This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Lusztig sheaves and integrable highest weight modules in symmetrizable case

Yixin Lan, Yumeng Wu, Jie Xiao Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100190, P.R. China [email protected] (Y. Lan) Department of Mathematical Sciences, Tsinghua University, Beijing 100084, P. R. China [email protected] (Y. Wu) School of mathematical seciences, Beijing Normal University, Beijing 100875, P. R. China [email protected](J.Xiao)
Abstract.

The present paper continues the work of [2]. For any symmetrizable generalized Cartan Matrix CC and the corresponding quantum group 𝐔\mathbf{U}, we consider the associated quiver QQ with an admissible automorphism aa. We construct the category 𝒬/𝒩~\widetilde{\mathcal{Q}/\mathcal{N}} of the localization of Lusztig sheaves for the quiver with the automorphism. Its Grothendieck group gives a realization of the integrable highest weight 𝐔\mathbf{U}-module Λλ\Lambda_{\lambda}, and modulo the traceless ones Lusztig sheaves provide the (signed) canonical basis of Λλ\Lambda_{\lambda}. As an application, the symmetrizable crystal structures on Nakajima’s quiver varieties and Lusztig’s nilpotent varieties of preprojective algebras are deduced.

2000 Mathematics Subject Classification:
16G20, 17B37

1. Introduction

1.1. Symmetrizable Cartan data and quantum groups

Given a symmetrizable generalized Cartan matrix C=(cij)i,jI=DBC=(c_{ij})_{i,j\in I^{\prime}}=DB, where D=𝐝𝐢𝐚𝐠(1si)iID=\mathbf{diag}(\frac{1}{s_{i}})_{i\in I^{\prime}} is a diagonal matrix and B=(bij)i,jIB=(b_{ij})_{i,j\in I^{\prime}} is a symmetric matrix. It determines a bilinear form ,:[I]×[I]\langle,\rangle:{\mathbb{Z}}[I^{\prime}]\times{\mathbb{Z}}[I^{\prime}]\rightarrow{\mathbb{Z}} by i,j=cij\langle i,j\rangle=c_{ij}. Then the quantum group 𝐔\mathbf{U} is the (v)\mathbb{Q}(v)-algebra generated by Ei,Fi,KνE_{i},F_{i},K_{\nu} for iI,ν[I]i\in I^{\prime},\nu\in{\mathbb{Z}}[I^{\prime}], subject to the following relations:

  • (a) K0=1K_{0}=1, and KνKν=Kν+νK_{\nu}K_{\nu^{\prime}}=K_{\nu+\nu^{\prime}} for any ν,ν[I]\nu,\nu^{\prime}\in{\mathbb{Z}}[I^{\prime}];

  • (b) KνEi=vν,iEiKνK_{\nu}E_{i}=v^{\langle\nu,i\rangle}E_{i}K_{\nu} for iI,ν[I]i\in I^{\prime},\nu\in{\mathbb{Z}}[I^{\prime}];

  • (c) KνFi=vν,iFiKνK_{\nu}F_{i}=v^{-\langle\nu,i\rangle}F_{i}K_{\nu} for iI,ν[I]i\in I^{\prime},\nu\in{\mathbb{Z}}[I^{\prime}];

  • (d) EiFjFjEi=δijK~iK~ivivi1,E_{i}F_{j}-F_{j}E_{i}=\delta_{ij}\frac{\tilde{K}_{i}-\tilde{K}_{-i}}{v_{i}-v_{i}^{-1}},

  • (e) p+q=1cij(1)pEi(p)EjEi(q)=0\sum\limits_{p+q=1-c_{ij}}(-1)^{p}E_{i}^{(p)}E_{j}E_{i}^{(q)}=0;

  • (f) p+q=1cij(1)pFi(p)FjFi(q)=0\sum\limits_{p+q=1-c_{ij}}(-1)^{p}F_{i}^{(p)}F_{j}F_{i}^{(q)}=0.

Here we use the notation vi:=vsiv_{i}:=v^{s_{i}}. For ν=iIνii[I]\nu=\sum\limits_{i\in I^{\prime}}\nu_{i}i\in{\mathbb{Z}}[I^{\prime}], we denote Kν~:=iIKsiνii\tilde{K_{\nu}}:=\prod\limits_{i\in I^{\prime}}K_{s_{i}\nu_{i}i}, and Ei(n):=Ein/[n]i!E_{i}^{(n)}:=E_{i}^{n}/[n]^{!}_{i}, Fi(n):=Fin/[n]i!F_{i}^{(n)}:=F_{i}^{n}/[n]^{!}_{i} are the divided power of Ei,FiE_{i},F_{i}, where [n]i:=vinvinvivi1[n]_{i}:=\frac{v_{i}^{n}-v_{i}^{-n}}{v_{i}-v_{i}^{-1}}, [n]i!:=Πs=1n[s]i[n]^{!}_{i}:=\Pi_{s=1}^{n}[s]_{i}.

Let MM be a 𝐔\mathbf{U}-module with a decomposition M=λ[I]MλM=\oplus_{\lambda\in{\mathbb{Z}}[I^{\prime}]}M^{\lambda}, such that for any mMλm\in M^{\lambda}, Kνm=vν,λmK_{\nu}m=v^{\langle\nu,\lambda\rangle}m. We say MM is integrable, if for any mMm\in M, there exists n0n_{0}\in{\mathbb{N}} such that for any n>n0n>n_{0}, Ei(n)m=0E_{i}^{(n)}m=0 and Fi(n)m=0F_{i}^{(n)}m=0. One can see details in [3] and [6].

Given a dominant weight λ\lambda (i.e. i,λ\langle i,\lambda\rangle\in{\mathbb{N}} for any iIi\in I^{\prime}), it is well-known that irreducible integrable highest weight module Λλ\Lambda_{\lambda} is defined by

Λλ:=𝐔/(iI𝐔Ei+ν[I]𝐔(Kνvν,λ)+iI𝐔Fii,λ+1).\Lambda_{\lambda}:=\mathbf{U}/\left(\sum\limits_{i\in I^{\prime}}\mathbf{U}E_{i}+\sum\limits_{\nu\in{\mathbb{Z}}[I^{\prime}]}\mathbf{U}(K_{\nu}-v^{\langle\nu,\lambda\rangle})+\sum\limits_{i\in I^{\prime}}\mathbf{U}F_{i}^{\langle i,\lambda\rangle+1}\right).

1.2. Lusztig’s sheaves for quivers with automophisms

In [7, Chapter 11, 12 and 14], Lusztig consider a finite quiver Q=(I,H,Ω)Q=(I,H,\Omega) with admissible automorphism aa for a given Cartan matrix.

A finite quiver Q=(I,H,Ω)Q=(I,H,\Omega) consists of finite sets I,HI,H and Ω\Omega, where II is the set of vertices, HH is the set of all oriented arrows s(h)t(h)s(h)\xrightarrow{h}t(h) and ΩH\Omega\subset H is a subset such that HH is the disjoint union of Ω\Omega and Ω¯\bar{\Omega}. Here ¯:HH\bar{}:H\rightarrow H is the map taking the opposite orientation. We call such Ω\Omega an orientation of QQ.

For a finite quiver (I,H,Ω)(I,H,\Omega) without loops, an admissible automorphism aa consists of two permutations a:IIa:I\rightarrow I and a:HHa:H\rightarrow H such that
(a) a(s(h))=s(a(h))a(s(h))=s(a(h)) and a(t(h))=t(a(h))a(t(h))=t(a(h)) for any hHh\in H;
(b) s(h)s(h) and t(h)It(h)\in I belong to different aa-orbits for any hHh\in H.

Given a symmetrizable generalized Cartan matrix C=(cij)i,jI=DBC=(c_{ij})_{i,j\in I^{\prime}}=DB, there is (not unique in general) a finite quiver Q=(I,H,Ω)Q=(I,H,\Omega) with addmissible automorphism aa such that the aa-orbits of II are bijective to II^{\prime}, the order of an aa-orbit corresponding to iIi\in I^{\prime} equals to sis_{i}, and the number of arrows between aa-orbits corresponding to i,jIi,j\in I^{\prime} equals to ci,jsi=cj,isjc_{i,j}s_{i}=c_{j,i}s_{j}. For such a quiver with automorphism, Lusztig has considered a full subcategory 𝒬=𝐕𝒬𝐕\mathcal{Q}=\coprod\limits_{\mathbf{V}}\mathcal{Q}_{\mathbf{V}} of perverse sheaves on the moduli space 𝐕𝐄𝐕,Ω\coprod\limits_{\mathbf{V}}\mathbf{E}_{\mathbf{V},\Omega} of quiver representations. Via the periodic functor aa^{\ast}, Lusztig has constructed a certain graded additive category 𝒬~\tilde{\mathcal{Q}} from 𝒬\mathcal{Q}, consisting of pairs (B,ϕ)(B,\phi) of semisimple complexes BB and morphisms ϕ:aBB\phi:a^{\ast}B\rightarrow B. He has also defined the induction functor and restriction functors for the category 𝒬~\tilde{\mathcal{Q}}. (See details in [7, Chapter 9 and 12] or Section 2.1.) The (generalized) Grothendieck group of 𝒬~\tilde{\mathcal{Q}} has a bialgebra structure, which is canonically isomorphic to the positive (or negative) part 𝐔±\mathbf{U}^{\pm} of 𝐔\mathbf{U}. Moreover, certain objects of 𝒬~\tilde{\mathcal{Q}} provide an integral basis of 𝐔±\mathbf{U}^{\pm}, which is called the signed basis by Lusztig.

When the generalized Cartan matrix CC is symmetric, or equivalently, the admissible automorphism aa is trivial, the simple perverse sheaves in 𝒬\mathcal{Q} form a basis of 𝐔±\mathbf{U}^{\pm}, which is called the canonical basis and shares many remarkable properties.

1.3. Localizations of Lusztig’s sheaves for symmetric cases

When the generalized Cartan matrix CC is symmetric, the authors of [2] have considered certain localizations of Lusztig’s category 𝒬\mathcal{Q} to realize the irreducible integrable highest weight module Λλ\Lambda_{\lambda}.

For a dominant weight λ\lambda, they have considered the moduli space 𝐄𝐕,𝐖,Ω\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega} of a framed quiver and certain thick subcategory 𝒩𝐕\mathcal{N}_{\mathbf{V}} of the G𝐕G_{\mathbf{V}}-equivariant derived category 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) of constructible ¯l\overline{\mathbb{Q}}_{l}-sheaves. They have defined functors i\mathcal{E}_{i} and i,iI\mathcal{F}_{i},i\in I for the Verdier quotient 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/\mathcal{N}_{\mathbf{V}} and showed that Lusztig’s sheaves 𝒬𝐕,𝐖\mathcal{Q}_{\mathbf{V},\mathbf{W}} for the framed quiver are preserved by these functors i\mathcal{E}_{i} and i\mathcal{F}_{i} (up to isomorphisms in localizations). (See details in [2] or Section 2.2.) The Grothendieck group of 𝐕𝒬𝐕,𝐖/𝒩𝐕\coprod\limits_{\mathbf{V}}\mathcal{Q}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}} becomes a 𝐔\mathbf{U}-module with the action of functors i\mathcal{E}_{i} and i\mathcal{F}_{i}, and is isomorphic to Λλ\Lambda_{\lambda} canonically. Moreover, the nonzero simple perverse sheaves in 𝒬𝐕,𝐖/𝒩𝐕\mathcal{Q}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}} form the canonical basis of Λλ\Lambda_{\lambda}.

1.4. Localizations of Lusztig’s sheaves for quiver with automorphisms

It is natural to ask how to generalize the construction of [2] to symmetrizable cases. More precisely, given a symmetrizable generalized Cartan matrix CC, our goal is to realize the irreducible highest weight module Λλ\Lambda_{\lambda} of 𝐔\mathbf{U} associated with CC via sheaves on moduli spaces of a quiver with automorphism.

In this article, we apply Lusztig’s method of periodic functors to localizations, then we obtain a category 𝒬/𝒩~\widetilde{\mathcal{Q}/\mathcal{N}}. Its Grothendieck group is a 𝐔\mathbf{U}-module, which is canonically isomorphic to Λλ\Lambda_{\lambda} of the symmetrizable quantum group associated to CC. See Theorem 4.6 and 4.10 for details.

Moreover, the signed basis of Λλ\Lambda_{\lambda} is naturally provided by those (B,ϕ)(B,\phi) in 𝒬/𝒩~\widetilde{\mathcal{Q}/\mathcal{N}} under the canonical isomorphism, which is almost orthogonal with respect to a contravariant geometric pairing. See Proposition 4.8 for details.

As an application, we also deduce a symmetrizable crystal structure on Nakajima’s quiver varieties and Lusztig’s nilpotent varieties of preprojective algebras. See Theorem 4.15 and its corollary for details.

1.5. Structure of the article

In section 2, we introduce the notations and results of [2]. In section 3, we define our category 𝒬/𝒩~\widetilde{\mathcal{Q}/\mathcal{N}} and its functors, and prove the commutative relations of these functors. There is a great discrepancy between the proofs of commutative relations in symmetric cases and those in symmetrizable cases. In section 4, we determine the module structure of the Grothendieck group of 𝒬/𝒩~\widetilde{\mathcal{Q}/\mathcal{N}}, construct the signed basis of Λλ\Lambda_{\lambda} and deduce the crystal structure on quiver varieties and nilpotent varieties.

Acknowledgement

Y. Lan is partially supported by the National Natural Science Foundation of China [Grant No. 1288201], Yumeng Wu and J. Xiao are partially supported by Natural Science Foundation of China [Grant No. 12031007]. This paper is a continuation of a collaborative work [2] with Jiepeng Fang, we are very grateful to his detailed discussion.

2. The category 𝒬𝐕,𝐖/𝒩𝐕\mathcal{Q}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}} and functors i(n),i(n)\mathcal{E}^{(n)}_{i},\mathcal{F}^{(n)}_{i}

2.1. Lusztig’s sheaves

In this subsection, we introduce the category of Lusztig’s sheaves for quivers. Assume 𝐤\mathbf{k} is a algebraic closed field with char(𝐤)=p>0char(\mathbf{k})=p>0. Given a quiver Q=(I,H,Ω)Q=(I,H,\Omega) and an II-graded 𝐤\mathbf{k}-space 𝐕=iI𝐕i\mathbf{V}=\bigoplus\limits_{i\in I}\mathbf{V}_{i} of dimension vector ν\nu, the affine variety 𝐄𝐕,Ω\mathbf{E}_{\mathbf{V},\Omega} is defined by

𝐄𝐕,Ω=hΩHom(𝐕s(h),𝐕t(h)).\mathbf{E}_{\mathbf{V},\Omega}=\bigoplus\limits_{h\in\Omega}\mbox{\rm Hom}(\mathbf{V}_{s(h)},\mathbf{V}_{t(h)}).

The algebraic group G𝐕=iIGL(𝐕i)G_{\mathbf{V}}=\prod\limits_{i\in I}GL(\mathbf{V}_{i}) acts on 𝐄𝐕,Ω\mathbf{E}_{\mathbf{V},\Omega} by composition naturally. We denote the G𝐕G_{\mathbf{V}}-equivariant derived category of constructible ¯l\overline{{\mathbb{Q}}}_{l}-sheaves on 𝐄𝐕,Ω\mathbf{E}_{\mathbf{V},\Omega} by 𝒟G𝐕b(𝐄𝐕,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\Omega}). We denote [n][n] the nn-times shift functor and 𝐃\mathbf{D} the Verdier duality.

Let 𝒮\mathcal{S} be the set of finite sequences 𝝂=(ν1,ν2,,νs)\boldsymbol{\nu}=(\nu^{1},\nu^{2},\cdots,\nu^{s}) of dimension vectors such that each νl=alil\nu^{l}=a_{l}i_{l} for some al1a_{l}\in{\mathbb{N}}_{\geqslant 1} and ilIi_{l}\in I. If 1lsνl=ν\sum\limits_{1\leqslant l\leqslant s}\nu^{l}=\nu, we say 𝝂\boldsymbol{\nu} is a flag type of ν\nu or 𝐕\mathbf{V}.

For a flag type 𝝂\boldsymbol{\nu} of 𝐕\mathbf{V}, the flag variety 𝝂,Ω\mathcal{F}_{\boldsymbol{\nu},\Omega} is the smooth variety which consists of pairs (x,f)(x,f), where x𝐄𝐕,Ωx\in\mathbf{E}_{\mathbf{V},\Omega} and f=(𝐕=𝐕s𝐕s1𝐕0=0)f=(\mathbf{V}=\mathbf{V}^{s}\subseteq\mathbf{V}^{s-1}\subseteq\cdots\subseteq\mathbf{V}^{0}=0) is a filtration of II-graded space such that x(𝐕l)𝐕lx(\mathbf{V}^{l})\subseteq\mathbf{V}^{l} and the dimension vector of 𝐕l1/𝐕l=νl\mathbf{V}^{l-1}/\mathbf{V}^{l}=\nu^{l} for any ll. There is a proper map π𝝂:𝝂,Ω𝐄𝐕,Ω;(x,f)x.\pi_{\boldsymbol{\nu}}:\mathcal{F}_{\boldsymbol{\nu},\Omega}\rightarrow\mathbf{E}_{\mathbf{V},\Omega};(x,f)\mapsto x. Hence by the decomposition theorem in [1], the complex L𝝂=(π𝝂,Ω)!¯l[dim𝝂,Ω]L_{\boldsymbol{\nu}}=(\pi_{\boldsymbol{\nu},\Omega})_{!}\bar{\mathbb{Q}}_{l}[\mbox{\rm dim}\mathcal{F}_{\boldsymbol{\nu},\Omega}] is a semisimple complex on 𝐄𝐕,Ω\mathbf{E}_{\mathbf{V},\Omega}, where ¯l\bar{\mathbb{Q}}_{l} is the constant sheaf on 𝝂,Ω\mathcal{F}_{\boldsymbol{\nu},\Omega}.

Definition 2.1.

Let 𝒫𝐕\mathcal{P}_{\mathbf{V}} be the set consisting of those simple perverse sheaves LL in 𝒟G𝐕b(𝐄𝐕,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\Omega}) such that LL is a direct summand (up to shifts) of L𝛎L_{\boldsymbol{\nu}} for some flag type 𝛎\boldsymbol{\nu} of 𝐕\mathbf{V}. Let 𝒬𝐕{\mathcal{Q}}_{\mathbf{V}} be the full subcategory consisting of 𝒟G𝐕b(𝐄𝐕,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\Omega}), which consists of direct sums of shifted simple perverse sheaves in 𝒫𝐕\mathcal{P}_{\mathbf{V}}. Following [12], we call 𝒬𝐕{\mathcal{Q}}_{\mathbf{V}} the category of Lusztig’s sheaves for QQ.

The framed quiver Q~=(II,H~,Ω~)\tilde{Q}=(I\cup I^{*},\tilde{H},\tilde{\Omega}) of Q=(I,H,Ω)Q=(I,H,\Omega) is defined by letting II^{*} be a copy of II, Ω~\tilde{\Omega} be Ω{ii,iI}\Omega\cup\{i\rightarrow i^{*},i\in I\}, and H~=Ω~Ω~¯\tilde{H}=\tilde{\Omega}\cup\bar{\tilde{\Omega}}. Given II-graded spaces 𝐕\mathbf{V} and 𝐖\mathbf{W}, the moduli space of the framed quiver is defined by 𝐄𝐕,𝐖,Ω:=iIHom(𝐕i,𝐖i)𝐄𝐕,Ω\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}:=\bigoplus\limits_{i\in I}\mbox{\rm Hom}(\mathbf{V}_{i},\mathbf{W}_{i})\oplus\mathbf{E}_{\mathbf{V},\Omega}. The algebraic group G𝐕G_{\mathbf{V}} acts canonically on 𝐄𝐕,𝐖,Ω\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega} and we can consider the G𝐕G_{\mathbf{V}}-equivariant derived category 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}).

Definition 2.2.

We define the category 𝒬𝐕,𝐖{\mathcal{Q}}_{\mathbf{V},\mathbf{W}} of Lusztig’s sheaves for Q~\tilde{Q} to be the full subcategory of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) consisting of objects in (π𝐕,𝐖)(𝒬𝐕)(\pi_{\mathbf{V},\mathbf{W}})^{\ast}({\mathcal{Q}}_{\mathbf{V}}), where π𝐕,𝐖:𝐄𝐕,𝐖,Ω𝐄𝐕,Ω\pi_{\mathbf{V},\mathbf{W}}:\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}\rightarrow\mathbf{E}_{\mathbf{V},\Omega} is the natural projection.

Assume I={i1,i2,,it}I=\{i_{1},i_{2},\cdots,i_{t}\} and dil=dim𝐖ild_{i_{l}}=\mbox{\rm dim}\mathbf{W}_{i_{l}} for each ll, let 𝒅=(di1i1,di2i2,ditit)\boldsymbol{d}=(d_{i_{1}}i_{1},d_{i_{2}}i_{2},\cdots d_{i_{t}}i_{t}) be a flag type of 𝐖\mathbf{W}. For any flag type 𝝂=(ν1,ν2,,νs)\boldsymbol{\nu}=(\nu^{1},\nu^{2},\cdots,\nu^{s}), the flag type

𝝂𝒅=(ν1,ν2,,νs,di1i1,di2i2,ditit)\boldsymbol{\nu}\boldsymbol{d}=(\nu^{1},\nu^{2},\cdots,\nu^{s},d_{i_{1}}i_{1},d_{i_{2}}i_{2},\cdots d_{i_{t}}i_{t})

is a flag type of 𝐕𝐖\mathbf{V}\oplus\mathbf{W}. Then by [2, Proposition 3.1] simple objects in 𝒬𝐕,𝐖{\mathcal{Q}}_{\mathbf{V},\mathbf{W}} are exactly simple direct summands (up to shifts) of those L𝝂𝒅L_{\boldsymbol{\nu}\boldsymbol{d}}, where 𝝂\boldsymbol{\nu} runs over all flag types of 𝐕\mathbf{V}.

2.2. Localizations

In this subsection, we introduce the localization of Lusztig’s sheaves in [2]. Choose an orientation Ωi\Omega_{i} such that iIi\in I is a source in Ωi\Omega_{i}. Let 𝐄𝐕,𝐖,i0\mathbf{E}_{\mathbf{V},\mathbf{W},i}^{0} be the open subset of 𝐄𝐕,𝐖,Ωi\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega_{i}} defined by

𝐄𝐕,𝐖,i0={x𝐄𝐕,𝐖,Ωi|dimker(hΩ~i,s(h)=ixh)=0},\mathbf{E}_{\mathbf{V},\mathbf{W},i}^{0}=\{x\in\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega_{i}}|\mbox{\rm dim}\ker(\bigoplus\limits_{h\in\tilde{\Omega}_{i},s(h)=i}x_{h})=0\},

and 𝐄𝐕,𝐖,i1\mathbf{E}_{\mathbf{V},\mathbf{W},i}^{\geqslant 1} be its complement. Let 𝒩𝐕,i\mathcal{N}_{\mathbf{V},i} be the thick subcategory of 𝒟G𝐕b(𝐄𝐕,𝐖,Ωi)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega_{i}}), consisting of complexes supported on 𝐄𝐕,𝐖,i1\mathbf{E}_{\mathbf{V},\mathbf{W},i}^{\geqslant 1}.

Recall that the Fourier-Deligne transform Ωi,Ω:𝒟G𝐕b(𝐄𝐕,𝐖,Ωi)𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{F}_{\Omega_{i},\Omega}:\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega_{i}})\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) induces derived equivalence between different orientations. (See details in [2, Section 2.2] or [7, Chapter 10].) We can consider the thick subcategory Ωi,Ω(𝒩𝐕,i)\mathcal{F}_{\Omega_{i},\Omega}(\mathcal{N}_{\mathbf{V},i}) of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) for each ii, and they generate a thick subcategory denoted by 𝒩𝐕{\mathcal{N}}_{\mathbf{V}}.

Definition 2.3.

Let 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}} be the Verdier quotient of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) with respect to the thick subcategory 𝒩𝐕{\mathcal{N}}_{\mathbf{V}}, the localization 𝒬𝐕,𝐖/𝒩𝐕{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} is defined to be the full subcategory of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}, which consists of objects isomorphic to those of 𝒬𝐕,𝐖{\mathcal{Q}}_{\mathbf{V},\mathbf{W}} in 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}.

The category 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}} inherits a perverse tt-structure from 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}), and the Verdier Duality 𝐃\mathbf{D} of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) also acts on 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}} and 𝒬𝐕,𝐖/𝒩𝐕{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}.

If we denote the localization functor by L:𝒟G𝐕b(𝐄𝐕,𝐖,Ω)𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕L:\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}, then LL restricts to an additive functor L:𝒬𝐕,𝐖𝒬𝐕,𝐖/𝒩𝐕L:{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}\rightarrow{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}.

2.3. Functors of localizations

In this subsection, we recall the definition of i(n){\mathcal{E}}^{(n)}_{i} and i(n){\mathcal{F}}^{(n)}_{i} in [2].

2.3.1. The functor i(n){\mathcal{E}}^{(n)}_{i}

For ν,ν[I]\nu,\nu^{\prime}\in{\mathbb{N}}[I] such that ν=νni\nu^{\prime}=\nu-ni, take graded spaces 𝐕,𝐕\mathbf{V},\mathbf{V}^{\prime} of dimension vectors ν\nu and ν\nu^{\prime} respectively. We Define 𝐄˙𝐕,𝐖,i=hΩi,s(h)iHom(𝐕s(h),𝐕t(h))ji,jIHom(𝐕j,𝐖j)\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}=\bigoplus\limits_{h\in\Omega_{i},s(h)\not=i}\mbox{\rm Hom}(\mathbf{V}_{s(h)},\mathbf{V}_{t(h)})\oplus\bigoplus\limits_{j\not=i,j\in I}\mbox{\rm Hom}(\mathbf{V}_{j},\mathbf{W}_{j}) and denote νi~=hΩi,s(h)=iνt(h)+dim𝐖i\tilde{\nu_{i}}=\sum\limits_{h\in\Omega_{i},s(h)=i}\nu_{t(h)}+\mbox{\rm dim}\mathbf{W}_{i}. Then we consider the following diagram

𝐄𝐕,𝐖,Ωi\textstyle{\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega_{i}}}𝐄𝐕,𝐖,Ωi\textstyle{\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\Omega_{i}}}𝐄𝐕,𝐖,i0\textstyle{\mathbf{E}^{0}_{\mathbf{V},\mathbf{W},i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕,i\scriptstyle{\phi_{\mathbf{V},i}}j𝐕,i\scriptstyle{j_{\mathbf{V},i}}𝐄𝐕,𝐖,i0\textstyle{\mathbf{E}^{0}_{\mathbf{V}^{\prime},\mathbf{W},i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕,i\scriptstyle{\phi_{\mathbf{V}^{\prime},i}}j𝐕,i\scriptstyle{j_{\mathbf{V}^{\prime},i}}𝐄˙𝐕,𝐖,i×𝐆𝐫(νi,ν~i)\textstyle{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Gr}(\nu_{i},\tilde{\nu}_{i})}𝐄˙𝐕,𝐖,i×𝐅𝐥(νi,νi,ν~i)\textstyle{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Fl}(\nu^{\prime}_{i},\nu_{i},\tilde{\nu}_{i})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q2\scriptstyle{q_{2}}q1\scriptstyle{q_{1}}𝐄˙𝐕,𝐖,i×𝐆𝐫(νi,ν~i);\textstyle{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Gr}(\nu^{\prime}_{i},\tilde{\nu}_{i});}

where 𝐆𝐫(νi,ν~i)\mathbf{Gr}(\nu_{i},\tilde{\nu}_{i}) is the Grassmannian consisting of νi\nu_{i}-dimensional subspaces of ν~i\tilde{\nu}_{i}-dimensional space (s(h)=i,hΩi𝐕t(h))𝐖i(\bigoplus\limits_{s(h)=i,h\in\Omega_{i}}\mathbf{V}_{t(h)})\oplus\mathbf{W}_{i}, and

𝐅𝐥(νin,νi,ν~i)={𝐒1𝐒2(s(h)=i𝐕t(h))𝐖i)|dim𝐒1=νin,dim𝐒2=νi}.\mathbf{Fl}(\nu_{i}-n,\nu_{i},\tilde{\nu}_{i})=\{\mathbf{S}_{1}\subset\mathbf{S}_{2}\subset(\bigoplus\limits_{s(h)=i}\mathbf{V}_{t(h)})\oplus\mathbf{W}_{i})|{\rm{dim}}\mathbf{S}_{1}=\nu_{i}-n,{\rm{dim}}\mathbf{S}_{2}=\nu_{i}\}.

is the flag variety. The morphisms are defined by

ϕ𝐕,i:𝐄𝐕,𝐖,i0𝐄˙𝐕,𝐖,i×𝐆𝐫(νi,ν~i);x(x˙,Im(hΩ~i,s(h)=ixh)),\displaystyle\phi_{\mathbf{V},i}:\mathbf{E}^{0}_{\mathbf{V},\mathbf{W},i}\longrightarrow\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Gr}(\nu_{i},\tilde{\nu}_{i});x\mapsto(\dot{x},{\rm{Im}}(\bigoplus\limits_{h\in\tilde{\Omega}_{i},s(h)=i}x_{h})),

j𝐕,i,j𝐕,ij_{\mathbf{V},i},j_{\mathbf{V}^{\prime},i} are open inclusions and q1,q2q_{1},q_{2} are natural projections

q1(x˙,𝐒1,𝐒2)=(x˙,𝐒2);q_{1}(\dot{x},\mathbf{S}_{1},\mathbf{S}_{2})=(\dot{x},\mathbf{S}_{2});
q2(x˙,𝐒1,𝐒2)=(x˙,𝐒1).q_{2}(\dot{x},\mathbf{S}_{1},\mathbf{S}_{2})=(\dot{x},\mathbf{S}_{1}).
Definition 2.4.

The functor ~i(n):𝒟G𝐕b(𝐄𝐕,𝐖,Ωi)𝒟G𝐕b(𝐄𝐕,𝐖,Ωi)\tilde{{\mathcal{E}}}^{(n)}_{i}:\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega_{i}})\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}^{\prime}}}(\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\Omega_{i}}) is defined by

~i(n):=(j𝐕,i)!(ϕ𝐕,i)(q2)!(q1)(ϕ𝐕,i)(j𝐕,i)[nνi],\tilde{{\mathcal{E}}}^{(n)}_{i}:=(j_{\mathbf{V}^{\prime},i})_{!}(\phi_{\mathbf{V}^{\prime},i})^{*}(q_{2})_{!}(q_{1})^{*}(\phi_{\mathbf{V},i})_{\flat}(j_{\mathbf{V},i})^{*}[-n\nu_{i}],

where ff_{\flat} means the inverse of ff^{\ast} for any principle bundle ff. We also denote i(1){\mathcal{E}}^{(1)}_{i} by i{\mathcal{E}}_{i}.

For a general orientation Ω\Omega, the functor i(n):𝒟G𝐕b(𝐄𝐕,𝐖,Ω)𝒟G𝐕b(𝐄𝐕,𝐖,Ω){\mathcal{E}}^{(n)}_{i}:\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}^{\prime}}}(\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\Omega}) is defined by

i(n)=Ωi,Ω~i(n)Ω,Ωi.{\mathcal{E}}^{(n)}_{i}={\mathcal{F}}_{\Omega_{i},\Omega}\tilde{{\mathcal{E}}}^{(n)}_{i}{\mathcal{F}}_{\Omega,\Omega_{i}}.

By [2, Proposition 3.18 and Corollary 3.19], the functor i(n){\mathcal{E}}^{(n)}_{i} induces a functor

i(n):𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕,{\mathcal{E}}^{(n)}_{i}:\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}^{\prime}}}(\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}^{\prime}},

and restricts to a functor i(n):𝒬𝐕,𝐖/𝒩𝐕𝒬𝐕,𝐖/𝒩𝐕.{\mathcal{E}}^{(n)}_{i}:{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}\rightarrow{\mathcal{Q}}_{\mathbf{V}^{\prime},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}^{\prime}}. Notice that i(n){\mathcal{E}}^{(n)}_{i} is independent of the choice of Ωi\Omega_{i} and Ω\Omega up to Fourier-Deligne transforms. (One can see a proof in [2, Lemma 3.6].)

2.3.2. The functor i(n){\mathcal{F}}^{(n)}_{i}

For graded spaces 𝐕,𝐕′′\mathbf{V},\mathbf{V}^{\prime\prime} of dimension vectors ν,ν′′\nu,\nu^{\prime\prime} respectively, such that ni+ν′′=νni+\nu^{\prime\prime}=\nu.

Let 𝐄𝐕,𝐖,Ω\mathbf{E}^{\prime}_{\mathbf{V},\mathbf{W},\Omega} be the variety consisting of (x,𝐒,ρ)(x,\mathbf{S},\rho), where x𝐄𝐕,𝐖,Ωx\in\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega} and 𝐒\mathbf{S} is an subspace of 𝐕𝐖\mathbf{V}\oplus\mathbf{W} such that x(𝐒)𝐒x(\mathbf{S})\subseteq\mathbf{S}, and ρ:𝐒𝐕′′𝐖\rho:\mathbf{S}\simeq\mathbf{V}^{\prime\prime}\oplus\mathbf{W} is a linear isomorphism of graded spaces such that ρ|𝐖=id𝐖\rho|_{\mathbf{W}}=id_{\mathbf{W}}. Notice that it is equivalent to forget 𝐖\mathbf{W} and just take a linear isomorphism 𝐒𝐕𝐕′′\mathbf{S}\cap\mathbf{V}\simeq\mathbf{V}^{\prime\prime}, hence sometimes we also say ρ\rho is a linear isomorphism between 𝐒\mathbf{S} and 𝐕′′\mathbf{V}^{\prime\prime}. Let 𝐄𝐕,𝐖,Ω′′\mathbf{E}^{\prime\prime}_{\mathbf{V},\mathbf{W},\Omega} be the variety consisting of (x,𝐒)(x,\mathbf{S}) with the same conditions as above. Consider the following diagram

𝐄𝐕′′,𝐖,Ωp1𝐄𝐕,𝐖,Ωp2𝐄𝐕,𝐖,Ω′′p3𝐄𝐕,𝐖,Ω,\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\Omega}\xleftarrow{p_{1}}\mathbf{E}^{\prime}_{\mathbf{V},\mathbf{W},\Omega}\xrightarrow{p_{2}}\mathbf{E}^{\prime\prime}_{\mathbf{V},\mathbf{W},\Omega}\xrightarrow{p_{3}}\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega},

where the morphisms are defined by

p1(x,𝐒,ρ)=(ρ(x|𝐒));p2(x,𝐒,ρ)=(x,𝐒);p3(x,𝐒)=x.p_{1}(x,\mathbf{S},\rho)=(\rho_{\ast}(x|_{\mathbf{S}}));~{}p_{2}(x,\mathbf{S},\rho)=(x,\mathbf{S});~{}p_{3}(x,\mathbf{S})=x.

Notice that p1p_{1} is smooth with connected fiber, p2p_{2} is a principle G𝐕′′G_{\mathbf{V}^{\prime\prime}} bundle and p3p_{3} is proper.

Definition 2.5.

The functor i(n):𝒟G𝐕′′b(𝐄𝐕′′,𝐖,Ω)𝒟G𝐕b(𝐄𝐕,𝐖,Ω){\mathcal{F}}^{(n)}_{i}:\mathcal{D}^{b}_{G_{\mathbf{V}^{\prime\prime}}}(\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\Omega})\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) is defined by

i(n)=(p3)!(p2)(p1)[d1d2],{\mathcal{F}}^{(n)}_{i}=(p_{3})_{!}(p_{2})_{\flat}(p_{1})^{\ast}[d_{1}-d_{2}],

where d1d_{1} and d2d_{2} are the relative dimensions of p1p_{1} and p2p_{2} respectively. We also denote i(1){\mathcal{F}}^{(1)}_{i} by i{\mathcal{F}}_{i}.

Notice that i(n){\mathcal{F}}^{(n)}_{i} is isomorphic to Lusztig’s induction functor 𝐈𝐧𝐝ni,𝐕′′𝐖𝐕𝐖(¯l)\mathbf{Ind}^{\mathbf{V}\oplus\mathbf{W}}_{ni,\mathbf{V^{\prime\prime}\oplus\mathbf{W}}}(\overline{\mathbb{Q}}_{l}\boxtimes-) (See [2, Section 3]), so for any flag type 𝝂\boldsymbol{\nu},

(1) i(n)(L𝝂𝒅)=L(ni,𝝂𝒅).{\mathcal{F}}^{(n)}_{i}(L_{\boldsymbol{\nu}\boldsymbol{d}})=L_{(ni,\boldsymbol{\nu}\boldsymbol{d})}.

By [2, Lemma 3.7], the functor i(n){\mathcal{F}}^{(n)}_{i} induces a functor

i(n):𝒟G𝐕′′b(𝐄𝐕′′,𝐖,Ω)/𝒩𝐕′′𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕,{\mathcal{F}}^{(n)}_{i}:\mathcal{D}^{b}_{G_{\mathbf{V}^{\prime\prime}}}(\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}^{\prime\prime}}\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}},

and restricts to a functor i(n):𝒬𝐕′′,𝐖/𝒩𝐕′′𝒬𝐕,𝐖/𝒩𝐕.{\mathcal{F}}^{(n)}_{i}:{\mathcal{Q}}_{\mathbf{V}^{\prime\prime},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}^{\prime\prime}}\rightarrow{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}.

3. The category 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}} and functors i¯(n),i¯(n)\mathcal{E}^{(n)}_{\underline{i}},\mathcal{F}^{(n)}_{\underline{i}}

3.1. Periodic functor

In this subsection, we review the definition of the periodic functor and refer [7, Chapter 11] for more details. Let oo be a fixed positive integer.

Let 𝒞{\mathcal{C}} be a ¯l\overline{{\mathbb{Q}}}_{l}-linear additive category, a periodic functor on 𝒞{\mathcal{C}} is a linear functor a:𝒞𝒞a^{*}:{\mathcal{C}}\rightarrow{\mathcal{C}} such that (a)o(a^{*})^{o} is the identity functor on 𝒞{\mathcal{C}}.

Definition 3.1.

Let aa^{*} be a periodic functor on 𝒞{\mathcal{C}}, we define the additive category 𝒞~\tilde{{\mathcal{C}}} as the following:
\bullet Its objects are pairs (A,φ)(A,\varphi), where A𝒞A\in{\mathcal{C}} and ϕ:a(A)A\phi:a^{*}(A)\rightarrow A is an isomorphism in 𝒞{\mathcal{C}} such that the composition

A=an(A)a(n1)(φ)a(n1)(A).a(A)𝜑AA=a^{*n}(A)\xrightarrow{a^{*(n-1)(\varphi)}}a^{*(n-1)}(A)\rightarrow....\rightarrow a^{*}(A)\xrightarrow{\varphi}A

is the identity morphism on AA.
\bullet For any (A,φ),(A,φ)𝒞~(A,\varphi),(A^{\prime},\varphi^{\prime})\in\tilde{{\mathcal{C}}}, the morphism space

Hom𝒞~((A,φ),(A,φ))={fHom𝒞(A,A)|fφ=φ(a(f))}.\mbox{\rm Hom}_{\tilde{{\mathcal{C}}}}((A,\varphi),(A^{\prime},\varphi^{\prime}))=\{f\in\mbox{\rm Hom}_{{\mathcal{C}}}(A,A^{\prime})|f\varphi=\varphi^{\prime}(a^{*}(f))\}.

\bullet The direct sum of (A,φ),(A,φ)𝒞~(A,\varphi),(A^{\prime},\varphi^{\prime})\in\tilde{{\mathcal{C}}} is defined naturally by (AA,φφ)(A\oplus A^{\prime},\varphi\oplus\varphi^{\prime}).

Definition 3.2.

An object (A,ϕ)𝒞~(A,\phi)\in\tilde{{\mathcal{C}}} is called traceless, if there exists an object B𝒞B\in{\mathcal{C}} and an integer t2t\geqslant 2 dividing oo such that at(B)Ba^{*t}(B)\cong B, ABa(B)a(t1)(B)A\cong B\oplus a^{*}(B)\oplus...\oplus a^{*(t-1)}(B) and φ:a(A)A\varphi:a^{*}(A)\rightarrow A corresponds to the isomorphism a(B)a2(B)at(B)a^{*}(B)\oplus a^{*2}(B)\oplus...\oplus a^{*t}(B) taking as(B)a^{*s}(B) onto a(as1(B))a^{*}(a^{*s-1}(B)) for 1st11\leqslant s\leqslant t-1 and the taking at(B)a^{*t}(B) onto BB, giving a permutation between the direct summands of AA and aAa^{*}A.

Definition 3.3.

We say AA and BB in 𝒞~\tilde{{\mathcal{C}}} are isomorphic up to traceless elements if there exist traceless objects CC and DD such that ACBD.A\oplus C\cong B\oplus D.

Lemma 3.4 ([7, Section 11.1.3]).

Let (A,φ),(A,φ),(A′′,φ′′)(A,\varphi),(A^{\prime},\varphi^{\prime}),(A^{\prime\prime},\varphi^{\prime\prime}) be objects in 𝒞~\tilde{{\mathcal{C}}} and i:(A,φ)(A,φ),p′′:(A,φ)(A′′,φ′′)i^{\prime}:(A^{\prime},\varphi^{\prime})\rightarrow(A,\varphi),\ p^{\prime\prime}:(A,\varphi)\rightarrow(A^{\prime\prime},\varphi^{\prime\prime}) be morphisms in 𝒞~\tilde{{\mathcal{C}}}, if there exists morphisms i′′:A′′A,p:AAi^{\prime\prime}:A^{\prime\prime}\rightarrow A,\ p^{\prime}:A\rightarrow A^{\prime} in 𝒞{\mathcal{C}} such that

pi=1A,pi′′=0,p′′i=0,p′′i′′=1A′′,ip+i′′p′′=1A,p^{\prime}i^{\prime}=1_{A^{\prime}},\ p^{\prime}i^{\prime\prime}=0,\ p^{\prime\prime}i^{\prime}=0,\ p^{\prime\prime}i^{\prime\prime}=1_{A^{\prime\prime}},\ i^{\prime}p^{\prime}+i^{\prime\prime}p^{\prime\prime}=1_{A},

then (A,φ)(A,φ)(A′′,φ′′)(A,\varphi)\cong(A^{\prime},\varphi^{\prime})\oplus(A^{\prime\prime},\varphi^{\prime\prime}) in 𝒞~\tilde{{\mathcal{C}}}.

3.2. Localization and its functors for a quiver with automorphism

For a given generalized Cartan matrix C=DBC=DB, we construct a finite quiver Q=(I,H,Ω)Q=(I,H,\Omega) with an admissible automorphism aa by using [7, Proposition 14.1.2] and let oo be the order of aa. The automorphism aa induces naturally an admissible automorphism of the framed quiver Q~\tilde{Q}. We denote the set of aa-orbits of II by I¯=I/a\underline{I}=I/\langle a\rangle, and denote the set of aa-invariant dimension vectors by Ia{\mathbb{N}}I^{a}. Then any i¯I¯\underline{i}\in\underline{I} determines a dimension vector i¯=ii¯i\underline{i}=\sum\limits_{i\in\underline{i}}i in Ia{\mathbb{N}}I^{a}.

Let 𝒮\mathcal{S}^{\prime} be the set of finite sequences 𝝂=(ν1,ν2,,νs)\boldsymbol{\nu}=(\nu^{1},\nu^{2},\cdots,\nu^{s}) of dimension vectors such that each νl=ali¯l\nu^{l}=a_{l}\underline{i}_{l} for some al1a_{l}\in{\mathbb{N}}_{\geqslant 1} and i¯lI¯Ia\underline{i}_{l}\in\underline{I}\subset{\mathbb{N}}I^{a}. For νIa\nu\in{\mathbb{N}}I^{a}, we say 𝝂𝒮\boldsymbol{\nu}\in\mathcal{S}^{\prime} is an aa-flag type of ν\nu if 1lsνl=ν\sum\limits_{1\leqslant l\leqslant s}\nu^{l}=\nu. We can also define semisimple complex L𝝂L_{\boldsymbol{\nu}} for 𝝂𝒮\boldsymbol{\nu}\in\mathcal{S}^{\prime} like Section 2.1.

Given a II-graded space 𝐖\mathbf{W} with dimension vectors ωIa\omega\in{\mathbb{N}}I^{a}, we assume ω=i¯I¯ei¯i¯\omega=\sum\limits_{\underline{i}\in\underline{I}}e_{\underline{i}}\underline{i}. Assume I={i¯1,i¯2,,i¯m}I^{\prime}=\{\underline{i}_{1},\underline{i}_{2},\cdots,\underline{i}_{m}\}, we fix an aa-flag type 𝒆=(ei¯1i¯1,ei¯2i¯2,,ei¯mi¯m)\boldsymbol{e}=(e_{\underline{i}_{1}}\underline{i}_{1},e_{\underline{i}_{2}}\underline{i}_{2},\cdots,e_{\underline{i}_{m}}\underline{i}_{m}). If 𝝂\boldsymbol{\nu} in 𝒮\mathcal{S}^{\prime} is an aa-flag type of 𝐕\mathbf{V}, then 𝝂𝒆\boldsymbol{\nu}\boldsymbol{e} is an aa-flag type of 𝐕𝐖\mathbf{V}\oplus\bf W.

Definition 3.5.

Take an II-graded space 𝐕\mathbf{V} with dimension vectors νIa\nu\in{\mathbb{N}}I^{a}.
(1) Let 𝒫𝐕,𝐖a\mathcal{P}^{a}_{\mathbf{V},\mathbf{W}} be the set consisting of those simple perverse sheaves LL in 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}) such that LL is a direct summand (up to shifts) of L𝝂𝒆L_{\boldsymbol{\nu}\boldsymbol{e}} for some aa-flag type 𝝂\boldsymbol{\nu} of 𝐕\mathbf{V}. Let 𝒬𝐕,𝐖a{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}} be the full subcategory consisting of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}), which consists of direct sums of shifted simple perverse sheaves in 𝒫𝐕,𝐖a\mathcal{P}^{a}_{\mathbf{V},\mathbf{W}}.
(2) The localization 𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} is defined to be the full subcategory of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}, which consists of objects isomorphic to those of 𝒬𝐕,𝐖a{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}} in 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}.

Definition 3.6.

For i¯I¯\underline{i}\in\underline{I} and II-graded spaces 𝐕,𝐕\mathbf{V},\mathbf{V}^{\prime} with dimension vectors ν,ν=νni¯\nu,\nu^{\prime}=\nu-n\underline{i}, the functor i¯(n):𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕{\mathcal{E}}^{(n)}_{\underline{i}}:\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}^{\prime}}}(\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}^{\prime}} is defined by

i¯(n)=ii¯i(n).{\mathcal{E}}^{(n)}_{\underline{i}}=\prod\limits_{i\in\underline{i}}{\mathcal{E}}^{(n)}_{i}.
Definition 3.7.

For i¯I¯\underline{i}\in\underline{I} and II-graded spaces 𝐕,𝐕′′\mathbf{V},\mathbf{V}^{\prime\prime} with dimension vectors ν,ν′′=νni¯\nu,\nu^{\prime\prime}=\nu-n\underline{i}, the functor i¯(n):𝒟G𝐕′′b(𝐄𝐕′′,𝐖,Ω)/𝒩𝐕′′𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕{\mathcal{F}}^{(n)}_{\underline{i}}:\mathcal{D}^{b}_{G_{\mathbf{V}^{\prime\prime}}}(\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}^{\prime\prime}}\rightarrow\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}} is defined by

i¯(n)=ii¯i(n).{\mathcal{F}}^{(n)}_{\underline{i}}=\prod\limits_{i\in\underline{i}}{\mathcal{F}}^{(n)}_{i}.

Since the automorphism aa is addmissible, there are no arrows between vertices in i¯\underline{i}, hence the functors i(n){\mathcal{E}}^{(n)}_{i} (or i(n){\mathcal{F}}^{(n)}_{i}), iI¯i\in\underline{I} commute with each other. The compositions of functors above are well-defined.

Proposition 3.8.

The functors i¯(n){\mathcal{F}}^{(n)}_{\underline{i}} and i¯(n){\mathcal{E}}^{(n)}_{\underline{i}} restrict to functors between 𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} and 𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} for 𝐕\mathbf{V}, 𝐕′′\mathbf{V}^{\prime\prime} with dimension vectors ν,ν′′=νni¯\nu,\nu^{\prime\prime}=\nu-n\underline{i},

i¯(n):𝒬𝐕,𝐖a/𝒩𝐕𝒬𝐕,𝐖a/𝒩𝐕,{\mathcal{E}}^{(n)}_{\underline{i}}:{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}\rightarrow{\mathcal{Q}}^{a}_{\mathbf{V}^{\prime},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}^{\prime}},
i¯(n):𝒬𝐕′′,𝐖a/𝒩𝐕′′𝒬𝐕,𝐖a/𝒩𝐕.{\mathcal{F}}^{(n)}_{\underline{i}}:{\mathcal{Q}}^{a}_{\mathbf{V}^{\prime\prime},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}^{\prime\prime}}\rightarrow{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}.
Proof.

It suffices to show that for any 𝝂𝒮\boldsymbol{\nu}\in\mathcal{S}^{\prime}, the functors i¯(n){\mathcal{E}}^{(n)}_{\underline{i}} and i¯(n){\mathcal{F}}^{(n)}_{\underline{i}} send L𝝂𝒆L_{\boldsymbol{\nu}\boldsymbol{e}} to a finite direct sum of those shifted L𝝂′′𝒆L_{\boldsymbol{\nu^{\prime\prime}}\boldsymbol{e}} with 𝝂′′𝒮\boldsymbol{\nu^{\prime\prime}}\in\mathcal{S}^{\prime}.

By equation (1), we can see that i¯(n)L𝝂𝒆=L(ni¯,𝝂𝒆),{\mathcal{F}}^{(n)}_{\underline{i}}L_{\boldsymbol{\nu}\boldsymbol{e}}=L_{(n\underline{i},\boldsymbol{\nu}\boldsymbol{e})}, hence i¯(n):𝒬𝐕′′,𝐖a/𝒩𝐕′′𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{F}}^{(n)}_{\underline{i}}:{\mathcal{Q}}^{a}_{\mathbf{V}^{\prime\prime},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}^{\prime\prime}}\rightarrow{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} is well-defined.

As for i¯(n){\mathcal{E}}^{(n)}_{\underline{i}}, i¯(n)L𝒆=0{\mathcal{E}}^{(n)}_{\underline{i}}L_{\boldsymbol{e}}=0 by definition. By [2, Corollary 3.20], i¯(n)j¯j¯i¯(n){\mathcal{E}}^{(n)}_{\underline{i}}{\mathcal{F}}_{\underline{j}}\cong{\mathcal{F}}_{\underline{j}}{\mathcal{E}}^{(n)}_{\underline{i}} if i¯j¯\underline{i}\neq\underline{j}. Otherwise, i¯(n)i¯{\mathcal{E}}^{(n)}_{\underline{i}}{\mathcal{F}}_{\underline{i}} and i¯i¯(n){\mathcal{F}}_{\underline{i}}{\mathcal{E}}^{(n)}_{\underline{i}} differ by a direct sum of some shifts of 𝐈𝐝\mathbf{Id}. Then by induction on the length of 𝝂\boldsymbol{\nu}, we can prove that i¯(n):𝒬𝐕,𝐖a/𝒩𝐕𝒬𝐕′′,𝐖a/𝒩𝐕′′{\mathcal{E}}^{(n)}_{\underline{i}}:{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}\rightarrow{\mathcal{Q}}^{a}_{\mathbf{V}^{\prime\prime},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}^{\prime\prime}} is well-defined. ∎

Notice that the autormorphism aa induces a natural map a:𝐄𝐕,𝐖,Ω𝐄𝐕,𝐖,Ωa:\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}\rightarrow\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}, then aa^{*} is a periodic functor on 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}). Since a(𝒩𝐕,i)𝒩𝐕,a1(i)a^{\ast}({\mathcal{N}}_{\mathbf{V},i})\subseteq{\mathcal{N}}_{\mathbf{V},a^{-1}(i)}, aa^{\ast} preserves 𝒩𝐕{\mathcal{N}}_{\mathbf{V}} and induces a periodic functor on 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}. Since aa^{\ast} also preserves 𝒬𝐕,𝐖a{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}, it restricts to a periodic functor on 𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}.

Definition 3.9.

Given II-graded spaces 𝐕,𝐖\mathbf{V},\mathbf{W} with dimension vectors ν,ωIa\nu,\omega\in{\mathbb{N}}I^{a} respectively, the category 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}} is defined by applying Definition 3.1 to the periodic functor aa^{\ast} on 𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}. The Verdier duality is defined by 𝐃(L,ϕ)=(𝐃L,𝐃(ϕ)1).\mathbf{D}(L,\phi)=(\mathbf{D}L,\mathbf{D}(\phi)^{-1}).

Notice that for any 𝝂𝒮\boldsymbol{\nu}\in\mathcal{S}^{\prime}, there is a canonical choice ϕ0\phi_{0} for L𝝂𝒆L_{\boldsymbol{\nu}\boldsymbol{e}}, given by ϕ0:aπ𝝂𝒆,Ω~¯l=π𝝂𝒆,Ω~a¯lπ𝝂𝒆,Ω~¯l.\phi_{0}:a^{\ast}\pi_{\boldsymbol{\nu}\boldsymbol{e},\tilde{\Omega}}\overline{{\mathbb{Q}}}_{l}=\pi_{\boldsymbol{\nu}\boldsymbol{e},\tilde{\Omega}}a^{\ast}\overline{{\mathbb{Q}}}_{l}\cong\pi_{\boldsymbol{\nu}\boldsymbol{e},\tilde{\Omega}}\overline{{\mathbb{Q}}}_{l}.

Since 𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} consists of shifted semisimple objects, then by Lemma 3.4, any object in 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}} is isomorphic up to traceless elements to L,ϕ,n(L,ϕ)[n]mL,ϕ,n\bigoplus\limits_{L,\phi,n}(L,\phi)[n]^{\oplus m_{L,\phi,n}}, where LL runs over simple objects in 𝒬𝐕,𝐖a\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}} and mL,ϕ,nm_{L,\phi,n} is the multiplicity of (L,ϕ)[n](L,\phi)[n].

The functors i¯(n){\mathcal{E}}^{(n)}_{\underline{i}} and i¯(n){\mathcal{F}}^{(n)}_{\underline{i}} induces functors between those 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}. Since i(n)aaa1(i)(n){\mathcal{E}}^{(n)}_{i}a^{\ast}\cong a^{\ast}{\mathcal{E}}^{(n)}_{a^{-1}(i)}, we know that i¯(n)aai¯(n).{\mathcal{E}}^{(n)}_{\underline{i}}a^{\ast}\cong a^{\ast}{\mathcal{E}}^{(n)}_{\underline{i}}. Similarly, i¯(n)aai¯(n).{\mathcal{F}}^{(n)}_{\underline{i}}a^{\ast}\cong a^{\ast}{\mathcal{F}}^{(n)}_{\underline{i}}. Assume ϕ:aLL\phi:a^{\ast}L\rightarrow L, we denote the composition ai¯(n)Li¯(n)aLi¯(n)ϕi¯(n)La^{\ast}{\mathcal{E}}^{(n)}_{\underline{i}}L\cong{\mathcal{E}}^{(n)}_{\underline{i}}a^{\ast}L\xrightarrow{{\mathcal{E}}^{(n)}_{\underline{i}}\phi}{\mathcal{E}}^{(n)}_{\underline{i}}L by i¯(n)ϕ{\mathcal{E}}^{(n)}_{\underline{i}}\phi, and denote the composition ai¯(n)Li¯(n)aLi¯(n)ϕi¯(n)La^{\ast}{\mathcal{F}}^{(n)}_{\underline{i}}L\cong{\mathcal{F}}^{(n)}_{\underline{i}}a^{\ast}L\xrightarrow{{\mathcal{F}}^{(n)}_{\underline{i}}\phi}{\mathcal{F}}^{(n)}_{\underline{i}}L by i¯(n)ϕ{\mathcal{F}}^{(n)}_{\underline{i}}\phi.

Definition 3.10.

The functors i¯(n):𝒬𝐕,𝐖a/𝒩𝐕~𝒬𝐕,𝐖a/𝒩𝐕~{\mathcal{E}}^{(n)}_{\underline{i}}:\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}\rightarrow\widetilde{\mathcal{Q}^{a}_{\mathbf{V}^{\prime},\mathbf{W}}/\mathcal{N}_{\mathbf{V}^{\prime}}} and i¯(n):𝒬𝐕′′,𝐖a/𝒩𝐕′′~𝒬𝐕,𝐖a/𝒩𝐕~{\mathcal{F}}^{(n)}_{\underline{i}}:\widetilde{\mathcal{Q}^{a}_{\mathbf{V}^{\prime\prime},\mathbf{W}}/\mathcal{N}_{\mathbf{V}^{\prime\prime}}}\rightarrow\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}} are defined respectively by

i¯(n)((L,ϕ))=(i¯(n)L,i¯(n)ϕ),{\mathcal{E}}^{(n)}_{\underline{i}}((L,\phi))=({\mathcal{E}}^{(n)}_{\underline{i}}L,{\mathcal{E}}^{(n)}_{\underline{i}}\phi),
i¯(n)((L,ϕ))=(i¯(n)L,i¯(n)ϕ).{\mathcal{F}}^{(n)}_{\underline{i}}((L,\phi))=({\mathcal{F}}^{(n)}_{\underline{i}}L,{\mathcal{F}}^{(n)}_{\underline{i}}\phi).

The functor 𝒦i¯,𝒦i¯:𝒬𝐕,𝐖a/𝒩𝐕~𝒬𝐕,𝐖a/𝒩𝐕~{\mathcal{K}}_{\underline{i}},{\mathcal{K}}_{-\underline{i}}:\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}\rightarrow\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}} is defined by

𝒦i¯=𝐈𝐝[si(ν~i2νi)],𝒦i¯=𝐈𝐝[si(ν~i2νi)],{\mathcal{K}}_{\underline{i}}=\mathbf{Id}[s_{i}(\tilde{\nu}_{i}-2\nu_{i})],~{}{\mathcal{K}}_{-\underline{i}}=\mathbf{Id}[-s_{i}(\tilde{\nu}_{i}-2\nu_{i})],

where sis_{i} is the order of i¯\underline{i} and νi~=hΩi,s(h)=iνt(h)+dim𝐖i\tilde{\nu_{i}}=\sum\limits_{h\in\Omega_{i},s(h)=i}\nu_{t(h)}+\mbox{\rm dim}\mathbf{W}_{i} as before.

3.3. Commutative relations of functors

In this subsection, we study the commutative relations of our functors.

3.3.1. Relations of divided powers

Take the aa orbit i¯\underline{i} of iIi\in I, let sis_{i} be the order of i¯\underline{i} and choose an orientation Ω\Omega such that every ii¯i\in\underline{i} are sources in Ω\Omega.

Proposition 3.11.

For any (L,ϕ)(L,\phi) in 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}, there are isomorphisms up to traceless elements

i¯(n1)i¯(L,ϕ)i¯i¯(n1)(L,ϕ)r=0n1i¯(n)[2sir+si(n1)](L,ϕ),{\mathcal{E}}_{\underline{i}}^{(n-1)}{\mathcal{E}}_{\underline{i}}(L,\phi)\cong{\mathcal{E}}_{\underline{i}}{\mathcal{E}}_{\underline{i}}^{(n-1)}(L,\phi)\cong\bigoplus_{r=0}^{n-1}{\mathcal{E}}_{\underline{i}}^{(n)}[-2s_{i}r+s_{i}(n-1)](L,\phi),
i¯(n1)i¯(L,ϕ)i¯i¯(n1)(L,ϕ)r=0n1i¯(n)[2sir+si(n1)](L,ϕ),{\mathcal{F}}_{\underline{i}}^{(n-1)}{\mathcal{F}}_{\underline{i}}(L,\phi)\cong{\mathcal{F}}_{\underline{i}}{\mathcal{F}}_{\underline{i}}^{(n-1)}(L,\phi)\cong\bigoplus_{r=0}^{n-1}{\mathcal{F}}_{\underline{i}}^{(n)}[-2s_{i}r+s_{i}(n-1)](L,\phi),

here the functor i¯(n1){\mathcal{E}}_{\underline{i}}^{(n-1)} and i¯(n1){\mathcal{F}}_{\underline{i}}^{(n-1)} for n=1n=1 is defined as the identity functor 𝐈𝐝\mathbf{Id}

Proof.

Notice that the functor i(n){\mathcal{F}}^{(n)}_{i} is isomorphic to Lusztig’s induction functor, the second relation follows from the proof of [7, Lemma 12.3.4]. We only need to prove the first relation.

We take 𝐕,𝐕,𝐕′′\mathbf{V},\mathbf{V}^{\prime},\mathbf{V}^{\prime\prime} of dimensions ν\nu, ν=ν(n1)i¯\nu^{\prime}=\nu-(n-1)\underline{i}, and ν′′=νni¯\nu^{\prime\prime}=\nu-n\underline{i}, respectively. Let 𝐄𝐕,𝐖,i¯0\mathbf{E}_{\mathbf{V},\mathbf{W},\underline{i}}^{0} be the open subset of 𝐄𝐕,𝐖,Ω\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega} where xx satisfies dimker(hΩ~,s(h)=ixh)=0\mbox{\rm dim}\ker(\bigoplus\limits_{h\in\tilde{\Omega},s(h)=i}x_{h})=0 for any iI¯i\in\underline{I}. We also define

𝐄˙𝐕,𝐖,i¯=hΩ,s(h)i¯Hom(𝐕s(h),𝐕t(h))ji¯,jIHom(𝐕j,𝐖j),\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}=\bigoplus_{h\in\Omega,s(h)\notin\underline{i}}\mbox{\rm Hom}(\mathbf{V}_{s(h)},\mathbf{V}_{t(h)})\oplus\bigoplus_{j\notin\underline{i},j\in I}\mbox{\rm Hom}(\mathbf{V}_{j},\mathbf{W}_{j}),

and denote ν~i=hH,s(h)=iνt(h)+dim𝐖i.\tilde{\nu}_{i}=\sum_{h\in H,s(h)=i}\nu_{t(h)}+\mbox{\rm dim}\mathbf{W}_{i}. We consider the following diagram

𝐄𝐕,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕,i¯\scriptstyle{j_{\mathbf{V},\underline{i}}}𝐄𝐕′′,𝐖\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕′′,i¯\scriptstyle{j_{\mathbf{V}^{\prime\prime},\underline{i}}}𝐄𝐕,𝐖,i¯0\textstyle{\mathbf{E}_{\mathbf{V},\mathbf{W},\underline{i}}^{0}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕,i¯\scriptstyle{\phi_{\mathbf{V},\underline{i}}}𝐄𝐕′′,𝐖,i¯0\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\underline{i}}^{0}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕′′,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime\prime},\underline{i}}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νi,ν~i)\textstyle{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q1\scriptstyle{q_{1}}𝐄˙𝐕,𝐖,i¯×𝐅𝐥si(νin,νi,ν~i)\textstyle{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Fl}^{s_{i}}(\nu_{i}-n,\nu_{i},\tilde{\nu}_{i})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q2\scriptstyle{q_{2}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νin,ν~i),\textstyle{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Gr}^{s_{i}}(\nu_{i}-n,\tilde{\nu}_{i}),}

where those j𝐕,i¯j_{\mathbf{V},\underline{i}} are open inclusions, and ϕ𝐕,i¯\phi_{\mathbf{V},\underline{i}} is defined by

ϕ𝐕,i¯:𝐄𝐕,𝐖,i¯0𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νi,ν~i);x(x˙,(Im(hΩ~,s(h)=ixh)ii¯)),\displaystyle\phi_{\mathbf{V},\underline{i}}:\mathbf{E}^{0}_{\mathbf{V},\mathbf{W},\underline{i}}\longrightarrow\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i});x\mapsto(\dot{x},({\rm{Im}}(\bigoplus\limits_{h\in\tilde{\Omega},s(h)=i}x_{h})_{i\in\underline{i}})),

and q1,q2q_{1},q_{2} are the natural projections, which are smooth and proper. Then the functor i¯(n){\mathcal{E}}_{\underline{i}}^{(n)} is isomorphic to (j𝐕′′,i¯)!(ϕ𝐕′′,i¯)(q2)!(q1)(ϕ𝐕,i¯)(j𝐕,i¯)[nsiνi+si(n1)](j_{\mathbf{V}^{\prime\prime},\underline{i}})_{!}(\phi_{\mathbf{V}^{\prime\prime},\underline{i}})^{*}(q_{2})_{!}(q_{1})^{*}(\phi_{\mathbf{V},\underline{i}})_{\flat}(j_{\mathbf{V},\underline{i}})^{*}[-ns_{i}\nu_{i}+s_{i}(n-1)].

Similarly, the functor i¯(n1)i¯{\mathcal{E}}_{\underline{i}}^{(n-1)}{\mathcal{E}}_{\underline{i}} is isomorphic to

(j𝐕′′,i¯)!(ϕ𝐕′′,i¯)(q2′′)!(q1′′)(ϕ𝐕,i¯)(j𝐕,i¯)(j𝐕,i¯)!(ϕ𝐕,i¯)(q2)!(q1)(ϕ𝐕,i¯)(j𝐕,i¯)[nsiνi],(j_{\mathbf{V}^{\prime\prime},\underline{i}})_{!}(\phi_{\mathbf{V}^{\prime\prime},\underline{i}})^{*}(q^{\prime\prime}_{2})_{!}(q^{\prime\prime}_{1})^{*}(\phi_{\mathbf{V}^{\prime},\underline{i}})_{\flat}(j_{\mathbf{V}^{\prime},\underline{i}})^{*}(j_{\mathbf{V}^{\prime},\underline{i}})_{!}(\phi_{\mathbf{V}^{\prime},\underline{i}})^{*}(q^{\prime}_{2})_{!}(q^{\prime}_{1})^{*}(\phi_{\mathbf{V},\underline{i}})_{\flat}(j_{\mathbf{V},\underline{i}})^{*}[-ns_{i}\nu_{i}],

where the morphisms are defined in the following diagram:

𝐄𝐕,𝐖\textstyle{\mathbf{E}_{\mathbf{V},\mathbf{W}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕,i¯\scriptstyle{j_{\mathbf{V},\underline{i}}}𝐄𝐕,𝐖\textstyle{\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕,i¯\scriptstyle{j_{\mathbf{V}^{\prime},\underline{i}}}𝐄𝐕′′,𝐖\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕′′,i¯\scriptstyle{j_{\mathbf{V}^{\prime\prime},\underline{i}}}𝐄𝐕,𝐖,i¯0\textstyle{\mathbf{E}^{0}_{\mathbf{V},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕,i¯\scriptstyle{\phi_{\mathbf{V},\underline{i}}}𝐄𝐕,𝐖,i¯0\textstyle{\mathbf{E}^{0}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime},\underline{i}}}𝐄𝐕′′,𝐖,i¯0\textstyle{\mathbf{E}^{0}_{\mathbf{V}^{\prime\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕′′,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime\prime},\underline{i}}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}) q1\scriptstyle{q_{1}^{\prime}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νi,νi,ν~i)\mathbf{Fl}^{s_{i}}(\nu^{\prime}_{i},\nu_{i},\tilde{\nu}_{i}) q2\scriptstyle{q_{2}^{\prime}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu^{\prime}_{i},\tilde{\nu}_{i}) q1′′\scriptstyle{q_{1}^{\prime\prime}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νi′′,νi,ν~i)\mathbf{Fl}^{s_{i}}(\nu^{\prime\prime}_{i},\nu^{\prime}_{i},\tilde{\nu}_{i}) q2′′\scriptstyle{q_{2}^{\prime\prime}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νi′′,ν~i).\textstyle{\hbox{\hbox{\kern 0.0pt\raise-13.62206pt\vbox{\halign{\relax\hfil\txtline@@{#}\hfil\cr\hbox{{\ignorespaces\leavevmode$\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}$}\\\hbox{{\ignorespaces$\times$}\\\hbox{{\ignorespaces$\mathbf{Gr}^{s_{i}}(\nu^{\prime\prime}_{i},\tilde{\nu}_{i})$}\crcr}}}}.}}}}

Since (ϕ𝐕,i¯)(j𝐕,i¯)(j𝐕,i)!(ϕ𝐕,i)𝐈𝐝(\phi_{\mathbf{V}^{\prime},\underline{i}})_{\flat}(j_{\mathbf{V}^{\prime},\underline{i}})^{*}(j_{\mathbf{V}^{\prime},i})_{!}(\phi_{\mathbf{V}^{\prime},i})^{*}\cong\mathbf{Id}, we can reduce to the following diagram

𝐄˙𝐕,𝐖,i×𝐆𝐫si(νi,ν~i)\textstyle{{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Gr}^{s_{i}}(\nu^{\prime}_{i},\tilde{\nu}_{i})}}𝐄˙𝐕,𝐖,i×𝐅𝐥si(νi′′,νi,ν~i)\textstyle{{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Fl}^{s_{i}}(\nu^{\prime\prime}_{i},\nu^{\prime}_{i},\tilde{\nu}_{i})}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q1′′\scriptstyle{q_{1}^{\prime\prime}}q2′′\scriptstyle{q_{2}^{\prime\prime}}𝐄˙𝐕,𝐖,i×𝐆𝐫si(νi′′,ν~i)\textstyle{{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Gr}^{s_{i}}(\nu^{\prime\prime}_{i},\tilde{\nu}_{i})}}𝐄˙𝐕,𝐖,i×𝐅𝐥si(νi,νi,ν~i)\textstyle{{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Fl}^{s_{i}}(\nu^{\prime}_{i},\nu_{i},\tilde{\nu}_{i})}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q2\scriptstyle{q_{2}^{\prime}}q1\scriptstyle{q_{1}^{\prime}}𝐄˙𝐕,𝐖,i×𝐅𝐥si(νi′′,νi,νi,ν~i)\textstyle{{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Fl}^{s_{i}}(\nu^{\prime\prime}_{i},\nu^{\prime}_{i},\nu_{i},\tilde{\nu}_{i})}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}r1\scriptstyle{r_{1}}r2\scriptstyle{r_{2}}π\scriptstyle{\pi}𝐄˙𝐕,𝐖,i×𝐆𝐫si(νi,ν~i)\textstyle{{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i})}}𝐄˙𝐕,𝐖,i×𝐅𝐥si(νi′′,νi,ν~i).\textstyle{{\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},i}\times\mathbf{Fl}^{s_{i}}(\nu^{\prime\prime}_{i},\nu_{i},\tilde{\nu}_{i})}.\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q2\scriptstyle{q_{2}}q1\scriptstyle{q_{1}}

Notice that π\pi is a local trivial fiberation with fiber isomorphic to sis_{i} product of n1{\mathbb{P}}^{n-1}. Thus (q2′′)!(q1′′)(q2)!(q1)(q2)!π!π(q1)(q2)!(π!¯l(q1)())(q2)!(𝐇((n1)si)(q1)())(q_{2}^{\prime\prime})_{!}(q_{1}^{\prime\prime})^{*}(q_{2}^{\prime})_{!}(q_{1}^{\prime})^{*}\cong(q_{2})_{!}\pi_{!}\pi^{*}(q_{1})^{*}\cong(q_{2})_{!}(\pi_{!}\overline{{\mathbb{Q}}}_{l}\otimes(q_{1})^{*}(-))\cong(q_{2})_{!}(\mathbf{H}^{\ast}(({\mathbb{P}}^{n-1})^{s_{i}})\otimes(q_{1})^{*}(-)).

We can identify 𝐇((n1)si)\mathbf{H}^{\ast}(({\mathbb{P}}^{n-1})^{s_{i}}) with 𝐇((n1))si\mathbf{H}^{\ast}(({\mathbb{P}}^{n-1}))^{\otimes s_{i}} via Kunneth formula and let ¯l[2r12r22rsi]𝐇((n1)si)\overline{{\mathbb{Q}}}_{l}[-2r_{1}-2r_{2}-\cdots-2r_{s_{i}}]\subseteqq\mathbf{H}^{\ast}(({\mathbb{P}}^{n-1})^{s_{i}}) be the direct summand contributed by 𝐇r1(n1)[2r1]𝐇r2(n1)[2r2]𝐇rsi(n1)[2rsi].\mathbf{H}^{r_{1}}({\mathbb{P}}^{n-1})[-2r_{1}]\otimes\mathbf{H}^{r_{2}}({\mathbb{P}}^{n-1})[-2r_{2}]\cdots\otimes\mathbf{H}^{r_{s_{i}}}({\mathbb{P}}^{n-1})[-2r_{s_{i}}]. Then for a morphism ϕ:aLL\phi:a^{*}L\rightarrow L, the morphism

(q2)!(π!¯l(q1)ϕ):(q2)!(π!¯l(q1)aL)(q2)!(π!¯l(q1)L)(q_{2})_{!}(\pi_{!}\overline{{\mathbb{Q}}}_{l}\otimes(q_{1})^{*}\phi):(q_{2})_{!}(\pi_{!}\overline{{\mathbb{Q}}}_{l}\otimes(q_{1})^{*}a^{*}L)\rightarrow(q_{2})_{!}(\pi_{!}\overline{{\mathbb{Q}}}_{l}\otimes(q_{1})^{*}L)

sends (q2)!(¯l[2r12r22rsi](q1)L)(q_{2})_{!}(\overline{{\mathbb{Q}}}_{l}[-2r_{1}-2r_{2}-\cdots-2r_{s_{i}}]\otimes(q_{1})^{*}L) to (q2)!(¯l[2r22r32rsi2r1](q1)L)(q_{2})_{!}(\overline{{\mathbb{Q}}}_{l}[-2r_{2}-2r_{3}-\cdots-2r_{s_{i}}-2r_{1}]\otimes(q_{1})^{*}L), where rk{0,1,2,,si1}r_{k}\in\{0,1,2,\dots,s_{i}-1\}. If rsrs′′r_{s^{\prime}}\neq r_{s^{\prime\prime}} for some s,s′′s^{\prime},s^{\prime\prime}, ϕ\phi acts by a cyclic permutation on the direct summands corresponding to {rk,0ksi1}\{r_{k},0\leqslant k\leqslant s_{i}-1\}, hence the direct sum of these direct summands contributes to a traceless object. As a result, only those ¯l[2rsi]\overline{{\mathbb{Q}}}_{l}[-2rs_{i}] with 0rn10\leqslant r\leqslant n-1 contribute to non-traceless object and we get the proof for i¯(n1)i¯{\mathcal{E}}_{\underline{i}}^{(n-1)}{\mathcal{E}}_{\underline{i}} . The proof for i¯i¯(n1){\mathcal{E}}_{\underline{i}}{\mathcal{E}}_{\underline{i}}^{(n-1)} is similar.

3.3.2. Relations of i¯(n){\mathcal{E}}_{\underline{i}}^{(n)} and j¯{\mathcal{F}}_{\underline{j}}

Take the aa orbits i¯,j¯\underline{i},\underline{j} of i,jIi,j\in I respectively, let sis_{i} be the order of i¯\underline{i}.

Proposition 3.12.

For any (L,ϕ)(L,\phi) in 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}, if i¯j¯\underline{i}\neq\underline{j}, there is an ismorphism

(2) i¯(n)j¯(L,ϕ)j¯i¯(n)(L,ϕ).{\mathcal{E}}_{\underline{i}}^{(n)}{\mathcal{F}}_{\underline{j}}(L,\phi)\cong{\mathcal{F}}_{\underline{j}}{\mathcal{E}}_{\underline{i}}^{(n)}(L,\phi).

Otherwise, denote m=n+νi~2νi1m=n+\tilde{\nu_{i}}-2\nu_{i}-1 and there is an ismorphism up to traceless elements

(3) i¯(n)i¯(L,ϕ)k=0m1i¯(n1)[(m12k)si](L,ϕ)i¯i¯(n)(L,ϕ)k=0m1i¯(n1)[(m12k)si](L,ϕ).\begin{split}&{\mathcal{E}}_{\underline{i}}^{(n)}{\mathcal{F}}_{\underline{i}}(L,\phi)\oplus\bigoplus_{k=0}^{-m-1}{\mathcal{E}}_{\underline{i}}^{(n-1)}[(-m-1-2k)s_{i}](L,\phi)\\ \cong&{\mathcal{F}}_{\underline{i}}{\mathcal{E}}_{\underline{i}}^{(n)}(L,\phi)\oplus\bigoplus_{k=0}^{m-1}{\mathcal{E}}_{\underline{i}}^{(n-1)}[(m-1-2k)s_{i}](L,\phi).\end{split}

.

Proof.

If i¯j¯\underline{i}\neq\underline{j}, by [2, Lemma 3.14 and Corollary 3.20] the functors i(n){\mathcal{E}}_{i}^{(n)} and j{\mathcal{F}}_{j} commute with each other on those 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}, so their products i¯(n){\mathcal{E}}_{\underline{i}}^{(n)} and j¯{\mathcal{F}}_{\underline{j}} also commute on those 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}. Hence they induce commutative functors on 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}. The equation (2) is proved.

Now, for i¯=j¯\underline{i}=\underline{j}, we consider ν′′′,ν,ν′′[I]a\nu^{\prime\prime\prime},\nu^{\prime},\nu^{\prime\prime}\in{\mathbb{N}}[I]^{a}, and take 𝐕,𝐕,𝐕′′\mathbf{V},\mathbf{V}^{\prime},\mathbf{V}^{\prime\prime} of dimensions ν′′′=ν+i¯,ν=ν(n1)i¯,ν′′=νni¯\nu^{\prime\prime\prime}=\nu+\underline{i},\nu^{\prime}=\nu-(n-1)\underline{i},\nu^{\prime\prime}=\nu-n\underline{i}, respectively. Choose an orientation Ω\Omega such that all vertex ii¯i\in\underline{i} are sources in Ω\Omega. Then we can draw the following diagrams for i¯(n)i¯{\mathcal{E}}_{\underline{i}}^{(n)}{\mathcal{F}}_{\underline{i}},

𝐄𝐕,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}}𝐄𝐕′′′,𝐖,Ω\textstyle{\mathbf{E}^{\prime}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}p1\scriptstyle{p_{1}}p2\scriptstyle{p_{2}}𝐄𝐕′′′,𝐖,Ω′′\textstyle{\mathbf{E}^{\prime\prime}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}p3\scriptstyle{p_{3}}𝐄𝐕′′′,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\Omega}}𝐄𝐕,𝐖,i¯0\textstyle{\mathbf{E}^{0}_{\mathbf{V},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕,i¯\scriptstyle{j_{\mathbf{V},\underline{i}}}ϕ𝐕,i¯\scriptstyle{\phi_{\mathbf{V},\underline{i}}}𝐄𝐕′′′,𝐖,i¯0,\textstyle{\mathbf{E}^{0,^{\prime}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝐄𝐕′′′,𝐖,i¯0,′′\textstyle{\mathbf{E}^{0,^{\prime\prime}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝐄𝐕′′′,𝐖,i¯0\textstyle{\mathbf{E}^{0}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕′′′,i¯\scriptstyle{j_{\mathbf{V}^{\prime\prime\prime},\underline{i}}}ϕ𝐕′′′,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime\prime\prime},\underline{i}}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}) 𝐄˙𝐕′′′,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νi,νi+1,ν~i)\mathbf{Fl}^{s_{i}}(\nu_{i},\nu_{i}+1,\tilde{\nu}_{i}) q1\scriptstyle{q_{1}} 𝐄˙𝐕′′′,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νi,νi+1,ν~i)\mathbf{Fl}^{s_{i}}(\nu_{i},\nu_{i}+1,\tilde{\nu}_{i}) q2\scriptstyle{q_{2}}𝐄˙𝐕′′′,𝐖,i¯×𝐆𝐫si(νi+1,ν~i),\textstyle{\hbox{\hbox{\kern 0.0pt\raise-13.66539pt\vbox{\halign{\relax\hfil\txtline@@{#}\hfil\cr\hbox{{\ignorespaces\leavevmode$\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}}$}\\\hbox{{\ignorespaces$\times$}\\\hbox{{\ignorespaces$\mathbf{Gr}^{s_{i}}(\nu_{i}+1,\tilde{\nu}_{i})$}\crcr}}}},}}}}
𝐄𝐕′′′,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕′′′,i¯\scriptstyle{j_{\mathbf{V}^{\prime\prime\prime},\underline{i}}}𝐄𝐕,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕,i¯\scriptstyle{j_{\mathbf{V}^{\prime},\underline{i}}}𝐄𝐕′′′,𝐖,i¯0\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}}^{0}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕′′′,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime\prime\prime},\underline{i}}}𝐄𝐕,𝐖,i¯0\textstyle{\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}}^{0}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime},\underline{i}}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi+1,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i}+1,\tilde{\nu}_{i}) q1\scriptstyle{q_{1}^{\prime}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νin+1,νi+1,ν~i)\mathbf{Fl}^{s_{i}}(\nu_{i}-n+1,\nu_{i}+1,\tilde{\nu}_{i}) q2\scriptstyle{q_{2}^{\prime}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νin+1,ν~i),\textstyle{\hbox{\hbox{\kern 0.0pt\raise-13.62206pt\vbox{\halign{\relax\hfil\txtline@@{#}\hfil\cr\hbox{{\ignorespaces\leavevmode$\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}$}\\\hbox{{\ignorespaces$\times$}\\\hbox{{\ignorespaces$\mathbf{Gr}^{s_{i}}(\nu_{i}-n+1,\tilde{\nu}_{i})$}\crcr}}}},}}}}

where 𝐄𝐕′′′,𝐖,i¯0,′′\mathbf{E}^{0,^{\prime\prime}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} is the preimage of 𝐄𝐕′′′,𝐖,i¯0\mathbf{E}^{0}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} under p3p_{3} and 𝐄𝐕′′′,𝐖,i¯0,\mathbf{E}^{0,^{\prime}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} is the preimage of 𝐄𝐕′′′,𝐖,i¯0,′′\mathbf{E}^{0,^{\prime\prime}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} under p2p_{2}. Then the functor i¯(n)i¯{\mathcal{E}}_{\underline{i}}^{(n)}{\mathcal{F}}_{\underline{i}} is isomorphic to

(j𝐕,i¯)!(ϕ𝐕,i¯)(q2)!(q1)(q2)!(q1)(ϕ𝐕,i¯)(j𝐕,i¯)[si(νi+ν~inνin)].(j_{\mathbf{V}^{\prime},\underline{i}})_{!}(\phi_{\mathbf{V}^{\prime},\underline{i}})^{*}(q_{2}^{\prime})_{!}(q_{1}^{\prime})^{*}(q_{2})_{!}(q_{1})^{*}(\phi_{\mathbf{V},\underline{i}})_{\flat}(j_{\mathbf{V},\underline{i}})^{*}[s_{i}(\nu_{i}+\tilde{\nu}_{i}-n\nu_{i}-n)].

Similarly, consider the following diagrams

𝐄𝐕,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕,i¯\scriptstyle{j_{\mathbf{V},\underline{i}}}𝐄𝐕′′,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕′′,i¯\scriptstyle{j_{\mathbf{V}^{\prime\prime},\underline{i}}}𝐄𝐕,𝐖,i¯0\textstyle{\mathbf{E}_{\mathbf{V},\mathbf{W},\underline{i}}^{0}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕,i¯\scriptstyle{\phi_{\mathbf{V},\underline{i}}}𝐄𝐕′′,𝐖,i¯0\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\underline{i}}^{0}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ𝐕′′,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime\prime},\underline{i}}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}) q1~\scriptstyle{\tilde{q_{1}}^{\prime}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νin,νi,ν~i)\mathbf{Fl}^{s_{i}}(\nu_{i}-n,\nu_{i},\tilde{\nu}_{i}) q2~\scriptstyle{\tilde{q_{2}}^{\prime}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νin,ν~i);\textstyle{\hbox{\hbox{\kern 0.0pt\raise-13.62206pt\vbox{\halign{\relax\hfil\txtline@@{#}\hfil\cr\hbox{{\ignorespaces\leavevmode$\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}$}\\\hbox{{\ignorespaces$\times$}\\\hbox{{\ignorespaces$\mathbf{Gr}^{s_{i}}(\nu_{i}-n,\tilde{\nu}_{i})$}\crcr}}}};}}}}
𝐄𝐕′′,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V}^{\prime\prime},\mathbf{W},\Omega}}𝐄𝐕,𝐖,Ω\textstyle{\mathbf{E}^{\prime}_{\mathbf{V}^{\prime},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}p1\scriptstyle{p_{1}}p2\scriptstyle{p_{2}}𝐄𝐕,𝐖,Ω′′\textstyle{\mathbf{E}^{\prime\prime}_{\mathbf{V}^{\prime},\mathbf{W},\Omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}p3\scriptstyle{p_{3}}𝐄𝐕,𝐖,Ω\textstyle{\mathbf{E}_{\mathbf{V}^{\prime},\mathbf{W},\Omega}}𝐄𝐕′′,𝐖,i¯0\textstyle{\mathbf{E}^{0}_{\mathbf{V}^{\prime\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕′′,i¯\scriptstyle{j_{\mathbf{V}^{\prime\prime},\underline{i}}}ϕ𝐕′′,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime\prime},\underline{i}}}𝐄𝐕,𝐖,i¯0,\textstyle{\mathbf{E}^{0,^{\prime}}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝐄𝐕,𝐖,i¯0,′′\textstyle{\mathbf{E}^{0,^{\prime\prime}}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝐄𝐕,𝐖,i¯0\textstyle{\mathbf{E}^{0}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}j𝐕,i¯\scriptstyle{j_{\mathbf{V}^{\prime},\underline{i}}}ϕ𝐕,i¯\scriptstyle{\phi_{\mathbf{V}^{\prime},\underline{i}}} 𝐄˙𝐕′′,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi′′,ν~i)\mathbf{Gr}^{s_{i}}(\nu^{\prime\prime}_{i},\tilde{\nu}_{i}) 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νi′′,νi,ν~i)\mathbf{Fl}^{s_{i}}(\nu^{\prime\prime}_{i},\nu^{\prime}_{i},\tilde{\nu}_{i}) q1~\scriptstyle{\tilde{q_{1}}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νi′′,νi,ν~i)\mathbf{Fl}^{s_{i}}(\nu^{\prime\prime}_{i},\nu^{\prime}_{i},\tilde{\nu}_{i}) q2~\scriptstyle{\tilde{q_{2}}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νi,ν~i).\textstyle{\hbox{\hbox{\kern 0.0pt\raise-13.66539pt\vbox{\halign{\relax\hfil\txtline@@{#}\hfil\cr\hbox{{\ignorespaces\leavevmode$\dot{\mathbf{E}}_{\mathbf{V}^{\prime},\mathbf{W},\underline{i}}$ }\\\hbox{{\ignorespaces$\times$ }\\\hbox{{\ignorespaces$\mathbf{Gr}^{s_{i}}(\nu^{\prime}_{i},\tilde{\nu}_{i})$}\crcr}}}}.}}}}

Then the functor i¯i¯(n){\mathcal{F}}_{\underline{i}}{\mathcal{E}}_{\underline{i}}^{(n)} is isomorphic to

(j𝐕,i¯)!(ϕ𝐕,i¯)(q2~)!(q1~)(q2~)!(q1~)(ϕ𝐕,i¯)(j𝐕,i¯)[si(νi+ν~inνin)].(j_{\mathbf{V}^{\prime},\underline{i}})_{!}(\phi_{\mathbf{V}^{\prime},\underline{i}})^{*}(\tilde{q_{2}})_{!}(\tilde{q_{1}})^{*}(\tilde{q_{2}}^{\prime})_{!}(\tilde{q_{1}}^{\prime})^{*}(\phi_{\mathbf{V},\underline{i}})_{\flat}(j_{\mathbf{V},\underline{i}})^{*}[s_{i}(\nu_{i}+\tilde{\nu}_{i}-n\nu_{i}-n)].

We now only need to study the relation between (q2)!(q1)(q2)!(q1)(q_{2}^{\prime})_{!}(q_{1}^{\prime})^{*}(q_{2})_{!}(q_{1})^{*} and (q2~)!(q1~)(q2~)!(q1~)(\tilde{q_{2}})_{!}(\tilde{q_{1}})^{*}(\tilde{q_{2}}^{\prime})_{!}(\tilde{q_{1}}^{\prime})^{*}. Let X1X_{1} be the pullback of q2,q1q_{2},q_{1}^{\prime}, and let X2X_{2} be the pullback of q2~,q1~\tilde{q_{2}}^{\prime},\tilde{q_{1}}, they are varieties

X1:=𝐄˙𝐕,𝐖,i¯×{(W,WW′′′W~dimW=νi,dimW=νi,dimW′′′=νi′′′,dimW~=ν~i}si,X_{1}:=\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\{(W,W^{\prime}\subset W^{\prime\prime\prime}\subset\tilde{W}\mid\mbox{\rm dim}W^{\prime}=\nu_{i}^{\prime},\mbox{\rm dim}W=\nu_{i},\mbox{\rm dim}W^{\prime\prime\prime}=\nu_{i}^{\prime\prime\prime},\mbox{\rm dim}\tilde{W}=\tilde{\nu}_{i}\}^{s_{i}},
X2:=𝐄˙𝐕,𝐖,i¯×{(W′′W,WW~dimW=νi,dimW=νi,dimW′′=νi′′,dimW~=ν~i}si.X_{2}:=\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\{(W^{\prime\prime}\subset W,W^{\prime}\subset\tilde{W}\mid\mbox{\rm dim}W^{\prime}=\nu_{i}^{\prime},\mbox{\rm dim}W=\nu_{i},\mbox{\rm dim}W^{\prime\prime}=\nu_{i}^{\prime\prime},\mbox{\rm dim}\tilde{W}=\tilde{\nu}_{i}\}^{s_{i}}.

Then (q2)!(q1)(q2)!(q1)(π2)!r!r(π1)(q_{2}^{\prime})_{!}(q_{1}^{\prime})^{*}(q_{2})_{!}(q_{1})^{*}\cong(\pi_{2})_{!}r_{!}r^{*}(\pi_{1})^{*}, where π1,π2\pi_{1},\pi_{2} and rr are the natural projections in the following diagram

𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}) 𝐄˙𝐕′′′,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νi,νi+1,ν~i)\mathbf{Fl}^{s_{i}}(\nu_{i},\nu_{i}+1,\tilde{\nu}_{i}) q1\scriptstyle{q_{1}}q2\scriptstyle{q_{2}} 𝐄˙𝐕′′′,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi+1,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i}+1,\tilde{\nu}_{i}) X1\textstyle{X_{1}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}r\scriptstyle{r} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νin+1,νi+1,ν~i)\mathbf{Fl}^{s_{i}}(\nu_{i}-n+1,\nu_{i}+1,\tilde{\nu}_{i}) q1\scriptstyle{q_{1}^{\prime}}q2\scriptstyle{q_{2}^{\prime}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}) ×\times 𝐆𝐫si(νin+1,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i}-n+1,\tilde{\nu}_{i}) π1\scriptstyle{\pi_{1}}π2\scriptstyle{\pi_{2}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νin+1,ν~i).\textstyle{\hbox{\hbox{\kern 0.0pt\raise-13.62206pt\vbox{\halign{\relax\hfil\txtline@@{#}\hfil\cr\hbox{{\ignorespaces\leavevmode$\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}$ }\\\hbox{{\ignorespaces$\times$ }\\\hbox{{\ignorespaces$\mathbf{Gr}^{s_{i}}(\nu_{i}-n+1,\tilde{\nu}_{i})$}\crcr}}}}.}}}}

Similarly, (q2~)!(q1~)(q2~)!(q1~)(π2)!r!r(π1)(\tilde{q_{2}})_{!}(\tilde{q_{1}})^{*}(\tilde{q_{2}}^{\prime})_{!}(\tilde{q_{1}}^{\prime})^{*}\cong(\pi_{2})_{!}r^{\prime}_{!}r^{\prime*}(\pi_{1})^{*}, where π1,π2\pi_{1},\pi_{2} and rr^{\prime} are the natural projections in the following diagram

𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}) 𝐄˙𝐕′′′,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νin,νi,ν~i)\mathbf{Fl}^{s_{i}}(\nu_{i}-n,\nu_{i},\tilde{\nu}_{i}) q1~\scriptstyle{\tilde{q_{1}}^{\prime}}q2~\scriptstyle{\tilde{q_{2}}^{\prime}} 𝐄˙𝐕′′′,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V}^{\prime\prime\prime},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νin,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i}-n,\tilde{\nu}_{i}) X2\textstyle{X_{2}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}r\scriptstyle{r^{\prime}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐅𝐥si(νin,νin+1,νi~)\mathbf{Fl}^{s_{i}}(\nu_{i}-n,\nu_{i}-n+1,\tilde{\nu_{i}}) q1~\scriptstyle{\tilde{q_{1}}}q2~\scriptstyle{\tilde{q_{2}}} 𝐄˙𝐕,𝐖,i¯\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}} ×\times 𝐆𝐫si(νi,ν~i)\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}) ×\times 𝐆𝐫si(νin+1,νi~)\mathbf{Gr}^{s_{i}}(\nu_{i}-n+1,\tilde{\nu_{i}}) π1\scriptstyle{\pi_{1}}π2\scriptstyle{\pi_{2}}𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νin+1,νi~).\textstyle{\hbox{\hbox{\kern 0.0pt\raise-13.62206pt\vbox{\halign{\relax\hfil\txtline@@{#}\hfil\cr\hbox{{\ignorespaces\leavevmode$\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}$ }\\\hbox{{\ignorespaces$\times$ }\\\hbox{{\ignorespaces$\mathbf{Gr}^{s_{i}}(\nu_{i}-n+1,\tilde{\nu_{i}})$}\crcr}}}}.}}}}

By projection formula, we have (π2)!r!r(π1)(π2)!(r!¯l(π1)())(\pi_{2})_{!}r_{!}r^{*}(\pi_{1})^{*}\cong(\pi_{2})_{!}(r_{!}\overline{{\mathbb{Q}}}_{l}\otimes(\pi_{1})^{*}(-)), and (π2)!r!r(π1)(π2)!(r!¯l(π1)())(\pi_{2})_{!}r^{\prime}_{!}r^{\prime*}(\pi_{1})^{*}\cong(\pi_{2})_{!}(r^{\prime}_{!}\overline{{\mathbb{Q}}}_{l}\otimes(\pi_{1})^{*}(-)). We need to study the difference between the semisimple complexes r!¯lr^{\prime}_{!}\overline{{\mathbb{Q}}}_{l} and r!¯lr_{!}\overline{{\mathbb{Q}}}_{l}. (They are semisimple since X1,X2X_{1},X_{2} are smooth and r,rr,r^{\prime} are proper.) It suffices to calculate restrictions of these complexes on a stratification.

Now we introduce subset X1tX_{1}^{\geqslant t} and X2tX_{2}^{\geqslant t} as the following: Define a map ϕ1:{(W,WW′′′W~dimW=νi,dimW=νi,dimW′′′=νi′′′,dimW~=ν~i}{0,1}\phi_{1}:\{(W,W^{\prime}\subset W^{\prime\prime\prime}\subset\tilde{W}\mid\mbox{\rm dim}W^{\prime}=\nu_{i}^{\prime},\mbox{\rm dim}W=\nu_{i},\mbox{\rm dim}W^{\prime\prime\prime}=\nu_{i}^{\prime\prime\prime},\mbox{\rm dim}\tilde{W}=\tilde{\nu}_{i}\}\rightarrow\{0,1\} by setting ϕ1((W,W,W′′′,W~))=1\phi_{1}((W,W^{\prime},W^{\prime\prime\prime},\tilde{W}))=1 if WWW^{\prime}\subset W and ϕ1((W,W,W′′′,W~))=0\phi_{1}((W,W^{\prime},W^{\prime\prime\prime},\tilde{W}))=0 otherwise. It induces a map Φ1:X1{k|0ksi}\Phi_{1}:X_{1}\rightarrow\{k\in\mathbb{N}|0\leqslant k\leqslant s_{i}\} by Φ1=1lsiϕ1,l,\Phi_{1}=\sum\limits_{1\leqslant l\leqslant s_{i}}\phi_{1,l}, where ϕ1,l\phi_{1,l} is defined as ϕ1\phi_{1} on the ll-th {(W,WW′′′W~dimW=νi,dimW=νi,dimW′′′=νi′′′,dimW~=ν~i}\{(W,W^{\prime}\subset W^{\prime\prime\prime}\subset\tilde{W}\mid\mbox{\rm dim}W^{\prime}=\nu_{i}^{\prime},\mbox{\rm dim}W=\nu_{i},\mbox{\rm dim}W^{\prime\prime\prime}=\nu_{i}^{\prime\prime\prime},\mbox{\rm dim}\tilde{W}=\tilde{\nu}_{i}\} component in X1X_{1}. Then X1tX_{1}^{\geqslant t} is defined to be the preimage of {t,t+1,,si}\{t,t+1,\cdots,s_{i}\} under Φ1\Phi_{1}. Denote X1tX1t+1X_{1}^{\geqslant t}-X_{1}^{\geqslant t+1} by X1tX^{t}_{1}. Similarly, we define ϕ2:{(W′′W,WW~dimW=νi,dimW=νi,dimW′′=νi′′,dimW~=ν~i}{0,1}\phi_{2}:\{(W^{\prime\prime}\subset W,W^{\prime}\subset\tilde{W}\mid\mbox{\rm dim}W^{\prime}=\nu_{i}^{\prime},\mbox{\rm dim}W=\nu_{i},\mbox{\rm dim}W^{\prime\prime}=\nu_{i}^{\prime\prime},\mbox{\rm dim}\tilde{W}=\tilde{\nu}_{i}\}\rightarrow\{0,1\} by setting ϕ2((W′′,W,W,W~))=1\phi_{2}((W^{\prime\prime},W,W^{\prime},\tilde{W}))=1 if WWW^{\prime}\subset W and ϕ2((W′′,W,W,W~))=0\phi_{2}((W^{\prime\prime},W,W^{\prime},\tilde{W}))=0 otherwise. The map ϕ2\phi_{2} induces a map Φ2=1lsiϕ2,l\Phi_{2}=\sum\limits_{1\leqslant l\leqslant s_{i}}\phi_{2,l} on X2X_{2}, and we define X2tX_{2}^{\geqslant t} to be the preimage of {t,t+1,,si}\{t,t+1,\cdots,s_{i}\} under Φ2\Phi_{2}. Denote X2tX2t+1X_{2}^{\geqslant t}-X_{2}^{\geqslant t+1} by X2tX^{t}_{2}.

Denote Yt:=r(X1t)r(X1t+1)Y_{t}:=r(X_{1}^{\geqslant t})-r(X_{1}^{\geqslant t+1}), then r!¯l=0tsiFt,r_{!}\overline{{\mathbb{Q}}}_{l}=\bigoplus\limits_{0\leqslant t\leqslant s_{i}}F_{t}, where each FtF_{t} is the direct summand of r!¯lr_{!}\overline{{\mathbb{Q}}}_{l} such that any simple direct summand (up to shift) LtL_{t} of FtF_{t} satisfies suppLtr(X1t){\rm supp}L_{t}\subset r(X_{1}^{\geqslant t}) and suppLtYt{\rm supp}L_{t}\cap Y_{t}\neq\emptyset. Simialrly, denote Zt:=r(X2t)r(X2t+1)Z_{t}:=r(X_{2}^{\geqslant t})-r(X_{2}^{\geqslant t+1}). There is also a decomposition r!¯l=0tsiFt,r^{\prime}_{!}\overline{{\mathbb{Q}}}_{l}=\bigoplus\limits_{0\leqslant t\leqslant s_{i}}F^{\prime}_{t}, where the simple direct summands (up to shift) LtL^{\prime}_{t} of FtF^{\prime}_{t} satisfy suppLtr(X2t){\rm supp}L^{\prime}_{t}\subset r^{\prime}(X_{2}^{\geqslant t}) and suppLtZt{\rm supp}L^{\prime}_{t}\cap Z_{t}\neq\emptyset.

(1)Case t0,sit\neq 0,s_{i} : Let jt:Ytr(X1t)j_{t}:Y_{t}\rightarrow r(X_{1}^{\geqslant t}), then any simple perverse sheaf of the form LtL_{t} satisfies Lt(jt)!(jt)LtL_{t}\cong(j_{t})_{*!}(j_{t})^{*}L_{t}. Notice that if t0,sit\neq 0,s_{i}, the connected components of YtY_{t} have nontrivial aa-orbits. The simple perverse sheaf (jt)Lt(j_{t})^{*}L_{t} is supported on a unique connected component and aa^{\ast} sends it to another simple perverse sheaf supported on another connected component of YtY_{t}, we may assume o(L)1o(L)\neq 1 is the least number such that (a)o(L)LtLt(a^{\ast})^{o(L)}L_{t}\cong L_{t}. By similar reason as in the proof of Proposition 3.11, after taking composition with \otimes and (π2)!(\pi_{2})_{!}, the direct sum of these LtaLt(a)2Lt(a)o(L)1LtL_{t}\oplus a^{\ast}L_{t}\oplus(a^{\ast})^{2}L_{t}\cdots\oplus(a^{\ast})^{o(L)-1}L_{t} is acted by ϕ\phi as a cyclic permutation, hence contributes to a traceless object. Since FtF_{t} is a direct sum of objects of the form LtaLt(a)2Lt(a)o(L)1LtL_{t}\oplus a^{\ast}L_{t}\oplus(a^{\ast})^{2}L_{t}\cdots\oplus(a^{\ast})^{o(L)-1}L_{t}, we know that FtF_{t} also contributes to traceless objects if t0,sit\neq 0,s_{i}. Similarly, those FtF^{\prime}_{t} also contribute to traceless objects if t0,sit\neq 0,s_{i}.

(2)Case t=0t=0 : When t=0t=0, r,rr,r^{\prime} restrict to a same isomorphism on the open subset Z0=Y0Z_{0}=Y_{0} of the images, since in this case W′′′=W+WW^{\prime\prime\prime}=W+W^{\prime} and W′′=WWW^{\prime\prime}=W\cap W^{\prime} are uniquely determined by W,WW,W^{\prime}. Then (j0)F0r!(j01)¯r!(j02)¯(j0)F0(j_{0})^{\ast}F_{0}\cong r_{!}(j^{1}_{0})^{\ast}\bar{\mathbb{Q}}\cong r^{\prime}_{!}(j^{2}_{0})^{\ast}\bar{\mathbb{Q}}\cong(j_{0})^{\ast}F^{\prime}_{0}, where j01,j02j^{1}_{0},j^{2}_{0} are the inclusion of X10,X20X^{0}_{1},X^{0}_{2} respectively. In particular, F0F0F_{0}\cong F^{\prime}_{0}.

(3)Case t=sit=s_{i} : Denote jsi:Ysi=Zsi𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νi,ν~i)×𝐆𝐫si(νin+1,νi~)j_{s_{i}}:Y_{s_{i}}=Z_{s_{i}}\rightarrow\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i})\times\mathbf{Gr}^{s_{i}}(\nu_{i}-n+1,\tilde{\nu_{i}}), then (jsi)r!¯lr!(¯l|X1si).(j_{s_{i}})^{\ast}r_{!}\overline{{\mathbb{Q}}}_{l}\cong r_{!}(\overline{\mathbb{Q}}_{l}|_{X^{s_{i}}_{1}}). Since the restriction of rr on X1siX^{s_{i}}_{1} is a local trivial fiberation and the fiber is sis_{i}-times product of the projective space νi~νi1{\mathbb{P}}^{\tilde{\nu_{i}}-\nu_{i}-1}, we have

(jsi)r!¯l0r1,,rsiνi~νi1¯l|Ysi[2r12r22rsi](j_{s_{i}})^{\ast}r_{!}\overline{{\mathbb{Q}}}_{l}\cong\bigoplus\limits_{0\leqslant r_{1},\dots,r_{s_{i}}\leqslant\tilde{\nu_{i}}-\nu_{i}-1}\overline{{\mathbb{Q}}}_{l}|_{Y_{s_{i}}}[-2r_{1}-2r_{2}-\dots-2r_{s_{i}}]

with respect to the Kunneth formula. Similarly,

(jsi)r!¯l0r1,,rsiνin¯l|Ysi[2r12r22rsi].(j_{s_{i}})^{\ast}r^{\prime}_{!}\overline{{\mathbb{Q}}}_{l}\cong\bigoplus\limits_{0\leqslant r^{\prime}_{1},\dots,r^{\prime}_{s_{i}}\leqslant\nu_{i}-n}\overline{{\mathbb{Q}}}_{l}|_{Y_{s_{i}}}[-2r^{\prime}_{1}-2r^{\prime}_{2}-\dots-2r^{\prime}_{s_{i}}].

By similar argument as in the proof of Proposition 3.11, only the direct summands with r1=r2==rsir_{1}=r_{2}=\dots=r_{s_{i}} and r1=r2==rsir^{\prime}_{1}=r^{\prime}_{2}=\dots=r^{\prime}_{s_{i}} contribute to non-traceless elements. Notice that for n2n\geqslant 2, (π2)!(¯l|Ysi(π1)())(q¯2)!(q¯1)(\pi_{2})_{!}(\overline{{\mathbb{Q}}}_{l}|_{Y_{s_{i}}}\otimes(\pi_{1})^{*}(-))\cong(\bar{q}_{2})_{!}(\bar{q}_{1})^{\ast}, where q¯1,q¯2\bar{q}_{1},\bar{q}_{2} are the natural projections appearing in the definition of i¯(n1)\mathcal{E}^{(n-1)}_{\underline{i}}

q¯1:𝐄˙𝐕,𝐖,i¯×𝐅𝐥si(νin+1,νi,ν~i)𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νi,ν~i),\bar{q}_{1}:\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Fl}^{s_{i}}(\nu_{i}-n+1,\nu_{i},\tilde{\nu}_{i})\rightarrow\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Gr}^{s_{i}}(\nu_{i},\tilde{\nu}_{i}),
q¯2:𝐄˙𝐕,𝐖,i¯×𝐅𝐥si(νin+1,νi,ν~i)𝐄˙𝐕,𝐖,i¯×𝐆𝐫si(νin+1,ν~i),\bar{q}_{2}:\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Fl}^{s_{i}}(\nu_{i}-n+1,\nu_{i},\tilde{\nu}_{i})\rightarrow\dot{\mathbf{E}}_{\mathbf{V},\mathbf{W},\underline{i}}\times\mathbf{Gr}^{s_{i}}(\nu_{i}-n+1,\tilde{\nu}_{i}),

we can obtain

i¯(n)i¯(L,ϕ)k=0m1i¯(n1)[(m12k)si](L,ϕ)i¯i¯(n)(L,ϕ)k=0m1i¯(n1)[(m12k)si](L,ϕ)\begin{split}&{\mathcal{E}}_{\underline{i}}^{(n)}{\mathcal{F}}_{\underline{i}}(L,\phi)\oplus\bigoplus_{k=0}^{-m-1}{\mathcal{E}}_{\underline{i}}^{(n-1)}[(-m-1-2k)s_{i}](L,\phi)\\ \cong&{\mathcal{F}}_{\underline{i}}{\mathcal{E}}_{\underline{i}}^{(n)}(L,\phi)\oplus\bigoplus_{k=0}^{m-1}{\mathcal{E}}_{\underline{i}}^{(n-1)}[(m-1-2k)s_{i}](L,\phi)\end{split}

up to traceless elements. For n=1n=1, since (j𝐕,i)!(j𝐕,i)𝐈𝐝(j_{\mathbf{V},i})_{!}(j_{\mathbf{V},i})^{*}\cong\mathbf{Id} on 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}, we can also get the proof. ∎

4. The Grothendieck group of 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}

In this section, we will determine the structure of the Grothendieck group of 𝒬𝐕,𝐖/𝒩𝐕~\widetilde{\mathcal{Q}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}.

4.1. Key Lemma

We first recall the key inductive lemma for framed quiver in [2], which generalizes Lusztig’s lemma in [6]. Assume Ω\Omega is an orientation such that ii is a source and 𝐕,𝐖\mathbf{V},\mathbf{W} are II-graded spaces with dimension vectors ν,ωI\nu,\omega\in\mathbb{N}I. The variety 𝐄𝐕,𝐖,Ω\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega} has a stratification 𝐄𝐕,𝐖,ir,0rνi\mathbf{E}^{r}_{\mathbf{V},\mathbf{W},i},0\leqslant r\leqslant\nu_{i}, where each 𝐄𝐕,𝐖,Ωr\mathbf{E}^{r}_{\mathbf{V},\mathbf{W},\Omega} is defined by

𝐄𝐕,𝐖,ir={x𝐄𝐕,𝐖,Ω|dimker(hΩ~,s(h)=ixh)=0}.\mathbf{E}_{\mathbf{V},\mathbf{W},i}^{r}=\{x\in\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}|\mbox{\rm dim}\ker(\bigoplus\limits_{h\in\tilde{\Omega},s(h)=i}x_{h})=0\}.

Then for each simple perverse sheaf LL, there exists a unique t:=ti(L)t:=t_{i}(L) such that suppL𝐄𝐕,𝐖,it{\rm supp}L\cap\mathbf{E}_{\mathbf{V},\mathbf{W},i}^{t} is dense in suppL{\rm supp}L.

Lemma 4.1.

As Corollary 4.15 in [2] For a simple perverse sheaf LL in 𝒬𝐕,𝐖{\mathcal{Q}}_{\mathbf{V},\mathbf{W}} with ti(L)=rt_{i}(L)=r, take 𝐕\mathbf{V}^{\prime} with dimension vector νri\nu-ri, then there exists a unique simple perverse sheaf KK in 𝒬𝐕,𝐖{\mathcal{Q}}_{\mathbf{V}^{\prime},\mathbf{W}} such that ti(K)=0t_{i}(K)=0 and

i(r)KLti(L)>rL[nL].{\mathcal{F}}^{(r)}_{i}K\cong L\oplus\bigoplus\limits_{t_{i}(L^{\prime})>r}L^{\prime}[n_{L^{\prime}}].

With the notation above, we denote by πi,r(L)=K,\pi_{i,r}(L)=K, then πi,r\pi_{i,r} is a bijection between the set {L𝒬𝐕,𝐖|\{L\in{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}| LL is a simple perverse sheaf with ti(L)=r}t_{i}(L)=r\} and the set {K𝒬𝐕,𝐖|\{K\in{\mathcal{Q}}_{\mathbf{V}^{\prime},\mathbf{W}}| KK is a simple perverse sheaf with ti(K)=0}t_{i}(K)=0\}.

Remark 4.2.

Notice that if 𝐕\mathbf{V} is not zero, a simple perverse sheaf LL in 𝒬𝐕,𝐖{\mathcal{Q}}_{\mathbf{V},\mathbf{W}} is always a direct summand of L𝛎𝐝L_{\boldsymbol{\nu}\boldsymbol{d}} with 𝛎\boldsymbol{\nu} nonempty. Assume 𝛎=(a1i1,,asis)\boldsymbol{\nu}=(a_{1}i_{1},\cdots,a_{s}i_{s}), then ti1(L)a1t_{i_{1}}(L)\geqslant a_{1}. More precisely, a simple object LL in 𝒬𝐕,𝐖{\mathcal{Q}}_{\mathbf{V},\mathbf{W}} has ti(L)=0t_{i}(L)=0 for any ii if and only if 𝐕\mathbf{V} is zero and LL is the constant sheaf on 𝐄0,𝐖,Ω\mathbf{E}_{0,\mathbf{W},\Omega}.

Now we apply the lemma above to a quiver with automorphism.

Corollary 4.3.

With the notation in Section 3.2, we assume LL is a simple perverse sheaf in 𝒬𝐕,𝐖a{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}} with ti(L)=rt_{i}(L)=r and aLLa^{\ast}L\cong L, take 𝐕\mathbf{V}^{\prime} with dimension vector νri¯\nu-r\underline{i}, then there exists a unique simple perverse sheaf KK in 𝒬𝐕,𝐖a{\mathcal{Q}}^{a}_{\mathbf{V}^{\prime},\mathbf{W}} such that ti(K)=0,aKKt_{i}(K)=0,a^{\ast}K\cong K and

i¯(r)KLti(L)>rL[nL].{\mathcal{F}}^{(r)}_{\underline{i}}K\cong L\oplus\bigoplus\limits_{t_{i}(L^{\prime})>r}L^{\prime}[n_{L^{\prime}}].
Proof.

We only need to check aKKa^{\ast}K\cong K. Notice that aa^{\ast} commutes with i¯(r){\mathcal{F}}^{(r)}_{\underline{i}}, we can see that

i¯(r)aKaLti(L)>raL[nL]Lti(L)>raL[nL].\begin{split}{\mathcal{F}}^{(r)}_{\underline{i}}a^{\ast}K\cong a^{\ast}L\oplus\bigoplus\limits_{t_{i}(L^{\prime})>r}a^{\ast}L^{\prime}[n_{L^{\prime}}]\cong L\oplus\bigoplus\limits_{t_{i}(L^{\prime})>r}a^{\ast}L^{\prime}[n_{L^{\prime}}].\end{split}

By the uniqueness of KK, we get aKKa^{\ast}K\cong K. ∎

4.2. The Grothendieck group 𝒦(ω){\mathcal{K}}(\omega)

Now let 𝒜=[v,v1],\mathcal{A}=\mathbb{Z}[v,v^{-1}], 𝒪=[ζ]\mathcal{O}=\mathbb{Z}[\zeta] and 𝒪=𝒪[v,v1]\mathcal{O}^{\prime}=\mathcal{O}[v,v^{-1}], where ζ\zeta is a primitive oo-th root of unity and oo is the order of the automorphism aa on the quiver QQ. There is a \mathbb{Z}-linear involution ¯\bar{\ } on 𝒪\mathcal{O}^{\prime} defined by v¯=v1\bar{v}=v^{-1} and ζ¯=ζ1\bar{\zeta}=\zeta^{-1}.

Let 𝐔𝒪{}_{\mathcal{O}^{\prime}}\mathbf{U} be the 𝒪\mathcal{O}^{\prime}-subalgebra of 𝐔\mathbf{U} generated by Lusztig’s divide powers and those K±iK_{\pm i^{\prime}} and Λλ𝒪{}_{\mathcal{O}^{\prime}}\Lambda_{\lambda} be the 𝐔𝒪{}_{\mathcal{O}^{\prime}}\mathbf{U}-module generated by the highest weight vector. We can also define 𝐔𝒜{}_{\mathcal{A}}\mathbf{U} and Λλ𝒜{}_{\mathcal{A}}\Lambda_{\lambda} in the same way.

Now assume 𝐕,𝐖\mathbf{V},\mathbf{W} are II-graded spaces with dimension vectors ν,ωIa\nu,\omega\in\mathbb{N}I^{a}. Since II^{\prime} is bijective to I¯\underline{I}, we can identify an aa-orbit i¯\underline{i} of ii with the element iIi^{\prime}\in I^{\prime}. Then ω\omega determines a dominant weight λ=iIωiβi,\lambda=\sum\limits_{i^{\prime}\in I^{\prime}}\omega_{i}\beta_{i^{\prime}}, where βi\beta_{i^{\prime}} is the ii^{\prime}-th fundamental weight.

Definition 4.4.

Following Lusztig, define 𝒦(ω)=ν[I]a𝒦(ν,ω){\mathcal{K}}(\omega)=\bigoplus\limits_{\nu\in{\mathbb{N}}[I]^{a}}{\mathcal{K}}(\nu,\omega) and the Grothendieck group 𝒦(ν,ω){\mathcal{K}}(\nu,\omega) of 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}} is defined to be the 𝒪\mathcal{O}^{\prime}-module spanned by [(L,ϕ)][(L,\phi)], (L,ϕ)(L,\phi) in 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}, subject to the following relations

[(L1,ϕ1)]+[(L2,ϕ2)]=[(L1L2,ϕ1ϕ2)];[(L_{1},\phi_{1})]+[(L_{2},\phi_{2})]=[(L_{1}\oplus L_{2},\phi_{1}\oplus\phi_{2})];
[(L,ϕ)[n]]=vn[(L,ϕ)];[(L,\phi)[n]]=v^{n}[(L,\phi)];
[(L,tϕ)]=t[(L,ϕ)]fort¯lwithto=1;[(L,t\phi)]=t[(L,\phi)]\quad\text{for}\quad t\in\overline{{\mathbb{Q}}}_{l}\quad\text{with}\quad t^{o}=1;
[(L,ϕ)]=0for a traceless object (L,ϕ).[(L,\phi)]=0\quad\text{for a traceless object $(L,\phi)$}.

Notice that by [7, 11.1.8], the 𝒦(ν,ω){\mathcal{K}}(\nu,\omega) is a 𝒪\mathcal{O}^{\prime}-module with a basis corresponding to the set of nonzero simple perverse sheaves LL in 𝒬𝐕,𝐖a/𝒩𝐕\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}} with aLLa^{\ast}L\cong L. Then the following Proposition is an analogy of [7, Proposition 12.5.2].

Proposition 4.5.

Given a nonzero simpe perverse sheaf LL in 𝒬𝐕,𝐖a/𝒩𝐕\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}} such that aLLa^{\ast}L\cong L, then there exists ϕ\phi such that (L,ϕ)(L,\phi) belongs to 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}} and (𝐃L,𝐃(ϕ)1)(L,ϕ)(\mathbf{D}L,\mathbf{D}(\phi)^{-1})\cong(L,\phi) in 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}. In particular, 𝐃LL\mathbf{D}L\cong L. Moreover, such ϕ\phi is unique if oo is odd and is unique up to ±1\pm 1 if oo is even. With the notation above, we denote by πi¯,r(L)=K,\pi_{\underline{i},r}(L)=K, then πi¯,r\pi_{\underline{i},r} is a bijection between the set {L𝒬𝐕,𝐖a/𝒩𝐕|\{L\in{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}| LL is a nonzero simple perverse sheaf with aLLa^{\ast}L\cong L and ti(L)=r}t_{i}(L)=r\} and the set {K𝒬𝐕,𝐖a/𝒩𝐕|\{K\in{\mathcal{Q}}^{a}_{\mathbf{V}^{\prime},\mathbf{W}}/\mathcal{N}_{\mathbf{V}^{\prime}}| KK is a nonzero simple perverse sheaf with aKKa^{\ast}K\cong K and ti(K)=0}t_{i}(K)=0\}.

Proof.

Since an isomorphism ϕ:aLL\phi:a^{\ast}L\rightarrow L in 𝒬𝐕,𝐖a\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}} naturally induces an isomorphism ϕ:aLL\phi:a^{\ast}L\rightarrow L in 𝒬𝐕,𝐖a/𝒩𝐕\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}, it suffices to find ϕ\phi in 𝒬𝐕,𝐖a\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}} such that (L,ϕ)(L,\phi) satisfies the same properties in 𝒬𝐕,𝐖a~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}}.

Given aLLa^{\ast}L\cong L, then by Remark 4.2, there exists an ii such that ti(L)=r>0t_{i}(L)=r>0. Then there exists a unique KK with ti(K)=0t_{i}(K)=0 and aKKa^{\ast}K\cong K such that

i¯(r)KLti(L)>rL[nL].{\mathcal{F}}^{(r)}_{\underline{i}}K\cong L\oplus\bigoplus\limits_{t_{i}(L^{\prime})>r}L^{\prime}[n_{L^{\prime}}].

By induction on dimension vector of 𝐕\mathbf{V} and descending induction on rr, we may assume that there exist ϕ\phi^{\prime} for KK and ϕL\phi_{L^{\prime}} for those LL^{\prime} with ti(L)>rt_{i}(L^{\prime})>r such that (K,ϕ)(K,\phi^{\prime}) and (L,ϕL)(L^{\prime},\phi_{L^{\prime}}) satisfy the condition (𝐃K,𝐃(ϕ)1)(K,ϕ),(𝐃L,𝐃(ϕL)1)(L,ϕL).(\mathbf{D}K,\mathbf{D}(\phi^{\prime})^{-1})\cong(K,\phi^{\prime}),(\mathbf{D}L^{\prime},\mathbf{D}(\phi_{L^{\prime}})^{-1})\cong(L^{\prime},\phi_{L^{\prime}}).

We assume that in 𝒦(ν,ω){\mathcal{K}}(\nu,\omega), i¯(r)[(K,ϕ)]=[(L,ϕ)]+ti(L)>rPL[(L,ϕL)]{\mathcal{F}}^{(r)}_{\underline{i}}[(K,\phi^{\prime})]=[(L,\phi)]+\sum\limits_{t_{i}(L^{\prime})>r}P_{L^{\prime}}[(L^{\prime},\phi_{L^{\prime}})] with PL𝒪P_{L^{\prime}}\in\mathcal{O}^{\prime}.

Apply i¯(r){\mathcal{F}}^{(r)}_{\underline{i}} to 𝐃[(K,ϕ)]=[(K,ϕ)]\mathbf{D}[(K,\phi^{\prime})]=[(K,\phi^{\prime})], we obtain

𝐃[(L,ϕ)]+ti(L)>rP¯L𝐃[(L,ϕL)]=[(L,ϕ)]+ti(L)>rPL[(L,ϕL)],\mathbf{D}[(L,\phi)]+\sum\limits_{t_{i}(L^{\prime})>r}\bar{P}_{L^{\prime}}\mathbf{D}[(L^{\prime},\phi_{L^{\prime}})]=[(L,\phi)]+\sum\limits_{t_{i}(L^{\prime})>r}P_{L^{\prime}}[(L^{\prime},\phi_{L^{\prime}})],

then 𝐃[(L,ϕ)][(L,ϕ)]=ti(L)>r(PLP¯L)𝐃[(L,ϕL)]\mathbf{D}[(L,\phi)]-[(L,\phi)]=\sum\limits_{t_{i}(L^{\prime})>r}(P_{L^{\prime}}\bar{P}_{L^{\prime}})\mathbf{D}[(L^{\prime},\phi_{L^{\prime}})].

Notice that those [(L,ϕ)][(L,\phi)] and [(L,ϕL)][(L^{\prime},\phi_{L^{\prime}})] are contained in an 𝒪\mathcal{O}^{\prime} basis of 𝒦(ν,ω){\mathcal{K}}(\nu,\omega) and 𝐃[(L,ϕ)]\mathbf{D}[(L,\phi)] must be a multiple of one of those elements. Since ti(L)ti(L)t_{i}(L)\neq t_{i}(L^{\prime}), 𝐃L\mathbf{D}L can never be isomorphic to L𝐃LL^{\prime}\cong\mathbf{D}L^{\prime}, the equation 𝐃[(L,ϕ)][(L,ϕ)]=ti(L)>r(PLP¯L)𝐃[(L,ϕL)]\mathbf{D}[(L,\phi)]-[(L,\phi)]=\sum\limits_{t_{i}(L^{\prime})>r}(P_{L^{\prime}}\bar{P}_{L^{\prime}})\mathbf{D}[(L^{\prime},\phi_{L^{\prime}})] forces 𝐃LL\mathbf{D}L\cong L and 𝐃[(L,ϕ)]=[(L,ϕ)]\mathbf{D}[(L,\phi)]=[(L,\phi)]. This proves the existence of ϕ\phi. If ηϕ\eta\phi is another such isomorphism with ηo=1\eta^{o}=1, then by applying 𝐃\mathbf{D}, we get η1=η\eta^{-1}=\eta. Hence η=1\eta=1 if oo is odd and η=±1\eta=\pm 1 if oo is even. The uniqueness follows. ∎

Theorem 4.6.

If we identify the orbit i¯I¯\underline{i}\in\underline{I} of ii with the element iIi^{\prime}\in I^{\prime}, then the functors i¯(n),i¯(n),𝒦±i¯{\mathcal{E}}^{(n)}_{\underline{i}},{\mathcal{F}}^{(n)}_{\underline{i}},{\mathcal{K}}_{\pm\underline{i}} act by Ei(n),Fi(n),K±iE^{(n)}_{i^{\prime}},F^{(n)}_{i^{\prime}},K_{\pm i^{\prime}} on the Grothendieck group 𝒦(ω){\mathcal{K}}(\omega) such that 𝒦(ω){\mathcal{K}}(\omega) becomes an integrable 𝐔𝒪{}_{\mathcal{O}^{\prime}}\mathbf{U}-module and is canonically isomorphic to Λλ𝒪{}_{\mathcal{O}^{\prime}}\Lambda_{\lambda}, where λ=iIωiβi\lambda=\sum\limits_{i^{\prime}\in I^{\prime}}\omega_{i^{\prime}}\beta_{i^{\prime}}.

Proof.

By Proposition 3.11 and 3.12, we know that 𝒦(ω){\mathcal{K}}(\omega) carries a 𝐔𝒪{}_{\mathcal{O}^{\prime}}\mathbf{U} module structure. For any x𝒦(ν,ω)x\in{\mathcal{K}}(\nu,\omega), if n>νin>\nu_{i}, we can see that Ei(n)(x)=0E^{(n)}_{i^{\prime}}(x)=0. If n>νi~,n>\tilde{\nu_{i}}, then Fi(n)(x)F^{(n)}_{i^{\prime}}(x) belongs to 𝒦(ν,ω){\mathcal{K}}(\nu^{\prime},\omega), where ν=ν+ni¯\nu^{\prime}=\nu+n\underline{i} satisfies νi>ν~i\nu^{\prime}_{i}>\tilde{\nu}^{\prime}_{i}. In this case, 𝐄𝐕,𝐖,i¯1=𝐄𝐕,𝐖,Ω\mathbf{E}^{\geqslant 1}_{\mathbf{V},\mathbf{W},\underline{i}}=\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}, we obtain 𝒦(ν,ω)=0{\mathcal{K}}(\nu^{\prime},\omega)=0. Hence 𝒦(ω){\mathcal{K}}(\omega) is an integrable module.

It suffices to show that 𝒦(ω){\mathcal{K}}(\omega) is a highest weight module. Let (ν,ω)\mathcal{M}(\nu,\omega) be the 𝒪\mathcal{O}^{\prime} module spanned by [(L𝝂𝒆,ϕ0)],𝝂𝒮.[(L_{\boldsymbol{\nu}\boldsymbol{e}},\phi_{0})],\boldsymbol{\nu}\in\mathcal{S}^{\prime}. Since i¯(n)(L𝝂𝒆,ϕ0)=(L(ni¯,𝝂𝒆),ϕ0),{\mathcal{F}}^{(n)}_{\underline{i}}(L_{\boldsymbol{\nu}\boldsymbol{e}},\phi_{0})=(L_{(n\underline{i},\boldsymbol{\nu}\boldsymbol{e})},\phi_{0}), ν[I]a(ν,ω)\bigoplus\limits_{\nu\in{\mathbb{N}}[I]^{a}}\mathcal{M}(\nu,\omega) is generated by (L(𝒆),ϕ0)(L_{(\boldsymbol{e})},\phi_{0}) under the action of i¯(n),n,i¯I¯{\mathcal{F}}^{(n)}_{\underline{i}},n\in\mathbb{N},\underline{i}\in\underline{I}.

We claim that if (L,ϕ)(L,\phi) satisfies Proposition 4.5, then [(L,ϕ)][(L,\phi)] belongs to (ν,ω)\mathcal{M}(\nu,\omega). If 𝐕=0\mathbf{V}=0, it is trivial. Otherwise, we can find ii such that ti(L)=r>0.t_{i}(L)=r>0. Then there exists a unique KK such that ti(K)=0t_{i}(K)=0 and

(4) i¯(r)[(K,ϕ)]=[(L,ϕ)]+ti(L)>rPL[(L,ϕL)].{\mathcal{F}}^{(r)}_{\underline{i}}[(K,\phi^{\prime})]=[(L,\phi)]+\sum\limits_{t_{i}(L^{\prime})>r}P_{L^{\prime}}[(L^{\prime},\phi_{L^{\prime}})].

By induction on dimension vector of 𝐕\mathbf{V} and descending induction on rr, we may assume [(K,ϕ)][(K,\phi^{\prime})] and [(L,ϕL)][(L^{\prime},\phi_{L^{\prime}})] belong to ν[I]a(ν,ω)\bigoplus\limits_{\nu\in{\mathbb{N}}[I]^{a}}\mathcal{M}(\nu,\omega). Since ν[I]a(ν,ω)\bigoplus\limits_{\nu\in{\mathbb{N}}[I]^{a}}\mathcal{M}(\nu,\omega) is closed under i¯(r){\mathcal{F}}^{(r)}_{\underline{i}}, we can see that [(L,ϕ)][(L,\phi)] also belongs to (ν,ω)\mathcal{M}(\nu,\omega).

In particular, 𝒦(ω)=ν[I]a(ν,ω){\mathcal{K}}(\omega)=\bigoplus\limits_{\nu\in{\mathbb{N}}[I]^{a}}\mathcal{M}(\nu,\omega) is an integrable highest weight module with the highest weight vector [(L𝒆,ϕ0)][(L_{\boldsymbol{e},\phi_{0}})], hence it is canonically isomorphic to Λλ𝒪{}_{\mathcal{O}^{\prime}}\Lambda_{\lambda}. ∎

4.3. Geometric pairing

Given (L1,ϕ1)(L_{1},\phi_{1}), (L2,ϕ2)(L_{2},\phi_{2}) in 𝒬𝐕,𝐖a/𝒩𝐕~\widetilde{\mathcal{Q}^{a}_{\mathbf{V},\mathbf{W}}/\mathcal{N}_{\mathbf{V}}}, ϕ1\phi_{1} and ϕ2\phi_{2} induce a linear map

Lϕ1,ϕ2:Hom𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕(a𝐃L1,aL2)Hom𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕(𝐃L1,L2)L_{\phi_{1},\phi_{2}}:\mbox{\rm Hom}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(a^{\ast}\mathbf{D}L_{1},a^{\ast}L_{2})\longrightarrow\mbox{\rm Hom}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(\mathbf{D}L_{1},L_{2})
fϕ2f𝐃(ϕ1),f\mapsto\phi_{2}\circ f\circ\mathbf{D}(\phi_{1}),

then ϕ1\phi_{1} and ϕ2\phi_{2} induces an endomorphism aϕ1,ϕ2a_{\phi_{1},\phi_{2}} of Hom𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕(𝐃L1,L2)\mbox{\rm Hom}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(\mathbf{D}L_{1},L_{2}) by

aϕ1,ϕ2:Hom𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕(𝐃L1,L2)Hom𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕(a𝐃L1,aL2)Lϕ1,ϕ2Hom𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕(𝐃L1,L2).\begin{split}a_{\phi_{1},\phi_{2}}:&\mbox{\rm Hom}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(\mathbf{D}L_{1},L_{2})\cong\mbox{\rm Hom}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(a^{\ast}\mathbf{D}L_{1},a^{\ast}L_{2})\\ &\xrightarrow{L_{\phi_{1},\phi_{2}}}\mbox{\rm Hom}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(\mathbf{D}L_{1},L_{2}).\end{split}

Similarly, ϕ1\phi_{1} and ϕ2\phi_{2} also induce linear maps on Ext𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕n(𝐃L1,L2)\mbox{\rm Ext}^{n}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(\mathbf{D}L_{1},L_{2}) for any nn, we still denote these endomorphisms by aϕ1,ϕ2a_{\phi_{1},\phi_{2}}.

Definition 4.7.

Define an 𝒪\mathcal{O}^{\prime}-bilinear form (,)(-,-) on 𝒦(ν,ω){\mathcal{K}}(\nu,\omega) by

([(L1,ϕ1)],[(L2,ϕ2)])λ=ntr(aϕ1,ϕ2,Ext𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕n(𝐃L1,L2))vn.([(L_{1},\phi_{1})],[(L_{2},\phi_{2})])^{\lambda}=\sum\limits_{n}tr(a_{\phi_{1},\phi_{2}},\mbox{\rm Ext}^{n}_{\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/{\mathcal{N}}_{\mathbf{V}}}(\mathbf{D}L_{1},L_{2}))v^{-n}.

We also extend (,)λ(-,-)^{\lambda} to (,)λ:𝒦(ω)𝒦(ω)𝒪(-,-)^{\lambda}:{\mathcal{K}}(\omega)\otimes{\mathcal{K}}(\omega)\rightarrow\mathcal{O}^{\prime} by setting (𝒦(ν,ω),𝒦(ν,ω))λ=0({\mathcal{K}}(\nu,\omega),{\mathcal{K}}(\nu^{\prime},\omega))^{\lambda}=0 for νν\nu\neq\nu^{\prime}.

Notice that if (L1,ϕ1)(L_{1},\phi_{1}) or (L2,ϕ2)(L_{2},\phi_{2}) is traceless, aϕ1,ϕ2a_{\phi_{1},\phi_{2}} acts by permutation and has trace zero, so (,)λ:𝒦(ω)𝒦(ω)𝒪(-,-)^{\lambda}:{\mathcal{K}}(\omega)\otimes{\mathcal{K}}(\omega)\rightarrow\mathcal{O}^{\prime} is well-defined.

Proposition 4.8.

The 𝒪\mathcal{O}^{\prime}-bilinear form (,)λ(-,-)^{\lambda} is contravariant with respect to EiE_{i^{\prime}} and FiF_{i^{\prime}} for any iIi^{\prime}\in I^{\prime},

(Fix,y)λ=(x,viKiEiy)λ.(F_{i^{\prime}}x,y)^{\lambda}=(x,v_{i^{\prime}}K_{-i^{\prime}}E_{i^{\prime}}y)^{\lambda}.

Moreover, for any (L1,ϕ1)(L_{1},\phi_{1}) and (L2,ϕ2)(L_{2},\phi_{2}) satisfy the condition in Proposition 4.5, we have
(1)If [(L1,ϕ1)][(L2,ϕ2)][(L_{1},\phi_{1})]\neq[(L_{2},\phi_{2})], then ([(L1,ϕ1)],[(L2,ϕ2)])λv1𝒪[[v1]]([(L_{1},\phi_{1})],[(L_{2},\phi_{2})])^{\lambda}\in v^{-1}\mathcal{O}[[v^{-1}]] ;
(2)If [(L1,ϕ1)]=[(L2,ϕ2)][(L_{1},\phi_{1})]=[(L_{2},\phi_{2})], then ([(L1,ϕ1)],[(L2,ϕ2)])λ1+v1𝒪[[v1]]([(L_{1},\phi_{1})],[(L_{2},\phi_{2})])^{\lambda}\in 1+v^{-1}\mathcal{O}[[v^{-1}]].

Proof.

Notice that (L1,ϕ1)(L_{1},\phi_{1}) and (L2,ϕ2)(L_{2},\phi_{2}) satisfying the condition in Proposition 4.5 are self-dual, the almost orthogonality follows from the perverse tt-structure of 𝒟G𝐕b(𝐄𝐕,𝐖,Ω)/𝒩𝐕\mathcal{D}^{b}_{G_{\mathbf{V}}}(\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega})/\mathcal{N}_{\mathbf{V}}. By the proof of [2, Proposition 3.31], the functor i{\mathcal{F}}_{i} is left adjoint to i𝒦i[1]{\mathcal{E}}_{i}{\mathcal{K}}_{i}[-1], so i¯{\mathcal{F}}_{\underline{i}} is left adjoint to i¯𝒦i¯[si]{\mathcal{E}}_{\underline{i}}{\mathcal{K}}_{\underline{i}}[-s_{i}]. By definition, (,)λ(-,-)^{\lambda} is contravariant with respect to EiE_{i^{\prime}} and FiF_{i^{\prime}}. ∎

Corollary 4.9.

For 𝛎,𝛎𝒮\boldsymbol{\nu},\boldsymbol{\nu^{\prime}}\in\mathcal{S}^{\prime}, we have ([(L𝛎𝐞,ϕ0)],[(L𝛎𝐞,ϕ0)])λ((v))(v)([(L_{\boldsymbol{\nu}\boldsymbol{e}},\phi_{0})],[(L_{\boldsymbol{\nu^{\prime}}\boldsymbol{e}},\phi_{0})])^{\lambda}\in\mathbb{Z}((v))\cap\mathbb{Q}(v).

Proof.

If one of 𝝂,𝝂𝒮\boldsymbol{\nu},\boldsymbol{\nu^{\prime}}\in\mathcal{S}^{\prime} is empty, the statement holds trivially. Using contravariant property, we can prove the statement by induction on the length of 𝝂,𝝂\boldsymbol{\nu},\boldsymbol{\nu^{\prime}}. ∎

4.4. Signed basis

We recall that a subset \mathcal{B} of a module MM is called a signed basis, if there exists a basis \mathcal{B}^{\prime} such that =\mathcal{B}=\mathcal{B}^{\prime}\cup-\mathcal{B}^{\prime}. In this section, we construct an 𝒜\mathcal{A}-signed basis of Λλ𝒜.{}_{\mathcal{A}}\Lambda_{\lambda}.

Proposition 4.10.

Let 𝒜(ν,ω){}_{\mathcal{A}}\mathcal{M}(\nu,\omega) be the 𝒜\mathcal{A}-submodule of 𝒦(ν,ω){\mathcal{K}}(\nu,\omega) spanned by those [(L𝛎𝐞,ϕ0)],𝛎𝒮[(L_{\boldsymbol{\nu}\boldsymbol{e}},\phi_{0})],\boldsymbol{\nu}\in\mathcal{S}^{\prime}, and 𝒦𝒜(ν,ω){}_{\mathcal{A}}\mathcal{K}(\nu,\omega) be the 𝒜\mathcal{A}-submodule of 𝒦(ν,ω){\mathcal{K}}(\nu,\omega) spanned by those [(L,ϕ)][(L,\phi)] in Proposition 4.5. Then 𝒜(ν,ω)=𝒦𝒜(ν,ω).{}_{\mathcal{A}}\mathcal{M}(\nu,\omega)={{}_{\mathcal{A}}\mathcal{K}(\nu,\omega)}.

Proof.

(1) We first show that 𝒜(ν,ω){}_{\mathcal{A}}\mathcal{M}(\nu,\omega) contains 𝒦𝒜(ν,ω).{}_{\mathcal{A}}\mathcal{K}(\nu,\omega). It suffices to show that any [(L,ϕ)][(L,\phi)] in Proposition 4.5 belongs to 𝒜(ν,ω){}_{\mathcal{A}}\mathcal{M}(\nu,\omega). Similarly as argument in Theorem 4.6, we can find (K,ϕ)(K,\phi^{\prime}) satisfies the following equation i¯(r)[(K,ϕ)]=[(L,ϕ)]+ti(L)>rPL[(L,ϕL)]{\mathcal{F}}^{(r)}_{\underline{i}}[(K,\phi^{\prime})]=[(L,\phi)]+\sum\limits_{t_{i}(L^{\prime})>r}P_{L^{\prime}}[(L^{\prime},\phi_{L^{\prime}})]. We claim that for any t0t\geqslant 0, the coefficient of vtv^{t} in those PLP_{L^{\prime}} are intergers. Otherwise, we take a maximal tt such that there exists a PLP_{L^{\prime}} such that its coefficient ct(L)c_{t}(L^{\prime}) of vtv^{t} is not an integer.

By induction on dimension vector of 𝐕\mathbf{V} and descending induction on rr, we may assume [(K,ϕ)][(K,\phi^{\prime})] and those [(L′′,ϕL′′)][(L^{\prime\prime},\phi_{L^{\prime\prime}})] with ti(L′′)>0t_{i}(L^{\prime\prime})>0 belong to ν[I]a𝒜(ν,ω)\bigoplus\limits_{\nu\in{\mathbb{N}}[I]^{a}}{{}_{\mathcal{A}}\mathcal{M}(\nu,\omega)}. Then (i¯(r)[(K,ϕ)],[(L,ϕL)])λ((v))(v).({\mathcal{F}}^{(r)}_{\underline{i}}[(K,\phi^{\prime})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda}\in\mathbb{Z}((v))\cap\mathbb{Q}(v). In particular, ti(L′′)>rPL′′([(L′′,ϕL′′)],[(L,ϕL)])λ+([(L,ϕ)],[(L,ϕL)])λ((v))(v)\sum\limits_{t_{i}(L^{\prime\prime})>r}P_{L^{\prime\prime}}([(L^{\prime\prime},\phi_{L^{\prime\prime}})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda}+([(L,\phi)],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda}\in\mathbb{Z}((v))\cap\mathbb{Q}(v).

We have ([(L,ϕ)],[(L,ϕL)])λv1𝒪[[v1]]([(L,\phi)],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda}\in v^{-1}\mathcal{O}[[v^{-1}]], then coefficient of vtv^{t} in this term must be zero. If [(L′′,ϕL′′)][(L,ϕL)][(L^{\prime\prime},\phi_{L^{\prime\prime}})]\neq[(L^{\prime},\phi_{L^{\prime}})], then ([(L′′,ϕL′′)],[(L,ϕL)])λv1𝒪[[v1]]((v))=v1[[v1]]([(L^{\prime\prime},\phi_{L^{\prime\prime}})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda}\in v^{-1}\mathcal{O}[[v^{-1}]]\cap\mathbb{Z}((v))=v^{-1}\mathbb{Z}[[v^{-1}]]. By maximality of tt, we can see that the coefficient of vtv^{t} in PL′′([(L′′,ϕL′′)],[(L,ϕL)])λP_{L^{\prime\prime}}([(L^{\prime\prime},\phi_{L^{\prime\prime}})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda} are intergers. Similarly, ([(L,ϕL)],[(L,ϕL)])λ1+v1[[v1]]([(L^{\prime},\phi_{L^{\prime}})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda}\in 1+v^{-1}\mathbb{Z}[[v^{-1}]], hence we can see that the coefficient of vtv^{t} in PL([(L,ϕL)],[(L,ϕL)])λP_{L^{\prime}}([(L^{\prime},\phi_{L^{\prime}})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda} is an interger plus ct(L)c_{t}(L^{\prime}). By consider the coefficient of vtv^{t} in (i¯(r)[(K,ϕ)],[(L,ϕL)])λ({\mathcal{F}}^{(r)}_{\underline{i}}[(K,\phi^{\prime})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda}, we get a contradiction.

Applying the Verdier duality 𝐃\mathbf{D} to Equation (4), we get

i¯(r)[(K,ϕ)]=[(L,ϕ)]+ti(L′′)>rP¯L′′[(L′′,ϕL′′)],{\mathcal{F}}^{(r)}_{\underline{i}}[(K,\phi^{\prime})]=[(L,\phi)]+\sum\limits_{t_{i}(L^{\prime\prime})>r}\bar{P}_{L^{\prime\prime}}[(L^{\prime\prime},\phi_{L^{\prime\prime}})],

so P¯L′′=PL′′\bar{P}_{L^{\prime\prime}}=P_{L^{\prime\prime}} for any L′′L^{\prime\prime}. In particular, the coefficients of vtv^{t} and vtv^{-t} in PL′′P_{L^{\prime\prime}} are the same and PL′′𝒜P_{L^{\prime\prime}}\in\mathcal{A} for any L′′L^{\prime\prime}. Hence [(L,ϕ)][(L,\phi)] belongs to 𝒜(ν,ω){}_{\mathcal{A}}\mathcal{M}(\nu,\omega).

(2) Now we show that 𝒜(ν,ω){}_{\mathcal{A}}\mathcal{M}(\nu,\omega) is contained in 𝒦𝒜(ν,ω).{}_{\mathcal{A}}\mathcal{K}(\nu,\omega). Take [(L𝝂𝒆,ϕ0)][(L_{\boldsymbol{\nu}\boldsymbol{e}},\phi_{0})], it can be 𝒪\mathcal{O}^{\prime}-spanned by those [(L,ϕ)][(L,\phi)], so we may assume

[(L𝝂𝒆,ϕ0)]=(L,ϕL)PL[(L,ϕL)],[(L_{\boldsymbol{\nu}\boldsymbol{e}},\phi_{0})]=\sum\limits_{(L,\phi_{L})}P_{L}[(L,\phi_{L})],

where PL𝒪P_{L}\in\mathcal{O}^{\prime} and (L,ϕL)(L,\phi_{L}) runs over all pairs satisfying the conditions in Proposition 4.5.

We claim the PL𝒜P_{L}\in\mathcal{A}. Otherwise, there exists tt^{\prime} such that the coefficient of vtv^{t^{\prime}} in some PLP_{L} is not a integer. We take the maximal tt^{\prime} with this property, and assume that the coefficient ct(L)c_{t^{\prime}}(L^{\prime}) of vtv^{t^{\prime}} in PLP_{L^{\prime}} is not an integer. Notice that the coefficient of vtv^{t^{\prime}} in ([(L𝝂𝒆,ϕ0)],[(L,ϕL)])λ([(L_{\boldsymbol{\nu}\boldsymbol{e}},\phi_{0})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda} belongs to \mathbb{Z}, but the coefficient of vtv^{t^{\prime}} in (L,ϕL)PL([(L,ϕL)],[(L,ϕL)])λ\sum\limits_{(L,\phi_{L})}P_{L}([(L,\phi_{L})],[(L^{\prime},\phi_{L^{\prime}})])^{\lambda} is ct(L)c_{t^{\prime}}(L^{\prime}) plus a integer by the maximality of tt^{\prime}, then we get a contradiction.

In a conclusion, 𝒜(ν,ω)=𝒦𝒜(ν,ω).{}_{\mathcal{A}}\mathcal{M}(\nu,\omega)={{}_{\mathcal{A}}\mathcal{K}(\nu,\omega)}.

Combine Proposition 4.10 with Theorem 4.6, we get the following theorem.

Theorem 4.11.

The 𝒜\mathcal{A}-submodule 𝒦𝒜(ω)=ν[I]a𝒦𝒜(ν,ω){{}_{\mathcal{A}}\mathcal{K}(\omega)}=\bigoplus\limits_{\nu\in{\mathbb{N}}[I]^{a}}{{}_{\mathcal{A}}\mathcal{K}(\nu,\omega)} is canonically isomorphic to Λλ𝒜.{}_{\mathcal{A}}\Lambda_{\lambda}. Moreover, those [(L,ϕ)][(L,\phi)] in Proposition 4.5 form a signed 𝒜\mathcal{A}-basis of Λλ𝒜{}_{\mathcal{A}}\Lambda_{\lambda}, which is almost orthogonal with respect to the contravariant form (,)λ(-,-)^{\lambda}.

Remark 4.12.

There is a natural vector bundle π𝐕,𝐖:𝐄𝐕,𝐖,Ω𝐄𝐕,Ω\pi_{\mathbf{V},\mathbf{W}}:\mathbf{E}_{\mathbf{V},\mathbf{W},\Omega}\rightarrow\mathbf{E}_{\mathbf{V},\Omega}, its pull-back (π𝐕,𝐖)(\pi_{\mathbf{V},\mathbf{W}})^{\ast} defines a functor (π𝐕,𝐖):𝒬𝐕a𝒬𝐕,𝐖a(\pi_{\mathbf{V},\mathbf{W}})^{\ast}:{\mathcal{Q}}^{a}_{\mathbf{V}}\rightarrow{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}. Compose (π𝐕,𝐖)(\pi_{\mathbf{V},\mathbf{W}})^{\ast} with the localization functor L:𝒬𝐕,𝐖a𝒬𝐕,𝐖a/𝒩𝐕L:{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}\rightarrow{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}, we obtain an additive functor L:𝒬𝐕a𝒬𝐕,𝐖a/𝒩𝐕L^{\prime}:{\mathcal{Q}}^{a}_{\mathbf{V}}\rightarrow{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}. Notice that LL^{\prime} commutes with aa^{\ast}, then LL^{\prime} induces a functor L~:𝒬𝐕a~𝒬𝐕,𝐖a/𝒩𝐕~\tilde{L}^{\prime}:\widetilde{{\mathcal{Q}}^{a}_{\mathbf{V}}}\rightarrow\widetilde{{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}}. This functor categorifies the canonical quotient π:𝐔MλΛλ\pi:\mathbf{U}^{-}\cong M_{\lambda}\rightarrow\Lambda_{\lambda}, where MλM_{\lambda} is the Verma module. In particular, if b=[(L,ϕ)]b=[(L,\phi)] is an element in the signed basis of 𝐔\mathbf{U}^{-} constructed by Lusztig in [7], then π(b)0\pi(b)\neq 0 if and only if LL is nonzero in 𝒬𝐕,𝐖a/𝒩𝐕{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} if and only if [(L,ϕ)][(L,\phi)] is an element in the signed basis of Λλ\Lambda_{\lambda}.

4.5. Application: symmetrizable crystal structure

As shown by Nakajima in [8] and [9], the highest weight integral module of a symmetric Kac-Moody Lie algebra can be realized by the Borel-Moore cohomology groups of his quiver varieties. In this subsection, we will deduce the symmetrizable crystal structure on the set of aa-fixed irreducible components of Nakajima’s quiver variety from our signed basis, which is one of the main result of [11].

In this subsection, all varieties are over the complex field \mathbb{C}. Consider the quiver Q=(I,H,Ω)Q=(I,H,\Omega) with an admissible automorphism aa, and II-graded spaces 𝐖\mathbf{W} and 𝐕\mathbf{V}, where their dimensions are given by vectors vv and ww as described above. Let Ω¯\bar{\Omega} be the set of arrows h¯\bar{h} that have the opposite orientation of hΩh\in\Omega.

Definition 4.13.

Forget the admissible automorphism aa, we define 𝐌=𝐌𝐕,𝐖\mathbf{M}=\mathbf{M}_{\mathbf{V},\mathbf{W}} to be the vector space

hΩHom(𝐕s(h),𝐕t(h))hΩ¯Hom(𝐕s(h),𝐕t(h))iIHom(𝐕i,𝐖i)iIHom(𝐖i,𝐕i),\bigoplus_{h\in\Omega}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{V}_{s(h)},\mathbf{V}_{t(h)})\oplus\bigoplus_{h\in\bar{\Omega}}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{V}_{s(h)},\mathbf{V}_{t(h)})\oplus\bigoplus_{i\in I}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{V}_{i},\mathbf{W}_{i})\oplus\bigoplus_{i\in I}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{W}_{i},\mathbf{V}_{i}),

with 𝐆=𝐆v:=kIGL(𝐕k)\mathbf{G}=\mathbf{G}_{v}:=\prod_{k\in I}\text{GL}(\mathbf{V}_{k}) acting it naturally. The moment map is defined by

μ:𝐌kIHom(𝐕k,𝐕k).\mu:\mathbf{M}\rightarrow\bigoplus_{k\in I}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{V}_{k},\mathbf{V}_{k}).

For BhΩHom(𝐕s(h),𝐕t(h))hΩ¯Hom(𝐕s(h),𝐕t(h))B\in\bigoplus_{h\in\Omega}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{V}_{s(h)},\mathbf{V}_{t(h)})\oplus\bigoplus_{h\in\bar{\Omega}}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{V}_{s(h)},\mathbf{V}_{t(h)}), jkIHom(𝐕k,𝐖k)j\in\bigoplus_{k\in I}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{V}_{k},\mathbf{W}_{k}), and ikIHom(𝐖k,𝐕k)i\in\bigoplus_{k\in I}\mbox{\rm Hom}_{{\mathbb{C}}}(\mathbf{W}_{k},\mathbf{V}_{k}), the moment map can be denoted by:

μ(B,i,j):=hΩΩ¯ε(h)BhBh¯+ij,\mu(B,i,j):=\sum_{h\in\Omega\cup\bar{\Omega}}\varepsilon(h)B_{h}B_{\bar{h}}+ij,

where ε(h)=1\varepsilon(h)=1 if hΩh\in\Omega and ε(h)=1\varepsilon(h)=-1 if hΩ¯h\in\bar{\Omega}.

The affine variety μ1(0)\mu^{-1}(0) defines an affine GIT quotient 0=0(v,w):=SpecA(μ1(0))𝐆{\mathcal{M}}_{0}={\mathcal{M}}_{0}(v,w):=\textbf{Spec}A(\mu^{-1}(0))^{\mathbf{G}}. By fixing a character χ\chi of 𝐆\mathbf{G} such that χ(g)=kIdet(gk)1\chi(g)=\prod_{k\in I}\det(g_{k})^{-1}, the quiver verity is defined by =(v,w):=Projn0A(μ1(0))𝐆,χn{\mathcal{M}}={\mathcal{M}}(v,w):=\textbf{Proj}\bigoplus_{n\geqslant 0}A(\mu^{-1}(0))^{\mathbf{G},\chi^{n}}.

Remark 4.14.

When 𝐖=𝟎\bf W=0 , the variety μ1(0)𝐌𝐕\mu^{-1}(0)\subset\mathbf{M}_{\mathbf{V}} is the moduli space of the preprojective algebra. The Lusztig’s nilpotent variety Λ𝐕\Lambda_{\mathbf{V}} is the subset of μ1(0)𝐌𝐕\mu^{-1}(0)\subset\mathbf{M}_{\mathbf{V}} consisting of nilpotent representations of the preprojective algebra. (See details in [6, Section 12].)

Lift the action of 𝐆\mathbf{G} to μ1(0)×\mu^{-1}(0)\times{\mathbb{C}} by g(B,i,j,z)=(gBg1,gi,jg1,χ(g)1z)g(B,i,j,z)=(gBg^{-1},gi,jg^{-1},\chi(g)^{-1}z), then μ1(0)s:={(B,i,j)μ1(0)𝐆(B,i,j,z)¯(μ1(0)×{0})=,z0}\mu^{-1}(0)^{s}:=\{(B,i,j)\in\mu^{-1}(0)\mid\overline{\mathbf{G}(B,i,j,z)}\cap(\mu^{-1}(0)\times\{0\})=\emptyset,z\neq 0\}. According to Nakajima, 𝐆\mathbf{G} acts transitively on μ1(0)s\mu^{-1}(0)^{s}, and the GIT quotient μ1(0)s/𝐆\mu^{-1}(0)^{s}/\mathbf{G} coincides with the geometric points of {\mathcal{M}}. There is a natural projective map π:0\pi:{\mathcal{M}}\rightarrow{\mathcal{M}}_{0}, with π1(0)\pi^{-1}(0) being a Lagrangian subvariety denoted as =(v,w){\mathcal{L}}={\mathcal{L}}(v,w), which is homotopy equivalent to {\mathcal{M}}.

Nakajima’s quiver variety admits a stratification

(v,w)=r0k,r(v,w),{\mathcal{M}}(v,w)=\bigcup_{r\geqslant 0}{\mathcal{M}}_{k,r}(v,w),

where

k,r(v,w):={(B,i,j)(v,w)codim(imhH,t(h)=kBh+imik)=r}.{\mathcal{M}}_{k,r}(v,w):=\{(B,i,j)\in{\mathcal{M}}(v,w)\mid\text{codim}(\text{im}\oplus_{h\in H,t(h)=k}B_{h}+\text{im}i_{k})=r\}.

There is a natural smooth map, p:k,r(v,w)k,0(vrk,w)p:{\mathcal{M}}_{k,r}(v,w)\rightarrow{\mathcal{M}}_{k,0}(v-rk,w) whose fiber is a connected Grassmannian. It establishes a bijection between the irreducible components of k,r(v,w){\mathcal{L}}_{k,r}(v,w) and k,0(vrk,w){\mathcal{L}}_{k,0}(v-rk,w), where k,r(v,w)=k,r(v,w){\mathcal{L}}_{k,r}(v,w)={\mathcal{L}}\cap{\mathcal{M}}_{k,r}(v,w). For an irreducible component Xk,r(v,w)X\in{\mathcal{L}}_{k,r}(v,w), then the closure X¯\bar{X} of XX is an irreducible component of (v,w){\mathcal{L}}(v,w) and we denote tk(X¯)=rt_{k}(\bar{X})=r. Hence pp induces a bijection by Xp(Xk,r(v,w))¯X\mapsto\overline{p(X\cap{\mathcal{M}}_{k,r}(v,w))} and we denot this bijection by

ρk,r:{XIrr((v,w))tk(X)=r}{XIrr((vrk,w))tk(X)=0}.\rho_{k,r}:\{X\in\text{Irr}({\mathcal{L}}(v,w))\mid t_{k}(X)=r\}\rightarrow\{X\in\text{Irr}({\mathcal{L}}(v-rk,w))\mid t_{k}(X)=0\}.

The admissible automorphism aa acts naturally on (v,w){\mathcal{L}}(v,w) and induces a permutation, which is still denoted by aa, on the set of its irreducible components. Let Irra(v,w)\text{Irr}^{a}{\mathcal{L}}(v,w) be the set of aa-fixed irreducible components. For any irreducible XIrra(v,w)X\in\text{Irr}^{a}{\mathcal{L}}(v,w), if tk(X)=rt_{k}(X)=r, then tk=rt_{k^{\prime}}=r for any kk¯k^{\prime}\in\underline{k}. Then the map ρk¯,r(X)=kk¯ρk,r(X)\rho_{\underline{k},r}(X)=\prod\limits_{k\in\underline{k}}\rho_{k,r}(X) is well-defined. Since each ρk,r\rho_{k,r} is bijective, we can see that ρk¯,r(X)\rho_{\underline{k},r}(X) is also fixed by aa. Hence ρk¯,r\rho_{\underline{k},r} is a bijection between aa-fixed irreducible components.

These maps ρk¯,r\rho_{\underline{k},r} induce a crystal operator on the set vIrra(v,w)\bigcup\limits_{v}\text{Irr}^{a}{\mathcal{L}}(v,w) in the following way:

e~k¯(X)={ρk¯,r11ρk¯,r(X)if tk(X)=r>0,0otherwise;\begin{split}\tilde{e}_{\underline{k}}(X)=\begin{cases}\rho^{-1}_{\underline{k},r-1}\rho_{\underline{k},r}(X)&\text{if }t_{k}(X)=r>0,\\ 0&\text{otherwise};\end{cases}\end{split}

and

f~k¯(X)=ρk¯,r+11ρk¯,r(X),if tk(X)=r.\tilde{f}_{\underline{k}}(X)=\rho^{-1}_{\underline{k},r+1}\rho_{\underline{k},r}(X),\text{if }t_{k}(X)=r.

(The detailed definition of the crystal operators and crystal structures can be seen in [4]. )

For XIrra(v,w)X\in\text{Irr}^{a}{\mathcal{L}}(v,w), we also set ϵk¯(X)=tk(X)\epsilon_{\underline{k}}(X)=t_{k}(X) and φk¯(X)=tk(X)+hk¯,wt(X)\varphi_{\underline{k}}(X)=t_{k}(X)+\langle h_{\underline{k}},wt(X)\rangle, then we have the following theorem.

Theorem 4.15.

Given a symmetrizable generalized Cartan matrix C=(cij)i,jIC=(c_{ij})_{i,j\in I^{\prime}}, let Q=(I,H,Ω)Q=(I,H,\Omega) be an associated quiver with an admissible automorphism aa. We identify I¯\underline{I} with II^{\prime}, then the crystal structure of (vIaIrr((v,w))a,f~k¯,e~k¯,ϵk¯,φk¯,k¯I)(\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w))^{a},\tilde{f}_{\underline{k}},\tilde{e}_{\underline{k}},\epsilon_{\underline{k}},\varphi_{\underline{k}},\underline{k}\in I^{\prime}) is isomorphic to the crystal structure B(λ)B(\lambda) of the irreducible highest weight module Λλ\Lambda_{\lambda} of the quantum group associated to CC.

Proof.

Let \mathcal{B} be the set of [(L,ϕ)][(L,\phi)], where (L,ϕ)𝐃((L,ϕ))(L,\phi)\cong\mathbf{D}((L,\phi)). We say (L,ϕ)(L,\phi) is equivalent to (L,ϕ)(L^{\prime},\phi^{\prime}) if [(L,ϕ)]=±[(L,ϕ)][(L,\phi)]=\pm[(L^{\prime},\phi^{\prime})], and denote the set of equivalent classes of \mathcal{B} by \mathcal{B}^{\prime}. A crystal structure on \mathcal{B}^{\prime} can be described as the following:

e~i¯([(L,ϕ)])={πi¯,r11πi¯,r([(L,ϕ)])if ti(L)=r>0,0otherwise;\begin{split}\tilde{e}_{\underline{i}}([(L,\phi)])=\begin{cases}\pi^{-1}_{\underline{i},r-1}\pi_{\underline{i},r}([(L,\phi)])&\text{if }t_{i}(L)=r>0,\\ 0&\text{otherwise};\end{cases}\end{split}

and

f~i¯([(L,ϕ)])=πi¯,r+11πi¯,r([(L,ϕ)]),if ti(L)=r.\tilde{f}_{\underline{i}}([(L,\phi)])=\pi^{-1}_{\underline{i},r+1}\pi_{\underline{i},r}([(L,\phi)]),\text{if }t_{i}(L)=r.

Apply [7, Thoerem 18.3.8] to the signed basis of 𝒦𝒜(ν,ω)Λλ𝒜{{}_{\mathcal{A}}\mathcal{K}(\nu,\omega)}\cong{{}_{\mathcal{A}}\Lambda_{\lambda}}, we can see that there is an isomorphism of crystals B(λ)\mathcal{B}^{\prime}\cong B(\lambda). Hence it suffices to show vIaIrr((v,w))a\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w))^{a} is isomorphic to \mathcal{B}^{\prime} as crystals.

Denote e~i¯max([(L,ϕ)])=e~i¯r([(L,ϕ)])=[πi¯,r((L,ϕ))]\tilde{e}_{\underline{i}}^{max}([(L,\phi)])=\tilde{e}_{\underline{i}}^{r}([(L,\phi)])=[\pi_{\underline{i},r}((L,\phi))] for those pairs with ti(L)=rt_{i}(L)=r. Similarly, we can define e~i¯max(X)=ρi¯,r(X)\tilde{e}_{\underline{i}}^{max}(X)=\rho_{\underline{i},r}(X) for an irreducible component XX with ti(X)=rt_{i}(X)=r. By [2, Theorem 4.20], there is a bijection Φ\Phi from the set 𝒫𝐖\mathcal{P}_{\mathbf{W}} of nonzero simple perverse sheaves in 𝐕𝒬𝐕,𝐖/𝒩𝐕\coprod\limits_{\mathbf{V}}{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} to the set vIaIrr((v,w))\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w)) of irreducible components such that πi,r(L)=K\pi_{i,r}(L)=K if and only if ρi,r(Φ(L))=Φ(K).\rho_{i,r}(\Phi(L))=\Phi(K).

If we forget those ϕ\phi, then \mathcal{B}^{\prime} is bijective to the set of nonzero simple perverse sheaves in 𝐕𝒬𝐕,𝐖a/𝒩𝐕𝐕𝒬𝐕,𝐖/𝒩𝐕\coprod\limits_{\mathbf{V}}{\mathcal{Q}}^{a}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}}\subset\coprod\limits_{\mathbf{V}}{\mathcal{Q}}_{\mathbf{V},\mathbf{W}}/{\mathcal{N}}_{\mathbf{V}} and πi¯,r(L)=ii¯πi,r(L)\pi_{\underline{i},r}(L)=\prod\limits_{i\in\underline{i}}\pi_{i,r}(L) for any simple perverse sheaf LL with aLLa^{\ast}L\cong L. In particular, Φ\Phi restricts to a bijection Φa\Phi^{a} on the aa-fixed points of 𝒫𝐖\mathcal{P}_{\mathbf{W}} and vIaIrr((v,w))\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w)) such that πi¯,r(L)=K\pi_{\underline{i},r}(L)=K if and only if ρi¯,r(Φa(L))=Φa(K).\rho_{\underline{i},r}(\Phi^{a}(L))=\Phi^{a}(K). Notice that the aa-fixed points of 𝒫𝐖\mathcal{P}_{\mathbf{W}} is naturally bijective to \mathcal{B}^{\prime} and the aa-fixed points of vIaIrr((v,w))\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w)) is vIaIrr((v,w))a\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w))^{a}, we get a bijection Φa:vIaIrr((v,w))a\Phi^{a}:\mathcal{B}^{\prime}\rightarrow\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w))^{a} which intertwines πi¯,r\pi_{\underline{i},r} and ρi¯,r\rho_{\underline{i},r}. Notice that the crystal operator e~i¯,f~i¯\tilde{e}_{\underline{i}},\tilde{f}_{\underline{i}} are uniquely determined by πi¯,r\pi_{\underline{i},r} and ρi¯,r\rho_{\underline{i},r}, we can see that Φa:vIaIrr((v,w))a\Phi^{a}:\mathcal{B}^{\prime}\rightarrow\bigcup\limits_{v\in\mathbb{N}I^{a}}\text{Irr}({\mathcal{L}}(v,w))^{a} is an isomorphism of crystal which is defined by Kashiwara in [4].

Let the dimension vector of 𝐖\mathbf{W} large enough that λ=iIωiβi\lambda=\sum\limits_{i^{\prime}\in I^{\prime}}\omega_{i^{\prime}}\beta_{i^{\prime}} satisfies ωi>νi\omega_{i^{\prime}}>\nu_{i}, for each iIi\in I^{\prime} correspondent to the aorbita-orbit of iIi^{\prime}\in I and B(λ)νB(\lambda)_{\nu} is bijective to B()νB(\infty)_{\nu} for a fixed weight ν\nu. Then we obtain the following corollary, which is another main result in [11].

Corollary 4.16.

Given a symmetrizable Cartan matrix C=(cij)i,jIC=(c_{ij})_{i,j\in I^{\prime}}, let Q=(I,H,Ω)Q=(I,H,\Omega) be an associated quiver with an admissible automorphism aa, let 𝐕Irr(Λ𝐕)a\bigcup\limits_{\mathbf{V}}\text{Irr}(\Lambda_{\bf V})^{a} be the set of aa-fixed irreducible components of Lusztig’s nilpotent variety 𝐕Λ𝐕\coprod\limits_{\mathbf{V}}\Lambda_{\bf V} of the preprojective algebra of QQ. Then 𝐕Irr(Λ𝐕)a\bigcup\limits_{\mathbf{V}}\text{Irr}(\Lambda_{\bf V})^{a} carries a natural crystal structure and is canonically isomorphic to B()B(\infty) of the quantum group associated to CC.

Proof.

Recall that by [8, Lemma 5.8], the set of irreducible components of (v,w){\mathcal{L}}(v,w) is bijective to a subset of irreducible components of Λ𝐕\Lambda_{\bf V} via the projection (B,i,0)B(B,i,0)\mapsto B. Notice that if 𝐖\mathbf{W} is large enough, the map (B,i,0)B(B,i,0)\mapsto B induces a bijection from Irr((v,w))\text{Irr}({\mathcal{L}}(v,w)) to Irr(Λ𝐕)\text{Irr}(\Lambda_{\bf V}), which is compatible with the crystal operator. Restrict this bijection to all aa-fixed irreducible components, we get a proof. ∎

Remark 4.17.

When the auotomorphism aa is trivial, i.e, the Cartan matrix is symmetric, we also agree with the main result of [5] and [10].

References

  • [1] A. A. Beĭlinson, J. Bernstein, and P. Deligne. Faisceaux pervers. In Analysis and topology on singular spaces, I (Luminy, 1981), volume 100 of Astérisque, pages 5–171. Soc. Math. France, Paris, 1982.
  • [2] J. Fang, Y. Lan, and J. Xiao. Lusztig sheaves and integrable highest weight modules, 2023.
  • [3] J. C. Jantzen. Lectures on quantum groups, volume 6 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 1996.
  • [4] M. Kashiwara. On crystal bases of the QQ-analogue of universal enveloping algebras. Duke Math. J., 63(2):465–516, 1991.
  • [5] M. Kashiwara and Y. Saito. Geometric construction of crystal bases. Duke Math. J., 89(1):9–36, 1997.
  • [6] G. Lusztig. Quivers, perverse sheaves, and quantized enveloping algebras. J. Amer. Math. Soc., 4(2):365–421, 1991.
  • [7] G. Lusztig. Introduction to quantum groups, volume 110 of Progress in Mathematics. Birkhäuser Boston, Inc., Boston, MA, 1993.
  • [8] H. Nakajima. Instantons on ALE spaces, quiver varieties, and Kac-Moody algebras. Duke Math. J., 76(2):365–416, 1994.
  • [9] H. Nakajima. Quiver varieties and Kac-Moody algebras. Duke Math. J., 91(3):515–560, 1998.
  • [10] Y. Saito. Crystal bases and quiver varieties. Mathematische Annalen, 324(4):675–688, Dec 2002.
  • [11] A. Savage. A geometric construction of crystal graphs using quiver varieties: extension to the non-simply laced case. In Infinite-dimensional aspects of representation theory and applications, volume 392 of Contemp. Math., pages 133–154. Amer. Math. Soc., Providence, RI, 2005.
  • [12] O. Schiffmann. Lectures on canonical and crystal bases of Hall algebras. In Geometric methods in representation theory. II, volume 24 of Sémin. Congr., pages 143–259. Soc. Math. France, Paris, 2012.