 $(\lambda :{\mathbb G}_m\to S)\in \Lambda _S^+$
 is a dominant real coweight. By [Reference NadlerNad05, Proposition 3.6.1], each orbit
$(\lambda :{\mathbb G}_m\to S)\in \Lambda _S^+$
 is a dominant real coweight. By [Reference NadlerNad05, Proposition 3.6.1], each orbit 
 ${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 is a real vector bundle over the quaternionic flag manifold
${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 is a real vector bundle over the quaternionic flag manifold 
 $G_{n,{\mathbb H}}/P^\lambda _{n,{\mathbb H}}$
 of real dimension
$G_{n,{\mathbb H}}/P^\lambda _{n,{\mathbb H}}$
 of real dimension 
 $2\langle \lambda ,\rho _{2n}\rangle$
. We denote by
$2\langle \lambda ,\rho _{2n}\rangle$
. We denote by 
 $D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 the dg-category of
$D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 the dg-category of 
 $\mathfrak{L}^+G_{n,{\mathbb H}}$
-equivariant constructible complexes on
$\mathfrak{L}^+G_{n,{\mathbb H}}$
-equivariant constructible complexes on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
. Since
${\textrm{Gr}}_{n,{\mathbb H}}$
. Since 
 $\langle \lambda ,\rho _{2n}\rangle =4\langle \lambda ,\rho _n\rangle \in 2{\mathbb Z}$
 for all
$\langle \lambda ,\rho _{2n}\rangle =4\langle \lambda ,\rho _n\rangle \in 2{\mathbb Z}$
 for all 
 $\lambda \in \Lambda _S^+$
 (in the second paring we regard
$\lambda \in \Lambda _S^+$
 (in the second paring we regard 
 $\lambda$
 as an element in
$\lambda$
 as an element in 
 $\Lambda _n$
), all the orbits
$\Lambda _n$
), all the orbits 
 ${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 have real even dimension, and hence middle perversity makes sense and we denote by
${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 have real even dimension, and hence middle perversity makes sense and we denote by 
 $\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 the category of
$\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 the category of 
 $\mathfrak{L}^+G_{n,{\mathbb H}}$
-equivariant perverse sheaves on
$\mathfrak{L}^+G_{n,{\mathbb H}}$
-equivariant perverse sheaves on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
. Note also that, as
${\textrm{Gr}}_{n,{\mathbb H}}$
. Note also that, as 
 $P^\lambda _{n,{\mathbb H}}$
 is connected, all the
$P^\lambda _{n,{\mathbb H}}$
 is connected, all the 
 $G_{n,{\mathbb H}}$
-equivariant local systems on
$G_{n,{\mathbb H}}$
-equivariant local systems on 
 ${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 are trivial and hence the irreducible objects in
${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 are trivial and hence the irreducible objects in 
 $\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 are intersection cohomology complexes
$\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 are intersection cohomology complexes 
 $\textrm{IC}_\lambda =\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\big)$
,
$\textrm{IC}_\lambda =\textrm{IC}\big(\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\big)$
, 
 $\lambda \in \Lambda _S^+$
, for the closure
$\lambda \in \Lambda _S^+$
, for the closure 
 $\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\subset {\textrm{Gr}}_{n,{\mathbb H}}$
.
$\overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}}\subset {\textrm{Gr}}_{n,{\mathbb H}}$
.
 Like in the case of complex reductive groups, there is a natural monoidal structure on 
 $D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 given by the convolution product: consider the convolution diagram
$D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 given by the convolution product: consider the convolution diagram
 \begin{equation*}{\textrm{Gr}}_{n,{\mathbb H}}\times {\textrm{Gr}}_{n,{\mathbb H}}\stackrel {\textrm{p}}\leftarrow \mathfrak{L}G_{n,{\mathbb H}} \times {\textrm{Gr}}_{n,{\mathbb H}} \stackrel {q}\rightarrow {\textrm{Gr}}_{n,{\mathbb H}}\tilde \times {\textrm{Gr}}_{n,{\mathbb H}}:=\mathfrak{L}G_{n,{\mathbb H}}\times ^{\mathfrak{L}^+G_{n,{\mathbb H}}}{\textrm{Gr}}_{n,{\mathbb H}}\stackrel {m}\to {\textrm{Gr}}_{n,{\mathbb H}}\end{equation*}
\begin{equation*}{\textrm{Gr}}_{n,{\mathbb H}}\times {\textrm{Gr}}_{n,{\mathbb H}}\stackrel {\textrm{p}}\leftarrow \mathfrak{L}G_{n,{\mathbb H}} \times {\textrm{Gr}}_{n,{\mathbb H}} \stackrel {q}\rightarrow {\textrm{Gr}}_{n,{\mathbb H}}\tilde \times {\textrm{Gr}}_{n,{\mathbb H}}:=\mathfrak{L}G_{n,{\mathbb H}}\times ^{\mathfrak{L}^+G_{n,{\mathbb H}}}{\textrm{Gr}}_{n,{\mathbb H}}\stackrel {m}\to {\textrm{Gr}}_{n,{\mathbb H}}\end{equation*}
where 
 $p$
 and
$p$
 and 
 $q$
 are the natural quotient maps and
$q$
 are the natural quotient maps and 
 $m(x,y\,\bmod\, \mathfrak{L}^+G_{n,{\mathbb H}})= xy\,\bmod\, \mathfrak{L}^+G_{n,{\mathbb H}}$
. For any
$m(x,y\,\bmod\, \mathfrak{L}^+G_{n,{\mathbb H}})= xy\,\bmod\, \mathfrak{L}^+G_{n,{\mathbb H}}$
. For any 
 $\mathcal {F}_1,\mathcal {F}_2\in D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
, the convolution is defined as
$\mathcal {F}_1,\mathcal {F}_2\in D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
, the convolution is defined as
 \begin{equation*}\mathcal {F}_1\star \mathcal {F}_2=m_!(\mathcal {F}_1\tilde \boxtimes \mathcal {F}_2)\end{equation*}
\begin{equation*}\mathcal {F}_1\star \mathcal {F}_2=m_!(\mathcal {F}_1\tilde \boxtimes \mathcal {F}_2)\end{equation*}
where 
 $\mathcal {F}_1\tilde \boxtimes \mathcal {F}_2\in D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}\tilde \times {\textrm{Gr}}_{n,{\mathbb H}})$
 is the unique complex such that
$\mathcal {F}_1\tilde \boxtimes \mathcal {F}_2\in D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}\tilde \times {\textrm{Gr}}_{n,{\mathbb H}})$
 is the unique complex such that 
 $q^*(\mathcal {F}_1\tilde \boxtimes \mathcal {F}_2)\simeq p^*(\mathcal {F}_1\boxtimes \mathcal {F}_2)$
.
$q^*(\mathcal {F}_1\tilde \boxtimes \mathcal {F}_2)\simeq p^*(\mathcal {F}_1\boxtimes \mathcal {F}_2)$
.
4.5 Real nearby cycles functor
 We shall recall the construction of the real nearby cycles functor in [Reference NadlerNad05]. Consider the Beilinson–Drinfeld Grassmannian 
 ${\textrm{Gr}}^{(2)}_{2n}\to {\mathbb C}$
 over the complex line
${\textrm{Gr}}^{(2)}_{2n}\to {\mathbb C}$
 over the complex line 
 $\mathbb C$
 classifying a
$\mathbb C$
 classifying a 
 $G_{2n}$
-bundle
$G_{2n}$
-bundle 
 $\mathcal {E}\to {\mathbb C}$
, a point
$\mathcal {E}\to {\mathbb C}$
, a point 
 $x\in {\mathbb C}$
, and a section
$x\in {\mathbb C}$
, and a section 
 $\nu :{\mathbb C}\setminus \{\pm x\}\to \mathcal {E}|_{{\mathbb C}\setminus \{\pm x\}}$
. It is well known that there are canonical isomorphisms
$\nu :{\mathbb C}\setminus \{\pm x\}\to \mathcal {E}|_{{\mathbb C}\setminus \{\pm x\}}$
. It is well known that there are canonical isomorphisms
 \begin{equation*}{\textrm{Gr}}^{(2)}_{2n}\big|_{\{0\}}\simeq {\textrm{Gr}}_{2n},\end{equation*}
\begin{equation*}{\textrm{Gr}}^{(2)}_{2n}\big|_{\{0\}}\simeq {\textrm{Gr}}_{2n},\end{equation*}
 \begin{equation*}{\textrm{Gr}}^{(2)}_{2n}\big|_{{\mathbb C}\setminus \{0\}}\simeq {\textrm{Gr}}_{2n}\times {\textrm{Gr}}_{2n}\times {\mathbb C}\setminus \{0\}.\end{equation*}
\begin{equation*}{\textrm{Gr}}^{(2)}_{2n}\big|_{{\mathbb C}\setminus \{0\}}\simeq {\textrm{Gr}}_{2n}\times {\textrm{Gr}}_{2n}\times {\mathbb C}\setminus \{0\}.\end{equation*}
It is shown in [Reference NadlerNad05, Proposition 4.3.1] that the real form 
 $G_{n,{\mathbb H}}$
 of
$G_{n,{\mathbb H}}$
 of 
 $G_{2n}$
 together with real form
$G_{2n}$
 together with real form 
 $i{\mathbb R}$
 of
$i{\mathbb R}$
 of 
 $\mathbb C$
 (corresponding to the complex conjugation
$\mathbb C$
 (corresponding to the complex conjugation 
 $x\to -\bar x$
 on
$x\to -\bar x$
 on 
 $\mathbb C$
) defines a real form
$\mathbb C$
) defines a real form 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{(2)}\to i{\mathbb R}$
 of
${\textrm{Gr}}_{n,{\mathbb H}}^{(2)}\to i{\mathbb R}$
 of 
 ${\textrm{Gr}}_{2n}^{(2)}$
 such that there are canonical isomorphisms
${\textrm{Gr}}_{2n}^{(2)}$
 such that there are canonical isomorphisms
 \begin{equation*}{\textrm{Gr}}^{(2)}_{n,{\mathbb H}}\big|_{\{0\}}\simeq {\textrm{Gr}}_{n,{\mathbb H}}\end{equation*}
\begin{equation*}{\textrm{Gr}}^{(2)}_{n,{\mathbb H}}\big|_{\{0\}}\simeq {\textrm{Gr}}_{n,{\mathbb H}}\end{equation*}
 \begin{equation*}{\textrm{Gr}}^{(2)}_{n,{\mathbb H}}\big|_{i{\mathbb R}\setminus \{0\}}\simeq {\textrm{Gr}}_{2n}\times (i{\mathbb R}\setminus \{0\})\end{equation*}
\begin{equation*}{\textrm{Gr}}^{(2)}_{n,{\mathbb H}}\big|_{i{\mathbb R}\setminus \{0\}}\simeq {\textrm{Gr}}_{2n}\times (i{\mathbb R}\setminus \{0\})\end{equation*}
Consider the following diagram.

 Note that the maps in the above diagram are all 
 $K_c$
-equivariant, and we define the functor
$K_c$
-equivariant, and we define the functor
 \begin{equation*} \mathrm R': D^b_{}(K_c\backslash {\textrm{Gr}}_{2n})\to D^b_{}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \end{equation*}
\begin{equation*} \mathrm R': D^b_{}(K_c\backslash {\textrm{Gr}}_{2n})\to D^b_{}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \end{equation*}
by the formula
 \begin{equation} \mathrm R'(\mathcal {F})=i^*j_*(\mathcal {F}\boxtimes {\mathbb C}_{i{\mathbb R}_{\geq 0}}). \end{equation}
\begin{equation} \mathrm R'(\mathcal {F})=i^*j_*(\mathcal {F}\boxtimes {\mathbb C}_{i{\mathbb R}_{\geq 0}}). \end{equation}
By [Reference NadlerNad05, Proposition 4.5.1], the functor 
 $\mathrm R'$
 takes
$\mathrm R'$
 takes 
 $\mathfrak{L}^+G_{2n}$
-constructible complexes to
$\mathfrak{L}^+G_{2n}$
-constructible complexes to 
 $\mathfrak{L}^+G_{n,{\mathbb H}}$
-constructible complexes. Introduce the subcategory
$\mathfrak{L}^+G_{n,{\mathbb H}}$
-constructible complexes. Introduce the subcategory 
 $D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 (respectively,
$D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 (respectively, 
 $D^b_{\{\mathfrak{L}^+G_{2n}\}}(G_c\backslash {\textrm{Gr}}_{2n})$
 and
$D^b_{\{\mathfrak{L}^+G_{2n}\}}(G_c\backslash {\textrm{Gr}}_{2n})$
 and 
 $D^b_{\{\mathfrak{L}^+G_{2n}\}}(K_c\backslash {\textrm{Gr}}_{2n})$
) of
$D^b_{\{\mathfrak{L}^+G_{2n}\}}(K_c\backslash {\textrm{Gr}}_{2n})$
) of 
 $D^b(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 (respectively,
$D^b(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 (respectively, 
 $D^b(G_c\backslash {\textrm{Gr}}_{2n})$
 and
$D^b(G_c\backslash {\textrm{Gr}}_{2n})$
 and 
 $D^b(K_c\backslash {\textrm{Gr}}_{2n})$
) consisting
$D^b(K_c\backslash {\textrm{Gr}}_{2n})$
) consisting 
 $\mathfrak{L}^+G_{n,{\mathbb H}}$
-constructible complexes (respectively,
$\mathfrak{L}^+G_{n,{\mathbb H}}$
-constructible complexes (respectively, 
 $\mathfrak{L}^+G_{2n}$
-constructible complexes). Since the quotients
$\mathfrak{L}^+G_{2n}$
-constructible complexes). Since the quotients 
 $\mathfrak{L}^+G_{2n}/G_c$
 and
$\mathfrak{L}^+G_{2n}/G_c$
 and 
 $\mathfrak{L}^+G_{n,{\mathbb H}}/K_c$
 are contractible, we have natural equivalences
$\mathfrak{L}^+G_{n,{\mathbb H}}/K_c$
 are contractible, we have natural equivalences 
 $D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 and
$D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 and 
 $D^b_{\{\mathfrak{L}^+G_{2n}\}}(G_c\backslash {\textrm{Gr}}_{2n})\simeq D^b(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})$
, and the nearby cycles functor
$D^b_{\{\mathfrak{L}^+G_{2n}\}}(G_c\backslash {\textrm{Gr}}_{2n})\simeq D^b(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})$
, and the nearby cycles functor 
 $\mathrm R'$
 above induces a functor
$\mathrm R'$
 above induces a functor
 \begin{equation} \mathrm R':D^b_{\{\mathfrak{L}^+G_{2n}\}}(K_c\backslash {\textrm{Gr}}_{2n})\to D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}). \end{equation}
\begin{equation} \mathrm R':D^b_{\{\mathfrak{L}^+G_{2n}\}}(K_c\backslash {\textrm{Gr}}_{2n})\to D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}(K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}). \end{equation}
Finally, the real nearby cycles functor is defined as
 \begin{equation} \mathrm R: D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})\simeq D^b_{\{\mathfrak{L}^+G_{2n}\}}(G_c\backslash {\textrm{Gr}}_{2n})\to D^b_{\{\mathfrak{L}^+G_{2n}\}}(K_c\backslash {\textrm{Gr}}_{2n})\stackrel {\mathrm R'}\to D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \end{equation}
\begin{equation} \mathrm R: D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})\simeq D^b_{\{\mathfrak{L}^+G_{2n}\}}(G_c\backslash {\textrm{Gr}}_{2n})\to D^b_{\{\mathfrak{L}^+G_{2n}\}}(K_c\backslash {\textrm{Gr}}_{2n})\stackrel {\mathrm R'}\to D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \end{equation}
where the middle arrow is the natural forgetful functor.
 The following properties of 
 $\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 and
$\textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 and 
 $\mathrm R$
 can be deduced from [Reference NadlerNad05].
$\mathrm R$
 can be deduced from [Reference NadlerNad05].
Proposition 4.5.
- 
(1) There is a tensor equivalence  $\textrm{Rep}(G_n)\simeq \textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 sending the irreducible representation $\textrm{Rep}(G_n)\simeq \textrm{Perv}_{}({\textrm{Gr}}_{n,{\mathbb H}})$
 sending the irreducible representation $V_\lambda$
 of $V_\lambda$
 of $G_n$
 with highest weight $G_n$
 with highest weight $\lambda \in \Lambda _S^+$
 to $\lambda \in \Lambda _S^+$
 to $\textrm{IC}_\lambda$
. $\textrm{IC}_\lambda$
.
- 
(2) The real nearby cycle functor  $\mathrm R$
 preserves semisimplicity, that is, we have
for any semisimple complex $\mathrm R$
 preserves semisimplicity, that is, we have
for any semisimple complex \begin{equation*}\mathrm R(\mathcal {F})\simeq \bigoplus _{n\in {\mathbb Z}} {^pH^n}\mathrm R(\mathcal {F})[-n]\end{equation*} \begin{equation*}\mathrm R(\mathcal {F})\simeq \bigoplus _{n\in {\mathbb Z}} {^pH^n}\mathrm R(\mathcal {F})[-n]\end{equation*} $\mathcal {F}$
 in $\mathcal {F}$
 in $D_{}^b(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})$
. $D_{}^b(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})$
.
- 
(3) Consider the monoidal subcategory The real nearby cycle functor restricts to a monoidal functor \begin{equation*}\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})_{\mathbb Z}:=\bigoplus _{n\in {\mathbb Z}}\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})[n]\subset D_{\mathfrak{L}^+G_{n,{\mathbb H}}}^b({\textrm{Gr}}_{n,{\mathbb H}}).\end{equation*} 
such that there is a commutative diagram \begin{equation*}\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})_{\mathbb Z}:=\bigoplus _{n\in {\mathbb Z}}\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})[n]\subset D_{\mathfrak{L}^+G_{n,{\mathbb H}}}^b({\textrm{Gr}}_{n,{\mathbb H}}).\end{equation*} 
such that there is a commutative diagram \begin{equation*}^p\mathrm R=\bigoplus _{n\in {\mathbb Z}} {^pH^n}\mathrm R(\mathcal {F}):\textrm{Perv}({\textrm{Gr}}_{2n})\to \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})_{\mathbb Z}\end{equation*} \begin{equation*}^p\mathrm R=\bigoplus _{n\in {\mathbb Z}} {^pH^n}\mathrm R(\mathcal {F}):\textrm{Perv}({\textrm{Gr}}_{2n})\to \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})_{\mathbb Z}\end{equation*} where the vertical tensor equivalences come from the complex and quaternionic Satake isomorphisms (part (1)) and the bottom arrow is the restriction map to the subgroup  $G_n\times {\mathbb G}_m\subset G_{2n}$
 as in § 
3.5
. $G_n\times {\mathbb G}_m\subset G_{2n}$
 as in § 
3.5
.
Proof.
 Part (1) is proved in [Reference NadlerNad05, Theorem 1.2.2], and part (3) is proved in [Reference NadlerNad05, § 10.3]. To prove part (2), it suffices to show that 
 $\mathrm R(\textrm{IC}_\lambda )$
 is semisimple for all dominant
$\mathrm R(\textrm{IC}_\lambda )$
 is semisimple for all dominant 
 $\lambda$
. It is shown in [Reference NadlerNad05, Corollary 1.2.1 and § 6.4] that
$\lambda$
. It is shown in [Reference NadlerNad05, Corollary 1.2.1 and § 6.4] that 
 $\mathrm R$
 is monoidal and that given two semisimple objects
$\mathrm R$
 is monoidal and that given two semisimple objects 
 $\mathcal {F}_1\mbox{ and }\mathcal {F}_2$
 in the essential image of
$\mathcal {F}_1\mbox{ and }\mathcal {F}_2$
 in the essential image of 
 $\mathrm R$
, the convolution
$\mathrm R$
, the convolution 
 $\mathcal {F}_1\star \mathcal {F}_2$
 is again semisimple. Let
$\mathcal {F}_1\star \mathcal {F}_2$
 is again semisimple. Let 
 $\omega _{1}$
 and
$\omega _{1}$
 and 
 $\epsilon$
 (respectively,
$\epsilon$
 (respectively, 
 $\omega _{1}'$
 and
$\omega _{1}'$
 and 
 $\epsilon '$
) be the highest weights of the standard representation and determinant character of
$\epsilon '$
) be the highest weights of the standard representation and determinant character of 
 $G_{n}$
 (respectively,
$G_{n}$
 (respectively, 
 $G_{2n}$
), respectively. Since
$G_{2n}$
), respectively. Since 
 $\textrm{Rep}(G_{2n})$
 is tensor-generated by the standard representation
$\textrm{Rep}(G_{2n})$
 is tensor-generated by the standard representation 
 $V_{\omega _{1}'}$
 and the determinant character
$V_{\omega _{1}'}$
 and the determinant character 
 $V_{\epsilon '}$
, it suffices to show that
$V_{\epsilon '}$
, it suffices to show that 
 $\mathrm R(\textrm{IC}_{\omega _{1}'})$
 and
$\mathrm R(\textrm{IC}_{\omega _{1}'})$
 and 
 $\mathrm R(\textrm{IC}_{\epsilon '})$
 are semisimple. It follows from part (3) that
$\mathrm R(\textrm{IC}_{\epsilon '})$
 are semisimple. It follows from part (3) that 
 ${^p\mathrm R}(\textrm{IC}_{\epsilon '})\simeq \textrm{IC}_{2\epsilon }$
 and
${^p\mathrm R}(\textrm{IC}_{\epsilon '})\simeq \textrm{IC}_{2\epsilon }$
 and 
 $^p\mathrm R(\textrm{IC}_{\omega _{1}'})\simeq \textrm{IC}_{\omega _1}\oplus \textrm{IC}_{\omega _1}$
 where
$^p\mathrm R(\textrm{IC}_{\omega _{1}'})\simeq \textrm{IC}_{\omega _1}\oplus \textrm{IC}_{\omega _1}$
 where 
 $\textrm{IC}_{\omega _1}$
 is the
$\textrm{IC}_{\omega _1}$
 is the 
 $\textrm{IC}$
-complex of
$\textrm{IC}$
-complex of 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{\omega _1}\simeq \mathbb {HP}^{n-1}$
. Since the
${\textrm{Gr}}_{n,{\mathbb H}}^{\omega _1}\simeq \mathbb {HP}^{n-1}$
. Since the 
 ${\mathbb G}_m$
-weights of
${\mathbb G}_m$
-weights of 
 $\textrm{det}_{2n}$
 and
$\textrm{det}_{2n}$
 and 
 $V_{\omega _{1}'}$
 with respect to the co-character
$V_{\omega _{1}'}$
 with respect to the co-character 
 ${\mathbb G}_m\simeq \{e\}\times {\mathbb G}_m\subset G_n\times {\mathbb G}_m\subset G_{2n}$
 in part (3) are equal to
${\mathbb G}_m\simeq \{e\}\times {\mathbb G}_m\subset G_n\times {\mathbb G}_m\subset G_{2n}$
 in part (3) are equal to 
 $0$
 and
$0$
 and 
 $\{1,-1\}$
, respectively, this implies that
$\{1,-1\}$
, respectively, this implies that 
 $\mathrm R(\textrm{IC}_{\epsilon '})\simeq {^p\mathrm R}(\textrm{IC}_{\epsilon '})\simeq \textrm{IC}_{2\epsilon }$
 is a simple perverse sheaf, and
$\mathrm R(\textrm{IC}_{\epsilon '})\simeq {^p\mathrm R}(\textrm{IC}_{\epsilon '})\simeq \textrm{IC}_{2\epsilon }$
 is a simple perverse sheaf, and 
 $\mathrm R(\textrm{IC}_{\omega _1'})$
 admits a filtration with associated graded given by
$\mathrm R(\textrm{IC}_{\omega _1'})$
 admits a filtration with associated graded given by 
 $\textrm{IC}_{\omega _1}[1]\oplus \textrm{IC}_{\omega _1}[-1]$
. Since
$\textrm{IC}_{\omega _1}[1]\oplus \textrm{IC}_{\omega _1}[-1]$
. Since 
 $\textrm{Ext}^1(\textrm{IC}_{\omega _1}[-1],\textrm{IC}_{\omega _1}[1])= \textrm{Ext}^3({\mathbb C}_{\mathbb {HP}^{n-1}},{\mathbb C}_{\mathbb {HP}^{n-1}})\simeq H^3_{G_{n,{\mathbb H}}}(\mathbb {HP}^{n-1}) \subset H^3_{T_n}(\mathbb {HP}^{n-1})=0$
, it follows that the filtration splits and hence that
$\textrm{Ext}^1(\textrm{IC}_{\omega _1}[-1],\textrm{IC}_{\omega _1}[1])= \textrm{Ext}^3({\mathbb C}_{\mathbb {HP}^{n-1}},{\mathbb C}_{\mathbb {HP}^{n-1}})\simeq H^3_{G_{n,{\mathbb H}}}(\mathbb {HP}^{n-1}) \subset H^3_{T_n}(\mathbb {HP}^{n-1})=0$
, it follows that the filtration splits and hence that 
 $\mathrm R(\textrm{IC}_{\omega _1'})\simeq \textrm{IC}_{\omega _1}[1]\oplus \textrm{IC}_{\omega _1}[-1]$
 is semisimple.
$\mathrm R(\textrm{IC}_{\omega _1'})\simeq \textrm{IC}_{\omega _1}[1]\oplus \textrm{IC}_{\omega _1}[-1]$
 is semisimple.
In the course of the proof, together with Lemma 4.3, we have shown the following.
Corollary 4.6. 
There is an isomorphism 
 $\mathrm R(\textrm{IC}_{\omega _1'})\simeq f_*(\textrm{IC}_{\mathbb P^{2n-1}})\simeq \textrm{IC}_{\mathbb {HP}^{n-1}}[1]\oplus \textrm{IC}_{\mathbb {HP}^{n-1}}[-1]$
.
$\mathrm R(\textrm{IC}_{\omega _1'})\simeq f_*(\textrm{IC}_{\mathbb P^{2n-1}})\simeq \textrm{IC}_{\mathbb {HP}^{n-1}}[1]\oplus \textrm{IC}_{\mathbb {HP}^{n-1}}[-1]$
.
Remark 4.7. In Theorem 5.7, we will give a spectral description of the nearby cycle functor 
 $\mathrm R$
 on the whole derived category (not just its restriction
$\mathrm R$
 on the whole derived category (not just its restriction 
 $^p\mathrm R$
 to the subcategory of perverse sheaves).
$^p\mathrm R$
 to the subcategory of perverse sheaves).
 Recall the nearby cycles functor 
 $\mathrm R': D_{\{\mathfrak{L}^+G_{2n}\}}^b(K_c\backslash {\textrm{Gr}}_{2n})\to D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}( K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 in (4.6). It extends to the ind-completion (denoted again by
$\mathrm R': D_{\{\mathfrak{L}^+G_{2n}\}}^b(K_c\backslash {\textrm{Gr}}_{2n})\to D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}( K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 in (4.6). It extends to the ind-completion (denoted again by 
 $\mathrm R'$
)
$\mathrm R'$
)
 \begin{equation*}\mathrm R': \textrm{Ind}\ D_{\{\mathfrak{L}^+G_{2n}\}}^b(K_c\backslash {\textrm{Gr}}_{2n})\to \textrm{Ind}\ D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}( K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}}).\end{equation*}
\begin{equation*}\mathrm R': \textrm{Ind}\ D_{\{\mathfrak{L}^+G_{2n}\}}^b(K_c\backslash {\textrm{Gr}}_{2n})\to \textrm{Ind}\ D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}( K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}}).\end{equation*}
Lemma 4.8. 
The functor 
 $\mathrm R'$
 admits the left adjoint
$\mathrm R'$
 admits the left adjoint
 \begin{equation*}^L\mathrm R': \textrm{Ind}\ D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}( K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \to \textrm{Ind}\ D_{\{\mathfrak{L}^+G_{2n}\}}^b(K_c\backslash {\textrm{Gr}}_{2n}). \end{equation*}
\begin{equation*}^L\mathrm R': \textrm{Ind}\ D^b_{\{\mathfrak{L}^+G_{n,{\mathbb H}}\}}( K_c\backslash {\textrm{Gr}}_{n,{\mathbb H}}) \to \textrm{Ind}\ D_{\{\mathfrak{L}^+G_{2n}\}}^b(K_c\backslash {\textrm{Gr}}_{2n}). \end{equation*}
Moreover, we have 
 $^L\mathrm R'\big({\mathbb C}_{{\textrm{Gr}}_{n,{\mathbb H}}}\big)\simeq {\mathbb C}_{{\textrm{Gr}}_{2n}}$
.
$^L\mathrm R'\big({\mathbb C}_{{\textrm{Gr}}_{n,{\mathbb H}}}\big)\simeq {\mathbb C}_{{\textrm{Gr}}_{2n}}$
.
Proof.
 By [Reference NadlerNad05, Proposition 4.5.1], the ind-proper family 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{(2)}\to i{\mathbb R}_{\geq 0}$
 is a Thom stratified map with respect to a Whitney stratification
${\textrm{Gr}}_{n,{\mathbb H}}^{(2)}\to i{\mathbb R}_{\geq 0}$
 is a Thom stratified map with respect to a Whitney stratification 
 $\mathcal T$
 on
$\mathcal T$
 on 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{(2)}$
 and the stratification
${\textrm{Gr}}_{n,{\mathbb H}}^{(2)}$
 and the stratification 
 $i{\mathbb R}_{\gt 0}\cup \{0\}$
 on
$i{\mathbb R}_{\gt 0}\cup \{0\}$
 on 
 $i{\mathbb R}_{\geq 0}$
 such that
$i{\mathbb R}_{\geq 0}$
 such that 
 $\mathcal T$
 restricts to the
$\mathcal T$
 restricts to the 
 $\mathfrak{L}^+G_{2n}$
-orbits stratification on the generic fiber
$\mathfrak{L}^+G_{2n}$
-orbits stratification on the generic fiber 
 ${\textrm{Gr}}_{2n}$
 and to the
${\textrm{Gr}}_{2n}$
 and to the 
 $\mathfrak{L}^+G_{n,{\mathbb H}}$
-orbits stratification on the special fiber
$\mathfrak{L}^+G_{n,{\mathbb H}}$
-orbits stratification on the special fiber 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
. The construction in [Reference Goresky and MacPhersonGM83, § 6], together with the results in [Reference Pflaum and WilkinPW19, Theorem 1.1] (extending Mather’s theory of control data to the equivariant setting), implies that the nearby cycles functor
${\textrm{Gr}}_{n,{\mathbb H}}$
. The construction in [Reference Goresky and MacPhersonGM83, § 6], together with the results in [Reference Pflaum and WilkinPW19, Theorem 1.1] (extending Mather’s theory of control data to the equivariant setting), implies that the nearby cycles functor 
 $\mathrm R'$
 is isomorphic to the functor given by
$\mathrm R'$
 is isomorphic to the functor given by 
 $*$
-pushforward along a
$*$
-pushforward along a 
 $K_c$
-equivariant specialization map
$K_c$
-equivariant specialization map 
 $\psi :{\textrm{Gr}}_{2n}\to {\textrm{Gr}}_{n,{\mathbb H}}$
, and hence admits a left adjoint given by the
$\psi :{\textrm{Gr}}_{2n}\to {\textrm{Gr}}_{n,{\mathbb H}}$
, and hence admits a left adjoint given by the 
 $*$
-pullback
$*$
-pullback 
 $\psi ^*$
. It is clear that
$\psi ^*$
. It is clear that 
 $\psi ^*$
 sends constant sheaf to constant sheaf. The lemma follows.
$\psi ^*$
 sends constant sheaf to constant sheaf. The lemma follows.
4.6 Equivariant homology and cohomology of affine Grassmannians
4.6.1
 We reviewed the description of the equivariant homologies 
 $H_*^{T_{2n}}({\textrm{Gr}}_{2n})$
 and
$H_*^{T_{2n}}({\textrm{Gr}}_{2n})$
 and 
 $H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})$
 of
$H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})$
 of 
 ${\textrm{Gr}}_{2n}$
 and
${\textrm{Gr}}_{2n}$
 and 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 in [Reference O’BrienO’Br23, Reference Yun and ZhuYZ11]. Recall that for an ind-proper semi-analytic set
${\textrm{Gr}}_{n,{\mathbb H}}$
 in [Reference O’BrienO’Br23, Reference Yun and ZhuYZ11]. Recall that for an ind-proper semi-analytic set 
 $Y={\textrm{colim}}_I Y_i$
 acting real analytically by a Lie group
$Y={\textrm{colim}}_I Y_i$
 acting real analytically by a Lie group 
 $G$
, the
$G$
, the 
 $G$
-equivariant homology
$G$
-equivariant homology 
 $H_*^G(Y)$
 of
$H_*^G(Y)$
 of 
 $Y$
 is defined as
$Y$
 is defined as 
 $H_*^G(Y):={\textrm{colim}}_I H^*_G(Y_i,\omega _{i})$
, where
$H_*^G(Y):={\textrm{colim}}_I H^*_G(Y_i,\omega _{i})$
, where 
 $\omega _{i}\in D(G\backslash Y_i)$
 is the dualizing sheaf of
$\omega _{i}\in D(G\backslash Y_i)$
 is the dualizing sheaf of 
 $G\backslash Y_i$
 and the colimit is induced by the natural adjunction map
$G\backslash Y_i$
 and the colimit is induced by the natural adjunction map 
 $(\iota _{i,i'})_*\omega _{i}\simeq (\iota _{i,i'})_!\omega _{i}\to \omega _{i'}$
, and the
$(\iota _{i,i'})_*\omega _{i}\simeq (\iota _{i,i'})_!\omega _{i}\to \omega _{i'}$
, and the 
 $G$
-equivariant cohomology
$G$
-equivariant cohomology 
 $H^*_G(Y)$
 of
$H^*_G(Y)$
 of 
 $Y$
 is defined as
$Y$
 is defined as 
 $H^*_G(Y):={\textrm{lim}}_I H^*_G(Y_i,{\mathbb C})$
, where the limit is induced by the natural restriction map
$H^*_G(Y):={\textrm{lim}}_I H^*_G(Y_i,{\mathbb C})$
, where the limit is induced by the natural restriction map 
 $H^*_G(Y_{i'},{\mathbb C})\to H^*_G(Y_i,{\mathbb C})$
.
$H^*_G(Y_{i'},{\mathbb C})\to H^*_G(Y_i,{\mathbb C})$
.
 Let 
 $\mathcal {L}$
 be the determinant line bundle on
$\mathcal {L}$
 be the determinant line bundle on 
 ${\textrm{Gr}}_{2n}$
 and let
${\textrm{Gr}}_{2n}$
 and let 
 $c_1^{T_{2n}}(\mathcal {L})\in H^2_{T_{2n}}({\textrm{Gr}}_{2n})$
 be its equivariant first Chern class. It is shown in [Reference Yun and ZhuYZ11, Lemma 2.2] that there is an isomorphism of functors
$c_1^{T_{2n}}(\mathcal {L})\in H^2_{T_{2n}}({\textrm{Gr}}_{2n})$
 be its equivariant first Chern class. It is shown in [Reference Yun and ZhuYZ11, Lemma 2.2] that there is an isomorphism of functors
 \begin{equation} H^*_{T_{2n}}({\textrm{Gr}}_{2n},-)\simeq H^*({\textrm{Gr}}_{2n},-)\otimes R_{T_{2n}}: \textrm{Perv}({\textrm{Gr}}_{2n})\to R_{T_{2n}}\textrm{-mod} \end{equation}
\begin{equation} H^*_{T_{2n}}({\textrm{Gr}}_{2n},-)\simeq H^*({\textrm{Gr}}_{2n},-)\otimes R_{T_{2n}}: \textrm{Perv}({\textrm{Gr}}_{2n})\to R_{T_{2n}}\textrm{-mod} \end{equation}
induced by the canonical splitting of the MV-filtration associated to the semi-infinite orbits 
 $S^\lambda _{2n}$
, the
$S^\lambda _{2n}$
, the 
 $\mathfrak{L} N_{2n}$
-orbits through
$\mathfrak{L} N_{2n}$
-orbits through 
 $\lambda \in \Lambda _{2n}$
. (Recall that
$\lambda \in \Lambda _{2n}$
. (Recall that 
 $N_{2n} \subset G_{2n}$
 denotes the subgroup of upper triangular unipotent matrices.) Moreover, the isomorphism respects the natural monoidal structures on
$N_{2n} \subset G_{2n}$
 denotes the subgroup of upper triangular unipotent matrices.) Moreover, the isomorphism respects the natural monoidal structures on 
 $H^*_{T_{2n}}({\textrm{Gr}}_{2n},-)$
 coming from fusion and the one on
$H^*_{T_{2n}}({\textrm{Gr}}_{2n},-)$
 coming from fusion and the one on 
 $H^*({\textrm{Gr}}_{2n},-)\otimes R_{T_{2n}}$
 induced from
$H^*({\textrm{Gr}}_{2n},-)\otimes R_{T_{2n}}$
 induced from 
 $H^*({\textrm{Gr}}_{2n},-)$
. The cup product action
$H^*({\textrm{Gr}}_{2n},-)$
. The cup product action 
 $\wedge c_1^{T_{2n}}(\mathcal {L})$
 on
$\wedge c_1^{T_{2n}}(\mathcal {L})$
 on 
 $H^*_{T_{2n}}({\textrm{Gr}}_{2n},\mathcal {F})$
 for
$H^*_{T_{2n}}({\textrm{Gr}}_{2n},\mathcal {F})$
 for 
 $\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{2n})$
 gives rise to a tensor endomorphism of
$\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{2n})$
 gives rise to a tensor endomorphism of 
 $H^*_{T_{2n}}({\textrm{Gr}}_{2n},-)$
 and hence, by the Tannakian formalism, gives rise to an element
$H^*_{T_{2n}}({\textrm{Gr}}_{2n},-)$
 and hence, by the Tannakian formalism, gives rise to an element 
 $c^{T_{2n}}\in {\mathfrak g}_{2n}\otimes R_{T_{2n}}$
. One can regard the element
$c^{T_{2n}}\in {\mathfrak g}_{2n}\otimes R_{T_{2n}}$
. One can regard the element 
 $c^{T_{2n}}$
 as a map
$c^{T_{2n}}$
 as a map
 \begin{equation} c^{T_{2n}}:{\mathfrak t}_{2n}\to {\mathfrak g}_{2n}. \end{equation}
\begin{equation} c^{T_{2n}}:{\mathfrak t}_{2n}\to {\mathfrak g}_{2n}. \end{equation}
The equivariant homology 
 $H_*^{T_{2n}}({\textrm{Gr}}_{2n})$
 is a commutative and cocommutative Hopf algebra over
$H_*^{T_{2n}}({\textrm{Gr}}_{2n})$
 is a commutative and cocommutative Hopf algebra over 
 $R_{T_{2n}}$
, and there is an isomorphism of group schemes
$R_{T_{2n}}$
, and there is an isomorphism of group schemes
 \begin{equation} \textrm{Spec}\big(H_*^{T_{2n}}({\textrm{Gr}}_{2n})\big)\simeq (G_{2n}\times {\mathfrak t}_{2n})^{c^{T_{2n}}} \end{equation}
\begin{equation} \textrm{Spec}\big(H_*^{T_{2n}}({\textrm{Gr}}_{2n})\big)\simeq (G_{2n}\times {\mathfrak t}_{2n})^{c^{T_{2n}}} \end{equation}
where 
 $(G_{2n}\times {\mathfrak t}_{2n})^{c^{T_{2n}}}$
 is the centralizer of
$(G_{2n}\times {\mathfrak t}_{2n})^{c^{T_{2n}}}$
 is the centralizer of 
 $c^{T_{2n}}$
 in
$c^{T_{2n}}$
 in 
 $G_{2n}\times {\mathfrak t}_{2n}$
.
$G_{2n}\times {\mathfrak t}_{2n}$
.
 We have a similar result for quaternionic Grassmannians. Let 
 $\mathcal {L}_{\mathbb H}$
 be the quaternionic determinant line bundle on
$\mathcal {L}_{\mathbb H}$
 be the quaternionic determinant line bundle on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 and let
${\textrm{Gr}}_{n,{\mathbb H}}$
 and let 
 $p^{T}(\mathcal {L}_{\mathbb H})\in H^4_{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be its equivariant Pontryagin class. It is shown in [Reference O’BrienO’Br23, Theorem 3] that there is an isomorphism of functors
$p^{T}(\mathcal {L}_{\mathbb H})\in H^4_{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be its equivariant Pontryagin class. It is shown in [Reference O’BrienO’Br23, Theorem 3] that there is an isomorphism of functors
 \begin{equation} H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},-)\simeq H^*({\textrm{Gr}}_{n,{\mathbb H}},-)\otimes _{\mathbb C} R_{T_{}}: \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\to R_{T_{}}\textrm{-mod} \end{equation}
\begin{equation} H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},-)\simeq H^*({\textrm{Gr}}_{n,{\mathbb H}},-)\otimes _{\mathbb C} R_{T_{}}: \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\to R_{T_{}}\textrm{-mod} \end{equation}
induced by the canonical splitting of the real MV-filtration associated to the real semi-infinite orbits 
 $S_{n,{\mathbb H}}^\lambda$
, the
$S_{n,{\mathbb H}}^\lambda$
, the 
 $\mathfrak{L} N_{n,{\mathbb H}}$
-orbits through
$\mathfrak{L} N_{n,{\mathbb H}}$
-orbits through 
 $\lambda \in \Lambda _S$
. Moreover, the isomorphism above respects the natural monoidal structures on
$\lambda \in \Lambda _S$
. Moreover, the isomorphism above respects the natural monoidal structures on 
 $H^*_{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}},-)$
 coming from fusion and the one on
$H^*_{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}},-)$
 coming from fusion and the one on 
 $H^*({\textrm{Gr}}_{n,{\mathbb H}},-)\otimes R_{T_{}}$
 induced from
$H^*({\textrm{Gr}}_{n,{\mathbb H}},-)\otimes R_{T_{}}$
 induced from 
 $H^*({\textrm{Gr}}_{n,{\mathbb H}},-)$
. The cup product action of
$H^*({\textrm{Gr}}_{n,{\mathbb H}},-)$
. The cup product action of 
 $p^{T}(\mathcal {L}_{\mathbb H})$
 on
$p^{T}(\mathcal {L}_{\mathbb H})$
 on 
 $H^*_{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 for
$H^*_{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 for 
 $\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Rep}(G_n)$
 gives rise to a tensor endomorphism of
$\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Rep}(G_n)$
 gives rise to a tensor endomorphism of 
 $H^*_{T_{}}({\textrm{Gr}}_{n,{\mathbb H}},-)$
 and hence an element
$H^*_{T_{}}({\textrm{Gr}}_{n,{\mathbb H}},-)$
 and hence an element 
 $p^T_X\in {\mathfrak g}_n\otimes R_T$
. Let
$p^T_X\in {\mathfrak g}_n\otimes R_T$
. Let
 \begin{equation} p^{T}_X:{\mathfrak t}_{}\to {\mathfrak g}_n \end{equation}
\begin{equation} p^{T}_X:{\mathfrak t}_{}\to {\mathfrak g}_n \end{equation}
be the corresponding map. The main result in [Reference O’BrienO’Br23, Theorem 9 and Corollary 1] says that there an isomorphism of group schemes
 \begin{equation} \textrm{Spec}\big(H_*^{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t}_{})^{p^{T_{}}_X} \end{equation}
\begin{equation} \textrm{Spec}\big(H_*^{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t}_{})^{p^{T_{}}_X} \end{equation}
where 
 $(G_n\times {\mathfrak t}_{})^{p^{T_{}}_X}$
 is the centralizer of
$(G_n\times {\mathfrak t}_{})^{p^{T_{}}_X}$
 is the centralizer of 
 $p^{T_{}}_X$
 in
$p^{T_{}}_X$
 in 
 $G_n\times {\mathfrak t}_{}$
.
$G_n\times {\mathfrak t}_{}$
.
 Recall the maps 
 $e^{T_{2n}}$
 and
$e^{T_{2n}}$
 and 
 $e^T_X$
 introduced in (3.13) and (3.14), respectively.
$e^T_X$
 introduced in (3.13) and (3.14), respectively.
Lemma 4.9. 
We have 
 $c^{T_{2n}}=e^{T_{2n}}$
 and
$c^{T_{2n}}=e^{T_{2n}}$
 and 
 $p^T_X=-e^T_X$
. Thus there are isomorphisms of group schemes
$p^T_X=-e^T_X$
. Thus there are isomorphisms of group schemes
 \begin{equation*}\textrm{Spec}\big(H_*^{T_{2n}}({\textrm{Gr}}_{2n})\big)\simeq (G_{2n}\times {\mathfrak t}_{2n})^{e^{T_{2n}}}\simeq J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}_{2n}, \end{equation*}
\begin{equation*}\textrm{Spec}\big(H_*^{T_{2n}}({\textrm{Gr}}_{2n})\big)\simeq (G_{2n}\times {\mathfrak t}_{2n})^{e^{T_{2n}}}\simeq J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}_{2n}, \end{equation*}
 \begin{equation*} \textrm{Spec}\big(H_*^{T_{}}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_{n}\times {\mathfrak t}_{})^{e_X^{T_{}}}\simeq J_{n}\times _{{\mathfrak c}_n}{\mathfrak t}\end{equation*}
\begin{equation*} \textrm{Spec}\big(H_*^{T_{}}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_{n}\times {\mathfrak t}_{})^{e_X^{T_{}}}\simeq J_{n}\times _{{\mathfrak c}_n}{\mathfrak t}\end{equation*}
over 
 ${\mathfrak t}_{2n}$
 and
${\mathfrak t}_{2n}$
 and 
 $\mathfrak t$
, respectively, and isomorphisms of group schemes
$\mathfrak t$
, respectively, and isomorphisms of group schemes
 \begin{equation*} H_*^{G_{c}}({\textrm{Gr}}_{2n})\simeq H_*^{T_{2n}}({\textrm{Gr}}_{2n})^{{\mathrm {W}}_{2n}}\simeq \big(J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}_{2n}\big)^{{\mathrm {W}}_{2n}}\simeq J_{2n}, \end{equation*}
\begin{equation*} H_*^{G_{c}}({\textrm{Gr}}_{2n})\simeq H_*^{T_{2n}}({\textrm{Gr}}_{2n})^{{\mathrm {W}}_{2n}}\simeq \big(J_{2n}\times _{{\mathfrak c}_{2n}}{\mathfrak t}_{2n}\big)^{{\mathrm {W}}_{2n}}\simeq J_{2n}, \end{equation*}
 \begin{equation*} H_*^{K_{c}}({\textrm{Gr}}_{n,{\mathbb H}})\simeq H_*^{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}})^{{\mathrm {W}}_{}}\simeq \big(J_{n}\times _{{\mathfrak c}_{n}}{\mathfrak t}_{}\big)^{{\mathrm {W}}_{}}\simeq J_{n}\end{equation*}
\begin{equation*} H_*^{K_{c}}({\textrm{Gr}}_{n,{\mathbb H}})\simeq H_*^{T_{c}}({\textrm{Gr}}_{n,{\mathbb H}})^{{\mathrm {W}}_{}}\simeq \big(J_{n}\times _{{\mathfrak c}_{n}}{\mathfrak t}_{}\big)^{{\mathrm {W}}_{}}\simeq J_{n}\end{equation*}
over 
 ${\mathfrak c}_{2n}={\mathfrak t}_{2n}//{\mathrm {W}}_{2n}$
 and
${\mathfrak c}_{2n}={\mathfrak t}_{2n}//{\mathrm {W}}_{2n}$
 and 
 ${\mathfrak t}//{\mathrm {W}}\simeq {\mathfrak c}_n$
, respectively.
${\mathfrak t}//{\mathrm {W}}\simeq {\mathfrak c}_n$
, respectively.
Proof. The result follows from the computations in [Reference Yun and ZhuYZ11, § 5] and [Reference O’BrienO’Br23]. We give an alternative (and more direct) proof using the computation in § 4.3.
It suffices to show that the element
 \begin{align*}c^{T_{2n}}=c_1^{T_{2n}}(\mathcal {L})\cup (-)\in \textrm{End}\big(H^*_{T_{2n}}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'}\big)\big)& \simeq \textrm{End}(H^*_{}({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'})\otimes R_{T_{2n}})\\ & \simeq \textrm{End}(V_{\omega _1'}\otimes R_{T_{2n}}) \simeq {\mathfrak g}_{2n}\otimes R_T \\ \big(\textrm{respectively,}\ p^T_X=p^{T}_1(\mathcal {L}_{\mathbb H})\cup (-)\in \textrm{End}\big(H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\big) &\simeq \textrm{End}(H^*_{}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\otimes R_T) \\ & \simeq \textrm{End}(V_{\omega _1}\otimes R_T) \simeq {\mathfrak g}_{n}\otimes R_T\big) \end{align*}
\begin{align*}c^{T_{2n}}=c_1^{T_{2n}}(\mathcal {L})\cup (-)\in \textrm{End}\big(H^*_{T_{2n}}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'}\big)\big)& \simeq \textrm{End}(H^*_{}({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'})\otimes R_{T_{2n}})\\ & \simeq \textrm{End}(V_{\omega _1'}\otimes R_{T_{2n}}) \simeq {\mathfrak g}_{2n}\otimes R_T \\ \big(\textrm{respectively,}\ p^T_X=p^{T}_1(\mathcal {L}_{\mathbb H})\cup (-)\in \textrm{End}\big(H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\big) &\simeq \textrm{End}(H^*_{}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\otimes R_T) \\ & \simeq \textrm{End}(V_{\omega _1}\otimes R_T) \simeq {\mathfrak g}_{n}\otimes R_T\big) \end{align*}
is given by 
 $e^{T_{2n}}$
$e^{T_{2n}}$
 
 $\big($
respectively,
$\big($
respectively, 
 $-e^T_X\big)$
. We have the following observations:
$-e^T_X\big)$
. We have the following observations:
- 
(1) there is a  $T_{2n}$
-equivariant (respectively, $T_{2n}$
-equivariant (respectively, $T_c$
-equivariant) isomorphism $T_c$
-equivariant) isomorphism ${\textrm{Gr}}^{\omega '_1}_{2n}\simeq \mathbb {P}^{2n-1}$ ${\textrm{Gr}}^{\omega '_1}_{2n}\simeq \mathbb {P}^{2n-1}$ $\big($
respectively, $\big($
respectively, ${\textrm{Gr}}^{\omega _1}_{n,{\mathbb H}}\simeq \mathbb {HP}^{n-1}\big)$
, where ${\textrm{Gr}}^{\omega _1}_{n,{\mathbb H}}\simeq \mathbb {HP}^{n-1}\big)$
, where $T_{2n}$
 (respectively, $T_{2n}$
 (respectively, $T_c$
) acts on $T_c$
) acts on $\mathbb {P}^{2n-1}$
 (respectively, $\mathbb {P}^{2n-1}$
 (respectively, $\mathbb {HP}^{n-1}$
) via the inverse of the natural action;Footnote 
6 $\mathbb {HP}^{n-1}$
) via the inverse of the natural action;Footnote 
6
- 
(2) the restriction  $\mathcal {L}|_{{\textrm{Gr}}_{2n}^{\omega '_1}}$ $\mathcal {L}|_{{\textrm{Gr}}_{2n}^{\omega '_1}}$ $\big($
respectively, $\big($
respectively, $\mathcal {L}_{{\mathbb H}}|_{{\textrm{Gr}}_{n,{\mathbb H}}^{\omega _1}}\big)$
 is isomorphic to $\mathcal {L}_{{\mathbb H}}|_{{\textrm{Gr}}_{n,{\mathbb H}}^{\omega _1}}\big)$
 is isomorphic to $\mathcal {O}(1)$
 (respectively, $\mathcal {O}(1)$
 (respectively, $\mathcal {O}_{\mathbb H}(1)$
, the $\mathcal {O}_{\mathbb H}(1)$
, the $\mathbb H$
-dual of the $\mathbb H$
-dual of the $\mathcal {O}_{\mathbb H}(-1)$
); $\mathcal {O}_{\mathbb H}(-1)$
);
- 
(3) the composed isomorphism  \begin{equation*}R_{T_{2n}}\otimes V_{\omega _1'}\simeq R_{T_{2n}}\otimes H^*_{}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'}\big)\simeq H^*_{T_{2n}}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'}\big)\simeq H^*_{T_{2n}}\big(\mathbb P^{2n-1},\textrm{IC}_{\mathbb P^{2n-1}}\big)\end{equation*} \begin{equation*}R_{T_{2n}}\otimes V_{\omega _1'}\simeq R_{T_{2n}}\otimes H^*_{}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'}\big)\simeq H^*_{T_{2n}}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1'}\big)\simeq H^*_{T_{2n}}\big(\mathbb P^{2n-1},\textrm{IC}_{\mathbb P^{2n-1}}\big)\end{equation*} \begin{equation*}(\textrm{respectively,}\ R_{T_{}}\otimes V_{\omega _1}\simeq R_{T_{}}\otimes H^*_{}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\simeq H^*_{T_{}}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\simeq H^*_{T_{}}(\mathbb {HP}^{n-1},\textrm{IC}_{\mathbb {HP}^{n-1}}))\end{equation*} \begin{equation*}(\textrm{respectively,}\ R_{T_{}}\otimes V_{\omega _1}\simeq R_{T_{}}\otimes H^*_{}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\simeq H^*_{T_{}}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_{\omega _1})\simeq H^*_{T_{}}(\mathbb {HP}^{n-1},\textrm{IC}_{\mathbb {HP}^{n-1}}))\end{equation*}
 sends the vectors 
 $1\otimes e_i, \:i=1,\ldots ,2n,$
 to the fundamental class
$1\otimes e_i, \:i=1,\ldots ,2n,$
 to the fundamental class
 \begin{equation*}[\mathbb {P}^{i-1}]\in H^{2n+1-2i}_{T_{2n}}(\mathbb P^{2n-1},\textrm{IC}_{\mathbb P^{2n-1}})=H^{4n-2i}_{T_{2n}}(\mathbb P^{2n-1})\end{equation*}
\begin{equation*}[\mathbb {P}^{i-1}]\in H^{2n+1-2i}_{T_{2n}}(\mathbb P^{2n-1},\textrm{IC}_{\mathbb P^{2n-1}})=H^{4n-2i}_{T_{2n}}(\mathbb P^{2n-1})\end{equation*}
(respectively, 
 $1\otimes e_i,\: i=1,\ldots ,n,$
 to the fundamental class
$1\otimes e_i,\: i=1,\ldots ,n,$
 to the fundamental class
 \begin{equation*}[\mathbb {HP}^{i-1}]\in H^{2n+2-4i}_{T_{}}(\mathbb {HP}^{n-1},\textrm{IC}_{\mathbb {HP}^{n-1}})=H^{4n-4i}_{T_{}}(\mathbb {HP}^{n-1})).\end{equation*}
\begin{equation*}[\mathbb {HP}^{i-1}]\in H^{2n+2-4i}_{T_{}}(\mathbb {HP}^{n-1},\textrm{IC}_{\mathbb {HP}^{n-1}})=H^{4n-4i}_{T_{}}(\mathbb {HP}^{n-1})).\end{equation*}
From the above observations, we see that 
 $c^{T_{2n}}\in {\mathfrak g}_{2n}\otimes R_{T_{2n}}$
 (respectively,
$c^{T_{2n}}\in {\mathfrak g}_{2n}\otimes R_{T_{2n}}$
 (respectively, 
 $p^T_X\in {\mathfrak g}_n\otimes R_{T}$
) is the matrix presentation of the cup product action
$p^T_X\in {\mathfrak g}_n\otimes R_{T}$
) is the matrix presentation of the cup product action 
 $c_1^{T_{2n}}(\mathcal {O}(1))\cup (-)$
 (respectively,
$c_1^{T_{2n}}(\mathcal {O}(1))\cup (-)$
 (respectively, 
 $e^T(\mathcal {O}_{\mathbb H}(1))\cup (-)$
)Footnote 
7
 in the basis
$e^T(\mathcal {O}_{\mathbb H}(1))\cup (-)$
)Footnote 
7
 in the basis 
 $\{[\mathbb P^{i-1}]\}_{i=1,\ldots ,2n}$
 (respectively,
$\{[\mathbb P^{i-1}]\}_{i=1,\ldots ,2n}$
 (respectively, 
 $\{[\mathbb {HP}^{i-1}]\}_{i=1,\ldots ,n}$
), and the desired claim follows from Lemma 4.4.
$\{[\mathbb {HP}^{i-1}]\}_{i=1,\ldots ,n}$
), and the desired claim follows from Lemma 4.4.
4.6.2
 Recall that for any Lie group 
 $G$
 and any ind-proper
$G$
 and any ind-proper 
 $G$
-variety
$G$
-variety 
 $Y$
 we have a paring
$Y$
 we have a paring
 \begin{equation*}H^*_{G}(Y)\times H_*^{G}(Y)\to H^*_{G}(\textrm{pt})\simeq R_{G}\end{equation*}
\begin{equation*}H^*_{G}(Y)\times H_*^{G}(Y)\to H^*_{G}(\textrm{pt})\simeq R_{G}\end{equation*}
induced by the action of cohomology on homology and then the pushforward map in the Borel–Moore homology 
 $H_*^{G}(Y)\to H^*_{G}(\textrm{pt})$
. On the other hand, for any commutative affine group scheme
$H_*^{G}(Y)\to H^*_{G}(\textrm{pt})$
. On the other hand, for any commutative affine group scheme 
 $H$
 over
$H$
 over 
 $S$
 there is a canonical paring
$S$
 there is a canonical paring
 \begin{equation*}U(\textrm{Lie} H)\times \mathcal {O}(H)\to \mathcal {O}(S),\quad (\xi ,f)\to \xi (f)\big|_e\end{equation*}
\begin{equation*}U(\textrm{Lie} H)\times \mathcal {O}(H)\to \mathcal {O}(S),\quad (\xi ,f)\to \xi (f)\big|_e\end{equation*}
between the relative universal enveloping algebra 
 $U({\textrm{Lie}} H)$
 and the ring of functions on
$U({\textrm{Lie}} H)$
 and the ring of functions on 
 $H$
. Here
$H$
. Here 
 $e:S\to H$
 is the unity map.
$e:S\to H$
 is the unity map.
According to [Reference Bezrukavnikov, Finkelberg and MirkovicBFM05, Remark 2.13], there are isomorphisms
 \begin{equation} H^*_{G_c}({\textrm{Gr}}_{2n})\simeq U({\textrm{Lie}} J_{2n})\quad \mbox{and} \quad H^*_{K_c}({\textrm{Gr}}_{n,{\mathbb H}})\simeq U({\textrm{Lie}} J_{n}) \end{equation}
\begin{equation} H^*_{G_c}({\textrm{Gr}}_{2n})\simeq U({\textrm{Lie}} J_{2n})\quad \mbox{and} \quad H^*_{K_c}({\textrm{Gr}}_{n,{\mathbb H}})\simeq U({\textrm{Lie}} J_{n}) \end{equation}
such that the paring above between the cohomology and homology of 
 $Y={\textrm{Gr}}_{2n}$
 (respectively,
$Y={\textrm{Gr}}_{2n}$
 (respectively, 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
) becomes the paring between the universal enveloping algebra and ring of functions for the group scheme
${\textrm{Gr}}_{n,{\mathbb H}}$
) becomes the paring between the universal enveloping algebra and ring of functions for the group scheme 
 $H=J_{2n}$
 (respectively,
$H=J_{2n}$
 (respectively, 
 $J_n$
).
$J_n$
).
4.7 Fully-faithfulness
 A key ingredient in the proof of the (complex) derived Satake theorem is the fully-faithfulness of the equivariant cohomology functor 
 $H_{\mathfrak{L}^+G_{2n}}^*({\textrm{Gr}}_{2n}, -)$
 into the category of modules over the global cohomology
$H_{\mathfrak{L}^+G_{2n}}^*({\textrm{Gr}}_{2n}, -)$
 into the category of modules over the global cohomology 
 $H_{\mathfrak{L}^+G_{2n}}^*({\textrm{Gr}}_{2n}, {\mathbb C})$
. In [Reference Bezrukavnikov and FinkelbergBF08], this was established using general results of Ginzburg [Reference GinzburgGin91]. Ginzburg’s arguments appeal to Hodge theory and therefore must be modified in the real setting. As in [Reference Achar and RiderAR15], we can use parity considerations in place of Hodge theory. More precisely, we will make use of the theory of parity sheaves [Reference Juteau, Mautner and WilliamsonJMW14]. Our first step, therefore, is to establish that the complexes
$H_{\mathfrak{L}^+G_{2n}}^*({\textrm{Gr}}_{2n}, {\mathbb C})$
. In [Reference Bezrukavnikov and FinkelbergBF08], this was established using general results of Ginzburg [Reference GinzburgGin91]. Ginzburg’s arguments appeal to Hodge theory and therefore must be modified in the real setting. As in [Reference Achar and RiderAR15], we can use parity considerations in place of Hodge theory. More precisely, we will make use of the theory of parity sheaves [Reference Juteau, Mautner and WilliamsonJMW14]. Our first step, therefore, is to establish that the complexes 
 $\textrm{IC}_\lambda$
 (for
$\textrm{IC}_\lambda$
 (for 
 $\lambda \in \Lambda _S^+$
) are even.
$\lambda \in \Lambda _S^+$
) are even.
Remark 4.10. In fact, our situation is simpler than the modular setting considered in [Reference Juteau, Mautner and WilliamsonJMW14] owing to the fact that the tensor category 
 $\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
 of spherical perverse sheaves on
$\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
 of spherical perverse sheaves on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 is semisimple (see Proposition 4.5).
${\textrm{Gr}}_{n,{\mathbb H}}$
 is semisimple (see Proposition 4.5).
4.7.1
 Recall that if a coweight 
 $\mu \in \Lambda _S^+$
 is minuscule, the orbit
$\mu \in \Lambda _S^+$
 is minuscule, the orbit 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{\mu }$
 is closed. Such an orbit is necessarily smooth.
${\textrm{Gr}}_{n,{\mathbb H}}^{\mu }$
 is closed. Such an orbit is necessarily smooth.
Lemma 4.11. 
Let 
 $\mu _1,\ldots ,\mu _k \in \Lambda _S^+$
 denote minuscule coweights. Consider the convolution morphism
$\mu _1,\ldots ,\mu _k \in \Lambda _S^+$
 denote minuscule coweights. Consider the convolution morphism
 \begin{equation*} m: {\textrm{Gr}}_{n,{\mathbb H}}^{\mu _{\bullet }} := {\textrm{Gr}}_{n,{\mathbb H}}^{\mu _1} {\widetilde {\times }} \cdots {\widetilde {\times }} {\textrm{Gr}}_{n,{\mathbb H}}^{\mu _k} \rightarrow {\textrm{Gr}}_{n,{\mathbb H}}. \end{equation*}
\begin{equation*} m: {\textrm{Gr}}_{n,{\mathbb H}}^{\mu _{\bullet }} := {\textrm{Gr}}_{n,{\mathbb H}}^{\mu _1} {\widetilde {\times }} \cdots {\widetilde {\times }} {\textrm{Gr}}_{n,{\mathbb H}}^{\mu _k} \rightarrow {\textrm{Gr}}_{n,{\mathbb H}}. \end{equation*}
Then the non-empty fibers of 
 $m$
 are paved by quaternionic affine spaces.
$m$
 are paved by quaternionic affine spaces.
Proof.
 In the complex setting, this result is due to [Reference HainesHai06]. We proceed by induction on 
 $k$
. When
$k$
. When 
 $k=1$
, there is nothing to prove. In general, we factor
$k=1$
, there is nothing to prove. In general, we factor 
 $m$
 as follows.
$m$
 as follows.

 Here, 
 $q$
 is induced by multiplying the first
$q$
 is induced by multiplying the first 
 $k-1$
 factors of
$k-1$
 factors of 
 ${\textrm{Gr}}^{\mu _\bullet }_{n,{\mathbb H}}$
. Since
${\textrm{Gr}}^{\mu _\bullet }_{n,{\mathbb H}}$
. Since 
 $m$
 is
$m$
 is 
 $\mathfrak {L}^+{G}_{n,{\mathbb H}}$
-equivariant, it suffices to show that each fiber
$\mathfrak {L}^+{G}_{n,{\mathbb H}}$
-equivariant, it suffices to show that each fiber 
 $m^{-1}(t^\lambda )$
 (for
$m^{-1}(t^\lambda )$
 (for 
 $\lambda \in \Lambda _S^+$
) is paved by quaternionic affine spaces. By the above diagram, we have
$\lambda \in \Lambda _S^+$
) is paved by quaternionic affine spaces. By the above diagram, we have
 \begin{equation*} m^{-1}(t^\lambda ) = q^{-1}(p^{-1}(t^\lambda )). \end{equation*}
\begin{equation*} m^{-1}(t^\lambda ) = q^{-1}(p^{-1}(t^\lambda )). \end{equation*}
Let 
 $\mu _\bullet ' = (\mu _1,\ldots ,\mu _{k-1})$
. Observe that we have the following commutative diagram.
$\mu _\bullet ' = (\mu _1,\ldots ,\mu _{k-1})$
. Observe that we have the following commutative diagram.

 Here, 
 $\pi$
 is the projection to the first factor and
$\pi$
 is the projection to the first factor and 
 $a$
 is the convolution map. Observe that both horizontal compositions are closed embeddings. Hence, we obtain a Cartesian diagram as follows.
$a$
 is the convolution map. Observe that both horizontal compositions are closed embeddings. Hence, we obtain a Cartesian diagram as follows.

 By induction, the fibers of 
 $a$
 are paved by quaternionic affine spaces. Hence, the same is true of
$a$
 are paved by quaternionic affine spaces. Hence, the same is true of 
 $\pi \circ q$
. It therefore suffices to show that
$\pi \circ q$
. It therefore suffices to show that 
 $\pi (p^{-1}(t^\lambda ))$
 is paved by quaternionic affine spaces over which
$\pi (p^{-1}(t^\lambda ))$
 is paved by quaternionic affine spaces over which 
 $a$
 is a trivial fibration. Since
$a$
 is a trivial fibration. Since 
 $\mathfrak {L}^+G_{n,{\mathbb H}}$
 acts transitively on the fiber
$\mathfrak {L}^+G_{n,{\mathbb H}}$
 acts transitively on the fiber 
 $p^{-1}(t^0)$
, we have
$p^{-1}(t^0)$
, we have 
 $\pi (p^{-1}(t^0))= \mathfrak {L}^+G_{n,{\mathbb H}} t^{-\mu _k} ={\textrm{Gr}}_{n,{\mathbb H}}^{-w_0(\mu _k)}$
 and hence
$\pi (p^{-1}(t^0))= \mathfrak {L}^+G_{n,{\mathbb H}} t^{-\mu _k} ={\textrm{Gr}}_{n,{\mathbb H}}^{-w_0(\mu _k)}$
 and hence
 \begin{equation*}\pi (p^{-1}(t^\lambda )) = t^\lambda \pi (p^{-1}(t^0)) = t^\lambda {\textrm{Gr}}_{n,{\mathbb H}}^{-w_0(\mu _k)},\end{equation*}
\begin{equation*}\pi (p^{-1}(t^\lambda )) = t^\lambda \pi (p^{-1}(t^0)) = t^\lambda {\textrm{Gr}}_{n,{\mathbb H}}^{-w_0(\mu _k)},\end{equation*}
where 
 $w_0$
 is the longest element of the Weyl group. Multiplication by
$w_0$
 is the longest element of the Weyl group. Multiplication by 
 $t^\lambda$
 is an isomorphism commuting with
$t^\lambda$
 is an isomorphism commuting with 
 $a$
, so it suffices to show that
$a$
, so it suffices to show that 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{-w_0(\mu _k)}$
 is paved by quaternionic affine spaces over which
${\textrm{Gr}}_{n,{\mathbb H}}^{-w_0(\mu _k)}$
 is paved by quaternionic affine spaces over which 
 $a$
 is a trivial fibration. Let
$a$
 is a trivial fibration. Let 
 $\mu = -w_0 (\mu _k)$
. The coweight
$\mu = -w_0 (\mu _k)$
. The coweight 
 $\mu$
 is once again minuscule. Recall that
$\mu$
 is once again minuscule. Recall that 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{\mu }$
 is a vector bundle over a partial flag variety of
${\textrm{Gr}}_{n,{\mathbb H}}^{\mu }$
 is a vector bundle over a partial flag variety of 
 $G_{n,{\mathbb H}}$
. On the other hand,
$G_{n,{\mathbb H}}$
. On the other hand, 
 ${\textrm{Gr}}_{n,{\mathbb H}}^{\mu }$
 is closed, so it is a partial flag variety of
${\textrm{Gr}}_{n,{\mathbb H}}^{\mu }$
 is closed, so it is a partial flag variety of 
 $G_{n,{\mathbb H}}$
. We claim that the orbits of
$G_{n,{\mathbb H}}$
. We claim that the orbits of 
 $P^{\mu }_{n,{\mathbb H}}$
 on
$P^{\mu }_{n,{\mathbb H}}$
 on 
 ${\textrm{Gr}}^\mu _{n,{\mathbb H}}$
 are the desired affine spaces.
${\textrm{Gr}}^\mu _{n,{\mathbb H}}$
 are the desired affine spaces.
 Each such orbit has the form 
 $P^{\mu }_{n,{\mathbb H}} t^{w(\mu )}$
 for
$P^{\mu }_{n,{\mathbb H}} t^{w(\mu )}$
 for 
 $w \in W_n$
 an element of the Weyl group. Let
$w \in W_n$
 an element of the Weyl group. Let 
 $^{w}P^{\mu }_{n,{\mathbb H}} = w\big(P^{\mu }_{n,{\mathbb H}}\big)$
. Then, by the real Bruhat decomposition, there exists a unipotent subgroup
$^{w}P^{\mu }_{n,{\mathbb H}} = w\big(P^{\mu }_{n,{\mathbb H}}\big)$
. Then, by the real Bruhat decomposition, there exists a unipotent subgroup 
 $^{w}N^{\mu }_{n,{\mathbb H}}$
 of
$^{w}N^{\mu }_{n,{\mathbb H}}$
 of 
 $^{w}P^{\mu }_{n,{\mathbb H}}$
 which acts freely and transitively on the orbit
$^{w}P^{\mu }_{n,{\mathbb H}}$
 which acts freely and transitively on the orbit 
 $P^{\mu }_{n,{\mathbb H}} t^{w(\mu )}$
. Hence
$P^{\mu }_{n,{\mathbb H}} t^{w(\mu )}$
. Hence 
 $P^{\mu }_{n,{\mathbb H}} t^{w(\mu )} = {^w}N^{\mu }_{n,{\mathbb H}} t^{w(\mu )}$
. By the
$P^{\mu }_{n,{\mathbb H}} t^{w(\mu )} = {^w}N^{\mu }_{n,{\mathbb H}} t^{w(\mu )}$
. By the 
 $\mathfrak {L}G_{n,{\mathbb H}}$
 equivariance of
$\mathfrak {L}G_{n,{\mathbb H}}$
 equivariance of 
 $a$
, we have the following commutative diagram.
$a$
, we have the following commutative diagram.

 As the diagram is Cartesian and the bottom arrow is an isomorphism, the top arrow is an isomorphism as well. By induction, 
 $a^{-1}(t^{w(\mu )})$
 is paved by quaternionic affine spaces. Therefore, it suffices to show that the unipotent subgroup
$a^{-1}(t^{w(\mu )})$
 is paved by quaternionic affine spaces. Therefore, it suffices to show that the unipotent subgroup 
 $^{w}N^\mu _{n,{\mathbb H}}$
 is a quaternionic affine space, which is clear.
$^{w}N^\mu _{n,{\mathbb H}}$
 is a quaternionic affine space, which is clear.
 We now recall the terminology of [Reference Juteau, Mautner and WilliamsonJMW14] that we will use. For 
 $\lambda \in \Lambda _S^+$
, let
$\lambda \in \Lambda _S^+$
, let
 \begin{equation*} i_\lambda : {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
\begin{equation*} i_\lambda : {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
denote the inclusion.
Definition 4.12. Let 
 $\mathcal {F} \in D^b_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
. We say that
$\mathcal {F} \in D^b_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
. We say that 
 $\mathcal {F}$
 is
$\mathcal {F}$
 is 
 $*$
-even (respectively,
$*$
-even (respectively, 
 $!$
-even) if for all
$!$
-even) if for all 
 $\lambda \in \Lambda _S^+$
, the
$\lambda \in \Lambda _S^+$
, the 
 $\mathfrak {L}^+G_{n,{\mathbb H}}$
-equivariant sheaf
$\mathfrak {L}^+G_{n,{\mathbb H}}$
-equivariant sheaf 
 $i_\lambda ^*\mathcal {F}$
 (respectively,
$i_\lambda ^*\mathcal {F}$
 (respectively, 
 $i_\lambda ^!\mathcal {F}$
) is a direct sum of constant sheaves appearing in even degrees. If
$i_\lambda ^!\mathcal {F}$
) is a direct sum of constant sheaves appearing in even degrees. If 
 $\mathcal {F}$
 is both
$\mathcal {F}$
 is both 
 $*$
-even and
$*$
-even and 
 $!$
-even, we simply say that it is even.
$!$
-even, we simply say that it is even.
 We say that 
 $\mathcal {F}$
 is
$\mathcal {F}$
 is 
 $*$
-odd (respectively,
$*$
-odd (respectively, 
 $!$
-odd) if
$!$
-odd) if 
 $\mathcal {F}[1]$
 is
$\mathcal {F}[1]$
 is 
 $*$
-even (respectively,
$*$
-even (respectively, 
 $!$
-even). If
$!$
-even). If 
 $\mathcal {F}$
 is both
$\mathcal {F}$
 is both 
 $*$
-odd and
$*$
-odd and 
 $!$
-odd, we simply say that it is odd.
$!$
-odd, we simply say that it is odd.
Proposition 4.13. 
For 
 $\lambda \in \Lambda _S^+$
, the complex
$\lambda \in \Lambda _S^+$
, the complex 
 $\textrm{IC}_\lambda$
 is even.
$\textrm{IC}_\lambda$
 is even.
Proof.
 Since 
 $\textrm{IC}_\lambda$
 is self-dual, it suffices to show that it is
$\textrm{IC}_\lambda$
 is self-dual, it suffices to show that it is 
 $*$
-even. Recall from Proposition 4.5(1) that we have an equivalence
$*$
-even. Recall from Proposition 4.5(1) that we have an equivalence 
 $\mathrm {Perv}({\textrm{Gr}}_{n,{\mathbb H}}) \simeq \mathrm {Rep}(G_n)$
 taking
$\mathrm {Perv}({\textrm{Gr}}_{n,{\mathbb H}}) \simeq \mathrm {Rep}(G_n)$
 taking 
 $\textrm{IC}_\lambda$
 to the highest-weight module
$\textrm{IC}_\lambda$
 to the highest-weight module 
 $V_\lambda$
. Then
$V_\lambda$
. Then 
 $V_\lambda$
 is a direct summand of a tensor product
$V_\lambda$
 is a direct summand of a tensor product 
 $V_{\epsilon }^{\otimes j} \otimes V_{\omega _1}^{\otimes k}$
 for some
$V_{\epsilon }^{\otimes j} \otimes V_{\omega _1}^{\otimes k}$
 for some 
 $j,k \geq 0$
. Hence
$j,k \geq 0$
. Hence 
 $\textrm{IC}_\lambda$
 is a direct summand of the convolution
$\textrm{IC}_\lambda$
 is a direct summand of the convolution 
 $\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}$
. It therefore suffices to show that
$\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}$
. It therefore suffices to show that 
 $\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}$
 is
$\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}$
 is 
 $*$
-even. We now apply Lemma 4.11 with
$*$
-even. We now apply Lemma 4.11 with 
 $\mu _1,\ldots ,\mu _j = \epsilon$
 and
$\mu _1,\ldots ,\mu _j = \epsilon$
 and 
 $\mu _{j+1} =\ldots = \mu _{j+k} = \omega _1$
. Let
$\mu _{j+1} =\ldots = \mu _{j+k} = \omega _1$
. Let
 \begin{equation*} m: {\textrm{Gr}}^{\mu _\bullet }_{n,{\mathbb H}} \rightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
\begin{equation*} m: {\textrm{Gr}}^{\mu _\bullet }_{n,{\mathbb H}} \rightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
denote the convolution map. We have
 \begin{equation*} \textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k} \simeq m_!\big(\textrm{IC}_\epsilon ^{\widetilde {\boxtimes } j} {\widetilde {\boxtimes }} \textrm{IC}_\epsilon ^{\widetilde {\boxtimes } k}\big). \end{equation*}
\begin{equation*} \textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k} \simeq m_!\big(\textrm{IC}_\epsilon ^{\widetilde {\boxtimes } j} {\widetilde {\boxtimes }} \textrm{IC}_\epsilon ^{\widetilde {\boxtimes } k}\big). \end{equation*}
Now let 
 $\nu \in \Lambda _S$
 and let
$\nu \in \Lambda _S$
 and let 
 $i_\nu : {\textrm{Gr}}^\nu _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}}$
 denote the inclusion. Firstly, we have
$i_\nu : {\textrm{Gr}}^\nu _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}}$
 denote the inclusion. Firstly, we have
 \begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)\big) \simeq H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)\big)^W. \end{equation*}
\begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)\big) \simeq H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)\big)^W. \end{equation*}
Next, by proper base change,
 \begin{equation*} H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)\big) \simeq H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), \textrm{IC}_\epsilon ^{\widetilde {\boxtimes } j} {\widetilde {\boxtimes }} \textrm{IC}_{\omega _1}^{\widetilde {\boxtimes } k}\big). \end{equation*}
\begin{equation*} H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)\big) \simeq H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), \textrm{IC}_\epsilon ^{\widetilde {\boxtimes } j} {\widetilde {\boxtimes }} \textrm{IC}_{\omega _1}^{\widetilde {\boxtimes } k}\big). \end{equation*}
Since 
 $\epsilon$
 and
$\epsilon$
 and 
 $\omega _1$
 are minuscule, the orbits
$\omega _1$
 are minuscule, the orbits 
 ${\textrm{Gr}}^{\omega _1}_{n,{\mathbb H}}$
 and
${\textrm{Gr}}^{\omega _1}_{n,{\mathbb H}}$
 and 
 ${\textrm{Gr}}^{\epsilon }_{n,{\mathbb H}}$
 are smooth. Therefore,
${\textrm{Gr}}^{\epsilon }_{n,{\mathbb H}}$
 are smooth. Therefore, 
 $\textrm{IC}_{\omega _1} \simeq \underline {{\mathbb C}}[2(n-1)]$
 and
$\textrm{IC}_{\omega _1} \simeq \underline {{\mathbb C}}[2(n-1)]$
 and 
 $\textrm{IC}_{\epsilon } \simeq \underline {{\mathbb C}}$
. Hence,
$\textrm{IC}_{\epsilon } \simeq \underline {{\mathbb C}}$
. Hence,
 \begin{equation*} H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), \textrm{IC}_\epsilon ^{\widetilde {\boxtimes } j} {\widetilde {\boxtimes }} \textrm{IC}_{\omega _1}^{\widetilde {\boxtimes } k}\big) \simeq H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), {\mathbb C}\big)[2k(n-1)]. \end{equation*}
\begin{equation*} H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), \textrm{IC}_\epsilon ^{\widetilde {\boxtimes } j} {\widetilde {\boxtimes }} \textrm{IC}_{\omega _1}^{\widetilde {\boxtimes } k}\big) \simeq H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), {\mathbb C}\big)[2k(n-1)]. \end{equation*}
By Lemma 4.11, the ordinary cohomology 
 $H^*\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), {\mathbb C}\big)$
 is concentrated in even degrees (in fact, in degrees divisible by
$H^*\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), {\mathbb C}\big)$
 is concentrated in even degrees (in fact, in degrees divisible by 
 $4$
). Hence,
$4$
). Hence, 
 $m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big)$
 is equivariantly formal with respect to the action of
$m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big)$
 is equivariantly formal with respect to the action of 
 $T_c$
. Therefore,
$T_c$
. Therefore, 
 $H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), {\mathbb C}\big)$
 is concentrated in even degrees.
$H^*_{T_c}\big(m^{-1}\big({\textrm{Gr}}^\nu _{n,{\mathbb H}}\big), {\mathbb C}\big)$
 is concentrated in even degrees.
 Now we may express 
 $i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)$
 as a direct sum of constant sheaves. We have
$i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big)$
 as a direct sum of constant sheaves. We have
 \begin{equation*} i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big) \simeq \underline {V} \end{equation*}
\begin{equation*} i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}\big) \simeq \underline {V} \end{equation*}
for a complex 
 $V \in D^b(\mathrm {Vect}_{{\mathbb C}})$
. Hence,
$V \in D^b(\mathrm {Vect}_{{\mathbb C}})$
. Hence,
 \begin{equation*} H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j}\big) \simeq H^*_{T_c}\big({\textrm{Gr}}^\lambda _{n,{\mathbb H}}, {\mathbb C}\big) \otimes V. \end{equation*}
\begin{equation*} H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j}\big) \simeq H^*_{T_c}\big({\textrm{Gr}}^\lambda _{n,{\mathbb H}}, {\mathbb C}\big) \otimes V. \end{equation*}
We have shown that 
 $H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j}\big)$
 is concentrated in even degrees. Since
$H^*_{T_c}\big(i_\nu ^*\big(\textrm{IC}_{\epsilon }^{\star j}\big)$
 is concentrated in even degrees. Since 
 $H^0_{T_c}\big({\textrm{Gr}}^\lambda _{n,{\mathbb H}}, {\mathbb C}\big) \neq 0$
, we conclude that
$H^0_{T_c}\big({\textrm{Gr}}^\lambda _{n,{\mathbb H}}, {\mathbb C}\big) \neq 0$
, we conclude that 
 $V$
 is concentrated in even degrees. The result follows.
$V$
 is concentrated in even degrees. The result follows.
As a corollary of the proof we obtain the following parity vanishing result.
Corollary 4.14. 
We have 
 $\mathscr H^{\,\,i-\langle \lambda ,\rho _{2n}\rangle }(\textrm{IC}_\lambda )=0$
 for
$\mathscr H^{\,\,i-\langle \lambda ,\rho _{2n}\rangle }(\textrm{IC}_\lambda )=0$
 for 
 $i\nmid 4$
.
$i\nmid 4$
.
Proof.
 We have shown that any direct summand 
 $\textrm{IC}_\lambda$
 of
$\textrm{IC}_\lambda$
 of 
 $\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}$
 satisfies
$\textrm{IC}_{\epsilon }^{\star j} \star \textrm{IC}_{\omega _1}^{\star k}$
 satisfies 
 $\mathscr H^{\,\,i-2k(n-1)}(\textrm{IC}_\lambda )Z=0$
 for
$\mathscr H^{\,\,i-2k(n-1)}(\textrm{IC}_\lambda )Z=0$
 for 
 $i\nmid 4$
. Since
$i\nmid 4$
. Since 
 $k\omega _1-\lambda$
 is a non-negative integral sum of positive coroots, we have
$k\omega _1-\lambda$
 is a non-negative integral sum of positive coroots, we have 
 $\langle k\omega _1-\lambda ,\rho _n\rangle \in {\mathbb Z}$
, and hence
$\langle k\omega _1-\lambda ,\rho _n\rangle \in {\mathbb Z}$
, and hence 
 $2k(n-1)-\langle \lambda ,\rho _{2n}\rangle = \langle k\omega _1-\lambda ,\rho _{2n}\rangle =4\langle k\omega _1-\lambda ,\rho _n\rangle$
 is divisible by four. The desired claim follows.
$2k(n-1)-\langle \lambda ,\rho _{2n}\rangle = \langle k\omega _1-\lambda ,\rho _{2n}\rangle =4\langle k\omega _1-\lambda ,\rho _n\rangle$
 is divisible by four. The desired claim follows.
4.7.2
Our goal is now to apply the parity vanishing result above to deduce the following faithfulness result.
Proposition 4.15. 
For any 
 $\lambda , \mu \in \Lambda _S^+$
, the natural map
$\lambda , \mu \in \Lambda _S^+$
, the natural map
 \begin{equation*}\textrm{Ext}^\bullet _{D^b_{\mathfrak{L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_\lambda ,\textrm{IC}_\mu )\to {\textrm{Hom}}^\bullet _{H^*_{\mathfrak{L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{\mathfrak{L}^+G_{n,{\mathbb H}}}(\textrm{IC}_\lambda ),H^*_{\mathfrak{L}^+G_{n,{\mathbb H}}}(\textrm{IC}_\mu )\big)\end{equation*}
\begin{equation*}\textrm{Ext}^\bullet _{D^b_{\mathfrak{L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_\lambda ,\textrm{IC}_\mu )\to {\textrm{Hom}}^\bullet _{H^*_{\mathfrak{L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{\mathfrak{L}^+G_{n,{\mathbb H}}}(\textrm{IC}_\lambda ),H^*_{\mathfrak{L}^+G_{n,{\mathbb H}}}(\textrm{IC}_\mu )\big)\end{equation*}
is an isomorphism of graded modules.
We will deduce Proposition 4.15 as a consequence of the following more general result.
Proposition 4.16. 
Let 
 $\mathcal {F}, \mathcal {G} \in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
. Assume that
$\mathcal {F}, \mathcal {G} \in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
. Assume that 
 $\mathcal {F}$
 and
$\mathcal {F}$
 and 
 $\mathcal {G}$
 are even. Then the natural map
$\mathcal {G}$
 are even. Then the natural map
 \begin{equation*} \mathrm {Ext}^\bullet _{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\mathcal {F}, \mathcal {G}) \rightarrow \mathrm {Hom}^\bullet _{H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}), H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {G})\big) \end{equation*}
\begin{equation*} \mathrm {Ext}^\bullet _{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\mathcal {F}, \mathcal {G}) \rightarrow \mathrm {Hom}^\bullet _{H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}), H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {G})\big) \end{equation*}
is an isomorphism of graded modules.
4.7.3
In the proof of Proposition 4.16, we will make use of the following terminology. Consider a triangulated functor
 \begin{equation*} \Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}) \rightarrow D^b(\mathrm {Vect}_{\mathbb C}). \end{equation*}
\begin{equation*} \Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}) \rightarrow D^b(\mathrm {Vect}_{\mathbb C}). \end{equation*}
We say that 
 $\Omega$
 is
$\Omega$
 is 
 $*$
-parity preserving (respectively,
$*$
-parity preserving (respectively, 
 $!$
-parity preserving) if it takes
$!$
-parity preserving) if it takes 
 $*$
-even (respectively,
$*$
-even (respectively, 
 $!$
-even) complexes of sheaves to even complexes of vector spaces. If
$!$
-even) complexes of sheaves to even complexes of vector spaces. If 
 $\Omega$
 is both
$\Omega$
 is both 
 $*$
-parity preserving and
$*$
-parity preserving and 
 $!$
-parity preserving, we will simply say that it is parity preserving. We use the same terminology for functors
$!$
-parity preserving, we will simply say that it is parity preserving. We use the same terminology for functors
 \begin{equation*} \Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})^{\mathrm {op}} \rightarrow D^b(\mathrm {Vect}_{\mathbb C}). \end{equation*}
\begin{equation*} \Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})^{\mathrm {op}} \rightarrow D^b(\mathrm {Vect}_{\mathbb C}). \end{equation*}
To check that functors are parity preserving, we will use the following criterion.
Lemma 4.17. Let
 \begin{equation*} \Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}) \rightarrow D^b(\mathrm {Vect}_{\mathbb C}) \end{equation*}
\begin{equation*} \Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}) \rightarrow D^b(\mathrm {Vect}_{\mathbb C}) \end{equation*}
be a triangulated functor. Then 
 $\Omega$
 is
$\Omega$
 is 
 $*$
-even if and only if for each
$*$
-even if and only if for each 
 $\lambda \in \Lambda _S^+$
 the complex
$\lambda \in \Lambda _S^+$
 the complex 
 $\Omega (j_{\lambda !} \underline {{\mathbb C}})$
 is even. Here,
$\Omega (j_{\lambda !} \underline {{\mathbb C}})$
 is even. Here, 
 $j_{\lambda }: {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}}$
 is the natural inclusion.
$j_{\lambda }: {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}}$
 is the natural inclusion.
Proof.
 We assume that the latter condition holds and prove that 
 $\Omega$
 is
$\Omega$
 is 
 $*$
-parity preserving. Let
$*$
-parity preserving. Let 
 $\mathcal {F} \in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be
$\mathcal {F} \in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be 
 $*$
-even. We must show that
$*$
-even. We must show that 
 $\Omega (\mathcal {F})$
 is even, which we do by induction on the support of
$\Omega (\mathcal {F})$
 is even, which we do by induction on the support of 
 $\mathcal {F}$
 (which is a finite union of
$\mathcal {F}$
 (which is a finite union of 
 $\mathfrak {L}^+ G_{n,{\mathbb H}}$
-orbits). Let
$\mathfrak {L}^+ G_{n,{\mathbb H}}$
-orbits). Let
 \begin{equation*} j: {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
\begin{equation*} j: {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
denote the inclusion of a 
 $\mathfrak {L}^+ G_{n,{\mathbb H}}$
-orbit open in the support of
$\mathfrak {L}^+ G_{n,{\mathbb H}}$
-orbit open in the support of 
 $\mathcal {F}$
. Let
$\mathcal {F}$
. Let
 \begin{equation*} i: \overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}} \setminus {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
\begin{equation*} i: \overline {{\textrm{Gr}}^\lambda _{n,{\mathbb H}}} \setminus {\textrm{Gr}}^\lambda _{n,{\mathbb H}} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
denote the complementary closed embedding. We have a triangle
 \begin{equation*} j_{ !}j_{}^! \mathcal {F} \rightarrow \mathcal {F} \rightarrow i_{ *}i^*_{} \mathcal {F} \rightarrow . \end{equation*}
\begin{equation*} j_{ !}j_{}^! \mathcal {F} \rightarrow \mathcal {F} \rightarrow i_{ *}i^*_{} \mathcal {F} \rightarrow . \end{equation*}
Applying 
 $\Omega$
 yields
$\Omega$
 yields
 \begin{equation*} \Omega (j_!j^! \mathcal {F}) \rightarrow \Omega (\mathcal {F}) \rightarrow \Omega (i_*i^* \mathcal {F}) \rightarrow . \end{equation*}
\begin{equation*} \Omega (j_!j^! \mathcal {F}) \rightarrow \Omega (\mathcal {F}) \rightarrow \Omega (i_*i^* \mathcal {F}) \rightarrow . \end{equation*}
By induction, we may assume that 
 $\Omega (i_*i^* \mathcal {F})$
 is even. On the other hand,
$\Omega (i_*i^* \mathcal {F})$
 is even. On the other hand, 
 $\mathcal {F}$
 is
$\mathcal {F}$
 is 
 $\mathfrak {L}^+G_{n,{\mathbb H}}$
 equivariant and
$\mathfrak {L}^+G_{n,{\mathbb H}}$
 equivariant and 
 $*$
-even. Hence,
$*$
-even. Hence, 
 $j^!\mathcal {F}$
 is a direct sum of complexes of the form
$j^!\mathcal {F}$
 is a direct sum of complexes of the form 
 $\underline {{\mathbb C}}[m]$
 for
$\underline {{\mathbb C}}[m]$
 for 
 $m$
 even. Therefore,
$m$
 even. Therefore, 
 $\Omega (j_!j^! \mathcal {F})$
 is
$\Omega (j_!j^! \mathcal {F})$
 is 
 $*$
-even,
$*$
-even, 
 $\Omega (\mathcal {F})$
 is
$\Omega (\mathcal {F})$
 is 
 $*$
-even, and
$*$
-even, and 
 $\Omega$
 is
$\Omega$
 is 
 $*$
-parity preserving.
$*$
-parity preserving.
 For the converse, observe that each 
 $j_{\lambda !}\underline {{\mathbb C}}$
 is
$j_{\lambda !}\underline {{\mathbb C}}$
 is 
 $*$
-even, as its only non-trivial stalk is isomorphic to
$*$
-even, as its only non-trivial stalk is isomorphic to 
 $H^*_{\mathfrak {L}^+ G_{n,{\mathbb H}}}(\mathrm {pt})$
. Hence
$H^*_{\mathfrak {L}^+ G_{n,{\mathbb H}}}(\mathrm {pt})$
. Hence 
 $\Omega (j_{\lambda !}\underline {{\mathbb C}})$
 is
$\Omega (j_{\lambda !}\underline {{\mathbb C}})$
 is 
 $*$
-even.
$*$
-even.
Corollary 4.18.
- 
(i) The functor  $\Gamma _{\mathfrak {L}^+ G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}, -)$
 is $\Gamma _{\mathfrak {L}^+ G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}, -)$
 is $*$
-parity preserving. $*$
-parity preserving.
- 
(ii) Let  $\mathcal {G} \in D^b_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be $\mathcal {G} \in D^b_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be $!$
-even. Then the functor $!$
-even. Then the functor $\mathrm {Hom}_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(-, \mathcal {G})$
 is $\mathrm {Hom}_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(-, \mathcal {G})$
 is $*$
-parity preserving. $*$
-parity preserving.
Proof.
- 
(i) We claim that the cohomology  $H^*_{\mathfrak {L}^+ G_{n,{\mathbb H}}}(j_{!\lambda }\underline {{\mathbb C}})$
 is even. By the long exact sequence, it suffices to show that each $H^*_{\mathfrak {L}^+ G_{n,{\mathbb H}}}(j_{!\lambda }\underline {{\mathbb C}})$
 is even. By the long exact sequence, it suffices to show that each $H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {{\mathbb C}}\big)$
 is even. The non-equivariant cohomology $H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {{\mathbb C}}\big)$
 is even. The non-equivariant cohomology $H^*\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {{\mathbb C}}\big)$
 is even because $H^*\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {{\mathbb C}}\big)$
 is even because $\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }$
 is paved by quaternionic affine spaces. Therefore $\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }$
 is paved by quaternionic affine spaces. Therefore $j_{\lambda !}\underline {{\mathbb C}}$
 is $j_{\lambda !}\underline {{\mathbb C}}$
 is $\mathfrak {L}^+G_{n,{\mathbb H}}$
-equivariantly formal, and so $\mathfrak {L}^+G_{n,{\mathbb H}}$
-equivariantly formal, and so $H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {\mathbb C}\big) \simeq H^*\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {\mathbb C}\big) \otimes H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathrm {pt}, {\mathbb C})$
 is even. By Lemma 4.17, we conclude that $H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {\mathbb C}\big) \simeq H^*\big(\overline {{\textrm{Gr}}_{n,{\mathbb H}}^\lambda }, {\mathbb C}\big) \otimes H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathrm {pt}, {\mathbb C})$
 is even. By Lemma 4.17, we conclude that $\Gamma _{\mathfrak {L}^+ G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}, -)$
 is $\Gamma _{\mathfrak {L}^+ G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}, -)$
 is $*$
-parity preserving. $*$
-parity preserving.
- 
(ii) By Lemma 4.17, we must check that for each \begin{equation*} \mathrm {Ext}^i_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(j_{\lambda !} \underline {{\mathbb C}}, \mathcal {G}) \simeq 0 \end{equation*} \begin{equation*} \mathrm {Ext}^i_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(j_{\lambda !} \underline {{\mathbb C}}, \mathcal {G}) \simeq 0 \end{equation*} $\lambda \in X_{A}^+$
 and $\lambda \in X_{A}^+$
 and $i$
 odd. By adjunction, $i$
 odd. By adjunction, \begin{equation*} \mathrm {Ext}^i_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(j_{\lambda !} \underline {{\mathbb C}}, \mathcal {G}) \simeq \mathrm {Ext}^i_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(\underline {{\mathbb C}}, j_{\lambda }^!\mathcal {G}\big) \simeq H_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}^i\big(j_{\lambda }^! \mathcal {G}\big). \end{equation*} \begin{equation*} \mathrm {Ext}^i_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(j_{\lambda !} \underline {{\mathbb C}}, \mathcal {G}) \simeq \mathrm {Ext}^i_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(\underline {{\mathbb C}}, j_{\lambda }^!\mathcal {G}\big) \simeq H_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}^i\big(j_{\lambda }^! \mathcal {G}\big). \end{equation*}
 The claim follows from the assumption that 
 $\mathcal {G}$
 is
$\mathcal {G}$
 is 
 $!$
-even.
$!$
-even.
Lemma 4.19. 
Let 
 $\mathcal {F}\in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be
$\mathcal {F}\in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
 be 
 $*$
-even. Suppose that
$*$
-even. Suppose that 
 $X \subseteq {\textrm{Gr}}_{n,{\mathbb H}}$
 is a closed finite union of
$X \subseteq {\textrm{Gr}}_{n,{\mathbb H}}$
 is a closed finite union of 
 $\mathfrak {L}^+{G_{n,{\mathbb H}}}$
-orbits such that
$\mathfrak {L}^+{G_{n,{\mathbb H}}}$
-orbits such that 
 $X$
 contains the support of
$X$
 contains the support of 
 $\mathcal {F}$
. Let
$\mathcal {F}$
. Let 
 $Z \subseteq X$
 denote a
$Z \subseteq X$
 denote a 
 $\mathfrak {L}^+{G_{n,{\mathbb H}}}$
-stable closed subset, and
$\mathfrak {L}^+{G_{n,{\mathbb H}}}$
-stable closed subset, and 
 $U = X \setminus Z$
 its open complement. Let
$U = X \setminus Z$
 its open complement. Let 
 $j: U \hookrightarrow X$
 and
$j: U \hookrightarrow X$
 and 
 $i: Z \hookrightarrow X$
 denote the natural inclusions. We have a triangle
$i: Z \hookrightarrow X$
 denote the natural inclusions. We have a triangle
 \begin{equation*} j_!j^! \mathcal {F} \rightarrow \mathcal {F} \rightarrow i_*i^* \mathcal {F} \rightarrow . \end{equation*}
\begin{equation*} j_!j^! \mathcal {F} \rightarrow \mathcal {F} \rightarrow i_*i^* \mathcal {F} \rightarrow . \end{equation*}
Now let 
 $\Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}) \rightarrow D^b(\mathrm {Vect}_{\mathbb C})$
 denote a
$\Omega : D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}}) \rightarrow D^b(\mathrm {Vect}_{\mathbb C})$
 denote a 
 $*$
-parity preserving functor. Then the triangle
$*$
-parity preserving functor. Then the triangle
 \begin{equation*} \Omega (j_!j^! \mathcal {F}) \rightarrow \Omega (\mathcal {F}) \rightarrow \Omega (i_*i^* \mathcal {F}) \rightarrow \end{equation*}
\begin{equation*} \Omega (j_!j^! \mathcal {F}) \rightarrow \Omega (\mathcal {F}) \rightarrow \Omega (i_*i^* \mathcal {F}) \rightarrow \end{equation*}
is split.
Proof.
 We must show that the boundary map 
 $\delta : \Omega (i_*i^* \mathcal {F}) \rightarrow \Omega (j_!j^! \mathcal {F})[1]$
 is zero. Observe that the functors
$\delta : \Omega (i_*i^* \mathcal {F}) \rightarrow \Omega (j_!j^! \mathcal {F})[1]$
 is zero. Observe that the functors 
 $j_!j^!$
 and
$j_!j^!$
 and 
 $i_*i^*$
 take
$i_*i^*$
 take 
 $*$
-even sheaves to
$*$
-even sheaves to 
 $*$
-even sheaves. Since
$*$
-even sheaves. Since 
 $\Omega$
 is
$\Omega$
 is 
 $*$
-parity preserving, the complexes
$*$
-parity preserving, the complexes 
 $\Omega (i_*i^* \mathcal {F})$
 and
$\Omega (i_*i^* \mathcal {F})$
 and 
 $\Omega (j_!j^! \mathcal {F})$
 are even. Hence,
$\Omega (j_!j^! \mathcal {F})$
 are even. Hence, 
 $\Omega (j_!j^! \mathcal {F})[1]$
 is odd. Therefore,
$\Omega (j_!j^! \mathcal {F})[1]$
 is odd. Therefore, 
 $\delta$
 induces the zero map in cohomology and so is zero.
$\delta$
 induces the zero map in cohomology and so is zero.
Lemma 4.20. 
Let 
 $\mathcal {F}, \mathcal {G} \in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
. We make the following assumptions:
$\mathcal {F}, \mathcal {G} \in D^b_{\mathfrak {L}^+G_{n, {\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
. We make the following assumptions:
- 
(1)  $\mathcal {F}$
 is $\mathcal {F}$
 is $*$
-even; $*$
-even;
- 
(2)  $\mathcal {G}$
 is $\mathcal {G}$
 is $!$
-even; $!$
-even;
- 
(3) for any  $\mu \in \Lambda _S^+$
, the map
is surjective; $\mu \in \Lambda _S^+$
, the map
is surjective; \begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*\mathcal {F}\big) \end{equation*} \begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*\mathcal {F}\big) \end{equation*}
- 
(4) for any  $\mu \in \Lambda _S^+$
, the map
is injective. $\mu \in \Lambda _S^+$
, the map
is injective. \begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_{\mu !}j_{\mu }^!\mathcal {G}\big) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {G}) \end{equation*} \begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_{\mu !}j_{\mu }^!\mathcal {G}\big) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {G}) \end{equation*}
Then the natural map
 \begin{equation*} \mathrm {Ext}^*_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\mathcal {F}, \mathcal {G}) \rightarrow \mathrm {Hom}^*_{H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}), H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {G})\big) \end{equation*}
\begin{equation*} \mathrm {Ext}^*_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\mathcal {F}, \mathcal {G}) \rightarrow \mathrm {Hom}^*_{H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}), H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {G})\big) \end{equation*}
is an isomorphism of graded modules.
Proof.
 Let 
 $Z$
 denote the union of the supports of
$Z$
 denote the union of the supports of 
 $\mathcal {F}$
 and
$\mathcal {F}$
 and 
 $\mathcal {G}$
. We proceed by induction on
$\mathcal {G}$
. We proceed by induction on 
 $Z$
. Certainly there exists an orbit
$Z$
. Certainly there exists an orbit 
 ${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 open in
${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 open in 
 $Z$
. Let
$Z$
. Let 
 $Y = Z \setminus {\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
, and let
$Y = Z \setminus {\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
, and let 
 $i_\lambda : Y \hookrightarrow Z$
 denote the inclusion. We claim that the pair
$i_\lambda : Y \hookrightarrow Z$
 denote the inclusion. We claim that the pair 
 $i_\lambda ^*\mathcal {F}$
 and
$i_\lambda ^*\mathcal {F}$
 and 
 $i_{\lambda }^!\mathcal {G}$
 satisfies the assumptions (1)–(4).
$i_{\lambda }^!\mathcal {G}$
 satisfies the assumptions (1)–(4).
 That 
 $i_{\lambda }^*\mathcal {F}$
 is
$i_{\lambda }^*\mathcal {F}$
 is 
 $*$
-even and that
$*$
-even and that 
 $i_{\lambda }^! \mathcal {G}$
 is
$i_{\lambda }^! \mathcal {G}$
 is 
 $!$
-even is evident. We verify (3) for
$!$
-even is evident. We verify (3) for 
 $i_{\lambda }^*{\mathcal {F}}$
. If
$i_{\lambda }^*{\mathcal {F}}$
. If 
 ${\textrm{Gr}}^\mu _{n,{\mathbb H}}$
 does not lie in the support of
${\textrm{Gr}}^\mu _{n,{\mathbb H}}$
 does not lie in the support of 
 $i_{\lambda }^*\mathcal {F}$
, then there is nothing to prove. So we may assume that
$i_{\lambda }^*\mathcal {F}$
, then there is nothing to prove. So we may assume that 
 ${\textrm{Gr}}^\mu _{n,{\mathbb H}} \subseteq Y$
. Therefore, we have the composition of maps
${\textrm{Gr}}^\mu _{n,{\mathbb H}} \subseteq Y$
. Therefore, we have the composition of maps
 \begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(i_{\lambda }^*\mathcal {F} \big) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*i_{\lambda }^*\mathcal {F}\big) \simeq H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*\mathcal {F}\big). \end{equation*}
\begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}(\mathcal {F}) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(i_{\lambda }^*\mathcal {F} \big) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*i_{\lambda }^*\mathcal {F}\big) \simeq H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*\mathcal {F}\big). \end{equation*}
The composite is surjective by assumption. Hence, the map
 \begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(i_{\lambda }^*\mathcal {F} \big) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*i_{\lambda }^*\mathcal {F}\big) \end{equation*}
\begin{equation*} H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(i_{\lambda }^*\mathcal {F} \big) \rightarrow H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}\big(j_\mu ^*i_{\lambda }^*\mathcal {F}\big) \end{equation*}
is surjective, as needed. The proof that 
 $i_{\lambda }^!\mathcal {G}$
 satisfies (4) is similar.
$i_{\lambda }^!\mathcal {G}$
 satisfies (4) is similar.
 Now we proceed with the induction. To avoid overly cumbersone notation, we will suppress the subscripts on 
 $\mathrm {Ext}^*_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}$
 and
$\mathrm {Ext}^*_{D^b_{\mathfrak {L}^{+}G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}$
 and 
 $\mathrm {Hom}^*_{H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}$
. Similarly, we will make use of the isomorphism
$\mathrm {Hom}^*_{H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}$
. Similarly, we will make use of the isomorphism 
 $H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}} \simeq H^*_G$
 to simplify notation. Lastly, we let
$H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}} \simeq H^*_G$
 to simplify notation. Lastly, we let 
 $H = H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
.
$H = H^*_{\mathfrak {L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})$
.
Consider the triangle
 \begin{equation} j_{\lambda !} j^{!}_{\lambda } \mathcal {F} \rightarrow \mathcal {F} \rightarrow i_{\lambda *} i_{\lambda }^*\mathcal {F} \rightarrow . \end{equation}
\begin{equation} j_{\lambda !} j^{!}_{\lambda } \mathcal {F} \rightarrow \mathcal {F} \rightarrow i_{\lambda *} i_{\lambda }^*\mathcal {F} \rightarrow . \end{equation}
By Lemma 4.19, Corollary 4.18, and adjunction, we have an exact sequence
 \begin{equation} 0 \rightarrow \mathrm {Ext}^*\big(i_{\lambda }^* \mathcal {F}, i_{\lambda }^!\mathcal {G}\big) \rightarrow \mathrm {Ext}^*(\mathcal {F}, \mathcal {G}) \rightarrow \mathrm {Ext}^*\big(j_{\lambda }^! \mathcal {F}, j_{\lambda }^!\mathcal {G}\big) \rightarrow 0. \end{equation}
\begin{equation} 0 \rightarrow \mathrm {Ext}^*\big(i_{\lambda }^* \mathcal {F}, i_{\lambda }^!\mathcal {G}\big) \rightarrow \mathrm {Ext}^*(\mathcal {F}, \mathcal {G}) \rightarrow \mathrm {Ext}^*\big(j_{\lambda }^! \mathcal {F}, j_{\lambda }^!\mathcal {G}\big) \rightarrow 0. \end{equation}
We can also apply the functor 
 $H^*$
 to (4.15) to obtain the exact sequence
$H^*$
 to (4.15) to obtain the exact sequence
 \begin{equation*} 0 \rightarrow H_G^*\big(j_{\lambda !}j_{\lambda }^! \mathcal {F}\big) \rightarrow H_G^*(\mathcal {F}) \rightarrow H_G^*\big(i_{\lambda }^*\mathcal {F}\big) \rightarrow 0. \end{equation*}
\begin{equation*} 0 \rightarrow H_G^*\big(j_{\lambda !}j_{\lambda }^! \mathcal {F}\big) \rightarrow H_G^*(\mathcal {F}) \rightarrow H_G^*\big(i_{\lambda }^*\mathcal {F}\big) \rightarrow 0. \end{equation*}
Similarly, we have the exact sequence
 \begin{equation} 0 \rightarrow H_G^*\big(i_{\lambda }^! \mathcal {G}\big) \rightarrow H_G^*(\mathcal {G}) \rightarrow H_G^*\big(j_{\lambda }^*\mathcal {G}\big) \rightarrow 0. \end{equation}
\begin{equation} 0 \rightarrow H_G^*\big(i_{\lambda }^! \mathcal {G}\big) \rightarrow H_G^*(\mathcal {G}) \rightarrow H_G^*\big(j_{\lambda }^*\mathcal {G}\big) \rightarrow 0. \end{equation}
These two exact sequences induce a sequence
 \begin{equation*} 0 \rightarrow \mathrm {Hom}^*\big(H^*_G\big(i_{\lambda }^*\mathcal {F}\big), H^*_G\big(i_{\lambda }^!\mathcal {G}\big)\big) \rightarrow \mathrm {Hom}^*\big(H^*_G(\mathcal {F}), H^*_G(\mathcal {G})\big) \rightarrow \mathrm {Hom}^*\big(H^*_G(\mathcal {F}), H^*_G\big(j_{\lambda }^*\mathcal {G}\big)\big). \end{equation*}
\begin{equation*} 0 \rightarrow \mathrm {Hom}^*\big(H^*_G\big(i_{\lambda }^*\mathcal {F}\big), H^*_G\big(i_{\lambda }^!\mathcal {G}\big)\big) \rightarrow \mathrm {Hom}^*\big(H^*_G(\mathcal {F}), H^*_G(\mathcal {G})\big) \rightarrow \mathrm {Hom}^*\big(H^*_G(\mathcal {F}), H^*_G\big(j_{\lambda }^*\mathcal {G}\big)\big). \end{equation*}
The second map is clearly an injection. We claim that the sequence is also exact in the middle. It suffices to show that any 
 $H$
-linear map
$H$
-linear map
 \begin{equation*} \alpha : H_G^*(\mathcal {F}) \rightarrow H_G^*\big(i_{\lambda }^! \mathcal {G}\big) \end{equation*}
\begin{equation*} \alpha : H_G^*(\mathcal {F}) \rightarrow H_G^*\big(i_{\lambda }^! \mathcal {G}\big) \end{equation*}
factors through 
 $H_G^*\big(i_{\lambda }^*\mathcal {F}\big)$
. Consider the compactly supported cohomology
$H_G^*\big(i_{\lambda }^*\mathcal {F}\big)$
. Consider the compactly supported cohomology 
 $H^*_{G,c}\big({\textrm{Gr}}^{\lambda }_{n,{\mathbb H}}\big)$
. Let
$H^*_{G,c}\big({\textrm{Gr}}^{\lambda }_{n,{\mathbb H}}\big)$
. Let 
 $\mathfrak {c}_\lambda \in H_{G,c}^{\langle 2\rho _{2n}, \lambda \rangle }\big({\textrm{Gr}}^{\lambda }_{n,{\mathbb H}}\big)$
 denote a lift of a generator to
$\mathfrak {c}_\lambda \in H_{G,c}^{\langle 2\rho _{2n}, \lambda \rangle }\big({\textrm{Gr}}^{\lambda }_{n,{\mathbb H}}\big)$
 denote a lift of a generator to 
 $G$
-equivariant cohomology; it maps to an element
$G$
-equivariant cohomology; it maps to an element 
 $\mathfrak {c}_{\lambda } \in H$
. Since
$\mathfrak {c}_{\lambda } \in H$
. Since 
 $\mathfrak {c}_{\lambda }$
 maps to
$\mathfrak {c}_{\lambda }$
 maps to 
 $0 \in H^*_G(Y)$
, it acts trivially on
$0 \in H^*_G(Y)$
, it acts trivially on 
 $H^*(i_{\lambda }^!\mathcal {G})$
. Since
$H^*(i_{\lambda }^!\mathcal {G})$
. Since 
 $\alpha$
 is
$\alpha$
 is 
 $H$
-linear, it suffices to show that
$H$
-linear, it suffices to show that 
 $H^*\big(j_{\lambda !}j_\lambda ^! \mathcal {F}\big)$
 lies in the image of
$H^*\big(j_{\lambda !}j_\lambda ^! \mathcal {F}\big)$
 lies in the image of
 \begin{equation*} \mathfrak {c}_{\lambda }: H_G^*(\mathcal {F}) \rightarrow H^*_G(\mathcal {F}[-\langle 2 \rho _{2n}, \lambda \rangle ]). \end{equation*}
\begin{equation*} \mathfrak {c}_{\lambda }: H_G^*(\mathcal {F}) \rightarrow H^*_G(\mathcal {F}[-\langle 2 \rho _{2n}, \lambda \rangle ]). \end{equation*}
To do so, we note that by Poincaré duality for the smooth manifold 
 ${\textrm{Gr}}_{n,{\mathbb H}}^\lambda$
, cupping with
${\textrm{Gr}}_{n,{\mathbb H}}^\lambda$
, cupping with 
 $\mathfrak {c}_{\lambda }$
 induces an isomorphism
$\mathfrak {c}_{\lambda }$
 induces an isomorphism
 \begin{equation*} \mathfrak {c}_{\lambda }: H_G^*\big(j_{\lambda !}j^!_{\lambda }\mathcal {F}\big) \rightarrow H_G^*\big(j_{\lambda *} j_{\lambda }^*\mathcal {F}[-\langle 2 \rho _{2n}, \lambda \rangle ]\big). \end{equation*}
\begin{equation*} \mathfrak {c}_{\lambda }: H_G^*\big(j_{\lambda !}j^!_{\lambda }\mathcal {F}\big) \rightarrow H_G^*\big(j_{\lambda *} j_{\lambda }^*\mathcal {F}[-\langle 2 \rho _{2n}, \lambda \rangle ]\big). \end{equation*}
Thus we obtain the following commutative diagram.

As the bottom arrow is an isomorphism, it suffices to show that the right vertical map is surjective. But this is assumed in (3).
 Next, we observe that any 
 $H$
-linear map
$H$
-linear map 
 $\beta : H_G^*(\mathcal {F}) \rightarrow H_G^*\big(j^*_{\lambda }\mathcal {G}\big)$
 factors through
$\beta : H_G^*(\mathcal {F}) \rightarrow H_G^*\big(j^*_{\lambda }\mathcal {G}\big)$
 factors through 
 $H_G^*\big(j^*_{\lambda } \mathcal {F}\big)$
. The proof is similar to that of the previous step, using (4) in place of (3), and is therefore omitted.
$H_G^*\big(j^*_{\lambda } \mathcal {F}\big)$
. The proof is similar to that of the previous step, using (4) in place of (3), and is therefore omitted.
Hence, we have an exact sequence
 \begin{equation*} 0 \rightarrow \mathrm {Hom}^*\big(H_G^*\big(i_{\lambda }^*\mathcal {F}\big), H_G^*\big(i_{\lambda }^!\mathcal {G}\big)\big) \rightarrow \mathrm {Hom}^*\big(H_G^*(\mathcal {F}), H_G^*(\mathcal {G})\big) \rightarrow \mathrm {Hom}^*\big(H_G^*\big(j_{\lambda }^*\mathcal {F}\big), H_G^*\big(j_{\lambda }^*\mathcal {G}\big)\big). \end{equation*}
\begin{equation*} 0 \rightarrow \mathrm {Hom}^*\big(H_G^*\big(i_{\lambda }^*\mathcal {F}\big), H_G^*\big(i_{\lambda }^!\mathcal {G}\big)\big) \rightarrow \mathrm {Hom}^*\big(H_G^*(\mathcal {F}), H_G^*(\mathcal {G})\big) \rightarrow \mathrm {Hom}^*\big(H_G^*\big(j_{\lambda }^*\mathcal {F}\big), H_G^*\big(j_{\lambda }^*\mathcal {G}\big)\big). \end{equation*}
To conclude, we observe that this exact sequence fits into the following commutative diagram with (4.16).

 The map 
 $f$
 is an isomorphism by induction, and
$f$
 is an isomorphism by induction, and 
 $h$
 is easily seen to be an isomorphism. Hence
$h$
 is easily seen to be an isomorphism. Hence 
 $g$
 is an isomorphism, as claimed.
$g$
 is an isomorphism, as claimed.
Proof of Proposition 4.16.
 It suffices to verify that 
 $\mathcal {F}$
 and
$\mathcal {F}$
 and 
 $\mathcal {G}$
 satisfy the hypotheses of Lemma 4.20. The properties (1) and (2) are assumed. We will show that (3) holds; the proof of (4) is similar. We must show that for each
$\mathcal {G}$
 satisfy the hypotheses of Lemma 4.20. The properties (1) and (2) are assumed. We will show that (3) holds; the proof of (4) is similar. We must show that for each 
 $\lambda \in \Lambda _S^+$
, the map
$\lambda \in \Lambda _S^+$
, the map
 \begin{equation*} H^*_G(\mathcal {F}) \rightarrow H^*_{G}\big(j_\lambda ^*\mathcal {F}\big) \end{equation*}
\begin{equation*} H^*_G(\mathcal {F}) \rightarrow H^*_{G}\big(j_\lambda ^*\mathcal {F}\big) \end{equation*}
is surjective. It identifies with
 \begin{equation*} H^*_{T_c}(\mathcal {F})^{W_n} \rightarrow H^*_{T_c}\big(j_\lambda ^*\mathcal {F}\big)^{W_n}. \end{equation*}
\begin{equation*} H^*_{T_c}(\mathcal {F})^{W_n} \rightarrow H^*_{T_c}\big(j_\lambda ^*\mathcal {F}\big)^{W_n}. \end{equation*}
Since the coefficient field has characteristic zero, the functor of 
 $W_n$
-invariants is exact, and it suffices to show that the restriction map
$W_n$
-invariants is exact, and it suffices to show that the restriction map 
 $H^*_{T_c}(\mathcal {F}) \rightarrow H^*_{T_c}\big(j_\lambda ^*\mathcal {F}\big)$
 is surjective. We let
$H^*_{T_c}(\mathcal {F}) \rightarrow H^*_{T_c}\big(j_\lambda ^*\mathcal {F}\big)$
 is surjective. We let
 \begin{equation*} k_{\lambda }: \big({\textrm{Gr}}^\lambda _{n,{\mathbb H}}\big)^{T_c} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
\begin{equation*} k_{\lambda }: \big({\textrm{Gr}}^\lambda _{n,{\mathbb H}}\big)^{T_c} \hookrightarrow {\textrm{Gr}}_{n,{\mathbb H}} \end{equation*}
denote the inclusion of the 
 $T_c$
-fixed locus in
$T_c$
-fixed locus in 
 ${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
. Now, consider the composition
${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
. Now, consider the composition
 \begin{equation} H^*_{T_c}(\mathcal {F}) \rightarrow H^*_{T_c}\big(j_\lambda ^*\mathcal {F}\big) \rightarrow H^*_{T_c}\big(k_\lambda ^*\mathcal {F}\big).\end{equation}
\begin{equation} H^*_{T_c}(\mathcal {F}) \rightarrow H^*_{T_c}\big(j_\lambda ^*\mathcal {F}\big) \rightarrow H^*_{T_c}\big(k_\lambda ^*\mathcal {F}\big).\end{equation}
Observe that 
 $j_\lambda ^*\mathcal {F}$
 is a constant sheaf and that
$j_\lambda ^*\mathcal {F}$
 is a constant sheaf and that 
 ${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 is an equivariantly formal
${\textrm{Gr}}^\lambda _{n,{\mathbb H}}$
 is an equivariantly formal 
 $T_c$
-manifold. Hence, the second map above is injective by the localization theorem. The surjectivity of the first map is then reduced to that of the composition. Now,
$T_c$
-manifold. Hence, the second map above is injective by the localization theorem. The surjectivity of the first map is then reduced to that of the composition. Now, 
 $k_\lambda$
 is a closed inclusion, so
$k_\lambda$
 is a closed inclusion, so 
 $k_{\lambda *}k_{\lambda }^{*} \mathcal {F}$
 is
$k_{\lambda *}k_{\lambda }^{*} \mathcal {F}$
 is 
 $*$
-even. The proof of Lemma 4.19, applied in the
$*$
-even. The proof of Lemma 4.19, applied in the 
 $T_c$
-equivariant derived category, shows that the restriction map
$T_c$
-equivariant derived category, shows that the restriction map 
 $H^*_{T_c}(\mathcal {F}) \rightarrow H^*_{T_c}\big(k_\lambda ^*\mathcal {F}\big)$
 is indeed surjective.
$H^*_{T_c}(\mathcal {F}) \rightarrow H^*_{T_c}\big(k_\lambda ^*\mathcal {F}\big)$
 is indeed surjective.
4.8 Ext algebras
 The tensor equivalence 
 $\textrm{Rep}(G_n)\simeq \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
 gives rise to a monoidal action of
$\textrm{Rep}(G_n)\simeq \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
 gives rise to a monoidal action of 
 $\textrm{Rep}(G_n)$
 on
$\textrm{Rep}(G_n)$
 on 
 $D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
. We compute the de-equivariantized extension algebra
$D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
. We compute the de-equivariantized extension algebra
 \begin{equation*}\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n)).\end{equation*}
\begin{equation*}\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n)).\end{equation*}
Strictly speaking, 
 $\mathcal {O}(G_n)$
 is not an object of
$\mathcal {O}(G_n)$
 is not an object of 
 $\textrm{Rep}(G_n)$
, but by the Peter–Weyl theorem, it is an increasing direct sum of objects. We understand the above extension algebra to mean the increasing direct sum of extensions.
$\textrm{Rep}(G_n)$
, but by the Peter–Weyl theorem, it is an increasing direct sum of objects. We understand the above extension algebra to mean the increasing direct sum of extensions.
Proposition 4.21. 
There is a 
 $G_n$
-equivariant isomorphism of graded algebras
$G_n$
-equivariant isomorphism of graded algebras
 \begin{equation*}\textrm{Ext}^*_{D^b_{\mathfrak{L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))\simeq \mathcal {O}({\mathfrak g}_{n}[4])\simeq \textrm{Sym}({\mathfrak g}_n[-4]).\end{equation*}
\begin{equation*}\textrm{Ext}^*_{D^b_{\mathfrak{L}^+G_{n,{\mathbb H}}}({\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))\simeq \mathcal {O}({\mathfrak g}_{n}[4])\simeq \textrm{Sym}({\mathfrak g}_n[-4]).\end{equation*}
Proof.
 By Proposition 4.15, taking equivariant cohomology induces a 
 $G_n$
-equivariant isomorphism of graded algebras
$G_n$
-equivariant isomorphism of graded algebras
 \begin{align*}\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))& \simeq \big({\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}(H^*_{T_c}(\textrm{pt}),H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_0\star \mathcal {O}(G_n))^{\mathrm {W}}\\ &\simeq \big({\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}(\mathcal {O}({\mathfrak t}),\mathcal {O}(G_n\times {\mathfrak t}))\big)^{\mathrm {W}}\nonumber\\ & \simeq \big(\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\big)^{\mathrm {W}},\end{align*}
\begin{align*}\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))& \simeq \big({\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}(H^*_{T_c}(\textrm{pt}),H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_0\star \mathcal {O}(G_n))^{\mathrm {W}}\\ &\simeq \big({\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}(\mathcal {O}({\mathfrak t}),\mathcal {O}(G_n\times {\mathfrak t}))\big)^{\mathrm {W}}\nonumber\\ & \simeq \big(\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\big)^{\mathrm {W}},\end{align*}
where 
 $\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\subset \mathcal {O}(G_n\times {\mathfrak t})$
 is the subspace consisting of functions that are invariant (relative over
$\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\subset \mathcal {O}(G_n\times {\mathfrak t})$
 is the subspace consisting of functions that are invariant (relative over 
 $\mathfrak t$
) with respect to the left action of the group scheme
$\mathfrak t$
) with respect to the left action of the group scheme 
 $\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t})^{e^T_X}$
 on
$\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t})^{e^T_X}$
 on 
 $G_n\times {\mathfrak t}$
. Since
$G_n\times {\mathfrak t}$
. Since 
 $\mathcal {O}\big({\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)=\mathcal {O}\big({\mathfrak g}_n^{\textrm{}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)$
 and the map
$\mathcal {O}\big({\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)=\mathcal {O}\big({\mathfrak g}_n^{\textrm{}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)$
 and the map
 \begin{equation} \nu :G_n\times {\mathfrak t}\to {\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t},\quad (g,t)\to \big(\textrm{Ad}_{g^{-1}}e^T_X(t),t\big) \end{equation}
\begin{equation} \nu :G_n\times {\mathfrak t}\to {\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t},\quad (g,t)\to \big(\textrm{Ad}_{g^{-1}}e^T_X(t),t\big) \end{equation}
realizes 
 $G_n\times {\mathfrak t}$
 as a
$G_n\times {\mathfrak t}$
 as a 
 $(G_n\times {\mathfrak t})^{e^T_X}$
-torsor over
$(G_n\times {\mathfrak t})^{e^T_X}$
-torsor over 
 ${\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}$
, we obtain an isomorphism of algebras
${\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}$
, we obtain an isomorphism of algebras
 \begin{align*}\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))& \simeq \big(\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\big)^{\mathrm {W}}\\ &\simeq \mathcal {O}\big({\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)^{\mathrm {W}} \simeq \mathcal {O}\big({\mathfrak g}_n^{\textrm{}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)^{\mathrm {W}}\simeq \mathcal {O}({\mathfrak g}_n). \end{align*}
\begin{align*}\textrm{Ext}^*_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))& \simeq \big(\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\big)^{\mathrm {W}}\\ &\simeq \mathcal {O}\big({\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)^{\mathrm {W}} \simeq \mathcal {O}\big({\mathfrak g}_n^{\textrm{}}\times _{{\mathfrak c}_n}{\mathfrak t}\big)^{\mathrm {W}}\simeq \mathcal {O}({\mathfrak g}_n). \end{align*}
It remains to check that the isomorphism above is compatible with the desired gradings. By [Reference NadlerNad05, Theorem 8.5.1], for any 
 $\lambda \in \Lambda _S$
 and
$\lambda \in \Lambda _S$
 and 
 $\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
, the compactly supported cohomology
$\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
, the compactly supported cohomology 
 $H^*_c\big(S_{n,{\mathbb H}}^\lambda ,\mathcal {F}\big)$
 along the real semi-infinite orbit
$H^*_c\big(S_{n,{\mathbb H}}^\lambda ,\mathcal {F}\big)$
 along the real semi-infinite orbit 
 $S^\lambda _{n,{\mathbb H}}$
 is non-zero only in degree
$S^\lambda _{n,{\mathbb H}}$
 is non-zero only in degree 
 $\langle \lambda ,\rho _{2n}\rangle$
. Note that
$\langle \lambda ,\rho _{2n}\rangle$
. Note that 
 $\langle \lambda ,\rho _{2n}\rangle =4\langle \lambda ,\rho _{n}\rangle$
, where in the second paring we regard
$\langle \lambda ,\rho _{2n}\rangle =4\langle \lambda ,\rho _{n}\rangle$
, where in the second paring we regard 
 $\lambda$
 as an element in
$\lambda$
 as an element in 
 $\Lambda _n$
. Thus the grading on
$\Lambda _n$
. Thus the grading on 
 $H^*({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_\lambda )$
 corresponds, under the geometric Satake equivalence, to the grading on
$H^*({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_\lambda )$
 corresponds, under the geometric Satake equivalence, to the grading on 
 $V_\lambda$
 given by co-character
$V_\lambda$
 given by co-character 
 $4\rho _n$
, and it follows that the grading on
$4\rho _n$
, and it follows that the grading on 
 $H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_0\star \mathcal {O}(G_n))\simeq \mathcal {O}(G_n\times {\mathfrak t})$
 is induced by the
$H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\textrm{IC}_0\star \mathcal {O}(G_n))\simeq \mathcal {O}(G_n\times {\mathfrak t})$
 is induced by the 
 ${\mathbb G}_m$
-action on
${\mathbb G}_m$
-action on 
 $G_n\times {\mathfrak t}$
 given by
$G_n\times {\mathfrak t}$
 given by 
 $x(g,t)=(4\rho _n(x)g,x^{-2}t)$
 (note that the generators of
$x(g,t)=(4\rho _n(x)g,x^{-2}t)$
 (note that the generators of 
 $\mathcal {O}({\mathfrak t})$
 are in degree
$\mathcal {O}({\mathfrak t})$
 are in degree 
 $2$
). We claim that the map
$2$
). We claim that the map 
 $\nu$
 in (4.19) is
$\nu$
 in (4.19) is 
 ${\mathbb G}_m$
-equivariant with respect to the above action on
${\mathbb G}_m$
-equivariant with respect to the above action on 
 $G_n\times {\mathfrak t}$
 and the action on
$G_n\times {\mathfrak t}$
 and the action on 
 ${\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}$
 given by
${\mathfrak g}_n^{\textrm{reg}}\times _{{\mathfrak c}_n}{\mathfrak t}$
 given by 
 $x(v,t)=(x^{-4}v,x^{-2}t)$
. Indeed, we have
$x(v,t)=(x^{-4}v,x^{-2}t)$
. Indeed, we have
 \begin{align*}\textrm{Ad}_{4\rho _n(x^{-1})}e_X^T(x^{-2}t) & = \textrm{Ad}_{4\rho _n(x^{-1})}\begin{pmatrix} x^{-4}t_1^2 & \quad 1 & \quad \\ \vdots & \quad x^{-4}t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad x^{-4}t_n^2 \end{pmatrix}\nonumber\\& =\begin{pmatrix} x^{-4}t_1^2 & \quad x^{-4} & \quad \\ \vdots & \quad x^{-4}t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad x^{-4} \\ 0& \quad 0& \quad \ldots & \quad x^{-4}t_n^2 \end{pmatrix}\\[5pt] & = x^{-4}e_X^T(t),\end{align*}
\begin{align*}\textrm{Ad}_{4\rho _n(x^{-1})}e_X^T(x^{-2}t) & = \textrm{Ad}_{4\rho _n(x^{-1})}\begin{pmatrix} x^{-4}t_1^2 & \quad 1 & \quad \\ \vdots & \quad x^{-4}t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad 1 \\ 0& \quad 0& \quad \ldots & \quad x^{-4}t_n^2 \end{pmatrix}\nonumber\\& =\begin{pmatrix} x^{-4}t_1^2 & \quad x^{-4} & \quad \\ \vdots & \quad x^{-4}t_2^2 & \quad \ddots & \quad \\ \vdots & \quad & \quad \ddots & \quad x^{-4} \\ 0& \quad 0& \quad \ldots & \quad x^{-4}t_n^2 \end{pmatrix}\\[5pt] & = x^{-4}e_X^T(t),\end{align*}
and hence
 \begin{align*}\nu (x(g,t)) &= \nu (4\rho _n(x)g,x^{-2}t)=\big(\textrm{Ad}_{g^{-1}}\textrm{Ad}_{4\rho _n(x)^{-1}}e_X^T(x^{-2}t),x^{-2}t\big)\\ & =\big(x^{-4}\textrm{Ad}_{g^{-1}}e_X^T(t),x^{-2}t\big)=x\big(\textrm{Ad}_{g^{-1}}e_X^T(t),t\big)=x\nu (g,t).\end{align*}
\begin{align*}\nu (x(g,t)) &= \nu (4\rho _n(x)g,x^{-2}t)=\big(\textrm{Ad}_{g^{-1}}\textrm{Ad}_{4\rho _n(x)^{-1}}e_X^T(x^{-2}t),x^{-2}t\big)\\ & =\big(x^{-4}\textrm{Ad}_{g^{-1}}e_X^T(t),x^{-2}t\big)=x\big(\textrm{Ad}_{g^{-1}}e_X^T(t),t\big)=x\nu (g,t).\end{align*}
Thus the pullback along the map 
 $\nu$
 induces an isomorphism of graded algebras
$\nu$
 induces an isomorphism of graded algebras
 \begin{equation*}\big(\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\big)^{\mathrm {W}}\simeq \mathcal {O}\big({\mathfrak g}_n^{\textrm{reg}}[4]\times _{{\mathfrak g}_n[4]//G_n}{\mathfrak t}[2]\big)^{\mathrm {W}}\simeq \mathcal {O}({\mathfrak g}_n[4]).\end{equation*}
\begin{equation*}\big(\mathcal {O}(G_n\times {\mathfrak t})^{\textrm{Spec}(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}}))}\big)^{\mathrm {W}}\simeq \mathcal {O}\big({\mathfrak g}_n^{\textrm{reg}}[4]\times _{{\mathfrak g}_n[4]//G_n}{\mathfrak t}[2]\big)^{\mathrm {W}}\simeq \mathcal {O}({\mathfrak g}_n[4]).\end{equation*}
This finishes the proof of the theorem.
4.9 IC-stalks, 
 $q$
-analogue of weight multiplicity, and Kostka–Foulkes polynomials
$q$
-analogue of weight multiplicity, and Kostka–Foulkes polynomials
In this section we shall prove Theorem1.9(2). We will follow Ginzburg’s approach [Reference GinzburgGin95] (see also [Reference ZhuZhu15, § 5]) using techniques of equivariant cohomology.
 Let 
 $V\in \textrm{Rep}(G_n)$
. Consider the Brylinski–Kostant filtration
$V\in \textrm{Rep}(G_n)$
. Consider the Brylinski–Kostant filtration 
 $F_iV:=\textrm{ker}\;\mathbf e_n^{i+1}, i\geq 0$
 on
$F_iV:=\textrm{ker}\;\mathbf e_n^{i+1}, i\geq 0$
 on 
 $V$
 associated to the regular nilpotent element
$V$
 associated to the regular nilpotent element 
 $\mathbf e_n$
. For any
$\mathbf e_n$
. For any 
 $\mu \in \Lambda _n$
, we denote by
$\mu \in \Lambda _n$
, we denote by 
 $V(\mu )$
 the
$V(\mu )$
 the 
 $\mu$
-weight space of
$\mu$
-weight space of 
 $V$
 (since
$V$
 (since 
 $G_n$
 is self-dual, we can view
$G_n$
 is self-dual, we can view 
 $\Lambda _n$
 as the weight lattice of
$\Lambda _n$
 as the weight lattice of 
 $G_n$
). The filtration
$G_n$
). The filtration 
 $F_iV$
 induces a filtration on the weight space:
$F_iV$
 induces a filtration on the weight space:
 \begin{equation*}F_iV(\mu )=F_iV\cap V(\mu ).\end{equation*}
\begin{equation*}F_iV(\mu )=F_iV\cap V(\mu ).\end{equation*}
Let
 \begin{equation*}P_\mu (V,q)=\sum _i\dim (F_iV(\mu )/F_{i-1}V(\mu ))q^i\end{equation*}
\begin{equation*}P_\mu (V,q)=\sum _i\dim (F_iV(\mu )/F_{i-1}V(\mu ))q^i\end{equation*}
be the 
 $q$
-analogue of the weight multiplicity polynomial.
$q$
-analogue of the weight multiplicity polynomial.
 From now on we will identify 
 $\Lambda _n$
 with the set
$\Lambda _n$
 with the set 
 $\Lambda _S$
 of real coweights and denote by
$\Lambda _S$
 of real coweights and denote by 
 $s_\mu :\{\mu \}\to \Lambda _n\simeq \Lambda _S\subset {\textrm{Gr}}_{n,{\mathbb H}}$
 the inclusion map.
$s_\mu :\{\mu \}\to \Lambda _n\simeq \Lambda _S\subset {\textrm{Gr}}_{n,{\mathbb H}}$
 the inclusion map.
Theorem 4.22. 
Let 
 $\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
 and let
$\mathcal {F}\in \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
 and let 
 $V=H^*({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 be the corresponding representation of
$V=H^*({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 be the corresponding representation of 
 $G_n$
. We have
$G_n$
. We have
 \begin{equation*}P_{\mu }(V,q)=\sum _i\dim H^{-4i-4\langle \mu ,\rho _n\rangle }\big(s_\mu ^*\mathcal {F}\big)q^i= \sum _i\dim H^{4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)q^i.\end{equation*}
\begin{equation*}P_{\mu }(V,q)=\sum _i\dim H^{-4i-4\langle \mu ,\rho _n\rangle }\big(s_\mu ^*\mathcal {F}\big)q^i= \sum _i\dim H^{4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)q^i.\end{equation*}
 The theorem above implies Theorem1.9(2) in the case of the quaternionic affine Grassmannian. Indeed, if 
 $\mu ,\lambda \in \Lambda _n^+$
 and
$\mu ,\lambda \in \Lambda _n^+$
 and 
 $V=V_\lambda$
 is the irreducible representation of highest weight
$V=V_\lambda$
 is the irreducible representation of highest weight 
 $\lambda$
, then it is known that
$\lambda$
, then it is known that 
 $P_\mu (V_\lambda ,q)=K_{\lambda ,\mu }(q)$
 is the Kostka–Foulkes polynomial associated to
$P_\mu (V_\lambda ,q)=K_{\lambda ,\mu }(q)$
 is the Kostka–Foulkes polynomial associated to 
 $\lambda \mbox{ and }\mu$
 (see e.g. [Reference BrylinskiBry89]). Thus, for any
$\lambda \mbox{ and }\mu$
 (see e.g. [Reference BrylinskiBry89]). Thus, for any 
 $x\in {\textrm{Gr}}_{n,{\mathbb H}}^\mu$
, we have
$x\in {\textrm{Gr}}_{n,{\mathbb H}}^\mu$
, we have
 \begin{equation*}K_{\lambda ,\mu }(q)=\sum _i\dim H^{-4i-4\langle \mu ,\rho _n\rangle }\big(s_\mu ^*\mathcal {F}\big)q^i=\sum _i\dim \mathscr H^{\,\,-4i-4\langle \mu ,\rho _n\rangle }_x(\textrm{IC}_\lambda )q^i,\end{equation*}
\begin{equation*}K_{\lambda ,\mu }(q)=\sum _i\dim H^{-4i-4\langle \mu ,\rho _n\rangle }\big(s_\mu ^*\mathcal {F}\big)q^i=\sum _i\dim \mathscr H^{\,\,-4i-4\langle \mu ,\rho _n\rangle }_x(\textrm{IC}_\lambda )q^i,\end{equation*}
and it follows that
 \begin{equation*}q^{\langle \lambda -\mu ,\rho _n\rangle }K_{\lambda ,\mu }(q^{-1})= \sum _i\dim \mathscr H^{\,\,-4i-4\langle \mu ,\rho _n\rangle }_x(\textrm{IC}_\lambda )q^{-i- \langle \mu ,\rho _n\rangle +\langle \lambda ,\rho _n\rangle }= \sum _i\dim \mathscr H^{\,\,4i-4\langle \lambda ,\rho _n\rangle }_x(\textrm{IC}_\lambda )q^{i}.\end{equation*}
\begin{equation*}q^{\langle \lambda -\mu ,\rho _n\rangle }K_{\lambda ,\mu }(q^{-1})= \sum _i\dim \mathscr H^{\,\,-4i-4\langle \mu ,\rho _n\rangle }_x(\textrm{IC}_\lambda )q^{-i- \langle \mu ,\rho _n\rangle +\langle \lambda ,\rho _n\rangle }= \sum _i\dim \mathscr H^{\,\,4i-4\langle \lambda ,\rho _n\rangle }_x(\textrm{IC}_\lambda )q^{i}.\end{equation*}
 The case of 
 $\mathfrak{L}K$
-orbits on
$\mathfrak{L}K$
-orbits on 
 ${\textrm{Gr}}_{2n}$
 follows from the fact [Reference Chen and NadlerCN24, Theorem 7.5] that there is a stratified
${\textrm{Gr}}_{2n}$
 follows from the fact [Reference Chen and NadlerCN24, Theorem 7.5] that there is a stratified 
 $K_c$
-equivariant homeomorphism between
$K_c$
-equivariant homeomorphism between 
 $\Omega K_c\backslash {\textrm{Gr}}_{2n}$
 and
$\Omega K_c\backslash {\textrm{Gr}}_{2n}$
 and 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
 (where
${\textrm{Gr}}_{n,{\mathbb H}}$
 (where 
 $\Omega K_c$
 is the based loop group of
$\Omega K_c$
 is the based loop group of 
 $K_c$
) with stratifications given by images of
$K_c$
) with stratifications given by images of 
 $\mathfrak{L} K$
-orbits on
$\mathfrak{L} K$
-orbits on 
 ${\textrm{Gr}}_{2n}$
 in the quotient
${\textrm{Gr}}_{2n}$
 in the quotient 
 $\Omega K_c\backslash {\textrm{Gr}}_{2n}$
 and the
$\Omega K_c\backslash {\textrm{Gr}}_{2n}$
 and the 
 $\mathfrak{L}^+G_{n,{\mathbb H}}$
-orbits on
$\mathfrak{L}^+G_{n,{\mathbb H}}$
-orbits on 
 ${\textrm{Gr}}_{n,{\mathbb H}}$
.
${\textrm{Gr}}_{n,{\mathbb H}}$
.
4.9.1 Proof of Theorem 4.22
 We follow closely the presentation in [Reference ZhuZhu15, § 5]. For any 
 $t\in {\mathfrak t}$
 we denote by
$t\in {\mathfrak t}$
 we denote by 
 $\kappa (t)$
 the residue field of
$\kappa (t)$
 the residue field of 
 $t$
. The specialized cohomology
$t$
. The specialized cohomology
 \begin{equation*}H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F}):=H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\otimes _{R_T}\kappa (t)\end{equation*}
\begin{equation*}H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F}):=H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\otimes _{R_T}\kappa (t)\end{equation*}
carries a canonical filtration
 \begin{equation*}H^{\leq i}_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F}):=\textrm{Im}\bigg(\sum _{j\leq i} H^j_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\to H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\bigg).\end{equation*}
\begin{equation*}H^{\leq i}_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F}):=\textrm{Im}\bigg(\sum _{j\leq i} H^j_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\to H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\bigg).\end{equation*}
Let us identify 
 $H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\simeq (H^*({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\otimes R_T)\otimes _{R_T}\kappa (t)\simeq V$
 via the canonical splitting in (4.11). As explained in the proof of Proposition 4.21, the cohomological grading on
$H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\simeq (H^*({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\otimes R_T)\otimes _{R_T}\kappa (t)\simeq V$
 via the canonical splitting in (4.11). As explained in the proof of Proposition 4.21, the cohomological grading on 
 $H^*({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 corresponds to the grading on the representation
$H^*({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 corresponds to the grading on the representation 
 $V$
 given by the eigenvalues of
$V$
 given by the eigenvalues of 
 $4\rho _n$
. It follows that the filtration
$4\rho _n$
. It follows that the filtration 
 $H^{\leq i}_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 corresponds to the increasing filtration on
$H^{\leq i}_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 corresponds to the increasing filtration on 
 $V$
 given by the eigenvalues of
$V$
 given by the eigenvalues of 
 $4\rho _n$
 (see e.g. [Reference GinzburgGin95, Theorem 5.2.1]).
$4\rho _n$
 (see e.g. [Reference GinzburgGin95, Theorem 5.2.1]).
 Fix a generic element 
 $t=(t_1,\ldots ,t_n)\in {\mathfrak t}$
 away from the root hyperplanes. The localization theorem implies that there is an isomorphism
$t=(t_1,\ldots ,t_n)\in {\mathfrak t}$
 away from the root hyperplanes. The localization theorem implies that there is an isomorphism
 \begin{equation} \bigoplus _{\mu \in \Lambda _n} H_t\big(s_\mu ^!\mathcal {F}\big)\simeq H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F}). \end{equation}
\begin{equation} \bigoplus _{\mu \in \Lambda _n} H_t\big(s_\mu ^!\mathcal {F}\big)\simeq H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F}). \end{equation}
Recall the description of the equivariant homology 
 $\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t})^{e_X^T}$
 in Lemma 4.9. The fiber of the group scheme
$\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t})^{e_X^T}$
 in Lemma 4.9. The fiber of the group scheme 
 $(G_n\times {\mathfrak t})^{e_X^T}$
 over
$(G_n\times {\mathfrak t})^{e_X^T}$
 over 
 $t$
 is the centralizer subgroup
$t$
 is the centralizer subgroup 
 $ (G_n)^{e_X^T(t)}\subset G_n$
 of the element
$ (G_n)^{e_X^T(t)}\subset G_n$
 of the element 
 $e_X^T(t)\in {\mathfrak g}_n$
 in (3.14) . Let
$e_X^T(t)\in {\mathfrak g}_n$
 in (3.14) . Let 
 $B_n\to T_n$
 be the natural projection. It is shown in [Reference O’BrienO’Br23, Lemma 4] (generalizing [Reference Yun and ZhuYZ11, Remark 3.4]) that the composition
$B_n\to T_n$
 be the natural projection. It is shown in [Reference O’BrienO’Br23, Lemma 4] (generalizing [Reference Yun and ZhuYZ11, Remark 3.4]) that the composition
 \begin{equation*}\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t})^{e_X^T}\subset B_n\times {\mathfrak t}\to T_n\times {\mathfrak t}\end{equation*}
\begin{equation*}\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\simeq (G_n\times {\mathfrak t})^{e_X^T}\subset B_n\times {\mathfrak t}\to T_n\times {\mathfrak t}\end{equation*}
can be identified with the map coming from equivariant localization,
 \begin{equation*}\textrm{Spec}(\mathrm{Loc}_*):\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\to \textrm{Spec}(R_T[\Lambda _n])=T_n\times {\mathfrak t}. \end{equation*}
\begin{equation*}\textrm{Spec}(\mathrm{Loc}_*):\textrm{Spec}\big(H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})\big)\to \textrm{Spec}(R_T[\Lambda _n])=T_n\times {\mathfrak t}. \end{equation*}
Over 
 $t$
, this is an isomorphism and therefore we obtain a canonical isomorphism
$t$
, this is an isomorphism and therefore we obtain a canonical isomorphism 
 $(G_n)^{e_X^T(t)}\simeq T_n\times \{t\}\simeq T_n$
. In addition, the action of
$(G_n)^{e_X^T(t)}\simeq T_n\times \{t\}\simeq T_n$
. In addition, the action of 
 $(G_n)^{e_X^T(t)}$
 on
$(G_n)^{e_X^T(t)}$
 on 
 $H_t\big(S^\mu _{n,{\mathbb H}} ,\mathcal {F}\big)$
 via
$H_t\big(S^\mu _{n,{\mathbb H}} ,\mathcal {F}\big)$
 via 
 $(G_n)^{e_X^T(t)}\simeq T_n$
 is identified with the natural action of
$(G_n)^{e_X^T(t)}\simeq T_n$
 is identified with the natural action of 
 $(G_n)^{e_X^T(t)}$
 on
$(G_n)^{e_X^T(t)}$
 on 
 $H_t\big(S^\mu _{n,{\mathbb H}} ,\mathcal {F}\big)\simeq H_t\big(s_\mu ^!\mathcal {F}\big)$
. Thus we conclude that the decomposition in (4.20) corresponds to the weight decomposition under
$H_t\big(S^\mu _{n,{\mathbb H}} ,\mathcal {F}\big)\simeq H_t\big(s_\mu ^!\mathcal {F}\big)$
. Thus we conclude that the decomposition in (4.20) corresponds to the weight decomposition under 
 $(G_n)^{e_X^T(t)}$
 (generalizing [Reference ZhuZhu15, Proposition 5.2]).
$(G_n)^{e_X^T(t)}$
 (generalizing [Reference ZhuZhu15, Proposition 5.2]).
Lemma 4.23. 
The decomposition in (
4.20
) corresponds, under the canonical isomorphism 
 $H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\simeq V$
, to the weight decomposition
$H_t({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\simeq V$
, to the weight decomposition 
 $V=\oplus _{\mu \in \Lambda _n} V(\mu _t)$
 with respect to the action of the maximal torus
$V=\oplus _{\mu \in \Lambda _n} V(\mu _t)$
 with respect to the action of the maximal torus 
 $(G_n)^{e_X^T(t)}$
. Here
$(G_n)^{e_X^T(t)}$
. Here 
 $V(\mu _t)$
 is the weight space associated to the character
$V(\mu _t)$
 is the weight space associated to the character 
 $\mu _t:(G_n)^{e_X^T(t)}\simeq T_n\stackrel {\mu }\to {\mathbb C}^\times$
.
$\mu _t:(G_n)^{e_X^T(t)}\simeq T_n\stackrel {\mu }\to {\mathbb C}^\times$
.
 Choose 
 $t\in {\mathfrak t}$
 such that
$t\in {\mathfrak t}$
 such that 
 $e_X^T(t)=\mathbf e_n+2\rho _n$
. Let
$e_X^T(t)=\mathbf e_n+2\rho _n$
. Let 
 $u$
 be the unique element in
$u$
 be the unique element in 
 $N_n$
 such that
$N_n$
 such that 
 $\textrm{Ad}_u(\mathbf e_n+2\rho _n)=2\rho _n$
.
$\textrm{Ad}_u(\mathbf e_n+2\rho _n)=2\rho _n$
.
Lemma 4.24. We have
 \begin{equation*}H_t^{\leq 4i+2m}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m}H_t\big(s^!_\mu \mathcal {F}\big)=F_iV\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m} V(\mu _t).\end{equation*}
\begin{equation*}H_t^{\leq 4i+2m}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m}H_t\big(s^!_\mu \mathcal {F}\big)=F_iV\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m} V(\mu _t).\end{equation*}
Proof.
 Let 
 $V=\bigoplus V^1(i)$
 and
$V=\bigoplus V^1(i)$
 and 
 $V=\bigoplus V^2(i)$
 be two gradings on
$V=\bigoplus V^2(i)$
 be two gradings on 
 $V$
 given by the cocharacters
$V$
 given by the cocharacters 
 $2\rho _n$
 and
$2\rho _n$
 and 
 $\textrm{Ad}_{u^{-1}}2\rho _n$
, respectively. Let
$\textrm{Ad}_{u^{-1}}2\rho _n$
, respectively. Let 
 $F^1_iV$
 and
$F^1_iV$
 and 
 $F^2_iV$
 be the two filtrations on
$F^2_iV$
 be the two filtrations on 
 $V$
 given by
$V$
 given by 
 $F^1_iV=\oplus _{j\leq i} V^1(j)$
 and
$F^1_iV=\oplus _{j\leq i} V^1(j)$
 and 
 $F^2_iV=\textrm{ker}\big(\mathbf e_n^{i+1}\big)$
. We have
$F^2_iV=\textrm{ker}\big(\mathbf e_n^{i+1}\big)$
. We have
 \begin{equation*}F_iV\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m} V(\mu _t)=F^2_iV\cap V^2(m)\end{equation*}
\begin{equation*}F_iV\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m} V(\mu _t)=F^2_iV\cap V^2(m)\end{equation*}
and
 \begin{equation*}H_t^{\leq 4i+2m}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m}H_t\big(s^!_\mu \mathcal {F}\big)=F^1_{2i+m}(V)\cap V^2(m),\end{equation*}
\begin{equation*}H_t^{\leq 4i+2m}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap \bigoplus _{\mu \in \Lambda _n,2\langle \mu ,\rho _n\rangle =m}H_t\big(s^!_\mu \mathcal {F}\big)=F^1_{2i+m}(V)\cap V^2(m),\end{equation*}
and the desired claim follows from [Reference ZhuZhu15, Lemma 5.5].
 Note that we have shown in (4.18) that the natural map 
 $H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\to H^*_{T_c}\big(s_\mu ^*\mathcal {F}\big)$
 is a surjective map of free
$H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\to H^*_{T_c}\big(s_\mu ^*\mathcal {F}\big)$
 is a surjective map of free 
 $R_T$
-modules, and it implies that the dual map
$R_T$
-modules, and it implies that the dual map 
 $ H^*_{T_c}\big(s_\mu ^!\mathcal {F}\big)\to H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 is a splitting injective map of free
$ H^*_{T_c}\big(s_\mu ^!\mathcal {F}\big)\to H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})$
 is a splitting injective map of free 
 $R_T$
-modules. Thus we have
$R_T$
-modules. Thus we have
 \begin{equation*}H^{\leq i}_{t}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap H_t\big(s_\mu ^!\mathcal {F}\big)=H^{\leq i}_t\big(s_\mu ^!\mathcal {F}\big).\end{equation*}
\begin{equation*}H^{\leq i}_{t}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap H_t\big(s_\mu ^!\mathcal {F}\big)=H^{\leq i}_t\big(s_\mu ^!\mathcal {F}\big).\end{equation*}
On the other hand, the element 
 $u\in N_n$
 above maps
$u\in N_n$
 above maps 
 $V(\mu _t)$
 to
$V(\mu _t)$
 to 
 $V(\mu )$
 and preserves the filtration
$V(\mu )$
 and preserves the filtration 
 $F_iV$
, and hence
$F_iV$
, and hence 
 $\textrm{dim}(F_iV(\mu )/F_{i-1}V(\mu ))=\dim (F_iV(\mu _t)/F_{i-1}V(\mu _t))$
. Now the lemma above implies
$\textrm{dim}(F_iV(\mu )/F_{i-1}V(\mu ))=\dim (F_iV(\mu _t)/F_{i-1}V(\mu _t))$
. Now the lemma above implies
 \begin{align*}P_\mu (V,q)&=\sum _i\dim (F_iV(\mu )/F_{i-1}V(\mu ))q^i= \sum _i\dim (F_iV(\mu _t)/F_{i-1}V(\mu _t))q^i\\ &=\sum _i \textrm{dim}\big(H^{\leq 4i+4\langle \mu ,\rho _n\rangle }_{t}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap H_t\big(s_\mu ^!\mathcal {F}\big)\big/H^{\leq 4(i-1)+4\langle \mu ,\rho _n\rangle }_{t}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap H_t\big(s_\mu ^!\mathcal {F}\big)\big)q^i\\ &=\sum _i \textrm{dim}\big(H_t^{\leq 4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big/H_t^{\leq 4(i-1)+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big)q^i. \end{align*}
\begin{align*}P_\mu (V,q)&=\sum _i\dim (F_iV(\mu )/F_{i-1}V(\mu ))q^i= \sum _i\dim (F_iV(\mu _t)/F_{i-1}V(\mu _t))q^i\\ &=\sum _i \textrm{dim}\big(H^{\leq 4i+4\langle \mu ,\rho _n\rangle }_{t}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap H_t\big(s_\mu ^!\mathcal {F}\big)\big/H^{\leq 4(i-1)+4\langle \mu ,\rho _n\rangle }_{t}({\textrm{Gr}}_{n,{\mathbb H}},\mathcal {F})\cap H_t\big(s_\mu ^!\mathcal {F}\big)\big)q^i\\ &=\sum _i \textrm{dim}\big(H_t^{\leq 4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big/H_t^{\leq 4(i-1)+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big)q^i. \end{align*}
To conclude the proof, we observe that under the canonical isomorphism 
 $H_t\big(s_\mu ^!\mathcal {F}\big)\simeq H^*\big(s_\mu ^!\mathcal {F}\big)$
, the canonical filtration on the left-hand side corresponds to the cohomological degree filtration on the right-hand side, and hence we obtain
$H_t\big(s_\mu ^!\mathcal {F}\big)\simeq H^*\big(s_\mu ^!\mathcal {F}\big)$
, the canonical filtration on the left-hand side corresponds to the cohomological degree filtration on the right-hand side, and hence we obtain
 \begin{equation*}P_\mu (V,q)= \sum _i \textrm{dim}\big(H_t^{\leq 4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big/H_t^{\leq 4(i-1)+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big)q^i=\sum _i\textrm{dim } H^{4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)q^i.\end{equation*}
\begin{equation*}P_\mu (V,q)= \sum _i \textrm{dim}\big(H_t^{\leq 4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big/H_t^{\leq 4(i-1)+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)\big)q^i=\sum _i\textrm{dim } H^{4i+4\langle \mu ,\rho _n\rangle }\big(s_\mu ^!\mathcal {F}\big)q^i.\end{equation*}
5. Main results
5.1 Formality
The goal of this section is to show that the dg-algebra
 \begin{equation*}\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))\end{equation*}
\begin{equation*}\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))\end{equation*}
is formal.
 The proof is based on the following key proposition. The existence of the left adjoint of the nearby cycles functor in Lemma 4.8 gives rise to a map between 
 $K_c$
-equivariant cohomology
$K_c$
-equivariant cohomology
 \begin{equation} H^*_{K_c}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Ext}^*\big({\mathbb C}_{{\textrm{Gr}}_{n,{\mathbb H}}},{\mathbb C}_{{\textrm{Gr}}_{n,{\mathbb H}}}\big)\stackrel {^L\mathrm R'}\longrightarrow \textrm{Ext}^*\big({\mathbb C}_{{\textrm{Gr}}_{2n}},{\mathbb C}_{{\textrm{Gr}}_{2n}}\big)\simeq H^*_{K_c}({\textrm{Gr}}_{2n}). \end{equation}
\begin{equation} H^*_{K_c}({\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Ext}^*\big({\mathbb C}_{{\textrm{Gr}}_{n,{\mathbb H}}},{\mathbb C}_{{\textrm{Gr}}_{n,{\mathbb H}}}\big)\stackrel {^L\mathrm R'}\longrightarrow \textrm{Ext}^*\big({\mathbb C}_{{\textrm{Gr}}_{2n}},{\mathbb C}_{{\textrm{Gr}}_{2n}}\big)\simeq H^*_{K_c}({\textrm{Gr}}_{2n}). \end{equation}
By taking the graded dual (see § 4.6.2), we get a map between equivariant homology
 \begin{equation} H_*^{K_c}({\textrm{Gr}}_{2n})\to H_*^{K_c}({\textrm{Gr}}_{n,{\mathbb H}}). \end{equation}
\begin{equation} H_*^{K_c}({\textrm{Gr}}_{2n})\to H_*^{K_c}({\textrm{Gr}}_{n,{\mathbb H}}). \end{equation}
Proposition 5.1. We have the commutative diagram

 
where the bottom arrow 
 $J_n\to J_{2n}|_{{\mathfrak c}_n}$
 is the morphism introduced in (
3.6
).
$J_n\to J_{2n}|_{{\mathfrak c}_n}$
 is the morphism introduced in (
3.6
).
Proof.
 We shall verify the statement for 
 $T_c$
-equivariant homology, that is, we have a commutative diagram
$T_c$
-equivariant homology, that is, we have a commutative diagram

 where the bottom arrow is the map (3.16). All the maps above are compatible with the natural 
 $\mathrm {W}$
-actions, and upon taking
$\mathrm {W}$
-actions, and upon taking 
 $\mathrm {W}$
-invariants we get the desired claim.
$\mathrm {W}$
-invariants we get the desired claim.
 Let 
 $V_{\omega _1}$
 and
$V_{\omega _1}$
 and 
 $V_{\omega _1'}$
 be the standard representations of
$V_{\omega _1'}$
 be the standard representations of 
 $G_n$
 and
$G_n$
 and 
 $G_{2n}$
, respectively. Recall the isomorphisms
$G_{2n}$
, respectively. Recall the isomorphisms
 \begin{equation}{\kern3pt} H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)\simeq H^*_{}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)\otimes R_T\simeq V_{\omega '_1}\otimes R_T,{\kern-3pt} \end{equation}
\begin{equation}{\kern3pt} H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)\simeq H^*_{}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)\otimes R_T\simeq V_{\omega '_1}\otimes R_T,{\kern-3pt} \end{equation}
 \begin{equation} {\kern-.5pt}H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\simeq H^*_{}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\otimes R_T\simeq V_{\omega '_1}\otimes R_T{\kern.5pt}\end{equation}
\begin{equation} {\kern-.5pt}H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\simeq H^*_{}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\otimes R_T\simeq V_{\omega '_1}\otimes R_T{\kern.5pt}\end{equation}
induced by the complex and real MV-filtrations. Together with the canonical isomorphism
 \begin{equation} H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\simeq H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big), \end{equation}
\begin{equation} H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\simeq H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big), \end{equation}
we get an automorphism
 \begin{equation} V_{\omega '_1}\otimes R_T\simeq H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\simeq H^*_{T_c}({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1})\simeq V_{\omega '_1}\otimes R_T\end{equation}
\begin{equation} V_{\omega '_1}\otimes R_T\simeq H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)\simeq H^*_{T_c}({\textrm{Gr}}_{2n},\textrm{IC}_{\omega _1})\simeq V_{\omega '_1}\otimes R_T\end{equation}
and hence an element
 \begin{equation} \Phi '\in {\textrm{GL}}\big(V_{\omega '_1}\big)\otimes R_T\simeq G_{2n}\otimes R_T.\end{equation}
\begin{equation} \Phi '\in {\textrm{GL}}\big(V_{\omega '_1}\big)\otimes R_T\simeq G_{2n}\otimes R_T.\end{equation}
Note that the isomorphisms (5.3) and (5.4) map the standard basis 
 $\{e_1\otimes 1,\ldots ,e_{2n}\otimes 1\}$
 of
$\{e_1\otimes 1,\ldots ,e_{2n}\otimes 1\}$
 of 
 $V_{\omega '_1}\otimes R_{T}$
 to the basis
$V_{\omega '_1}\otimes R_{T}$
 to the basis
 \begin{equation*}\{b_1,\ldots ,b_{2n}\}=\{[\mathbb P^{0}],\ldots ,[\mathbb P^{2n-1}]\}\end{equation*}
\begin{equation*}\{b_1,\ldots ,b_{2n}\}=\{[\mathbb P^{0}],\ldots ,[\mathbb P^{2n-1}]\}\end{equation*}
of 
 $H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)\simeq H^{*}_{T_c}(\mathbb P^{2n-1})$
 (up to a constant degree shift) and the basis
$H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)\simeq H^{*}_{T_c}(\mathbb P^{2n-1})$
 (up to a constant degree shift) and the basis
 \begin{equation*}\{c_1,\ldots ,c_{2n}\}=\{[\mathbb {HP}^0][2],[\mathbb {HP}^0],[\mathbb {HP}^1][2],[\mathbb {HP}^1], \ldots ,[\mathbb {HP}^{n-1}][2],[\mathbb {HP}^{n-1}]\}\end{equation*}
\begin{equation*}\{c_1,\ldots ,c_{2n}\}=\{[\mathbb {HP}^0][2],[\mathbb {HP}^0],[\mathbb {HP}^1][2],[\mathbb {HP}^1], \ldots ,[\mathbb {HP}^{n-1}][2],[\mathbb {HP}^{n-1}]\}\end{equation*}
of 
 (up to a constant degree shift), respectively, and the element
 (up to a constant degree shift), respectively, and the element 
 $\Phi '$
 is the matrix for the linear map sending
$\Phi '$
 is the matrix for the linear map sending 
 $c_i\to b_i$
 in the basis
$c_i\to b_i$
 in the basis 
 $c_1,\ldots ,c_{2n}$
 (which is not the identity element).
$c_1,\ldots ,c_{2n}$
 (which is not the identity element).
By Lemma 4.9, there is a commutative diagram

 where the upper and lower middle arrows are given by the co-action of 
 $H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})$
 and
$H_*^{T_c}({\textrm{Gr}}_{n,{\mathbb H}})$
 and 
 $H_*^{T_c}({\textrm{Gr}}_{2n})$
 on
$H_*^{T_c}({\textrm{Gr}}_{2n})$
 on 
 $H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)$
 and
$H^*_{T_c}\big({\textrm{Gr}}_{n,{\mathbb H}},\mathrm R\big(\textrm{IC}_{\omega '_1}\big)\big)$
 and 
 $H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)$
, and the right vertical isomorphism is given by the conjugation action
$H^*_{T_c}\big({\textrm{Gr}}_{2n},\textrm{IC}_{\omega '_1}\big)$
, and the right vertical isomorphism is given by the conjugation action
 \begin{equation*}\textrm{Ad}_{\Phi '}:G_{2n}\times {\mathfrak t}\to G_{2n}\times {\mathfrak t}, \quad (g,t)\to \big(\textrm{Ad}_{\Phi '(t)}g,t\big).\end{equation*}
\begin{equation*}\textrm{Ad}_{\Phi '}:G_{2n}\times {\mathfrak t}\to G_{2n}\times {\mathfrak t}, \quad (g,t)\to \big(\textrm{Ad}_{\Phi '(t)}g,t\big).\end{equation*}
 Note that in the above diagram the lower composed map 
 $(G_{2n}\times {\mathfrak t})^{e^T}\to G_{2n}\times {\mathfrak t}$
 is the natural embedding and the upper composed map
$(G_{2n}\times {\mathfrak t})^{e^T}\to G_{2n}\times {\mathfrak t}$
 is the natural embedding and the upper composed map 
 $(G_n\times {\mathfrak t})^{e^T_X}\to G_{2n}\times {\mathfrak t}$
 is the restriction of the map
$(G_n\times {\mathfrak t})^{e^T_X}\to G_{2n}\times {\mathfrak t}$
 is the restriction of the map
 \begin{equation*}\textrm{Ad}_P\circ \delta :G_n\times {\mathfrak t}\to G_{2n}\times {\mathfrak t}\to G_{2n}\times {\mathfrak t},\quad (g,t)\to (P\delta (g)P^{-1},t)\end{equation*}
\begin{equation*}\textrm{Ad}_P\circ \delta :G_n\times {\mathfrak t}\to G_{2n}\times {\mathfrak t}\to G_{2n}\times {\mathfrak t},\quad (g,t)\to (P\delta (g)P^{-1},t)\end{equation*}
to 
 $(G_n\times {\mathfrak t})^{e^T_X}$
, where
$(G_n\times {\mathfrak t})^{e^T_X}$
, where 
 $P\in G_{2n}$
 is the permutation matrix which sends the the ordered basis
$P\in G_{2n}$
 is the permutation matrix which sends the the ordered basis 
 $\{e_1,e_3,\ldots ,e_{2n-1},e_2,e_4,\ldots ,e_{2n}\}$
 to the ordered basis
$\{e_1,e_3,\ldots ,e_{2n-1},e_2,e_4,\ldots ,e_{2n}\}$
 to the ordered basis 
 $\{e_1,\ldots ,e_{2n}\}$
 (see § 3.4.3).
$\{e_1,\ldots ,e_{2n}\}$
 (see § 3.4.3).
 Thus, in view of the description of the map 
 $(G_{n}\times {\mathfrak t})^{e^T_X}\to (G_{2n}\times {\mathfrak t})^{e^T}$
 in (3.16), we need to show that the element
$(G_{n}\times {\mathfrak t})^{e^T_X}\to (G_{2n}\times {\mathfrak t})^{e^T}$
 in (3.16), we need to show that the element
 \begin{equation*}\Phi :=\Phi '\circ P\in G_{2n}\otimes R_T\end{equation*}
\begin{equation*}\Phi :=\Phi '\circ P\in G_{2n}\otimes R_T\end{equation*}
satisfies
 \begin{equation} e^T=\Phi \big(\tau \circ e^T_X\big) \Phi ^{-1}\in {\mathfrak g}_{2n}\otimes R_T. \end{equation}
\begin{equation} e^T=\Phi \big(\tau \circ e^T_X\big) \Phi ^{-1}\in {\mathfrak g}_{2n}\otimes R_T. \end{equation}
To this end, we observe that, by Lemma 4.4, the elements 
 $\tau \circ e^T_X$
 and
$\tau \circ e^T_X$
 and 
 $e^T$
 in
$e^T$
 in 
 ${\mathfrak g}_{2n}\otimes R_T$
 are the matrices of the cup product map
${\mathfrak g}_{2n}\otimes R_T$
 are the matrices of the cup product map 
 $c_1^T(\mathcal {L})\cup (-):H^*_{T_c}\big({\textrm{Gr}}_{2n}, \textrm{IC}_{\omega '_1}\big)\to H^*_{T_c}\big({\textrm{Gr}}_{2n}, \textrm{IC}_{\omega '_1}\big)$
 in the bases
$c_1^T(\mathcal {L})\cup (-):H^*_{T_c}\big({\textrm{Gr}}_{2n}, \textrm{IC}_{\omega '_1}\big)\to H^*_{T_c}\big({\textrm{Gr}}_{2n}, \textrm{IC}_{\omega '_1}\big)$
 in the bases 
 $\{d_1,\ldots ,d_{2n}\}= \{[\mathbb {HP}^0][2],\ldots ,[\mathbb {HP}^{n-1}][2],[\mathbb {HP}^0],\ldots ,[\mathbb {HP}^{n-1}]\}$
 and
$\{d_1,\ldots ,d_{2n}\}= \{[\mathbb {HP}^0][2],\ldots ,[\mathbb {HP}^{n-1}][2],[\mathbb {HP}^0],\ldots ,[\mathbb {HP}^{n-1}]\}$
 and 
 $\{b_1,\ldots ,b_{2n}\}$
, respectively. On the other hand, the element
$\{b_1,\ldots ,b_{2n}\}$
, respectively. On the other hand, the element 
 $\Phi =\Phi '\circ P$
 is the matrix for the linear map sending
$\Phi =\Phi '\circ P$
 is the matrix for the linear map sending 
 $d_i\stackrel {}\to c_i\stackrel {}\to b_i$
 in the basis
$d_i\stackrel {}\to c_i\stackrel {}\to b_i$
 in the basis 
 $d_1,\ldots ,d_{2n}$
, and hence (5.9) holds. This completes the proof of the proposition.
$d_1,\ldots ,d_{2n}$
, and hence (5.9) holds. This completes the proof of the proposition.
Remark 5.2. The proof gives a canonical construction of the element 
 $\Phi$
 in (3.17).
$\Phi$
 in (3.17).
 In the statement and proof of the next results, we continue with the following convention: the corresponding objects 
 $\mathcal {O}(G_n)\leftrightarrow \textrm{IC}_{\mathcal {O}(G_n)}$
 and their variants may not lie in
$\mathcal {O}(G_n)\leftrightarrow \textrm{IC}_{\mathcal {O}(G_n)}$
 and their variants may not lie in 
 $\textrm{Rep}(G_n)\simeq \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
, but they are an increasing direct sum of objects, and we understand all calculations to mean the increasing direct sum of calculations.
$\textrm{Rep}(G_n)\simeq \textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
, but they are an increasing direct sum of objects, and we understand all calculations to mean the increasing direct sum of calculations.
Proposition 5.3. 
The dg-algebra 
 $\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
 is formal.
$\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
 is formal.
Proof. Consider the dg-algebras
 \begin{align*}A&=\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_{2n})), \\ B&=\textrm{RHom}_{{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n\times {\mathbb G}_m)).\end{align*}
\begin{align*}A&=\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_{2n})), \\ B&=\textrm{RHom}_{{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n\times {\mathbb G}_m)).\end{align*}
Here we consider 
 $\mathcal {O}(G_n\times {\mathbb G}_m)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{IC}_{\mathcal {O}(G_n)}[j]$
 via the monoidal functor
$\mathcal {O}(G_n\times {\mathbb G}_m)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{IC}_{\mathcal {O}(G_n)}[j]$
 via the monoidal functor 
 $\textrm{Rep}(G_n\times {\mathbb G}_m)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})[j]\subset D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
. Proposition 4.5(3) implies that the nearby cycle functor gives rise to a map of dg-algebras
$\textrm{Rep}(G_n\times {\mathbb G}_m)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})[j]\subset D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
. Proposition 4.5(3) implies that the nearby cycle functor gives rise to a map of dg-algebras
 \begin{align*}\phi :A & =\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_{2n}))\\&\stackrel {\mathrm R}\to \textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}\big(\textrm{IC}_0,\textrm{IC}_0\star \big(\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}\mathcal {O}(G_{2n})\big)\big)\\ &\to B=\textrm{RHom}_{{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n\times {\mathbb G}_m)),\end{align*}
\begin{align*}\phi :A & =\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{2n}\backslash {\textrm{Gr}}_{2n})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_{2n}))\\&\stackrel {\mathrm R}\to \textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}\big(\textrm{IC}_0,\textrm{IC}_0\star \big(\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}\mathcal {O}(G_{2n})\big)\big)\\ &\to B=\textrm{RHom}_{{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n\times {\mathbb G}_m)),\end{align*}
where the last arrow is induced by the quotient map 
 $\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}\mathcal {O}(G_{2n})\to \mathcal {O}(G_n\times {\mathbb G}_m)$
 (in the category of
$\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}\mathcal {O}(G_{2n})\to \mathcal {O}(G_n\times {\mathbb G}_m)$
 (in the category of 
 $\textrm{Rep}(G_n\times {\mathbb G}_m)$
). The right regular representations of
$\textrm{Rep}(G_n\times {\mathbb G}_m)$
). The right regular representations of 
 $G_{2n}$
 on
$G_{2n}$
 on 
 $G_n\times {\mathbb G}_m$
 induce natural
$G_n\times {\mathbb G}_m$
 induce natural 
 $G_{2n}$
- and
$G_{2n}$
- and 
 $G_n\times {\mathbb G}_m$
-actions on
$G_n\times {\mathbb G}_m$
-actions on 
 $A$
 and
$A$
 and 
 $B$
, and their restrictions to the subgroup
$B$
, and their restrictions to the subgroup 
 ${\mathbb G}_m\subset G_n\times {\mathbb G}_m\subset G_{2n}$
 give rise to
${\mathbb G}_m\subset G_n\times {\mathbb G}_m\subset G_{2n}$
 give rise to 
 ${\mathbb G}_m$
-weight decompositions
${\mathbb G}_m$
-weight decompositions 
 $A=\oplus_{j\in {\mathbb Z}} A_j$
 and
$A=\oplus_{j\in {\mathbb Z}} A_j$
 and 
 $B=\oplus_{j\in {\mathbb Z}} B_j$
. Note that the zero-weight spaces
$B=\oplus_{j\in {\mathbb Z}} B_j$
. Note that the zero-weight spaces 
 $A_0$
 and
$A_0$
 and 
 $B_0$
 are dg-subalgebras of
$B_0$
 are dg-subalgebras of 
 $A$
 and
$A$
 and 
 $B$
 and that
$B$
 and that 
 $B_0=\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
.
$B_0=\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))$
.
 According to [Reference Bezrukavnikov and FinkelbergBF08], the dg-algebra 
 $A$
 is formal; moreover, we have
$A$
 is formal; moreover, we have 
 $A\simeq H^*(A)\simeq \mathcal {O}({\mathfrak g}_{2n}[2])$
. Note that the map
$A\simeq H^*(A)\simeq \mathcal {O}({\mathfrak g}_{2n}[2])$
. Note that the map 
 $\phi :A\to B$
 above respects the
$\phi :A\to B$
 above respects the 
 ${\mathbb G}_m$
-action and hence restricts to a map
${\mathbb G}_m$
-action and hence restricts to a map 
 $\phi _0:A_0\to B_0$
 fitting into the following diagram.
$\phi _0:A_0\to B_0$
 fitting into the following diagram.

 We claim that the map 
 $H^*(\phi _0):H^*(A_0)\to H^*(B_0)$
 between cohomology is surjective. Since
$H^*(\phi _0):H^*(A_0)\to H^*(B_0)$
 between cohomology is surjective. Since 
 $A_0$
 is formal with generators in even degree and
$A_0$
 is formal with generators in even degree and 
 $H^*(B_0)\simeq H^*\big(\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))\big)\simeq \mathcal {O}({\mathfrak g}_n[4])$
, which is a polynomial ring with generators in even degree (see Lemma 4.21), Lemma 5.4 below implies that
$H^*(B_0)\simeq H^*\big(\textrm{RHom}_{D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))\big)\simeq \mathcal {O}({\mathfrak g}_n[4])$
, which is a polynomial ring with generators in even degree (see Lemma 4.21), Lemma 5.4 below implies that 
 $B_0$
 is formal. The proposition follows.
$B_0$
 is formal. The proposition follows.
 To proof the claim and show the surjectivity of 
 $H^*(\phi _0):H^*(A_0)\to H^*(B_0)$
, we can ignore the grading and view
$H^*(\phi _0):H^*(A_0)\to H^*(B_0)$
, we can ignore the grading and view 
 $H^*(\phi _0)$
 as maps between ungraded algebras. We have a commutative diagram
$H^*(\phi _0)$
 as maps between ungraded algebras. We have a commutative diagram

 where the horizontal isomorphisms are given by the functor of equivariant cohomology; see Proposition 4.15. Note that 
 $\textrm{IC}_0\star \mathcal {O}(G_n\times {\mathbb G}_m)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{IC}_{\mathcal {O}(G_n)}[j]$
 is a direct sum of shifts of IC-complexes and hence Proposition 4.15 is applicable. On the other hand, using Proposition 5.1, we can identify the right vertical arrow as
$\textrm{IC}_0\star \mathcal {O}(G_n\times {\mathbb G}_m)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{IC}_{\mathcal {O}(G_n)}[j]$
 is a direct sum of shifts of IC-complexes and hence Proposition 4.15 is applicable. On the other hand, using Proposition 5.1, we can identify the right vertical arrow as

 where the right vertical map above is induced by the embeddings 
 $\tau :{\mathfrak c}_n\to {\mathfrak c}_{2n}$
 in (3.4) and
$\tau :{\mathfrak c}_n\to {\mathfrak c}_{2n}$
 in (3.4) and 
 $\delta \times 2\rho _L:G_n\times {\mathbb G}_m\to G_{2n}$
 in (3.19). The group schemes
$\delta \times 2\rho _L:G_n\times {\mathbb G}_m\to G_{2n}$
 in (3.19). The group schemes 
 $J_{2n}$
 and
$J_{2n}$
 and 
 $J_n$
 act on
$J_n$
 act on 
 $\mathcal {O}(G_{2n}\times {\mathfrak c}_{2n})$
 and
$\mathcal {O}(G_{2n}\times {\mathfrak c}_{2n})$
 and 
 $\mathcal {O}(G_{n}\times {\mathbb G}_m\times {\mathfrak c}_n)$
 via the identifications
$\mathcal {O}(G_{n}\times {\mathbb G}_m\times {\mathfrak c}_n)$
 via the identifications 
 $J_{2n}\simeq (G_{2n}\times {\mathfrak c}_{2n})^{\textrm{Ad}_P^{-1}\circ \kappa _{2n}}$
 and
$J_{2n}\simeq (G_{2n}\times {\mathfrak c}_{2n})^{\textrm{Ad}_P^{-1}\circ \kappa _{2n}}$
 and 
 $J_n\simeq (G_{n}\times {\mathfrak c}_n)^{\tau \circ \kappa _n}$
, where
$J_n\simeq (G_{n}\times {\mathfrak c}_n)^{\tau \circ \kappa _n}$
, where 
 $\textrm{Ad}_{P}^{-1}\circ \kappa _{2n}:{\mathfrak c}_{2n}\stackrel {\kappa _{2n}}\to {\mathfrak g}_{2n}^{\textrm{reg}}\stackrel {\textrm{Ad}_{P_{_{\!}}}^{-1}}\to {\mathfrak g}_{2n}^{\textrm{reg}}$
 and
$\textrm{Ad}_{P}^{-1}\circ \kappa _{2n}:{\mathfrak c}_{2n}\stackrel {\kappa _{2n}}\to {\mathfrak g}_{2n}^{\textrm{reg}}\stackrel {\textrm{Ad}_{P_{_{\!}}}^{-1}}\to {\mathfrak g}_{2n}^{\textrm{reg}}$
 and 
 $\tau \circ \kappa _n:{\mathfrak c}_n\stackrel {\kappa _n}\to {\mathfrak g}^{\textrm{reg}}_{n}\stackrel {\tau }\to {\mathfrak g}^{\textrm{reg}}_{2n}$
 are the maps in (3.9). Thus we have a commutative diagram
$\tau \circ \kappa _n:{\mathfrak c}_n\stackrel {\kappa _n}\to {\mathfrak g}^{\textrm{reg}}_{n}\stackrel {\tau }\to {\mathfrak g}^{\textrm{reg}}_{2n}$
 are the maps in (3.9). Thus we have a commutative diagram

where the right vertical arrow is given by pullback of functions along the map
 \begin{equation} {\mathfrak g}_n\times {\mathbb G}_m\to {\mathfrak g}_{2n},\ \ \ \ (C,t)\to \textrm{Ad}_{2\rho _L(t)^{-1}}\tau (C)= \begin{pmatrix}0& \quad t^{-2}\textrm{Id}_n\\ t^{2}C& \quad 0\end{pmatrix}. \end{equation}
\begin{equation} {\mathfrak g}_n\times {\mathbb G}_m\to {\mathfrak g}_{2n},\ \ \ \ (C,t)\to \textrm{Ad}_{2\rho _L(t)^{-1}}\tau (C)= \begin{pmatrix}0& \quad t^{-2}\textrm{Id}_n\\ t^{2}C& \quad 0\end{pmatrix}. \end{equation}
All together we can identify 
 $H^*(\phi ):H^*(A)\to H^*(B)$
 with the map
$H^*(\phi ):H^*(A)\to H^*(B)$
 with the map 
 $\mathcal {O}({\mathfrak g}_{2n})\to \mathcal {O}({\mathfrak g}_n\times {\mathbb G}_m)$
 (as map between non-graded algebras), and we need to show that the induced map
$\mathcal {O}({\mathfrak g}_{2n})\to \mathcal {O}({\mathfrak g}_n\times {\mathbb G}_m)$
 (as map between non-graded algebras), and we need to show that the induced map
 \begin{equation} \mathcal {O}({\mathfrak g}_{2n})_0\to \mathcal {O}({\mathfrak g}_n\times {\mathbb G}_m)_0=\mathcal {O}({\mathfrak g}_n) \end{equation}
\begin{equation} \mathcal {O}({\mathfrak g}_{2n})_0\to \mathcal {O}({\mathfrak g}_n\times {\mathbb G}_m)_0=\mathcal {O}({\mathfrak g}_n) \end{equation}
between the zero-
 ${\mathbb G}_m$
-weight spaces is surjective. For this we observe that the map
${\mathbb G}_m$
-weight spaces is surjective. For this we observe that the map
 \begin{equation} {\mathfrak g}_{2n}\to {\mathfrak g}_n,\quad \begin{pmatrix}A& \quad B\\ C& \quad D\end{pmatrix}\to BC \end{equation}
\begin{equation} {\mathfrak g}_{2n}\to {\mathfrak g}_n,\quad \begin{pmatrix}A& \quad B\\ C& \quad D\end{pmatrix}\to BC \end{equation}
is 
 ${\mathbb G}_m$
-equivariant (
${\mathbb G}_m$
-equivariant (
 ${\mathbb G}_m$
 acts trivially on
${\mathbb G}_m$
 acts trivially on 
 ${\mathfrak g}_n$
) and the composition
${\mathfrak g}_n$
) and the composition 
 is the projection map
 is the projection map 
 $(C,t)\to C$
. Thus the pullback map
$(C,t)\to C$
. Thus the pullback map 
 $\mathcal {O}({\mathfrak g}_n)\to \mathcal {O}({\mathfrak g}_{2n})_0$
 along (5.12) defines a section of (5.11). We are done.
$\mathcal {O}({\mathfrak g}_n)\to \mathcal {O}({\mathfrak g}_{2n})_0$
 along (5.12) defines a section of (5.11). We are done.
Lemma 5.4. 
Let 
 $\phi :A_1\to A_2$
 be a map of dg-algebras. Assume that (1)
$\phi :A_1\to A_2$
 be a map of dg-algebras. Assume that (1) 
 $H^*(A_1)$
 is commutative, (2)
$H^*(A_1)$
 is commutative, (2) 
 $H^*(A_2)$
 is isomorphic to a polynomial ring with generators in even degree, and (3) the map
$H^*(A_2)$
 is isomorphic to a polynomial ring with generators in even degree, and (3) the map 
 $H^*(\phi ):H^*(A_1)\to H^*(A_2)$
 is surjective. Then
$H^*(\phi ):H^*(A_1)\to H^*(A_2)$
 is surjective. Then 
 $A_1$
 being formal implies that
$A_1$
 being formal implies that 
 $A_2$
 is formal.
$A_2$
 is formal.
Proof.
 Let 
 $x_1,\ldots ,x_l$
 be the set of generators of
$x_1,\ldots ,x_l$
 be the set of generators of 
 $H^*(A_2)$
 in even degree such that
$H^*(A_2)$
 in even degree such that 
 ${\mathbb C}[x_1,\ldots ,x_l]\simeq H^*(A_2)$
. Since
${\mathbb C}[x_1,\ldots ,x_l]\simeq H^*(A_2)$
. Since 
 $H^*(\phi ):H^*(A_1)\to H^*(A_2)$
 is surjective, one can find homogeneous elements
$H^*(\phi ):H^*(A_1)\to H^*(A_2)$
 is surjective, one can find homogeneous elements 
 $ y_1,\ldots ,y_l\in H^*(A_1)$
 such that
$ y_1,\ldots ,y_l\in H^*(A_1)$
 such that 
 $H^*(\phi )(y_i)=x_i$
 for
$H^*(\phi )(y_i)=x_i$
 for 
 $i=1,\ldots ,l$
. Assume that
$i=1,\ldots ,l$
. Assume that 
 $H^*(A_1)\simeq A_1$
 is formal; then we have map of dg-algebras
$H^*(A_1)\simeq A_1$
 is formal; then we have map of dg-algebras 
 ${\mathbb C}[z_1,\ldots ,z_l]\to H^*(A_1)\simeq A_1$
 sending
${\mathbb C}[z_1,\ldots ,z_l]\to H^*(A_1)\simeq A_1$
 sending 
 $z_i$
 to
$z_i$
 to 
 $y_i$
. Then the composition
$y_i$
. Then the composition 
 $\gamma :{\mathbb C}[z_1,\ldots ,z_l]\to H^*(A_1)\simeq A_1\stackrel {\phi }\to A_2$
 defines a dg-algebra morphism such that
$\gamma :{\mathbb C}[z_1,\ldots ,z_l]\to H^*(A_1)\simeq A_1\stackrel {\phi }\to A_2$
 defines a dg-algebra morphism such that 
 $H^*(\gamma ):{\mathbb C}[z_1,\ldots ,z_l]\simeq {\mathbb C}[x_1,\ldots ,x_l]\simeq H^*(A_2)$
 is the isomorphism sending
$H^*(\gamma ):{\mathbb C}[z_1,\ldots ,z_l]\simeq {\mathbb C}[x_1,\ldots ,x_l]\simeq H^*(A_2)$
 is the isomorphism sending 
 $z_i$
 to
$z_i$
 to 
 $x_i$
. The lemma follows.
$x_i$
. The lemma follows.
5.2 Derived geometric Satake equivalence for the quaternionic groups
 Denote by 
 $D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 the dg-category of
$D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 the dg-category of 
 $G_n$
-equivariant dg-modules over the dg-algebra
$G_n$
-equivariant dg-modules over the dg-algebra 
 $\textrm{Sym}({\mathfrak g}_n[-4])$
 (equipped with trivial differential). It is known that
$\textrm{Sym}({\mathfrak g}_n[-4])$
 (equipped with trivial differential). It is known that 
 $D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 is compactly generated and the full subcategory
$D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 is compactly generated and the full subcategory 
 $D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))^c$
 of compact objects coincides with the full subcategory
$D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))^c$
 of compact objects coincides with the full subcategory 
 $D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))^c=D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 consisting of perfect modules. Denote by
$D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))^c=D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 consisting of perfect modules. Denote by 
 $D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
 and
$D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
 and 
 $D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
 the full subcategories of
$D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
 the full subcategories of 
 $D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 and
$D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 and 
 $D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))$
, respectively, consisting of modules that are set-theoretically supported on the nilpotent cone
$D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))$
, respectively, consisting of modules that are set-theoretically supported on the nilpotent cone 
 $\textrm{Nilp}({\mathfrak g}_n)$
 of
$\textrm{Nilp}({\mathfrak g}_n)$
 of 
 ${\mathfrak g}_n$
.
${\mathfrak g}_n$
.
 Note that the category 
 $D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 (respectively,
$D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 (respectively, 
 $D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))$
,
$D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))$
, 
 $ D^{G_n}(\textrm{Sym} ({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
, or
$ D^{G_n}(\textrm{Sym} ({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
, or 
 $ D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
) has a natural monoidal structure given by the (derived) tensor product:
$ D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
) has a natural monoidal structure given by the (derived) tensor product: 
 $(\mathcal {F}_1,\mathcal {F}_2)\to \mathcal {F}_1\otimes \mathcal {F}_2:= \mathcal {F}_1\otimes ^L_{\textrm{Sym}({\mathfrak g}_n[-4])}\mathcal {F}_2$
.
$(\mathcal {F}_1,\mathcal {F}_2)\to \mathcal {F}_1\otimes \mathcal {F}_2:= \mathcal {F}_1\otimes ^L_{\textrm{Sym}({\mathfrak g}_n[-4])}\mathcal {F}_2$
.
Theorem 5.5.
- 
(1) There is a canonical equivalence of monoidal categories which induces a monoidal equivalence \begin{equation*}\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))\end{equation*} 
between the corresponding (non-cocomplete) full subcategory of compact objects. \begin{equation*}\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))\end{equation*} 
between the corresponding (non-cocomplete) full subcategory of compact objects. \begin{equation*}D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))\end{equation*} \begin{equation*}D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))\end{equation*}
- 
(2) There is a canonical equivalence of monoidal categories which induces a monoidal equivalence \begin{equation*}D_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}\end{equation*} 
between the corresponding (non-cocomplete) full subcategory of compact objects. \begin{equation*}D_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}\end{equation*} 
between the corresponding (non-cocomplete) full subcategory of compact objects. \begin{equation*}D_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})^c\simeq D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}\end{equation*} \begin{equation*}D_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})^c\simeq D^{G_n}_{\textrm{perf}}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}\end{equation*}
Proof.
 To prove (1), write 
 ${\mathcal C}:=\textrm{Ind}({\textrm{D}}^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))$
. The dg-category
${\mathcal C}:=\textrm{Ind}({\textrm{D}}^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))$
. The dg-category 
 $\mathcal C$
 is a module category for the dg-category
$\mathcal C$
 is a module category for the dg-category 
 $D(\textrm{QCoh}(BG_n))$
 of quasi-coherent sheaves on
$D(\textrm{QCoh}(BG_n))$
 of quasi-coherent sheaves on 
 $BG_n$
, and we can form the de-equivariantized category
$BG_n$
, and we can form the de-equivariantized category 
 ${\mathcal C}_{\textrm{deeq}}:={\mathcal C}\times _{BG}\{\textrm{pt}\}$
 with objects
${\mathcal C}_{\textrm{deeq}}:={\mathcal C}\times _{BG}\{\textrm{pt}\}$
 with objects 
 $\textrm{Ob}({\mathcal C}_{\textrm{deeq}})=\textrm{Ob}({\mathcal C})$
 and (dg-)morphisms
$\textrm{Ob}({\mathcal C}_{\textrm{deeq}})=\textrm{Ob}({\mathcal C})$
 and (dg-)morphisms
 \begin{equation*}\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\mathcal {F}_1,\mathcal {F}_2)=\textrm{Hom}_{{\mathcal C}}(\mathcal {F}_1,\mathcal {F}_2\star \mathcal {O}(G_n))= \textrm{RHom}_{\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\mathcal {F}_1,\mathcal {F}_2\star \mathcal {O}(G_n)).\end{equation*}
\begin{equation*}\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\mathcal {F}_1,\mathcal {F}_2)=\textrm{Hom}_{{\mathcal C}}(\mathcal {F}_1,\mathcal {F}_2\star \mathcal {O}(G_n))= \textrm{RHom}_{\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\mathcal {F}_1,\mathcal {F}_2\star \mathcal {O}(G_n)).\end{equation*}
Every object 
 $\mathcal {F}\in {\mathcal C}_{\textrm{deeq}}$
 carries a natural action of
$\mathcal {F}\in {\mathcal C}_{\textrm{deeq}}$
 carries a natural action of 
 $G_n$
, and we can recover
$G_n$
, and we can recover 
 $\mathcal C$
 by taking
$\mathcal C$
 by taking 
 $G_n$
-equivariant objects in
$G_n$
-equivariant objects in 
 ${\mathcal C}_{\textrm{deeq}}$
. The fact that
${\mathcal C}_{\textrm{deeq}}$
. The fact that 
 $\textrm{IC}_0$
 is compact and generates
$\textrm{IC}_0$
 is compact and generates 
 $\mathcal C$
 under the action of
$\mathcal C$
 under the action of 
 $D\textrm{QCoh}(BG_n)$
 implies that
$D\textrm{QCoh}(BG_n)$
 implies that 
 $\textrm{IC}_0$
, viewed as an object in
$\textrm{IC}_0$
, viewed as an object in 
 ${\mathcal C}_{\textrm{deeq}}$
, is a compact generator. Hence the Barr–Beck–Lurie theorem [Reference LurieLur17, Theorem 4.7.3.5] implies that the assignment
${\mathcal C}_{\textrm{deeq}}$
, is a compact generator. Hence the Barr–Beck–Lurie theorem [Reference LurieLur17, Theorem 4.7.3.5] implies that the assignment 
 $\mathcal {F}\to \textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\mathcal {F})$
 defines an equivalence of categories
$\mathcal {F}\to \textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\mathcal {F})$
 defines an equivalence of categories
 \begin{equation*}{\mathcal C}_{\textrm{deeq}}\simeq D\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)\quad \big(\text{respectively,}\ {\mathcal C}_{} \simeq D^{G_n}\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)\big), \end{equation*}
\begin{equation*}{\mathcal C}_{\textrm{deeq}}\simeq D\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)\quad \big(\text{respectively,}\ {\mathcal C}_{} \simeq D^{G_n}\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)\big), \end{equation*}
where 
 $\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{op}$
 is the opposite of the dg-algebra of endomorphisms of
$\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{op}$
 is the opposite of the dg-algebra of endomorphisms of 
 $\textrm{IC}_0$
 and
$\textrm{IC}_0$
 and 
 $D\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)$
$D\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)$
 
 $\big($
respectively,
$\big($
respectively, 
 $D^{G_n}\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)\big)$
 are the corresponding dg-categories of dg-modules (respectively,
$D^{G_n}\big(\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}\big)\big)$
 are the corresponding dg-categories of dg-modules (respectively, 
 $G_n$
-equivariant dg-modules). Now Proposition 4.21 and Proposition 5.3 imply that the dg-algebra
$G_n$
-equivariant dg-modules). Now Proposition 4.21 and Proposition 5.3 imply that the dg-algebra 
 $\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}$
 is formal and there is a
$\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}}$
 is formal and there is a 
 $G_n$
-equivariant isomorphismFootnote 
8
$G_n$
-equivariant isomorphismFootnote 
8
 \begin{align*}\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}} & \simeq \textrm{RHom}_{\textrm{Ind}({\textrm{D}}^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))^{\rm op}\\ &\simeq \textrm{Ext}^*_{\textrm{Ind}({\textrm{D}}^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))^{\rm op} \\&\simeq \textrm{Sym}({\mathfrak g}_n[-4])^{\rm op}\simeq\textrm{Sym}({\mathfrak g}_n[-4])^{8},\end{align*}
\begin{align*}\textrm{Hom}_{{\mathcal C}_{\textrm{deeq}}}(\textrm{IC}_0,\textrm{IC}_0)^{\textrm{op}} & \simeq \textrm{RHom}_{\textrm{Ind}({\textrm{D}}^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))^{\rm op}\\ &\simeq \textrm{Ext}^*_{\textrm{Ind}({\textrm{D}}^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\textrm{IC}_0\star \mathcal {O}(G_n))^{\rm op} \\&\simeq \textrm{Sym}({\mathfrak g}_n[-4])^{\rm op}\simeq\textrm{Sym}({\mathfrak g}_n[-4])^{8},\end{align*}
and hence we conclude that there is an equivalence
 \begin{align*}\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4])), \\ \;\mathcal {F} \to \textrm{RHom}_{\textrm{Ind}(D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\mathcal {F}\star \mathcal {O}(G_n)).\!\end{align*}
\begin{align*}\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}(\textrm{Sym}({\mathfrak g}_n[-4])), \\ \;\mathcal {F} \to \textrm{RHom}_{\textrm{Ind}(D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\mathcal {F}\star \mathcal {O}(G_n)).\!\end{align*}
The monoidal structure on the constructed equivalence will be proved in § 5.4. This finishes the proof of part (1).
 Part (2) follows from the general discussion in [Reference ArinkinAG15, § 12] on derived geometric Satake for complex reductive groups. Let 
 $\widetilde {\textrm{IC}}_0=p_!(\mathbb C_{\textrm{pt}})\in D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 where
$\widetilde {\textrm{IC}}_0=p_!(\mathbb C_{\textrm{pt}})\in D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 where 
 $p:\textrm{pt}\to \mathfrak{L}^+G_{n,{\mathbb H}}\backslash \textrm{pt}\simeq \mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}^0\to \mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}$
. By [Reference ArinkinAG15, § 12.6.6],
$p:\textrm{pt}\to \mathfrak{L}^+G_{n,{\mathbb H}}\backslash \textrm{pt}\simeq \mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}^0\to \mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}$
. By [Reference ArinkinAG15, § 12.6.6], 
 $\widetilde {\textrm{IC}}_0$
 is in fact a compact object
$\widetilde {\textrm{IC}}_0$
 is in fact a compact object 
 $\widetilde {\textrm{IC}}_0\in D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})^c$
 and the category
$\widetilde {\textrm{IC}}_0\in D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})^c$
 and the category 
 $D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 is compactly generated by objects of the form
$D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 is compactly generated by objects of the form 
 $\widetilde {\textrm{IC}}_0*V$
 for
$\widetilde {\textrm{IC}}_0*V$
 for 
 $V\in \textrm{Rep}(G_n)$
. We claim that under the fully faithful embedding
$V\in \textrm{Rep}(G_n)$
. We claim that under the fully faithful embedding
 \begin{equation} D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Ind}(D_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})^c)\to \textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4])), \end{equation}
\begin{equation} D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})\simeq \textrm{Ind}(D_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})^c)\to \textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4])), \end{equation}
the compact generator 
 $\widetilde {\textrm{IC}}_0$
 goes to
$\widetilde {\textrm{IC}}_0$
 goes to 
 $\textrm{Sym}({\mathfrak g}_n[-4])\otimes _{\textrm{Sym}({\mathfrak g}_n[-4])^{G_n}}{\mathbb C}$
, where
$\textrm{Sym}({\mathfrak g}_n[-4])\otimes _{\textrm{Sym}({\mathfrak g}_n[-4])^{G_n}}{\mathbb C}$
, where 
 $\mathbb C$
 is the augmentation module of
$\mathbb C$
 is the augmentation module of 
 $\textrm{Sym}({\mathfrak g}_n[-4])^{G_n}$
. Since
$\textrm{Sym}({\mathfrak g}_n[-4])^{G_n}$
. Since 
 $\textrm{Sym}({\mathfrak g}_n)\otimes _{\textrm{Sym}({\mathfrak g}_n)^{G_n}}{\mathbb C}\simeq \mathcal {O}(\textrm{Nilp}({\mathfrak g}_n))$
, we conclude that
$\textrm{Sym}({\mathfrak g}_n)\otimes _{\textrm{Sym}({\mathfrak g}_n)^{G_n}}{\mathbb C}\simeq \mathcal {O}(\textrm{Nilp}({\mathfrak g}_n))$
, we conclude that 
 $D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 is equivalent to the full subcategory of
$D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}})$
 is equivalent to the full subcategory of 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 generated by objects of the form
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 generated by objects of the form 
 $\mathcal {O}(\textrm{Nilp}({\mathfrak g}_n))\otimes V$
 for
$\mathcal {O}(\textrm{Nilp}({\mathfrak g}_n))\otimes V$
 for 
 $V\in \textrm{Rep}(G_n)$
. It is clear that this subcategory is exactly
$V\in \textrm{Rep}(G_n)$
. It is clear that this subcategory is exactly 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
. To prove the claim, we observe that Proposition 4.15 implies that the image of
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))_{\textrm{Nilp}({\mathfrak g}_n)}$
. To prove the claim, we observe that Proposition 4.15 implies that the image of 
 $\widetilde {\textrm{IC}}_0$
 under (5.13) is given by
$\widetilde {\textrm{IC}}_0$
 under (5.13) is given by
 \begin{equation*}\textrm{RHom}_{\textrm{Ind}(D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\widetilde {\textrm{IC}}_0\star \mathcal {O}(G_n))\simeq {\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{T_c}(\textrm{IC}_0),H^*_{T_c}(\widetilde {\textrm{IC}}_0)\otimes \mathcal {O}(G_n)\big)^{\mathrm {W}}.\end{equation*}
\begin{equation*}\textrm{RHom}_{\textrm{Ind}(D^b(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))}(\textrm{IC}_0,\widetilde {\textrm{IC}}_0\star \mathcal {O}(G_n))\simeq {\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{T_c}(\textrm{IC}_0),H^*_{T_c}(\widetilde {\textrm{IC}}_0)\otimes \mathcal {O}(G_n)\big)^{\mathrm {W}}.\end{equation*}
Note that 
 $H^*_{T_c}\big(\widetilde {\textrm{IC}}_0\big)\simeq {\mathbb C}$
 is isomorphic to the augmented module of
$H^*_{T_c}\big(\widetilde {\textrm{IC}}_0\big)\simeq {\mathbb C}$
 is isomorphic to the augmented module of 
 $H^*_{T_c}(\textrm{IC}_0)\simeq \mathcal {O}({\mathfrak t}[2])$
, and the same computation as in Proposition 4.21 shows that
$H^*_{T_c}(\textrm{IC}_0)\simeq \mathcal {O}({\mathfrak t}[2])$
, and the same computation as in Proposition 4.21 shows that
 \begin{align*}{\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{T_c}(\textrm{IC}_0),H^*_{T_c}\big(\widetilde {\textrm{IC}}_0\big)\otimes \mathcal {O}(G_n)\big)^{\mathrm {W}} &\simeq \mathcal {O}\big({\mathfrak g}^{\textrm{reg}}_n[4]\times _{{\mathfrak g}^{}_n[4]//G_n}\{0\}\big)^{\mathrm {W}}\\& \simeq \mathcal {O}({\mathfrak g}_n[4]\times _{{\mathfrak g}_n[4]//G_n}\{0\})\\ &\simeq \textrm{Sym}({\mathfrak g}_n[-4])\otimes _{\textrm{Sym}({\mathfrak g}_n[-4])^{G_n}}{\mathbb C}.\end{align*}
\begin{align*}{\textrm{Hom}}^*_{H^*_{T_c}({\textrm{Gr}}_{n,{\mathbb H}})}\big(H^*_{T_c}(\textrm{IC}_0),H^*_{T_c}\big(\widetilde {\textrm{IC}}_0\big)\otimes \mathcal {O}(G_n)\big)^{\mathrm {W}} &\simeq \mathcal {O}\big({\mathfrak g}^{\textrm{reg}}_n[4]\times _{{\mathfrak g}^{}_n[4]//G_n}\{0\}\big)^{\mathrm {W}}\\& \simeq \mathcal {O}({\mathfrak g}_n[4]\times _{{\mathfrak g}_n[4]//G_n}\{0\})\\ &\simeq \textrm{Sym}({\mathfrak g}_n[-4])\otimes _{\textrm{Sym}({\mathfrak g}_n[-4])^{G_n}}{\mathbb C}.\end{align*}
The claim follows.
5.3 Spectral description of nearby cycles functors
5.3.1 Shift of grading
 Let 
 $(A=\oplus _{} A^i,d)$
 be a dg-algebra equipped with an action of
$(A=\oplus _{} A^i,d)$
 be a dg-algebra equipped with an action of 
 $G=H\times {\mathbb G}_m$
. We will write
$G=H\times {\mathbb G}_m$
. We will write 
 $A^i=\oplus A^i_j$
 where the lower index
$A^i=\oplus A^i_j$
 where the lower index 
 $j$
 refers to the
$j$
 refers to the 
 ${\mathbb G}_m$
-weights coming from the
${\mathbb G}_m$
-weights coming from the 
 ${\mathbb G}_m$
-action. Assume that the
${\mathbb G}_m$
-action. Assume that the 
 ${\mathbb G}_m$
-weights are even, that is, we have
${\mathbb G}_m$
-weights are even, that is, we have 
 $A^i_j=0$
 if
$A^i_j=0$
 if 
 $j\in 2{\mathbb Z}+1$
. Following [Reference ArinkinAG15, Appendix A.2], one can introduce a new dg-algebra
$j\in 2{\mathbb Z}+1$
. Following [Reference ArinkinAG15, Appendix A.2], one can introduce a new dg-algebra 
 $\big(\tilde A=\bigoplus \tilde A^i_j,d\big)$
 where
$\big(\tilde A=\bigoplus \tilde A^i_j,d\big)$
 where
 \begin{equation*}\tilde A^{i}_j=A^{i+j}_j,\end{equation*}
\begin{equation*}\tilde A^{i}_j=A^{i+j}_j,\end{equation*}
such that the map sending a 
 $G$
-equivariant dg-module
$G$
-equivariant dg-module 
 $\big(M=\bigoplus M^i_j,d\big)$
 over
$\big(M=\bigoplus M^i_j,d\big)$
 over 
 $A$
 to the dg-module
$A$
 to the dg-module 
 $\big(\tilde M=\bigoplus \tilde M^i_j,d\big)$
 over
$\big(\tilde M=\bigoplus \tilde M^i_j,d\big)$
 over 
 $\tilde A$
 with
$\tilde A$
 with
 \begin{equation*}\tilde M^i_j=M^{i+j}_j\end{equation*}
\begin{equation*}\tilde M^i_j=M^{i+j}_j\end{equation*}
induces an equivalence of triangulated categories
 \begin{equation*}D^G_{}(A)\simeq D^G_{}(\tilde A)\quad \big(\text{respectively,} \ D^G_{\textrm{perf}}(A)\simeq D^G_{\textrm{perf}}(\tilde A)\big).\end{equation*}
\begin{equation*}D^G_{}(A)\simeq D^G_{}(\tilde A)\quad \big(\text{respectively,} \ D^G_{\textrm{perf}}(A)\simeq D^G_{\textrm{perf}}(\tilde A)\big).\end{equation*}
Example 5.6. Consider the dg-algebra 
 $(A=\textrm{Sym}({\mathfrak g}_{2n}),d=0)$
. The subgroup
$(A=\textrm{Sym}({\mathfrak g}_{2n}),d=0)$
. The subgroup 
 $G_n \times {\mathbb G}_m\subset G_{2n}$
 as in § 3.5 acts on the generators
$G_n \times {\mathbb G}_m\subset G_{2n}$
 as in § 3.5 acts on the generators 
 ${\mathfrak g}_{2n}$
 of
${\mathfrak g}_{2n}$
 of 
 $A$
 via the adjoint action, and if we write the elements in
$A$
 via the adjoint action, and if we write the elements in 
 ${\mathfrak g}_{2n}$
 in the form
${\mathfrak g}_{2n}$
 in the form
 \begin{equation*}{\mathfrak g}_{2n}=\bigg\{\bigg(\begin{array}{c@{\quad}c} A& B\\ C& D \end{array}\bigg)\bigg| A,B,C,D\in {\mathfrak g}_n\bigg\}, \end{equation*}
\begin{equation*}{\mathfrak g}_{2n}=\bigg\{\bigg(\begin{array}{c@{\quad}c} A& B\\ C& D \end{array}\bigg)\bigg| A,B,C,D\in {\mathfrak g}_n\bigg\}, \end{equation*}
then 
 $A\mbox{ and }D$
 are of weight zero,
$A\mbox{ and }D$
 are of weight zero, 
 $B$
 is of weight
$B$
 is of weight 
 $2$
, and
$2$
, and 
 $C$
 is of weight
$C$
 is of weight 
 $-2$
. It follows that
$-2$
. It follows that
 \begin{equation*}\tilde A\simeq \textrm{Sym}(\tilde {\mathfrak g}_{2n})\end{equation*}
\begin{equation*}\tilde A\simeq \textrm{Sym}(\tilde {\mathfrak g}_{2n})\end{equation*}
where 
 $\tilde {\mathfrak g}_{2n}$
 consists of elements of the form
$\tilde {\mathfrak g}_{2n}$
 consists of elements of the form
 \begin{equation*}\tilde {\mathfrak g}_{2n}=\bigg\{\bigg(\begin{array}{c@{\quad}c} A[0]& B[-2]\\ C[2]& D[0] \end{array}\bigg) \bigg| A,B,C,D\in {\mathfrak g}_n\bigg\}.\end{equation*}
\begin{equation*}\tilde {\mathfrak g}_{2n}=\bigg\{\bigg(\begin{array}{c@{\quad}c} A[0]& B[-2]\\ C[2]& D[0] \end{array}\bigg) \bigg| A,B,C,D\in {\mathfrak g}_n\bigg\}.\end{equation*}
5.3.2
It follows from Example 5.6 that we have an equivalence of categories
 \begin{equation} D^{G_n\times {\mathbb G}_m}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2]))\simeq D^{G_n\times {\mathbb G}_m}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])). \end{equation}
\begin{equation} D^{G_n\times {\mathbb G}_m}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2]))\simeq D^{G_n\times {\mathbb G}_m}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])). \end{equation}
On the other hand, the natural 
 $G_n$
-equivariant map
$G_n$
-equivariant map 
 ${\mathfrak g}_{n}[4]\to \tilde {\mathfrak g}_{2n}[2]$
 sending
${\mathfrak g}_{n}[4]\to \tilde {\mathfrak g}_{2n}[2]$
 sending
 \begin{equation*}C[4]\to \bigg(\begin{array}{c@{\quad}c} 0& \textrm{Id}_n\\ C[4]& 0 \end{array}\bigg)\quad\mbox{for } C\in {\mathfrak g}_n \end{equation*}
\begin{equation*}C[4]\to \bigg(\begin{array}{c@{\quad}c} 0& \textrm{Id}_n\\ C[4]& 0 \end{array}\bigg)\quad\mbox{for } C\in {\mathfrak g}_n \end{equation*}
gives rise to a map of dg-algebras
 \begin{equation} \textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])\simeq \mathcal {O}(\tilde {\mathfrak g}_{2n}[2])\longrightarrow \mathcal {O}({\mathfrak g}_{n}[4])\simeq \textrm{Sym}({\mathfrak g}_{n}[-4]) \end{equation}
\begin{equation} \textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])\simeq \mathcal {O}(\tilde {\mathfrak g}_{2n}[2])\longrightarrow \mathcal {O}({\mathfrak g}_{n}[4])\simeq \textrm{Sym}({\mathfrak g}_{n}[-4]) \end{equation}
(here we identify the graded duals of 
 $\tilde {\mathfrak g}_{2n}[-2]$
 and
$\tilde {\mathfrak g}_{2n}[-2]$
 and 
 ${\mathfrak g}_{n}[4]$
 with
${\mathfrak g}_{n}[4]$
 with 
 $\tilde {\mathfrak g}_{2n}[2]$
 and
$\tilde {\mathfrak g}_{2n}[2]$
 and 
 ${\mathfrak g}_{n}[-4]$
 via the trace form) and hence a functor
${\mathfrak g}_{n}[-4]$
 via the trace form) and hence a functor
 \begin{equation} D^{{\textrm{GL}}_n}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2]))\to D^{{\textrm{GL}}_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4])),\quad M\to \textrm{Sym}({\mathfrak g}_{n}[-4])\otimes ^L_{\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])}M. \end{equation}
\begin{equation} D^{{\textrm{GL}}_n}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2]))\to D^{{\textrm{GL}}_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4])),\quad M\to \textrm{Sym}({\mathfrak g}_{n}[-4])\otimes ^L_{\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])}M. \end{equation}
Finally, let us consider the functor

where 
 $F$
 are the natural forgetful functors.
$F$
 are the natural forgetful functors.
Theorem 5.7. The following square is commutative

 
where 
 $\Psi$
 and
$\Psi$
 and 
 $\Psi _{\mathbb H}$
 are the complex and quaternionic Satake equivalences, respectively. It induces a similar commutative diagram for the subcategories of compact objects.
$\Psi _{\mathbb H}$
 are the complex and quaternionic Satake equivalences, respectively. It induces a similar commutative diagram for the subcategories of compact objects.
Proof.
 We shall construct a natural transformation 
 $\Phi \circ \Psi \to \Psi _{\mathbb H}\circ \mathrm R$
. Write
$\Phi \circ \Psi \to \Psi _{\mathbb H}\circ \mathrm R$
. Write 
 $A= \textrm{RHom}\big(\textrm{IC}_0, \textrm{IC}_{\mathcal {O}(G_{2n})}\big)\simeq \textrm{Sym}({\mathfrak g}_{2n}[-2])$
,
$A= \textrm{RHom}\big(\textrm{IC}_0, \textrm{IC}_{\mathcal {O}(G_{2n})}\big)\simeq \textrm{Sym}({\mathfrak g}_{2n}[-2])$
, 
 $B=\textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\mathcal {O}(G_{n})}\big)\simeq \textrm{Sym}({\mathfrak g}_n[-4])$
, and
$B=\textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\mathcal {O}(G_{n})}\big)\simeq \textrm{Sym}({\mathfrak g}_n[-4])$
, and 
 $A'= \textrm{RHom}\big(\textrm{IC}_0, \mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\big)$
. Since
$A'= \textrm{RHom}\big(\textrm{IC}_0, \mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\big)$
. Since 
 $\mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)$
 is an algebra object in
$\mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)$
 is an algebra object in 
 $\textrm{Ind}(D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))$
, the (dg) Hom space
$\textrm{Ind}(D(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))$
, the (dg) Hom space 
 $A'$
 is naturally a dg-algebra.
$A'$
 is naturally a dg-algebra.
 For any 
 $\mathcal {F}$
, we have a map of dg-modules for the dg-algebra
$\mathcal {F}$
, we have a map of dg-modules for the dg-algebra 
 $A$
,
$A$
,
 \begin{equation*}\Psi (\mathcal {F})\simeq \textrm{RHom}\big(\textrm{IC}_0,\mathcal {F}\star \textrm{IC}_{\mathcal {O}(G_{2n})}\big)\stackrel {\mathrm R}\longrightarrow \textrm{RHom}\big(\textrm{IC}_0,\mathrm R(\mathcal {F})\star \mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\big):= \Psi (\mathcal {F})',\end{equation*}
\begin{equation*}\Psi (\mathcal {F})\simeq \textrm{RHom}\big(\textrm{IC}_0,\mathcal {F}\star \textrm{IC}_{\mathcal {O}(G_{2n})}\big)\stackrel {\mathrm R}\longrightarrow \textrm{RHom}\big(\textrm{IC}_0,\mathrm R(\mathcal {F})\star \mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\big):= \Psi (\mathcal {F})',\end{equation*}
where 
 $A$
 acts on
$A$
 acts on 
 $\Psi (\mathcal {F})'$
 via the dg-algebra map
$\Psi (\mathcal {F})'$
 via the dg-algebra map
 \begin{equation*}A=\textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\stackrel {\mathrm R}\longrightarrow A'=\textrm{RHom}\big(\textrm{IC}_0,\mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\big).\end{equation*}
\begin{equation*}A=\textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\stackrel {\mathrm R}\longrightarrow A'=\textrm{RHom}\big(\textrm{IC}_0,\mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\big).\end{equation*}
The right regular 
 ${\mathbb G}_m$
-action on
${\mathbb G}_m$
-action on 
 $G_{2n}$
 via the co-character
$G_{2n}$
 via the co-character 
 $2\rho _L:{\mathbb G}_m\to G_n\times {\mathbb G}_m\subset G_{2n}$
 induces a
$2\rho _L:{\mathbb G}_m\to G_n\times {\mathbb G}_m\subset G_{2n}$
 induces a 
 ${\mathbb G}_m$
-action on the dg-algebras
${\mathbb G}_m$
-action on the dg-algebras 
 $A$
 and
$A$
 and 
 $A'$
 (with even weights) and also the dg-modules
$A'$
 (with even weights) and also the dg-modules 
 $\Psi (\mathcal {F})$
 and
$\Psi (\mathcal {F})$
 and 
 $\Psi (\mathcal {F})'$
. Thus we can perform the shift of grading operation in § 5.3.1 and obtain a map of dg-modules for the dg-algebra
$\Psi (\mathcal {F})'$
. Thus we can perform the shift of grading operation in § 5.3.1 and obtain a map of dg-modules for the dg-algebra 
 $\tilde A$
,
$\tilde A$
,
 \begin{equation} \widetilde {\Psi (\mathcal {F})}\to \widetilde {\Psi (\mathcal {F})'}, \end{equation}
\begin{equation} \widetilde {\Psi (\mathcal {F})}\to \widetilde {\Psi (\mathcal {F})'}, \end{equation}
where 
 $\tilde A$
 acts on
$\tilde A$
 acts on 
 $\widetilde {\Psi (\mathcal {F})'}$
 via the map
$\widetilde {\Psi (\mathcal {F})'}$
 via the map 
 $\tilde A\to \tilde A'$
. By Example 5.6, we have
$\tilde A\to \tilde A'$
. By Example 5.6, we have
 \begin{equation*}\tilde A\simeq \textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2]).\end{equation*}
\begin{equation*}\tilde A\simeq \textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2]).\end{equation*}
On the other hand, by Proposition 4.5(3), we haveFootnote 9
 \begin{equation*}\mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}\mathcal {O}(G_{2n})_j}[j]\end{equation*}
\begin{equation*}\mathrm R\big(\textrm{IC}_{\mathcal {O}(G_{2n})}\big)\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}\mathcal {O}(G_{2n})_j}[j]\end{equation*}
where
 \begin{equation*}\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}\mathcal {O}(G_{2n})_j \end{equation*}
\begin{equation*}\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))\simeq \bigoplus _{j\in {\mathbb Z}}\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}\mathcal {O}(G_{2n})_j \end{equation*}
is the 
 ${\mathbb G}_m$
-weight decomposition of
${\mathbb G}_m$
-weight decomposition of 
 $\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))$
, and it follows that
$\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))$
, and it follows that
 \begin{equation*}\tilde A'\simeq \textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))}\big),\quad \widetilde {\Psi (\mathcal {F})'}\simeq \textrm{RHom}\big(\textrm{IC}_0,\mathrm R(\mathcal {F})\star \textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))}\big).\end{equation*}
\begin{equation*}\tilde A'\simeq \textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))}\big),\quad \widetilde {\Psi (\mathcal {F})'}\simeq \textrm{RHom}\big(\textrm{IC}_0,\mathrm R(\mathcal {F})\star \textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))}\big).\end{equation*}
Since the natural algebra map 
 $\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))\to \mathcal {O}(G_n)$
 of algebra objects in
$\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))\to \mathcal {O}(G_n)$
 of algebra objects in 
 $\textrm{Rep}(G_n)$
 coming from the embedding
$\textrm{Rep}(G_n)$
 coming from the embedding 
 $G_n\to G_{n}\times {\mathbb G}_m, g\to (g,e)$
 induces a map
$G_n\to G_{n}\times {\mathbb G}_m, g\to (g,e)$
 induces a map
 \begin{equation*}\iota :\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}(\mathcal {O}(G_{2n}))}\to \textrm{IC}_{\mathcal {O}(G_n)}\end{equation*}
\begin{equation*}\iota :\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n\times {\mathbb G}_m}(\mathcal {O}(G_{2n}))}\to \textrm{IC}_{\mathcal {O}(G_n)}\end{equation*}
between the corresponding algebra objects in 
 $\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
, we obtain a map of dg-algebras,
$\textrm{Perv}({\textrm{Gr}}_{n,{\mathbb H}})$
, we obtain a map of dg-algebras,
 \begin{equation} \tilde A\to \tilde A'\simeq \textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))}\big)\stackrel {\iota }\to \textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\mathcal {O}(G_n)}\big)=B, \end{equation}
\begin{equation} \tilde A\to \tilde A'\simeq \textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\textrm{Res}^{G_{2n}}_{G_n}(\mathcal {O}(G_{2n}))}\big)\stackrel {\iota }\to \textrm{RHom}\big(\textrm{IC}_0,\textrm{IC}_{\mathcal {O}(G_n)}\big)=B, \end{equation}
and a map of dg-modules over the dg-algebra 
 $\tilde A$
,
$\tilde A$
,

where 
 $\tilde A$
 acts on
$\tilde A$
 acts on 
 $\Psi _{\mathbb H}\circ \mathrm R(\mathcal {F})$
 via the morphism (5.19). Moreover, the proof of Proposition 5.3 implies that the map (5.19) is equal to the map in (5.15). Thus, by the universal property of the tensor product, the map (5.20) gives rise to a map of dg-modules over the dg-algebra
$\Psi _{\mathbb H}\circ \mathrm R(\mathcal {F})$
 via the morphism (5.19). Moreover, the proof of Proposition 5.3 implies that the map (5.19) is equal to the map in (5.15). Thus, by the universal property of the tensor product, the map (5.20) gives rise to a map of dg-modules over the dg-algebra 
 $B$
,
$B$
,
 \begin{equation} \Phi \circ \Psi (\mathcal {F})\simeq B\otimes ^L_{\tilde A}\widetilde {\Psi (\mathcal {F})}\to \Psi _{\mathbb H}\circ \mathrm R(\mathcal {F}). \end{equation}
\begin{equation} \Phi \circ \Psi (\mathcal {F})\simeq B\otimes ^L_{\tilde A}\widetilde {\Psi (\mathcal {F})}\to \Psi _{\mathbb H}\circ \mathrm R(\mathcal {F}). \end{equation}
This finishes the construction of the desired natural transformation map.
 Now, to finish the proof, it suffices to check that (5.21) is an isomorphism when 
 $\mathcal {F}$
 is of the form
$\mathcal {F}$
 is of the form 
 $\mathcal {F}\simeq \textrm{IC}_{\textrm{V}}$
 with
$\mathcal {F}\simeq \textrm{IC}_{\textrm{V}}$
 with 
 $\textrm{V}\in \textrm{Rep}(G_{2n})$
. For this we observe that if
$\textrm{V}\in \textrm{Rep}(G_{2n})$
. For this we observe that if 
 $\textrm{V}=\oplus _{j\in {\mathbb Z}}\textrm{V}_j$
 is the
$\textrm{V}=\oplus _{j\in {\mathbb Z}}\textrm{V}_j$
 is the 
 ${\mathbb G}_m$
-weight decomposition, then we have
${\mathbb G}_m$
-weight decomposition, then we have
 \begin{equation} \Psi _{\mathbb H}\circ \mathrm R(\textrm{IC}_{\textrm{V}})\simeq \bigoplus _{j\in {\mathbb Z}} \Psi _{\mathbb H}(\textrm{IC}_{\textrm{V}_j})[-j]\simeq \bigoplus _{j\in {\mathbb Z}}B\otimes _{\mathbb C}\textrm{V}_j[-j]. \end{equation}
\begin{equation} \Psi _{\mathbb H}\circ \mathrm R(\textrm{IC}_{\textrm{V}})\simeq \bigoplus _{j\in {\mathbb Z}} \Psi _{\mathbb H}(\textrm{IC}_{\textrm{V}_j})[-j]\simeq \bigoplus _{j\in {\mathbb Z}}B\otimes _{\mathbb C}\textrm{V}_j[-j]. \end{equation}
On the other hand, we have
 \begin{equation*}\widetilde {\Psi (\textrm{IC}_{\textrm{V}})}\simeq \widetilde {(A\otimes _{\mathbb C}\textrm{V})}\simeq \bigoplus _{j\in {\mathbb Z}}\tilde {A}\otimes _{\mathbb C}\textrm{V}_j[-j]\end{equation*}
\begin{equation*}\widetilde {\Psi (\textrm{IC}_{\textrm{V}})}\simeq \widetilde {(A\otimes _{\mathbb C}\textrm{V})}\simeq \bigoplus _{j\in {\mathbb Z}}\tilde {A}\otimes _{\mathbb C}\textrm{V}_j[-j]\end{equation*}
and hence
 \begin{equation} \Phi \circ \Psi (\textrm{IC}_{\textrm{V}})\simeq B\otimes _{\tilde A}\widetilde {\Psi (\textrm{IC}_{\textrm{V}})}\simeq B\otimes _{\tilde A}\bigg(\bigoplus _{j\in {\mathbb Z}}\tilde {A}\otimes _{\mathbb C}\textrm{V}_j[-j]\bigg)\simeq \bigoplus _{j\in {\mathbb Z}}B\otimes _{\mathbb C}\textrm{V}_j[-j]. \end{equation}
\begin{equation} \Phi \circ \Psi (\textrm{IC}_{\textrm{V}})\simeq B\otimes _{\tilde A}\widetilde {\Psi (\textrm{IC}_{\textrm{V}})}\simeq B\otimes _{\tilde A}\bigg(\bigoplus _{j\in {\mathbb Z}}\tilde {A}\otimes _{\mathbb C}\textrm{V}_j[-j]\bigg)\simeq \bigoplus _{j\in {\mathbb Z}}B\otimes _{\mathbb C}\textrm{V}_j[-j]. \end{equation}
It follows from the construction that the map (5.21) is given by

and hence is an isomorphism. This completes the proof of the proposition.
5.4 Monoidal structures
 We construct a monoidal structure on the equivalence 
 $\Psi _{\mathbb H}:\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 in Theorem5.5. Consider the monoidal structure on
$\Psi _{\mathbb H}:\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))\simeq D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))$
 in Theorem5.5. Consider the monoidal structure on 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))$
,
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_n[-4]))$
,
 \begin{equation*}M_1\otimes 'M_2:=\Psi _{\mathbb H}\big(\Psi _{\mathbb H}^{-1}(M_1)\star \Psi _{\mathbb H}^{-1}(M_2)\big),\end{equation*}
\begin{equation*}M_1\otimes 'M_2:=\Psi _{\mathbb H}\big(\Psi _{\mathbb H}^{-1}(M_1)\star \Psi _{\mathbb H}^{-1}(M_2)\big),\end{equation*}
induced from the monoidal structure on 
 $\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))$
 via the equivalence
$\textrm{Ind}(D^b_{}(\mathfrak{L}^+G_{n,{\mathbb H}}\backslash {\textrm{Gr}}_{n,{\mathbb H}}))$
 via the equivalence 
 $\Psi _{\mathbb H}$
. We would like to show that
$\Psi _{\mathbb H}$
. We would like to show that 
 $\otimes '$
 is isomorphic to the natural tensor monoidal structure. The square in Theorem5.7, together with the fact that the derived Satake equivalence
$\otimes '$
 is isomorphic to the natural tensor monoidal structure. The square in Theorem5.7, together with the fact that the derived Satake equivalence 
 $\Psi$
 is monoidal, implies that the functor
$\Psi$
 is monoidal, implies that the functor 
 $\Phi :D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2]))\rightarrow D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 in loc. cit. is monoidal with respect to the natural tensor monoidal structure on
$\Phi :D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2]))\rightarrow D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 in loc. cit. is monoidal with respect to the natural tensor monoidal structure on 
 $D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2])$
 and the above monoidal structure
$D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2])$
 and the above monoidal structure 
 $\otimes '$
 on
$\otimes '$
 on 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
. Now the desired claim follows from the following lemma.
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
. Now the desired claim follows from the following lemma.
Lemma 5.8. 
Equip 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 with its natural tensor monoidal structure.
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 with its natural tensor monoidal structure.
 
Then the natural tensor monoidal structure on 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 is the unique (up to equivalence) monoidal structure on
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 is the unique (up to equivalence) monoidal structure on 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 such that the
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 such that the 
 $\textrm{Rep}(G_{2n})$
-module functor
$\textrm{Rep}(G_{2n})$
-module functor

 
may be compatibly lifted to a monoidal functor. Moreover, the compatible monoidal structure on 
 $\Phi$
 is unique (up to equivalence).
$\Phi$
 is unique (up to equivalence).
Proof.
 Returning to its construction in (5.17), recall that 
 $\Phi$
 factors into the sheared forgetful functor
$\Phi$
 factors into the sheared forgetful functor
 \begin{equation*} D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2])) \to D^{G_n}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])) \end{equation*}
\begin{equation*} D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2])) \to D^{G_n}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])) \end{equation*}
followed by the restriction
 \begin{equation*} D^{G_n}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])) \to D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4])), \quad M \mapsto \textrm{Sym}({\mathfrak g}_{n}[-4])\otimes ^L_{\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])}M. \end{equation*}
\begin{equation*} D^{G_n}_{}(\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])) \to D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4])), \quad M \mapsto \textrm{Sym}({\mathfrak g}_{n}[-4])\otimes ^L_{\textrm{Sym}(\tilde {\mathfrak g}_{2n}[-2])}M. \end{equation*}
 First, 
 $\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
 is the unit of
$\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
 is the unit of 
 $D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
, so
$D^{G_{2n}}_{}(\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
, so 
 $ \textrm{Sym}({\mathfrak g}_{n}[-4]) \simeq \Phi (\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
 must be the unit of
$ \textrm{Sym}({\mathfrak g}_{n}[-4]) \simeq \Phi (\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
 must be the unit of 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
.
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
.
 Next, recall that 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 is compactly generated by
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 is compactly generated by 
 $V \otimes \textrm{Sym}({\mathfrak g}_{n}[-4])$
, where
$V \otimes \textrm{Sym}({\mathfrak g}_{n}[-4])$
, where 
 $V$
 is a finite-dimensional representation of
$V$
 is a finite-dimensional representation of 
 $G_n$
. Note that every such
$G_n$
. Note that every such 
 $V$
 is a direct summand in the restriction of a finite-dimensional representation of
$V$
 is a direct summand in the restriction of a finite-dimensional representation of 
 $G_{2n}$
. Since
$G_{2n}$
. Since 
 $\Phi$
 is a
$\Phi$
 is a 
 $\textrm{Rep}(G_{2n})$
-module map, this determines the monoidal product on
$\textrm{Rep}(G_{2n})$
-module map, this determines the monoidal product on 
 $D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 as well as its coherent associativity structure.
$D^{G_n}_{}(\textrm{Sym}({\mathfrak g}_{n}[-4]))$
 as well as its coherent associativity structure.
 Finally, since the monoidal structures on 
 $\Phi$
 must be compatible with its
$\Phi$
 must be compatible with its 
 $\textrm{Rep}(G_{2n})$
-module structure, it is determined by its restriction to the unit
$\textrm{Rep}(G_{2n})$
-module structure, it is determined by its restriction to the unit 
 $\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
 where there are no choices.
$\textrm{Sym}({\mathfrak g}_{2n}[-2]))$
 where there are no choices.
Acknowledgments
We would like to thank David Ben-Zvi, Jonathan Wang, and Geordie Williamson for useful discussions. We also would like to thank the anonymous referee for their thoughtful and detailed suggestions. T-H Chen would also like to thank the Institute for Advanced Study in Princeton for support, hospitality, and a nice research environment.
Conflicts of interest
None.
Financial support
T-H Chen is supported by NSF grant DMS-2001257 and the S. S. Chern Foundation. D. Nadler is supported by NSF grant DMS-2101466. MM is supported by an NSF Graduate Research Fellowship.
Journal information
Compositio Mathematica is owned by the Foundation Compositio Mathematica and published by the London Mathematical Society in partnership with Cambridge University Press. All surplus income from the publication of Compositio Mathematica is returned to mathematics and higher education through the charitable activities of the Foundation, the London Mathematical Society and Cambridge University Press.
 
 



 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
