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

On bases of quantum affine algebras

Jie Xiao Department of Mathematics, Tsinghua University, Beijing 100084, P. R. China [email protected] (J. Xiao) Han Xu Department of Mathematics, Tsinghua University, Beijing 100084, P. R. China [email protected] (H. Xu)  and  Minghui Zhao School of Science, Beijing Forestry University, Beijing 100083, P. R. China [email protected] (M. Zhao)
This review is based on the report of Jie Xiao given in the conference ”Forty Years of Algebraic Groups, Algebraic Geometry, and Representation Theory in China” in Jan. 5th, 2020
supported by Tsinghua University Initiative Scientific Research Program (2019Z07L01006)

In Memory of Prof. Xihua Cao

1. Introduction

Let U+\textbf{U}^{+} be the positive part of the quantum group U associated with a generalized Cartan matrix AA.

In the case of finite type, Lusztig constructed the canonical basis B of U+\textbf{U}^{+} via two approaches ([8]). The first one is an elementary algebraic construction via Ringel-Hall algebra realization of U+\textbf{U}^{+}. The isomorphism classes of representations of the corresponding Dynkin quiver form a PBW-type basis of U+\textbf{U}^{+}. By a lemma (Lemma 3.1) of Lusztig, one can construct a bar-invariant basis, which is the canonical basis B. A remarkable characteristic in his construction is that Lusztig reveals the triangular relations among three kind of bases: PBW-basis, monomial basis and canonical basis.

The second one is a geometric construction. In [8] and [9], Lusztig gave a geometric realization of 𝐔+\mathbf{U}^{+} via the category of some semisimple complexes on the variety EνE_{\nu} consisting of representations with dimension vector νI\nu\in\mathbb{N}I of the corresponding quiver QQ. The set of the isomorphism classes of simple perverse sheaves gives the canonical basis 𝐁\mathbf{B} of 𝐔+\mathbf{U}^{+}.

The geometric construction of canonical basis was generalized to the cases of all types in [9] (see also [11]). Furthermore, Lusztig in [10] gave the construction of affine canonical bases by the perverse sheaves associated with tame quivers, an important feature is that those perverse sheaves are indexed by the classes of aperiodic modules of tame quivers and irreducible modules of symmetric groups.

The generalization of this elementary algebraic construction to affine type is an important problem.

In [1] and [2], Beck-Chari-Pressley and Beck-Nakajima defined a PBW-type basis and gave an algebraic construction of canonical basis. However, the meaning of this construction is not clear in terms of representation theory of quivers, in particular, it is not clear how to parametrize the canonical basis in terms of aperiodic modules of tame quivers and irreducible modules of symmetric groups.

In the case of Kronecker quiver, Chen ([3]) and Zhang ([17]) defined a PBW-type basis by using Ringel-Hall algebra realization of U+\textbf{U}^{+}, McGerty ([12]) gave the interpretation of the PBW-type basis of Beck-Nakajima in terms of representation theory of the Kronecker quiver. In the case of cyclic quiver, Deng-Du-Xiao ([4]) defined a PBW-type basis and gave a concrete algebraic construction of the canonical basis by using the triangular relations between the monomial basis and the PBW-basis.

Lin-Xiao-Zhang in [7] provided a process to construct a PBW-type basis of U+\textbf{U}^{+} and the canonical basis B by using Ringel-Hall algebra approach. Recently Xiao-Xu-Zhao ([16]) provided a direct method to construct a PBW-type basis of U+\textbf{U}^{+} and the canonical basis B. Compared with the construction Lin-Xiao-Zhang in [7], the PBW-type basis of Xiao-Xu-Zhao is a [v,v1]\mathbb{Z}[v,v^{-1}]-basis. Particularly the parametrization of the basis by aperiodic modules of tame quivers and irreducible characters of SnS_{n} naturally arises in the construction.

In this paper, we shall review these constructions. Since we shall use representations of quivers, we mainly consider the quantum group corresponding to a symmetric generalized Cartan matrix.

2. Preliminaries

Let II be a finite index set, A=(aij)i,jIA=(a_{ij})_{i,j\in I} be a symmetric generalized Cartan matrix. Denote by 𝐔\mathbf{U} the quantum group associated with the Cartan matrix AA and 𝐔+\mathbf{U}^{+} the positive part of 𝐔\mathbf{U} ([11]).

Fix an indeterminate vv. For any nn\in\mathbb{Z}, set

[n]v=vnvnvv1.[n]_{v}=\frac{v^{n}-v^{-n}}{v-v^{-1}}.

Let [0]v!=1[0]_{v}!=1 and [n]v!=[n]v[n1]v[1]v[n]_{v}!=[n]_{v}[n-1]_{v}\cdots[1]_{v} for any n>0n\in\mathbb{Z}_{>0}.

Note that 𝐔+\mathbf{U}^{+} is an associate algebra over (v)\mathbb{Q}(v) generated by the elements EiE_{i} for various iIi\in I subject to the quantum Serre relations

k=01aij(1)kEi(k)EjEi(1aijk)=0\sum_{k=0}^{1-a_{ij}}(-1)^{k}E_{i}^{(k)}E_{j}E_{i}^{(1-a_{ij}-k)}=0

for all ijIi\neq j\in I, where Ei(n)=Ein/[n]v!E_{i}^{(n)}=E_{i}^{n}/[n]_{v}!.

Let ¯:𝐔+𝐔+\,\,\bar{}:\mathbf{U}^{+}\rightarrow\mathbf{U}^{+} be the unique \mathbb{Q}-algebra involution such that

vn¯=vn and Ei¯=Ei.\overline{v^{n}}=v^{-n}\,\,\,\,\,\,\textrm{ and }\,\,\,\,\,\,\overline{E_{i}}=E_{i}.

Let Q=(I,H,s,t)Q=(I,H,s,t) be a quiver without loops, where II is the set of vertices, HH is the set of arrows and s,t:HIs,t:H\rightarrow I are two maps sending an arrow hHh\in H to the source s(h)s(h) and target t(h)t(h) respectively. Let

aij=|{hH|s(h)=i and t(h)=j}|+|{hH|s(h)=j and t(h)=i}|.a_{ij}=|\{h\in H\,\,|\,\,\textrm{$s(h)=i$ and $t(h)=j$}\}|+|\{h\in H\,\,|\,\,\textrm{$s(h)=j$ and $t(h)=i$}\}|.

Then A=(aij)i,jIA=(a_{ij})_{i,j\in I} is a symmetric generalized Cartan matrix and called the generalized Cartan matrix associated to the quiver QQ. The Euler form on I\mathbb{Z}I is defined as

ν,ν=iIνiνihHνs(h)νt(h)\langle\nu,\nu^{\prime}\rangle=\sum_{i\in I}\nu_{i}\nu^{\prime}_{i}-\sum_{h\in H}\nu_{s(h)}\nu^{\prime}_{t(h)}

and the symmetric Euler form is defined as (ν,ν)=ν,ν+ν,ν(\nu,\nu^{\prime})=\langle\nu,\nu^{\prime}\rangle+\langle\nu^{\prime},\nu\rangle for any ν=iIνii\nu=\sum_{i\in I}\nu_{i}i and ν=iIνii\nu^{\prime}=\sum_{i\in I}\nu^{\prime}_{i}i.

Let k=𝔽qk=\mathbb{F}_{q} be a finite field with qq elements and kQkQ be the path algebra. Denote by repkQ\textrm{rep}_{k}Q the abelian category of finite dimensional representations of QQ over kk. There is an isomorphism between repkQ\textrm{rep}_{k}Q and the category mod-kQ\textrm{mod-}kQ of finite dimensional kQkQ-modules.

Let 𝒫k\mathcal{P}_{k} be the set of isomorphism classes of objects in repkQ\textrm{rep}_{k}Q. For any α𝒫k\alpha\in\mathcal{P}_{k}, choose an object MαrepkQM_{\alpha}\in\textrm{rep}_{k}Q such that the isomorphism classes [Mα]=α[M_{\alpha}]=\alpha. Define the dimension vector of α𝒫k\alpha\in\mathcal{P}_{k} by dim¯α=dim¯Mα\underline{\dim}\,\alpha=\underline{\dim}M_{\alpha}.

Given three elements α1,α2\alpha_{1},\alpha_{2} and α\alpha in 𝒫k\mathcal{P}_{k}, denote by gα1α2αg^{\alpha}_{\alpha_{1}\alpha_{2}} the number of subrepresentations NN of MαM_{\alpha} such that NMα2N\simeq M_{\alpha_{2}} and Mα/NMα1M_{\alpha}/N\simeq M_{\alpha_{1}} in repkQ\textrm{rep}_{k}Q. Let vq=qv_{q}=\sqrt{q}\in\mathbb{C}. The twisted Ringel-Hall algebra (kQ)\mathcal{H}^{\ast}(kQ) is the (vq)\mathbb{Q}(v_{q})-space with basis

{uα|α𝒫k},\{u_{\alpha}\,\,|\,\,\alpha\in\mathcal{P}_{k}\},

whose multiplication is given by

uα1uα2=α𝒫vqdim¯α1,dim¯α2gα1,α2αuα.u_{\alpha_{1}}\ast u_{\alpha_{2}}=\sum_{\alpha\in\mathcal{P}}v_{q}^{\langle\underline{\dim}\alpha_{1},\underline{\dim}\alpha_{2}\rangle}g^{\alpha}_{\alpha_{1},\alpha_{2}}u_{\alpha}.

For any α𝒫k\alpha\in\mathcal{P}_{k}, let

Mα=vqdimMα+dimEndMαuα.\langle{M_{\alpha}}\rangle=v_{q}^{-\dim M_{\alpha}+\dim\textrm{End}M_{\alpha}}u_{\alpha}.

The set {Mα|α𝒫k}\{\langle{M_{\alpha}}\rangle\,\,|\,\,\alpha\in\mathcal{P}_{k}\} is also a (vq)\mathbb{Q}(v_{q})-basis of (kQ)\mathcal{H}^{\ast}(kQ). There is a bilinear form (,)(-,-) on (kQ)\mathcal{H}^{\ast}(kQ) defined in [5].

Denote by 𝒞(kQ)\mathcal{C}^{\ast}(kQ) the composition subalgebra of (kQ)\mathcal{H}^{\ast}(kQ) generated by ui=u[Si]u_{i}=u_{[S_{i}]} for all iIi\in I, where SiS_{i} is the simple kQkQ-module corresponding to iIi\in I.

Then we shall recall the generic form of 𝒞(kQ)\mathcal{C}^{\ast}(kQ). Let 𝒦\mathcal{K} be a set of some finite fields kk such that the set {qk=|k||k𝒦}\{q_{k}=|k|\,\,|\,\,k\in\mathcal{K}\} is an infinite set. Consider the direct product

(Q)=k𝒦(kQ)\mathcal{H}^{\ast}(Q)=\prod_{k\in\mathcal{K}}\mathcal{H}^{\ast}(kQ)

and the elements v=(vqk)k𝒦v=(v_{q_{k}})_{k\in\mathcal{K}}, v1=(vqk1)k𝒦v^{-1}=(v_{q_{k}}^{-1})_{k\in\mathcal{K}} and ui=(ui(k))k𝒦u_{i}=(u_{i}(k))_{k\in\mathcal{K}}. By 𝒞(Q)[v,v1]\mathcal{C}^{\ast}(Q)_{\mathbb{Q}[v,v^{-1}]} we denote the subalgebra of (Q)\mathcal{H}^{\ast}(Q) generated by vv, v1v^{-1} and uiu_{i} over \mathbb{Q}. We may regard it as a [v,v1]\mathbb{Q}[v,v^{-1}]-algebra generated by uiu_{i}, where vv is viewed as an indeterminate. Finally, define 𝒞(Q)=(v)[v,v1]𝒞(Q)[v,v1]\mathcal{C}^{\ast}(Q)=\mathbb{Q}(v)\otimes_{\mathbb{Q}[v,v^{-1}]}\mathcal{C}^{\ast}(Q)_{\mathbb{Q}[v,v^{-1}]}, which is called the generic twisted composition algebra of QQ.

Theorem 2.1 ([5][14]).

Let QQ be a quiver without loops, AA the corresponding generalized Cartan matrix and 𝐔+\mathbf{U}^{+} the positive part of quantum group of type AA. There is an isomorphism of (v)\mathbb{Q}(v)-algebras:

𝒞(Q)\displaystyle\mathcal{C}^{\ast}(Q) \displaystyle\cong 𝐔+\displaystyle\mathbf{U}^{+}
ui\displaystyle u_{i} \displaystyle\mapsto Ei.\displaystyle E_{i}.

Under this isomorphism, the bar involution on 𝐔+\mathbf{U}^{+} induces a bar involution ¯:𝒞(Q)𝒞(Q)\bar{}:\mathcal{C}^{\ast}(Q)\rightarrow\mathcal{C}^{\ast}(Q) such that vn¯=vn\overline{v^{n}}=v^{-n} and ui¯=ui.\overline{u_{i}}=u_{i}.

In [8, 9], Lusztig gave a geometric realization of 𝐔+\mathbf{U}^{+}. Let QQ be a quiver without loops with associated generalized Cartan matrix AA. Lusztig considered the variety EνE_{\nu} consisting of representations with dimension vector νI\nu\in\mathbb{N}I of the quiver QQ over algebraically closed field k¯\bar{k} and the category 𝒬ν\mathcal{Q}_{\nu} of some semisimple complexes of constructible sheaves on EνE_{\nu}. Let K(𝒬ν)K(\mathcal{Q}_{\nu}) be the Grothendieck group of 𝒬ν\mathcal{Q}_{\nu}. Considering all dimension vectors, let

K(𝒬)=νK(𝒬ν).K(\mathcal{Q})=\bigoplus_{\nu}K(\mathcal{Q}_{\nu}).

Lusztig define induction functors on 𝒬ν\mathcal{Q}_{\nu} and get a [v,v1]\mathbb{Z}[v,v^{-1}]-algebra structure on K(𝒬)K(\mathcal{Q}). This algebra is isomorphic to the integral form of 𝐔+\mathbf{U}^{+}. The set 𝐁\mathbf{B} of the isomorphism classes of simple perverse sheaves gives a basis of 𝐔+\mathbf{U}^{+}, which is called the canonical basis.

3. Canonical bases of finite types

Assume that QQ is a Dynkin quiver. In this case, the algebraic construction of canonical basis was introduced by Lusztig in [8] (see also [15]).

Let A=(aij)i,jIA=(a_{ij})_{i,j\in I} be the generalized Cartan matrix associated to the quiver QQ and denote by Δ+\Delta^{+} the set of positive roots of the Lie algebra 𝔤(A)\mathfrak{g}(A) with ii corresponding to simple roots. The dimension vector dim¯\underline{\dim} induces a bijection between the set of isomorphism classes of indecomposable objects ind-𝒫\mathcal{P} and the set Δ+\Delta^{+} by Gabriel theorem. Given a positive root α\alpha, choose an indecomposable representation MαM_{\alpha} of QQ such that [Mα]=α[M_{\alpha}]=\alpha.

Denote by Δ+\mathbb{N}^{\Delta^{+}} the set of all functions ϕ:Δ+\phi:\Delta^{+}\rightarrow\mathbb{N}. For each ϕ:Δ+\phi:\Delta^{+}\rightarrow\mathbb{N}, define a representation

Mϕ=αΔ+Mαϕ(α).M_{\phi}=\bigoplus_{\alpha\in\Delta^{+}}M_{\alpha}^{\oplus\phi(\alpha)}.

Then 𝒫={[Mϕ]|ϕΔ+}\mathcal{P}=\{[M_{\phi}]\,\,|\,\,\phi\in\mathbb{N}^{\Delta^{+}}\}.

Now (kQ)\mathcal{H}^{\ast}(kQ) is spanned by the set

{uϕ=u[Mϕ]|ϕ:Δ+}\{u_{\phi}=u_{[M_{\phi}]}\,\,|\,\,\phi:\Delta^{+}\rightarrow\mathbb{N}\}

as (v)\mathbb{Q}(v)-vector space, which is called a PBW-type basis.

Since QQ is representation-directed, we can define a total order on the set Δ+\Delta^{+} such that

Hom(Mα,Mβ)0αβ\textrm{Hom}(M_{\alpha},M_{\beta})\not=0\Rightarrow\alpha\leq\beta

for any α,βΔ+\alpha,\beta\in\Delta^{+}. This total order induces an order on Δ+\mathbb{N}^{\Delta^{+}}. For any ϕ,ψ:Δ+\phi,\psi:\Delta^{+}\rightarrow\mathbb{N}, define ϕ<ψ\phi<\psi if and only if there exists αΔ+\alpha\in\Delta^{+} such that ϕ(α)>ψ(α)\phi(\alpha)>\psi(\alpha) and ϕ(β)=ψ(β)\phi(\beta)=\psi(\beta) for all α>βΔ+\alpha>\beta\in\Delta^{+}.

For each ϕ:Δ+\phi:\Delta^{+}\rightarrow\mathbb{N}, there exists a monomial mϕm_{\phi} on the divided powers of Chevalley generators uiu_{i} satisfying

mϕ=Mϕ+ϕ<ϕaϕϕMϕ,m_{\phi}=\langle{M_{\phi}}\rangle+\sum_{\phi^{\prime}<\phi}a^{\phi^{\prime}}_{\phi}\langle{M_{\phi^{\prime}}}\rangle,

with aϕϕ[v,v1]a^{\phi^{\prime}}_{\phi}\in\mathbb{Z}[v,v^{-1}]. Since mϕ¯=mϕ\overline{m_{\phi}}=m_{\phi}, we have

Mϕ¯=Mϕ+ϕ<ϕbϕϕMϕ,\overline{\langle{M_{\phi}}\rangle}=\langle{M_{\phi}}\rangle+\sum_{\phi^{\prime}<\phi}b^{\phi^{\prime}}_{\phi}\langle{M_{\phi^{\prime}}}\rangle,

with bϕϕ[v,v1]b^{\phi^{\prime}}_{\phi}\in\mathbb{Z}[v,v^{-1}] such that

  1. (1)

    bϕϕ=1b^{\phi}_{\phi}=1 for all ϕ\phi in Δ+\mathbb{N}^{\Delta^{+}};

  2. (2)

    for all ϕϕ\phi^{\prime}\leq\phi in Δ+\mathbb{N}^{\Delta^{+}},

    ϕ′′,ϕϕ′′ϕbϕ′′ϕ¯bϕϕ′′=δϕ,ϕ.\sum_{\phi^{\prime\prime},\phi^{\prime}\leq\phi^{\prime\prime}\leq\phi}\overline{b^{\phi^{\prime}}_{\phi^{\prime\prime}}}b^{\phi^{\prime\prime}}_{\phi}=\delta_{\phi,\phi^{\prime}}.

Here we need a lemma by Lusztig, which can be obtained by an elementary linear algebra method.

Lemma 3.1 ([11]).

Let HH be a set with a partial order \leq such that for any hhh^{\prime}\leq h in HH, the set {h′′|hh′′h}\{h^{\prime\prime}\,\,|\,\,h^{\prime}\leq h^{\prime\prime}\leq h\} is finite. Assume that for each hhh^{\prime}\leq h in HH, we are given an element rhh[v,v1]r_{h}^{h^{\prime}}\in\mathbb{Z}[v,v^{-1}] such that

  1. (1)

    rhh=1r_{h}^{h}=1 for all hh in HH;

  2. (2)

    for all hhh^{\prime}\leq h in HH,

    h′′,hh′′hrh′′h¯rhh′′=δh,h.\sum_{h^{\prime\prime},h^{\prime}\leq h^{\prime\prime}\leq h}\overline{r^{h^{\prime}}_{h^{\prime\prime}}}r^{h^{\prime\prime}}_{h}=\delta_{h,h^{\prime}}.

Then there is a unique family of elements phh[v1]p_{h}^{h^{\prime}}\in\mathbb{Z}[v^{-1}] defined for all hhh^{\prime}\leq h in HH such that

  1. (1)

    phh=1p_{h}^{h}=1 for all hh in HH;

  2. (2)

    phhv1[v1]p_{h}^{h^{\prime}}\in v^{-1}\mathbb{Z}[v^{-1}] for all hhh^{\prime}\leq h in HH;

  3. (3)

    for all hhh^{\prime}\leq h in HH,

    phh=h′′,hh′′hph′′h¯rhh′′.p_{h}^{h^{\prime}}=\sum_{h^{\prime\prime},h^{\prime}\leq h^{\prime\prime}\leq h}\overline{p_{h^{\prime\prime}}^{h^{\prime}}}r_{h}^{h^{\prime\prime}}.

By Lemma 3.1, there exists a unique family of elements cϕϕ[v1]c^{\phi^{\prime}}_{\phi}\in\mathbb{Z}[v^{-1}] defined for all ϕϕ\phi^{\prime}\leq\phi in Δ+\mathbb{N}^{\Delta^{+}} such that

  1. (1)

    cϕϕ=1c^{\phi}_{\phi}=1 for all ϕ\phi in Δ+\mathbb{N}^{\Delta^{+}};

  2. (2)

    cϕϕv1[v1]c^{\phi^{\prime}}_{\phi}\in v^{-1}\mathbb{Z}[v^{-1}] for all ϕϕ\phi^{\prime}\leq\phi in Δ+\mathbb{N}^{\Delta^{+}};

  3. (3)

    for all ϕϕ\phi^{\prime}\leq\phi in Δ+\mathbb{N}^{\Delta^{+}},

    cϕϕ=ϕ′′,ϕϕ′′ϕcϕ′′ϕ¯bϕϕ′′.c^{\phi^{\prime}}_{\phi}=\sum_{\phi^{\prime\prime},\phi^{\prime}\leq\phi^{\prime\prime}\leq\phi}\overline{c^{\phi^{\prime}}_{\phi^{\prime\prime}}}b^{\phi^{\prime\prime}}_{\phi}.

For any ϕΔ+\phi\in\mathbb{N}^{\Delta^{+}}, let

Cϕ=Mϕ+ϕ<ϕcϕϕMϕ.C_{\phi}=\langle{M_{\phi}}\rangle+\sum_{\phi^{\prime}<\phi}c^{\phi^{\prime}}_{\phi}\langle{M_{\phi^{\prime}}}\rangle.

These formulas hold for every finite fields and may be viewed as formulas in (Q)=𝒞(Q)\mathcal{H}^{\ast}(Q)=\mathcal{C}^{\ast}(Q). Then we have the following theorem.

Theorem 3.2.

The set {Cϕ|ϕ:Δ+}\{C_{\phi}\,\,|\,\,\phi:\Delta^{+}\rightarrow\mathbb{N}\} is a [v,v1]\mathbb{Z}[v,v^{-1}]-basis of (Q)[v,v1]\mathcal{H}^{\ast}(Q)_{\mathbb{Z}[v,v^{-1}]} satisfying the following conditions.

  1. (1)

    Cϕ¯=Cϕ\overline{C_{\phi}}=C_{\phi};

  2. (2)

    (Cϕ,Cϕ)δϕ,ϕ+v1[[v1]](v)(C_{\phi},C_{\phi^{\prime}})\in\delta_{\phi,\phi^{\prime}}+v^{-1}\mathbb{Z}[[v^{-1}]]\cap\mathbb{Q}(v).

Under the isomorphism between (Q)\mathcal{H}^{\ast}(Q) and 𝐔+\mathbf{U}^{+}, the set

{Cϕ|ϕ:Δ+}\{C_{\phi}\,\,|\,\,\phi:\Delta^{+}\rightarrow\mathbb{N}\}

induces a basis of 𝐔+\mathbf{U}^{+}. This basis is just the canonical basis 𝐁\mathbf{B} of 𝐔+\mathbf{U}^{+}, by Theorem 3.2 and the uniqueness of canonical basis of 𝐔+\mathbf{U}^{+}.

Example 3.3.

Take the quiver QQ of type A3A_{3} for example.

Q:123.Q:\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 5.5pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&&\crcr}}}\ignorespaces{\hbox{\kern-5.5pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{1\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 29.5pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 29.5pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{2\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 64.5pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 64.5pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{3}$}}}}}}}\ignorespaces}}}}\ignorespaces.

The AR-quiver is as following.

M(111)\textstyle{M_{(111)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}M(110)\textstyle{M_{(110)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}M(011)\textstyle{M_{(011)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}M(100)\textstyle{M_{(100)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}M(010)\textstyle{M_{(010)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}M(001).\textstyle{M_{(001)}.}

For dimension vector ν=(111)\nu=(111), there exist isomorphism classes of the following modules

[M(111)],[M(110)M(001)],[M(100)M(011)],[M(100)M(010)M(001)].[M_{(111)}],[M_{(110)}\oplus M_{(001)}],[M_{(100)}\oplus M_{(011)}],[M_{(100)}\oplus M_{(010)}\oplus M_{(001)}].

Hence,

{M(111),M(110)M(001),M(100)M(011),M(100)M(010)M(001)}\{\langle{M_{(111)}}\rangle,\langle{M_{(110)}\oplus M_{(001)}}\rangle,\langle{M_{(100)}\oplus M_{(011)}}\rangle,\langle{M_{(100)}\oplus M_{(010)}\oplus M_{(001)}}\rangle\}

are the elements in the PBW-type basis with dimension vector ν=(111)\nu=(111).

By construction, the elements in the corresponding monomial basis are

u1u2u3=\displaystyle u_{1}u_{2}u_{3}= M(111)+v2M(110)M(001)+v2M(100)M(011)\displaystyle\langle{M_{(111)}}\rangle+v^{-2}\langle{M_{(110)}\oplus M_{(001)}}\rangle+v^{-2}\langle{M_{(100)}\oplus M_{(011)}}\rangle
+v3M(100)M(010)M(001),\displaystyle+v^{-3}\langle{M_{(100)}\oplus M_{(010)}\oplus M_{(001)}}\rangle,
u3u1u2=M(110)M(001)+v1M(100)M(010)M(001),u_{3}u_{1}u_{2}=\langle{M_{(110)}\oplus M_{(001)}}\rangle+v^{-1}\langle{M_{(100)}\oplus M_{(010)}\oplus M_{(001)}}\rangle,
u2u3u1=M(100)M(011)+v1M(100)M(010)M(001),u_{2}u_{3}u_{1}=\langle{M_{(100)}\oplus M_{(011)}}\rangle+v^{-1}\langle{M_{(100)}\oplus M_{(010)}\oplus M_{(001)}}\rangle,
u3u2u1=M(100)M(010)M(001),u_{3}u_{2}u_{1}=\langle{M_{(100)}\oplus M_{(010)}\oplus M_{(001)}}\rangle,

which are also the elements with dimension vector (111)(111) in the canonical basis.

It is clear that these PBW-type basis elements are the leading terms of the corresponding canonical basis elements.

4. Beck-Nakajima’s construction

In this section, we shall recall the construction of canonical basis given by Beck-Nakajima ([2][13]).

Let A=(aij)i,jIA=(a_{ij})_{i,j\in I} be a generalized Cartan matrix of affine type, where I={0,1,,n}I=\{0,1,\ldots,n\}, 0I0\in I is the exceptional point and I0=I\{0}I_{0}=I\backslash\{0\}. Let

D=diag(d0,d1,,dn)D=\textrm{diag}(d_{0},d_{1},\ldots,d_{n})

be a diagonal matrix such that DADA is symmetric. Let Δ+\Delta^{+} be the set of positive roots and RR the set of all positive real roots. Let {αi|iI}\{\alpha_{i}\,\,|\,\,i\in I\} be the set of simple roots. Let vi=vdiv_{i}=v^{d_{i}}.

We follow the notations of [2]. Denote by W^\hat{W} the affine Weyl group generated by simple reflections sis_{i} for iIi\in I. Let W~\tilde{W} be the extended affine Weyl group. Then W~=JW^\tilde{W}=J\ltimes\hat{W}, where JJ is a subgroup of the group of Dynkin diagram automorphism and τsi=sτ(i)τW~\tau s_{i}=s_{\tau(i)}\tau\in\tilde{W} for τJ,siW^\tau\in J,s_{i}\in\hat{W}. For any iI0i\in I_{0}, there exists tw~iW~t_{\tilde{w}_{i}}\in\tilde{W} such that

tw~i(αj)={αjif ji,0αidiδif j=iα0+aidiδif j=0,t_{\tilde{w}_{i}}(\alpha_{j})=\left\{\begin{array}[]{cc}\alpha_{j}&\textrm{if $j\neq i,0$}\\ \alpha_{i}-d_{i}\delta&\textrm{if $j=i$}\\ \alpha_{0}+a_{i}d_{i}\delta&\textrm{if $j=0$},\end{array}\right.

where the minimal imaginary positive root δ=aiαi\delta=\sum a_{i}\alpha_{i} and a0=1a_{0}=1.

Let si1si2siNτs_{i_{1}}s_{i_{2}}\cdots s_{i_{N}}\tau be a reduced expression of tw~ntw~n1tw~1t_{\tilde{w}_{n}}t_{\tilde{w}_{n-1}}\cdots t_{\tilde{w}_{1}}. Define an infinite sequence

h=(,i1,i0,i1,)h=(\cdots,i_{-1},i_{0},i_{1},\cdots)

in II such that ik+N=τ(ik)i_{k+N}=\tau(i_{k}) for any kk\in\mathbb{Z}. Let

R<={β0=αi0,β1=si0(αi1),β2=si0si1(αi2),}R_{<}=\{\beta_{0}=\alpha_{i_{0}},\beta_{-1}=s_{i_{0}}(\alpha_{i_{-1}}),\beta_{-2}=s_{i_{0}}s_{i_{-1}}(\alpha_{i_{-2}}),\cdots\}

and

R>={β1=αi1,β2=si1(αi2),β3=si1si2(αi2),}.R_{>}=\{\beta_{1}=\alpha_{i_{1}},\beta_{2}=s_{i_{1}}(\alpha_{i_{2}}),\beta_{3}=s_{i_{1}}s_{i_{2}}(\alpha_{i_{2}}),\cdots\}.

It is well-known that

R=R>R<.R=R_{>}\bigsqcup R_{<}.

For all jIj\in I, denote by TjT_{j} the Lusztig’s symmetries Tj,1′′T^{\prime\prime}_{j,1} in [11]. For any k>0k\in\mathbb{Z}_{>0}, let

Eβk=Ti1Ti2Tik1(Eik).E_{\beta_{k}}=T_{i_{1}}T_{i_{2}}\cdots T_{i_{k-1}}(E_{i_{k}}).

For any k0k\in\mathbb{Z}_{\leq 0}, let

Eβk=Ti01Ti11Tik+11(Eik).E_{\beta_{k}}=T^{-1}_{i_{0}}T^{-1}_{i_{-1}}\cdots T^{-1}_{i_{k+1}}(E_{i_{k}}).

Then EβkE_{\beta_{k}} are the root vectors for the real roots βkR\beta_{k}\in R.

Then we shall define imaginary root vectors. For k>0k>0 and iI0i\in I_{0}, let

Ψ~i,kdi=EkdiδαiEαivi2EαiEkdiδαi,\tilde{\Psi}_{i,kd_{i}}=E_{kd_{i}\delta-\alpha_{i}}E_{\alpha_{i}}-v^{-2}_{i}E_{\alpha_{i}}E_{kd_{i}\delta-\alpha_{i}},

P~i,0=1\tilde{P}_{i,0}=1 and

P~i,kdi={1[2k]ns=1kvn2(sk)Ψ~n,sP~n,ksif i=n and A is of type A2n(2)1[k]is=1kviskΨ~i,sdiP~i,(ks)diotherwise.\tilde{P}_{i,kd_{i}}=\left\{\begin{array}[]{cc}\frac{1}{[2k]_{n}}\sum_{s=1}^{k}v^{2(s-k)}_{n}\tilde{\Psi}_{n,s}\tilde{P}_{n,k-s}&\textrm{if $i=n$ and $A$ is of type $A^{(2)}_{2n}$}\\ \frac{1}{[k]_{i}}\sum_{s=1}^{k}v^{s-k}_{i}\tilde{\Psi}_{i,sd_{i}}\tilde{P}_{i,(k-s)d_{i}}&\textrm{otherwise}.\\ \end{array}\right.

Let 𝒫\mathscr{P} be the set of all partitions and c0:I0𝒫c_{0}:I_{0}\rightarrow\mathscr{P} be a map. For λ=(λ1λ2)𝒫\lambda=(\lambda_{1}\geq\lambda_{2}\geq\cdots)\in\mathscr{P}, define

Sλ=det(P~i,(λkk+m)di)1k,mt,S_{\lambda}=\det(\tilde{P}_{i,(\lambda_{k}-k+m)d_{i}})_{1\leq k,m\leq t},

where tt is the length of λ\lambda. Denote

Sc0=i=1nSc0(i).S_{c_{0}}=\prod_{i=1}^{n}S_{c_{0}(i)}.

Let \mathcal{E} be the set of all such c¯=(c,c0)\bar{c}=(c,c_{0}), where c0:I0𝒫c_{0}:I_{0}\rightarrow\mathscr{P} is a map and c:c:\mathbb{Z}\rightarrow\mathbb{N} is a function with finite support. For any c¯\bar{c}\in\mathcal{E} and pp\in\mathbb{Z}, let

L(c¯,p)\displaystyle L(\bar{c},p) =\displaystyle= (Eip(c(p))Tip1(Eip1(c(p1)))Tip1Tip11(Eip2(c(p2))))\displaystyle\left(E_{i_{p}}^{(c(p))}T_{i_{p}}^{-1}(E_{i_{p-1}}^{(c(p-1))})T_{i_{p}}^{-1}T_{i_{p-1}}^{-1}(E_{i_{p-2}}^{(c(p-2))})\cdots\right)
×\displaystyle\times Tip+1Tip+2Ti0(Sc0)\displaystyle T_{i_{p+1}}T_{i_{p+2}}\cdots T_{i_{0}}(S_{c_{0}})
×\displaystyle\times (Tip+1Tip+2(Eip+3(c(p+3)))Tip+1(Eip+2(c(p+2)))Eip+1(c(p+1))),\displaystyle\left(\cdots T_{i_{p+1}}T_{i_{p+2}}(E_{i_{p+3}}^{(c(p+3))})T_{i_{p+1}}(E_{i_{p+2}}^{(c(p+2))})E_{i_{p+1}}^{(c(p+1))}\right),

where ipi_{p} are from the sequence hh.

For any pp\in\mathbb{Z}, Beck-Nakajima defined a partial ordering <p<_{p} on \mathcal{E} such that the following Theorem holds.

Theorem 4.1 ([2]).

The set {L(c¯,p)|c¯,p}\{L(\bar{c},p)\,\,|\,\,\bar{c}\in\mathcal{E},p\in\mathbb{Z}\} is a [v,v1]\mathbb{Z}[v,v^{-1}]-basis of 𝐔[v,v1]+\mathbf{U}_{\mathbb{Z}[v,v^{-1}]}^{+} such that

  1. (1)

    (L(c¯,p),L(c¯,p))δc¯,c¯+v1[[v1]](v)(L(\bar{c},p),L(\bar{c}^{\prime},p))\in\delta_{\bar{c},\bar{c}^{\prime}}+v^{-1}\mathbb{Z}[[v^{-1}]]\cap\mathbb{Q}(v);

  2. (2)
    L(c¯,p)¯=L(c¯,p)+c¯<pc¯ac¯c¯L(c¯,p)\overline{L(\bar{c},p)}=L(\bar{c},p)+\sum_{\bar{c}^{\prime}<_{p}\bar{c}}a_{\bar{c}\bar{c}^{\prime}}L(\bar{c}^{\prime},p)

    with ac¯c¯(v)a_{\bar{c}\bar{c}^{\prime}}\in\mathbb{Q}(v).

The set {L(c¯,p)|c¯,p}\{L(\bar{c},p)\,\,|\,\,\bar{c}\in\mathcal{E},p\in\mathbb{Z}\} is called a PBW-type basis of 𝐔+\mathbf{U}^{+}.

Beck-Nakajima also proved the following Theorem.

Theorem 4.2 ([2]).

For any c¯\bar{c}\in\mathcal{E} and pp\in\mathbb{Z}, there exists a unique b(c¯,p)𝐔[v,v1]+b(\bar{c},p)\in\mathbf{U}_{\mathbb{Z}[v,v^{-1}]}^{+} satisfying the following conditions

  1. (1)

    b(c¯,p)¯=b(c¯,p)\overline{b(\bar{c},p)}=b(\bar{c},p);

  2. (2)

    (b(c¯,p),b(c¯,p))δc¯,c¯+v1[[v1]](v)(b(\bar{c},p),b(\bar{c}^{\prime},p))\in\delta_{\bar{c},\bar{c}^{\prime}}+v^{-1}\mathbb{Z}[[v^{-1}]]\cap\mathbb{Q}(v);

  3. (3)
    b(c¯,p)=L(c¯,p)+c¯<pc¯ξc¯c¯L(c¯,p)b(\bar{c},p)=L(\bar{c},p)+\sum_{\bar{c}^{\prime}<_{p}\bar{c}}\xi_{\bar{c}\bar{c}^{\prime}}L(\bar{c}^{\prime},p)

    with ξc¯c¯v1[v1]\xi_{\bar{c}\bar{c}^{\prime}}\in v^{-1}\mathbb{Z}[v^{-1}].

Moreover, the set {b(c¯,p)|c¯,p}\{b(\bar{c},p)\,\,|\,\,\bar{c}\in\mathcal{E},p\in\mathbb{Z}\} is the canonical basis of 𝐔+\mathbf{U}^{+}.

5. Kronecker quiver

Let QQ be the Kronecker quiver with I={0,1}I=\{0,1\} and H={ρ1,ρ2}H=\{\rho_{1},\rho_{2}\}:

0\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ρ1\scriptstyle{\rho_{1}}ρ2\scriptstyle{\rho_{2}}1\textstyle{1}

Let kQkQ be the path algebra of QQ over finite field kk.

The set of dimension vectors of indecomposable kQkQ-modules is

Δ+={(l+1,l),(m,m),(n,n+1)|l,m,n,l0,m1,n0}.\Delta^{+}=\{(l+1,l),(m,m),(n,n+1)\,\,|\,\,l,m,n\in\mathbb{Z},l\geq 0,m\geq 1,n\geq 0\}.

The dimension vectors (l+1,l)(l+1,l) and (n,n+1)(n,n+1) correspond to preprojective and preinjective indecomposable kQkQ-modules respectively. The subset consisting of (l+1,l)(l+1,l) (resp. (n,n+1)(n,n+1)) is denoted by PrepPrep (resp. PreiPrei).

For any nn\in\mathbb{N}, let M(n+1,n)M(n+1,n) and M(n,n+1)M(n,n+1) be the indecomposable kQkQ-module of dimension vectors (n+1,n)(n+1,n) and (n,n+1)(n,n+1) respectively. For real root vectors, define

E(n+1,n)=M(n+1,n)E_{(n+1,n)}=\langle{M(n+1,n)}\rangle

and

E(n,n+1)=M(n,n+1).E_{(n,n+1)}=\langle{M(n,n+1)}\rangle.

Then we shall define imaginary root vectors. Let δ=(1,1)\delta=(1,1). As a special case of the definition in Section 4, let

Ψ~k=Ψ~1,k=E(k1,k)E(1,0)v2E(1,0)E(k1,k),\tilde{\Psi}_{k}=\tilde{\Psi}_{1,k}=E_{(k-1,k)}E_{(1,0)}-v^{-2}E_{(1,0)}E_{(k-1,k)},

P~0=1\tilde{P}_{0}=1 and

P~k=1[k]s=1kvskΨ~sP~ks,\tilde{P}_{k}=\frac{1}{[k]}\sum_{s=1}^{k}v^{s-k}\tilde{\Psi}_{s}\tilde{P}_{k-s},

for k>0k>0.

Proposition 5.1 ([3][17][12]).

It holds that

P~k=[M]:dim¯M=kδM is regularvdimMu[M]\tilde{P}_{k}=\sum_{[M]:\underline{\dim}M=k\delta\atop\textrm{$M$ is regular}}v^{-\dim M}u_{[M]}

for k>0k\in\mathbb{Z}_{>0}.

For any partition λ=(λ1λ2λt)\lambda=(\lambda_{1}\geq\lambda_{2}\geq\cdots\geq\lambda_{t}), let

P~λ=P~λ1δP~λ2δP~λtδ\tilde{P}_{\lambda}=\tilde{P}_{\lambda_{1}\delta}\ast\tilde{P}_{\lambda_{2}\delta}\ast\cdots\ast\tilde{P}_{\lambda_{t}\delta}

and

Sλ=det(P(λkk+m)δ)1k,mt.S_{\lambda}=\det(P_{(\lambda_{k}-k+m)\delta})_{1\leq k,m\leq t}.

The relation between P~λ\tilde{P}_{\lambda} and SλS_{\lambda} is

P~λ=μPKλμSμ,\tilde{P}_{\lambda}=\sum_{\mu\in P}K_{\lambda\mu}S_{\mu},

where KλμK_{\lambda\mu} is the Kostka number associated to the partitions λ\lambda and μ\mu.

Theorem 5.2 ([12]).

The set {Sλ|λ is a partition}\{S_{\lambda}\,\,|\,\,\textrm{$\lambda$ is a partition}\} is a subset of the canonical basis 𝐁\mathbf{B}.

Let 𝒢\mathcal{G} be the set of (c=(c,c+),λ)(c=(c_{-},c_{+}),\lambda), where c:Prepc_{-}:Prep\rightarrow\mathbb{N}, c+:Preic_{+}:Prei\rightarrow\mathbb{N} are functions with finite support and λ\lambda is a partition.

For any (c,λ)𝒢(c,\lambda)\in\mathcal{G}, consider

N(c,λ)=M(c)P~λM(c+)N^{\prime}(c,\lambda)=\langle{M(c_{-})}\rangle\ast\tilde{P}_{\lambda}\ast\langle{M(c_{+})}\rangle

and

N(c,λ)=M(c)SλM(c+)N(c,\lambda)=\langle{M(c_{-})}\rangle\ast S_{\lambda}\ast\langle{M(c_{+})}\rangle

where

M(c)=αPrepMαc(α)M(c_{-})=\bigoplus_{\alpha\in Prep}M_{\alpha}^{\oplus c_{-}(\alpha)}

and

M(c+)=αPreiMαc+(α).M(c_{+})=\bigoplus_{\alpha\in Prei}M_{\alpha}^{\oplus c_{+}(\alpha)}.
Proposition 5.3 ([3][17]).

The sets {N(c,λ)|(c,λ)𝒢}\{N(c,\lambda)\,\,|\,\,(c,\lambda)\in\mathcal{G}\} and {N(c,λ)|(c,λ)𝒢}\{N^{\prime}(c,\lambda)\,\,|\,\,(c,\lambda)\in\mathcal{G}\} are two [v,v1]\mathbb{Z}[v,v^{-1}]-bases of 𝒞(Q)[v,v1]\mathcal{C}^{\ast}(Q)_{\mathbb{Z}[v,v^{-1}]}.

The sets {N(c,λ)|(c,λ)𝒢}\{N(c,\lambda)\,\,|\,\,(c,\lambda)\in\mathcal{G}\} and {N(c,λ)|(c,λ)𝒢}\{N^{\prime}(c,\lambda)\,\,|\,\,(c,\lambda)\in\mathcal{G}\} are called PBW-type bases of 𝒞(Q)\mathcal{C}^{\ast}(Q).

6. The construction for cyclic quivers

The construction of various bases of affine AA type was obtained in [4] by considering the Hall algebra of the cyclic quiver. Let QQ be the following cyclic quiver whose vertex set is I={0,1,2,,n}I=\{0,1,2,\ldots,n\}:

0\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}1\textstyle{1\ignorespaces\ignorespaces\ignorespaces\ignorespaces}2\textstyle{2\ignorespaces\ignorespaces\ignorespaces\ignorespaces}3\textstyle{3\ignorespaces\ignorespaces\ignorespaces\ignorespaces}4\textstyle{4\ignorespaces\ignorespaces\ignorespaces\ignorespaces}n\textstyle{n\ignorespaces\ignorespaces\ignorespaces\ignorespaces}

Denote by \mathcal{H}^{\ast} the twisted Ringel-Hall algebra of the category of nilpotent representations of QQ and 𝒞\mathcal{C}^{\ast} the twisted composition subalgebra of \mathcal{H}^{\ast}.

A multisegment is a formal sum

π=iI,l1πil[i,l],\pi=\sum_{i\in I,l\geq 1}\pi_{il}[i,l],

where πil\pi_{il}\in\mathbb{N} and {iI,l1|πil0}\{i\in I,l\geq 1\,\,|\,\,\pi_{il}\neq 0\} is a finite set. Let Π\Pi be the set of multisegments.

There is a bijection between the set Π\Pi and the isomorphism classes of nilpotent representations of QQ. The isomorphism classes corresponding to π\pi is

M(π)=iI,l1Si[l]πil,M(\pi)=\bigoplus_{i\in I,l\geq 1}S_{i}[l]^{\oplus\pi_{il}},

where Si[l]S_{i}[l] is the unique indecomposable kQkQ-module with top SiS_{i} and length ll.

An element πΠ\pi\in\Pi is called aperiodic, if

iIπil=0\prod_{i\in I}\pi_{il}=0

for each l1l\geq 1. The set of all aperiodic multisegments is denoted by Πa\Pi^{a}.

There is a partial order on Π\Pi defined as follows: for π,πΠ\pi^{\prime},\pi\in\Pi with the same dimension vector, π<π\pi^{\prime}<\pi if and only if dimHom(M,M(π))>dimHom(M,M(π))\dim\textrm{Hom}(M,M(\pi^{\prime}))>\dim\textrm{Hom}(M,M(\pi)) for all indecomposable nilpotent representations MM of QQ.

Proposition 6.1 ([4]).

For each πΠa\pi\in\Pi^{a}, there exists a monomial mωπm^{\omega_{\pi}} on the divided powers of uiu_{i} such that

mωπ=M(π)+π<πηππM(π)m^{\omega_{\pi}}=\langle{M(\pi)}\rangle+\sum_{\pi^{\prime}<\pi}\eta^{\pi^{\prime}}_{\pi}\langle{M(\pi^{\prime})}\rangle

with ηππ[v,v1]\eta_{\pi}^{\pi^{\prime}}\in\mathbb{Z}[v,v^{-1}].

Every non-empty subset of Πa\Pi^{a} contains a minimal element. Define EπE_{\pi} for all πΠa\pi\in\Pi^{a} inductively by the following relations. If πΠa\pi\in\Pi^{a} is minimal,

Eπ=mwπ=M(π)+π<π,πΠ\ΠaηππM(π),E_{\pi}={m}^{w_{\pi}}=\langle{M(\pi)}\rangle+\sum_{\pi^{\prime}<\pi,\pi^{\prime}\in\Pi\backslash\Pi^{a}}\eta_{\pi}^{\pi^{\prime}}\langle{M(\pi^{\prime})}\rangle,

If EπE_{\pi^{\prime}} have been defined for all π>πΠa\pi>\pi^{\prime}\in\Pi^{a}, then

Eπ\displaystyle E_{\pi} =\displaystyle= mwππ<π,πΠaηππEπ\displaystyle{m}^{w_{\pi}}-\sum_{\pi^{\prime}<\pi,\pi^{\prime}\in\Pi^{a}}\eta_{\pi}^{\pi^{\prime}}E_{\pi^{\prime}}
=\displaystyle= M(π)+π<π,πΠ\ΠaγππM(π).\displaystyle\langle{M(\pi)}\rangle+\sum_{\pi^{\prime}<\pi,\pi^{\prime}\in\Pi\backslash\Pi^{a}}\gamma_{\pi}^{\pi^{\prime}}\langle{M(\pi^{\prime})}\rangle.
Proposition 6.2 ([4]).

The set {Eπ|πΠa}\{{E}_{\pi}\,\,|\,\,\pi\in\Pi^{a}\} is a [v,v1]\mathbb{Z}[v,v^{-1}] basis of 𝒞[v,v1]\mathcal{C}^{\ast}_{\mathbb{Z}[v,v^{-1}]}, satisfying the following conditions

  1. (1)

    {Eπ|πΠa}\{{E}_{\pi}\,\,|\,\,\pi\in\Pi^{a}\} is independent of the choice of monomials mwπ{m}^{w_{\pi}};

  2. (2)
    Eπ¯=Eπ+π<πrππEπ\overline{{E}_{\pi}}={E}_{\pi}+\sum_{\pi^{\prime}<\pi}r^{\pi^{\prime}}_{\pi}{E}_{\pi^{\prime}}

    with rππ[v,v1]r^{\pi^{\prime}}_{\pi}\in\mathbb{Z}[v,v^{-1}].

The set {Eπ|πΠa}\{{E}_{\pi}\,\,|\,\,\pi\in\Pi^{a}\} is called a PBW-type basis of 𝒞\mathcal{C}^{\ast}.

By Lemma 3.1, there exists a unique family of elements pππ[v1]p^{\pi^{\prime}}_{\pi}\in\mathbb{Z}[v^{-1}] defined for all ππ\pi^{\prime}\leq\pi in Πa\Pi^{a} such that

  1. (1)

    pππ=1p^{\pi}_{\pi}=1 for all πΠa\pi\in\Pi^{a};

  2. (2)

    pππv1[v1]p^{\pi^{\prime}}_{\pi}\in v^{-1}\mathbb{Z}[v^{-1}] for all ππ\pi^{\prime}\leq\pi in Πa\Pi^{a};

  3. (3)

    for all ππ\pi^{\prime}\leq\pi in Πa\Pi^{a},

    pππ=π′′,ππ′′πpπ′′π¯rππ′′.p^{\pi^{\prime}}_{\pi}=\sum_{\pi^{\prime\prime},\pi^{\prime}\leq\pi^{\prime\prime}\leq\pi}\overline{p^{\pi^{\prime}}_{\pi^{\prime\prime}}}r^{\pi^{\prime\prime}}_{\pi}.

For any πΠa\pi\in\Pi^{a}, let

cπ=Eπ+π<πpππEπ.c_{\pi}=E_{\pi}+\sum_{\pi^{\prime}<\pi}p^{\pi^{\prime}}_{\pi}{E}_{\pi^{\prime}}.
Theorem 6.3 ([4]).

The set {cπ|πΠa}\{c_{\pi}\,\,|\,\,\pi\in\Pi^{a}\} is a [v,v1]\mathbb{Z}[v,v^{-1}]-basis of 𝒞[v,v1]\mathcal{C}^{\ast}_{\mathbb{Z}[v,v^{-1}]} satisfying the following conditions.

  1. (1)

    cπ¯=cπ\overline{c_{\pi}}=c_{\pi};

  2. (2)

    (cπ,cπ)δπ,π+v1[[v1]](v)(c_{\pi},c_{\pi^{\prime}})\in\delta_{\pi,\pi^{\prime}}+v^{-1}\mathbb{Z}[[v^{-1}]]\cap\mathbb{Q}(v).

Corollary 6.4 ([4]).

The set {cπ|πΠa}\{c_{\pi}\,\,|\,\,\pi\in\Pi^{a}\} is the canonical basis of 𝒞\mathcal{C}^{\ast}.

7. The construction for tame quivers I

This construction of bases of affine A,D,EA,D,E type was obtained in [7] by using the Ringel-Hall algebra approach. Let QQ be an acyclic quiver of affine type. Give an order on I={0,1,2,,n}I=\{0,1,2,\ldots,n\} such that i>ji>j implies that there doesn’t exist an arrow iji\rightarrow j. Define a double infinite sequence

h=(,i1,i0,i1,)h=(\cdots,i_{-1},i_{0},i_{1},\cdots)

such that ik=ki_{k}=k for all k=0,1,2,,nk=0,1,2,\ldots,n and ik+n+1=iki_{k+n+1}=i_{k} for all kk\in\mathbb{Z}. Then

{β0=αi0,β1=si0(αi1),β2=si0si1(αi2),}\{\beta_{0}=\alpha_{i_{0}},\beta_{-1}=s_{i_{0}}(\alpha_{i_{-1}}),\beta_{-2}=s_{i_{0}}s_{i_{-1}}(\alpha_{i_{-2}}),\cdots\}

is the set of dimension vectors of all indecomposable preprojective modules and

{β1=αi1,β2=si1(αi2),β3=si1si2(αi3),}\{\beta_{1}=\alpha_{i_{1}},\beta_{2}=s_{i_{1}}(\alpha_{i_{2}}),\beta_{3}=s_{i_{1}}s_{i_{2}}(\alpha_{i_{3}}),\cdots\}

is the set of dimension vectors of all indecomposable preinjective modules.

The category repkQ\textrm{rep}_{k}Q has direct sum decomposition

repkQ=PrepRegPrei\textrm{rep}_{k}Q=Prep\oplus Reg\oplus Prei

and each component is closed on taking extensions in repkQ\textrm{rep}_{k}Q and direct summands. Thus the Hall algebras of these components are subalgebras of (kQ)\mathcal{H}^{\ast}(kQ). Generic composition algebra 𝒞(Q)\mathcal{C}^{\ast}(Q) contains the Hall algebras of the components PrepPrep and PreiPrei as subalgebras. Under the isomorphism between 𝒞(Q)\mathcal{C}^{\ast}(Q) and 𝐔+\mathbf{U}^{+}, we can view them as subalgebras of 𝐔+\mathbf{U}^{+}.

Let

M(βk)={Ti01Ti11Tik+11(Eik)if k0,T1Ti2Tik1(Eik)if k>0.\langle{M(\beta_{k})}\rangle=\left\{\begin{array}[]{cc}T^{-1}_{i_{0}}T^{-1}_{i_{-1}}\cdots T^{-1}_{i_{k+1}}(E_{i_{k}})&\textrm{if $k\leq 0$,}\\ T_{1}T_{i_{2}}\cdots T_{i_{k-1}}(E_{i_{k}})&\textrm{if $k>0$.}\\ \end{array}\right.

For a support finite function a:0a:\mathbb{Z}^{\leq 0}\rightarrow\mathbb{N}, define

M(a)\displaystyle\langle{M(a)}\rangle =\displaystyle= k0M(βk)a(k)\displaystyle\langle{\oplus_{k\leq 0}M(\beta_{k})^{\oplus{a}(k)}}\rangle
=\displaystyle= Ei0(a(0))Ti01(Ei1(a(1)))Ti01Ti11(Ei2(a(2))).\displaystyle E_{i_{0}}^{(a(0))}T^{-1}_{i_{0}}(E_{i_{-1}}^{(a(-1))})T^{-1}_{i_{0}}T^{-1}_{i_{-1}}(E_{i_{-2}}^{(a(-2))})\cdots.

For a support finite function b:>0b:\mathbb{Z}^{>0}\rightarrow\mathbb{N}, define

M(b)\displaystyle\langle{M(b)}\rangle =\displaystyle= k>0M(βk)b(k)\displaystyle\langle{\oplus_{k>0}M(\beta_{k})^{\oplus{b}(k)}}\rangle
=\displaystyle= Ti1Ti2(Ei3(b(3)))Ti1(Ei2(b(2)))Ei1(b(1)).\displaystyle\cdots T_{i_{1}}T_{i_{2}}(E_{i_{3}}^{(b(3))})T_{i_{1}}(E_{i_{2}}^{(b(2))})E_{i_{1}}^{(b(1))}.

Note that M(a)\langle{M(a)}\rangle and M(b)\langle{M(b)}\rangle belong to the Hall algebras of the components PrepPrep and PreiPrei respectively.

For regular part, there exist s(s3)s(s\leq 3) non-homogeneous tubes J1,J2,,JsJ_{1},J_{2},\ldots,J_{s}. It is well-known that the full subcategory corresponding to the tube JtJ_{t} is equivalent to the category of nilpotent representations of the cyclic quiver with rtr_{t} vertices, where rtr_{t} is the rank of JtJ_{t}. Choose such an equivalence for each JtJ_{t}, which induces an algebra isomorphism

εt:rt(Jt),\varepsilon_{t}:\mathcal{H}_{r_{t}}^{\ast}\rightarrow\mathcal{H}^{\ast}(J_{t}),

where rt\mathcal{H}_{r_{t}}^{\ast} is the twisted Ringel-Hall algebra corresponding to the cyclic quiver with rtr_{t} vertices and (Jt)\mathcal{H}^{\ast}(J_{t}) is the twisted Ringel-Hall algebra of the full subcategory corresponding to the tube JtJ_{t}.

In Section 6, a PBW-type basis {Eπ|πΠta}\{E_{\pi}\,\,|\,\,\pi\in\Pi^{a}_{t}\} for the composition subalgebra 𝒞rt\mathcal{C}_{r_{t}}^{\ast} of the twisted Ringel-Hall algebra rt\mathcal{H}_{r_{t}}^{\ast} has been constructed. For any πΠta\pi\in\Pi^{a}_{t}, the image of EπE_{\pi} under εt\varepsilon_{t} is still denoted by EπE_{\pi}.

Let K2K_{2} be the path algebra of the Kronecker quiver and F:modkK2modkQF:\textrm{mod}kK_{2}\hookrightarrow\textrm{mod}kQ be the canonical embedding. This embedding induces a map F:(kK2)(kQ)F:\mathcal{H}^{\ast}(kK_{2})\rightarrow\mathcal{H}^{\ast}(kQ).

In Section 5, Ψ~nδ\tilde{\Psi}_{n\delta} and P~nδ\tilde{P}_{n\delta} have been defined. Let

Enδ=F(P~nδ)E_{n\delta}=F(\tilde{P}_{n\delta})

for n>0n\in\mathbb{Z}_{>0} and

Epδ=Ep1δEpsδE_{p\delta}=E_{p_{1}\delta}\ast\cdots\ast E_{p_{s}\delta}

for a partition p=(p1ps)p=(p_{1}\geq\cdots\geq p_{s}).

Let \mathcal{M} be the set of quadruples c=(ac,bc,πc,pc)c=(a_{c},b_{c},\pi_{c},p_{c}), where ac:0a_{c}:\mathbb{Z}^{\leq 0}\rightarrow\mathbb{N} and bc:>0b_{c}:\mathbb{Z}^{>0}\rightarrow\mathbb{N} are functions with finite support, πcΠ1a××Πsa\pi_{c}\in\Pi^{a}_{1}\times\cdots\times\Pi^{a}_{s} and pcp_{c} is a partition. For each cc\in\mathcal{M}, define

Ec=M(ac)Eπ1cEπscEpcδM(bc).E^{c}=\langle{M(a_{c})}\rangle\ast E_{\pi_{1}c}\ast\cdots\ast E_{\pi_{s}c}\ast E_{p_{c}\delta}\ast\langle{M(b_{c})}\rangle.

Recall that EνE_{\nu} is the variety consisting of representations with dimension vector νI\nu\in\mathbb{N}I of the quiver QQ over k¯\bar{k}. For subset 𝒜Eα\mathcal{A}\subset E_{\alpha} and Eβ\mathcal{B}\subset E_{\beta}, define the extension set 𝒜\mathcal{A}\star\mathcal{B} of 𝒜\mathcal{A} by \mathcal{B} to be the set of all zEα+βz\in E_{\alpha+\beta} such that M(z)M(z) is an extension of M(x)M(x) by M(y)M(y) for some x𝒜,yx\in\mathcal{A},y\in\mathcal{B}.

Define the subvariety of EνE_{\nu}

𝒪c=𝒪M(ac)𝒪M(π1c)𝒪M(π2c)𝒪M(πsc)𝒩pc𝒪M(bc)\mathcal{O}_{c}=\mathcal{O}_{M(a_{c})}\star\mathcal{O}_{M(\pi_{1}c)}\star\mathcal{O}_{M(\pi_{2}c)}\star\dots\star\mathcal{O}_{M(\pi_{s}c)}\star\mathcal{N}_{p_{c}}\star\mathcal{O}_{M(b_{c})}

for any cc\in\mathcal{M}, where 𝒩p=𝒩p1𝒩p2𝒩ps\mathcal{N}_{p}=\mathcal{N}_{p_{1}}\star\mathcal{N}_{p_{2}}\star\dots\star\mathcal{N}_{p_{s}} if p=(p1ps)p=(p_{1}\geq\cdots\geq p_{s}) and 𝒩pi\mathcal{N}_{p_{i}} are the union of orbits of images of all regular modules in kK2kK_{2} under FF with dimension vector piδp_{i}\delta.

Proposition 7.1 ([7]).

The set {Ec|c}\{E^{c}\,\,|\,\,c\in\mathcal{M}\} is a (v)\mathbb{Q}(v)-basis of 𝒞(kQ)\mathcal{C}^{\ast}(kQ).

The set {Ec|c}\{E^{c}\,\,|\,\,c\in\mathcal{M}\} is a PBW-type basis of 𝒞(kQ)\mathcal{C}^{\ast}(kQ).

Proposition 7.2 ([7]).

For each cc\in\mathcal{M}, there exists a monomial mcm_{c} on the divided powers of uiu_{i} such that

mc=Ec+dim𝒪c<dim𝒪chccEcm_{c}=E^{c}+\sum_{\dim\mathcal{O}_{c^{\prime}}<\dim\mathcal{O}_{c}}h_{c}^{c^{\prime}}E^{c^{\prime}}

with hcc[v,v1]h_{c}^{c^{\prime}}\in\mathbb{Q}[v,v^{-1}].

Similarly to the case of finite type, from this basis we can get a bar-invariant basis. But it is not the canonical basis considered by Lusztig. Hence in [7], another PBW-type basis is constructed.

There is a bilinear form (,)(-,-) on q(kQ)\mathcal{H}_{q}^{\ast}(kQ) defined in [5]. Consider the (v)\mathbb{Q}(v)-basis {Ec|c}\{E^{{c}}\,\,|\,\,{c}\in\mathcal{M}\}. Let R(𝒞(kQ))R(\mathcal{C}^{\ast}(kQ)) be the (v)\mathbb{Q}(v)-subspace of 𝒞(kQ)\mathcal{C}^{\ast}(kQ) with the basis {Eπ1𝐜Eπ2cEπscEpcδ}\{E_{\pi_{1\mathbf{c}}}\ast E_{\pi_{2{c}}}\ast\cdots\ast E_{\pi_{s{c}}}\ast E_{p_{{c}}\delta}\}, where πc=(π1c,π2c,,πsc)Π1a×Π2a××Πsa\pi_{{c}}=(\pi_{1{c}},\pi_{2{c}},\ldots,\pi_{s{c}})\in\Pi_{1}^{a}\times\Pi_{2}^{a}\times\cdots\times\Pi_{s}^{a}, and pc=(p1p2pt)p_{{c}}=({p_{1}}\geq{p_{2}}\geq\cdots\geq{p_{t}}) is a partition. Note that R(𝒞(kQ))R(\mathcal{C}^{\ast}(kQ)) is a subalgebra of 𝒞(kQ)\mathcal{C}^{\ast}(kQ).

Let Ra(𝒞(kQ))R^{a}(\mathcal{C}^{\ast}(kQ)) be the subalgebra of R(𝒞(kQ))R(\mathcal{C}^{\ast}(kQ)) with the basis {Eπ1cEπ2cEπsc|πc=(π1c,π2c,,πsc)Π1a×Π2a××Πsa}\{E_{\pi_{1{c}}}\ast E_{\pi_{2{c}}}\ast\cdots\ast E_{\pi_{s{c}}}\,\,|\,\,\pi_{{c}}=(\pi_{1{c}},\pi_{2{c}},\ldots,\pi_{s{c}})\in\Pi_{1}^{a}\times\Pi_{2}^{a}\times\cdots\times\Pi_{s}^{a}\}. For any α,βI\alpha,\beta\in\mathbb{N}I, define αβ\alpha\leq\beta if βαI\beta-\alpha\in\mathbb{N}I. If β<δ\beta<\delta, R(𝒞(kQ))β=Ra(𝒞(kQ))βR(\mathcal{C}^{\ast}(kQ))_{\beta}=R^{a}(\mathcal{C}^{\ast}(kQ))_{\beta}. Define δ={xR(𝒞(kQ))δ|(x,Ra(𝒞(kQ))δ)=0}\mathcal{F}_{\delta}=\{x\in R(\mathcal{C}^{\ast}(kQ))_{\delta}\,\,|\,\,(x,R^{a}(\mathcal{C}^{\ast}(kQ))_{\delta})=0\}.

In [7], it is proved that

R(𝒞(kQ))δ=Ra(𝒞(kQ))δδR(\mathcal{C}^{\ast}(kQ))_{\delta}=R^{a}(\mathcal{C}^{\ast}(kQ))_{\delta}\oplus\mathcal{F}_{\delta}

and dimδ=1\dim\mathcal{F}_{\delta}=1. By the method of Schmidt orthogonalization, we may set

Eδ=EδM(πi𝐜),dim¯M(πic)=δ,1isaπicEπic.E^{\prime}_{\delta}=E_{\delta}-\sum_{M(\pi_{i\mathbf{c}}),\underline{\dim}M(\pi_{i{c}})=\delta,1\leq i\leq s}a_{\pi_{i{c}}}E_{\pi_{i{c}}}.

Then δ=(v)Eδ\mathcal{F}_{\delta}=\mathbb{Q}(v)E^{\prime}_{\delta}.

Now let R(𝒞(kQ))(1)R(\mathcal{C}^{\ast}(kQ))(1) be the subalgebra of R(𝒞(kQ))R(\mathcal{C}^{\ast}(kQ)) generated by Ra(𝒞(kQ))R^{a}(\mathcal{C}^{\ast}(kQ)) and δ\mathcal{F}_{\delta}. If β<2δ\beta<2\delta, R(𝒞(kQ))(1)β=R(𝒞(kQ))βR(\mathcal{C}^{\ast}(kQ))(1)_{\beta}=R(\mathcal{C}^{\ast}(kQ))_{\beta}. Define

2δ={xR(𝒞(kQ))2δ|(x,R(𝒞(kQ))(1)2δ)=0}.\mathcal{F}_{2\delta}=\{x\in R(\mathcal{C}^{\ast}(kQ))_{2\delta}\,\,|\,\,(x,R(\mathcal{C}^{\ast}(kQ))(1)_{2\delta})=0\}.

Then dim2δ=1\dim\mathcal{F}_{2\delta}=1 and R(𝒞(kQ))2δ=R(𝒞(kQ))(1)2δ2δR(\mathcal{C}^{\ast}(kQ))_{2\delta}=R(\mathcal{C}^{\ast}(kQ))(1)_{2\delta}\oplus\mathcal{F}_{2\delta}.

Suppose R(𝒞(kQ))(n1)R(\mathcal{C}^{\ast}(kQ))(n-1) has been defined, we define

nδ={xR(𝒞(kQ))nδ|(x,R(𝒞(kQ))(n1)nδ)=0}.\mathcal{F}_{n\delta}=\{x\in R(\mathcal{C}^{\ast}(kQ))_{n\delta}\,\,|\,\,(x,R(\mathcal{C}^{\ast}(kQ))(n-1)_{n\delta})=0\}.

Let R(𝒞(kQ))(n)R(\mathcal{C}^{\ast}(kQ))(n) be the subalgebra of R(𝒞(kQ))R(\mathcal{C}^{\ast}(kQ)) generated by R(𝒞(kQ))(n1)R(\mathcal{C}^{\ast}(kQ))(n-1) and nδ\mathcal{F}_{n\delta}. Then dimnδ=1\dim\mathcal{F}_{n\delta}=1 and R(𝒞(kQ))nδ=R(𝒞(kQ))(n1)nδnδR(\mathcal{C}^{\ast}(kQ))_{n\delta}=R(\mathcal{C}^{\ast}(kQ))(n-1)_{n\delta}\oplus\mathcal{F}_{n\delta}. Similarly, choose EnδE^{\prime}_{n\delta} such that EnδEnδR(𝒞(kQ))(n1)nδE_{n\delta}-E^{\prime}_{n\delta}\in R(\mathcal{C}^{\ast}(kQ))(n-1)_{n\delta} and nδ=(v)Enδ\mathcal{F}_{n\delta}=\mathbb{Q}(v)E^{\prime}_{n\delta} for all n1n\geq 1.

Let Pnδ=nEnδP_{n\delta}=nE^{\prime}_{n\delta} and

Spcδ=det(P((pc)kk+m)δ)1k,mtS^{\prime}_{p_{c}\delta}=\det(P_{((p_{c})_{k}-k+m)\delta})_{1\leq k,m\leq t}

be the Schur functions corresponding to PnδP_{n\delta}.

For each cc\in\mathcal{M}, define

ec=M(ac)Eπ1cEπscSpcδM(bc).e^{c}=\langle{M(a_{c})}\rangle\ast E_{\pi_{1}c}\ast\cdots\ast E_{\pi_{s}c}\ast S^{\prime}_{p_{c}\delta}\ast\langle{M(b_{c})}\rangle.

The set {ec|c}\{e^{c}\,\,|\,\,c\in\mathcal{M}\} is another PBW-type basis of 𝒞(kQ)\mathcal{C}^{\ast}(kQ).

For two c,cc,c^{\prime}\in\mathcal{M}, define ec<ece^{c^{\prime}}<e^{c} if either dim𝒪c<dim𝒪c\dim\mathcal{O}_{c^{\prime}}<\dim\mathcal{O}_{c} or dim𝒪c=dim𝒪c\dim\mathcal{O}_{c^{\prime}}=\dim\mathcal{O}_{c} but pc<pcp_{c}<p_{c^{\prime}} under lexicographic order of partitions.

Proposition 7.3 ([7]).

The set {ec|c}\{e^{c}\,\,|\,\,c\in\mathcal{M}\} is a [v,v1]\mathbb{Q}[v,v^{-1}]-basis of 𝒞(kQ)[v,v1]\mathcal{C}^{\ast}(kQ)_{\mathbb{Q}[v,v^{-1}]} satisfying

  1. (1)

    (ec,ec)δc,c+v1[[v1]](v)(e^{c},e^{c^{\prime}})\in\delta_{c,c^{\prime}}+v^{-1}\mathbb{Q}[[v^{-1}]]\cap\mathbb{Q}(v);

  2. (2)
    mc=ec+ec<ecaccecm_{c}=e^{c}+\sum_{e^{c^{\prime}}<e^{c}}a_{c}^{c^{\prime}}e^{c^{\prime}}

    with acc[v,v1]a_{c}^{c^{\prime}}\in\mathbb{Q}[v,v^{-1}].

Similarly to the case of finite type, Lin-Xiao-Zhang proved the following Theorem by using Lemma 3.1.

Theorem 7.4 ([7]).

For any cc\in\mathcal{M}, there exists a unique c𝒞(Q)[v,v1]\mathcal{E}^{c}\in\mathcal{C}^{\ast}(Q)_{\mathbb{Q}[v,v^{-1}]} satisfying the following conditions

  1. (1)

    c¯=c\overline{\mathcal{E}^{c}}=\mathcal{E}^{c};

  2. (2)

    (c,c)δc,c+v1[[v1]](v)(\mathcal{E}^{c},\mathcal{E}^{c^{\prime}})\in\delta_{c,c^{\prime}}+v^{-1}\mathbb{Q}[[v^{-1}]]\cap\mathbb{Q}(v);

  3. (3)
    c=ec+ec<ecbccec\mathcal{E}^{c}=e^{c}+\sum_{e^{c^{\prime}}<e^{c}}b_{c}^{c^{\prime}}e^{c^{\prime}}

    with bcc[v,v1]b_{c}^{c^{\prime}}\in\mathbb{Q}[v,v^{-1}].

Moreover, the set {c|c}\{\mathcal{E}^{c}\,\,|\,\,c\in\mathcal{M}\} is the canonical basis of 𝒞(Q)\mathcal{C}^{\ast}(Q).

8. The construction for tame quivers II

Let QQ be an acyclic quiver of affine type. Denoted by J1,J2,,Js(s3)J_{1},J_{2},\ldots,J_{s}(s\leq 3) the non-homogeneous tubes. Let 0(Q)\mathcal{H}^{0}(Q) be the (v)\mathbb{Q}(v)-subalgebra of (Q)\mathcal{H}^{\ast}(Q) generated by uiu_{i} for iIi\in I and u[M]u_{[M]} for MJiM\in J_{i}. Note that 𝒞(Q)0(Q)\mathcal{C}^{\ast}(Q)\subset\mathcal{H}^{0}(Q) and 0(Q)\mathcal{H}^{0}(Q) is called the reductive extension of 𝒞(Q)\mathcal{C}^{\ast}(Q).

With the same notations in Section 7, there is a double infinite sequence

h=(,i1,i0,i1,)h=(\cdots,i_{-1},i_{0},i_{1},\cdots)

such that

{β0=αi0,β1=si0(αi1),β2=si0si1(αi2),}\{\beta_{0}=\alpha_{i_{0}},\beta_{-1}=s_{i_{0}}(\alpha_{i_{-1}}),\beta_{-2}=s_{i_{0}}s_{i_{-1}}(\alpha_{i_{-2}}),\cdots\}

is the set of dimension vectors of all indecomposable preprojective modules and

{β1=αi1,β2=si1(αi2),β3=si1si2(αi3),}\{\beta_{1}=\alpha_{i_{1}},\beta_{2}=s_{i_{1}}(\alpha_{i_{2}}),\beta_{3}=s_{i_{1}}s_{i_{2}}(\alpha_{i_{3}}),\cdots\}

is the set of dimension vectors of all indecomposable preinjective modules. We order these βt\beta_{t} for various tt\in\mathbb{Z} by

(β0<β1<β2<)<(β3<β2<β1).(\beta_{0}<\beta_{-1}<\beta_{-2}<\cdots)<(\cdots\beta_{3}<\beta_{2}<\beta_{1}).

For a support finite function c:0c_{-}:\mathbb{Z}^{\leq 0}\rightarrow\mathbb{N}, define

M(c)\displaystyle\langle{M(c_{-})}\rangle =\displaystyle= k0M(βk)c(k)\displaystyle\langle{\oplus_{k\leq 0}M(\beta_{k})^{\oplus{c_{-}}(k)}}\rangle
=\displaystyle= Ei0(c(0))Ti01(Ei1(c(1)))Ti01Ti11(Ei2(c(2))).\displaystyle E_{i_{0}}^{(c_{-}(0))}T^{-1}_{i_{0}}(E_{i_{-1}}^{(c_{-}(-1))})T^{-1}_{i_{0}}T^{-1}_{i_{-1}}(E_{i_{-2}}^{(c_{-}(-2))})\cdots.

For a support finite function c+:>0c_{+}:\mathbb{Z}^{>0}\rightarrow\mathbb{N}, define

M(c+)\displaystyle\langle{M(c_{+})}\rangle =\displaystyle= k>0M(βk)c+(k)\displaystyle\langle{\oplus_{k>0}M(\beta_{k})^{\oplus{c_{+}}(k)}}\rangle
=\displaystyle= Ti1Ti2(Ei3(c+(3)))Ti1(Ei2(c+(2)))Ei1(c+(1)).\displaystyle\cdots T_{i_{1}}T_{i_{2}}(E_{i_{3}}^{(c_{+}(3))})T_{i_{1}}(E_{i_{2}}^{(c_{+}(2))})E_{i_{1}}^{(c_{+}(1))}.

For any c0=(π1,,πs)Π1××Πsc_{0}=(\pi_{1},\ldots,\pi_{s})\in\Pi_{1}\times\cdots\times\Pi_{s}, let

M(c0)=ε1(M(π1))εs(M(πs)),M(c_{0})=\varepsilon_{1}(M(\pi_{1}))\ast\cdots\ast\varepsilon_{s}(M(\pi_{s})),

where

εt:rt(Jt)\varepsilon_{t}:\mathcal{H}_{r_{t}}^{\ast}\rightarrow\mathcal{H}^{\ast}(J_{t})

is the algebra isomorphism in Section 7.

Similarly to the case of Kronecker quiver, define

P~m=[M]:dim¯M=mδM is homogeneous regularvdimMu[M]\tilde{P}_{m}=\sum_{[M]:\underline{\dim}M=m\delta\atop\textrm{$M$ is homogeneous regular}}v^{-\dim M}u_{[M]}

for m0m\in\mathbb{Z}_{\geq 0} and

P~λ=1ktP~λk,\tilde{P}_{\lambda}=\prod_{1\leq k\leq t}\tilde{P}_{\lambda_{k}},
Sλ=det(Pλkk+m)1k,mtS_{\lambda}=\det(P_{\lambda_{k}-k+m})_{1\leq k,m\leq t}

for a partition λ=(λ1λ2λt)\lambda=(\lambda_{1}\geq\lambda_{2}\geq\cdots\geq\lambda_{t}).

We can write in terms of modules

P~λ=[M]:dim¯M=mδM is homogeneous regularAλ[M](v)M,\tilde{P}_{\lambda}=\sum_{[M]:\underline{\dim}M=m\delta\atop\textrm{$M$ is homogeneous regular}}A_{\lambda}^{[M]}(v)\langle{M}\rangle,
Sλ=[M]:dim¯M=mδM is homogeneous regularBλ[M](v)M,S_{\lambda}=\sum_{[M]:\underline{\dim}M=m\delta\atop\textrm{$M$ is homogeneous regular}}B_{\lambda}^{[M]}(v)\langle{M}\rangle,

where Aλ[M](v),Bλ[M](v)[v,v1]A_{\lambda}^{[M]}(v),B_{\lambda}^{[M]}(v)\in\mathbb{Z}[v,v^{-1}].

It is interesting to compute Aλ[M](v),Bλ[M](v)A_{\lambda}^{[M]}(v),B_{\lambda}^{[M]}(v) for some special homogeneous regular MM.

Let 𝒵k\mathcal{Z}_{k} be the set of all homogeneous tubes of modkQ\textrm{mod}\,kQ and by degz\deg z we denote the degree of the corresponding irreducible polynomial of z𝒵kz\in\mathcal{Z}_{k}. We denote by M(l,z)M(l,z) the indecomposable module in tube zz with dimension vector ldzδld_{z}\delta where dz=degzd_{z}=\deg z, that is, ll means the ”level” of the corresponding module.

For a partition μ=(μ1μ2μt)\mu=(\mu_{1}\geq\mu_{2}\geq\cdots\geq\mu_{t}) and z¯=(z1,z2,,zt)\underline{z}=(z_{1},z_{2},\cdots,z_{t}) such that zi𝒵kz_{i}\in\mathcal{Z}_{k} and degzi=1\deg z_{i}=1 for all ii, we denote

M(μ,z¯)=M(μ1,z1)M(μ2,z2)M(μt,zt).M(\mu,\underline{z})=M(\mu_{1},z_{1})\oplus M(\mu_{2},z_{2})\oplus\cdots\oplus M(\mu_{t},z_{t}).

For a partition μ=(μ1μ2μt)\mu=(\mu_{1}\geq\mu_{2}\geq\cdots\geq\mu_{t}) and z¯=(z1,z2,,zt)\underline{z}^{\prime}=(z^{\prime}_{1},z^{\prime}_{2},\cdots,z^{\prime}_{t}) such that zi𝒵kz^{\prime}_{i}\in\mathcal{Z}_{k} and degzi=μi\deg z^{\prime}_{i}=\mu_{i} for all ii, we denote

M[μ,z¯]=M(1,z1)M(1,z2)M(1,zt).M[\mu,\underline{z}^{\prime}]=M(1,z^{\prime}_{1})\oplus M(1,z^{\prime}_{2})\oplus\cdots\oplus M(1,z^{\prime}_{t}).

Note that both M(μ,z¯)M(\mu,\underline{z}) and M[μ,z¯]M[\mu,\underline{z}^{\prime}] have the dimension vector |μ|δ|\mu|\delta.

Let KλμK_{\lambda\mu}\in\mathbb{Z} be the Kostka numbers.

Proposition 8.1 ([16]).

For partitions λ,μ\lambda,\mu with |λ|=|μ||\lambda|=|\mu| and z¯=(z1,z2,,zt)\underline{z}=(z_{1},z_{2},\cdots,z_{t}) such that zi𝒵kz_{i}\in\mathcal{Z}_{k} and degzi=1\deg z_{i}=1 for all ii, Bλ[M(μ,z)](v)=v|λ|KμλB_{\lambda}^{[M(\mu,z)]}(v)=v^{-|\lambda|}K_{\mu\lambda}.

For a partition λ\lambda of mm, let SλS^{\lambda} be the Specht module for 𝒮m\mathcal{S}_{m}. Let tλ(μ)=χSλ(gμ)t_{\lambda}(\mu)=\chi_{S^{\lambda}}(g_{\mu}) be the complex character value of SλS^{\lambda} at gμ𝒮mg_{\mu}\in\mathcal{S}_{m} of cycle type μ\mu. Then (tλ(μ))λ,μ(t_{\lambda}(\mu))_{\lambda,\mu} is the character table of 𝒮m\mathcal{S}_{m}. Let tλt^{\prime}_{\lambda} be the character of the permutation module MλM^{\lambda}. It is known that tλ=μKλμtμt^{\prime}_{\lambda}=\sum_{\mu}K_{\lambda\mu}t_{\mu}.

Proposition 8.2 ([16]).

For partitions λ,μ\lambda,\mu with |λ|=|μ||\lambda|=|\mu| and z¯=(z1,z2,,zt)\underline{z}^{\prime}=(z^{\prime}_{1},z^{\prime}_{2},\cdots,z^{\prime}_{t}) such that zi𝒵kz^{\prime}_{i}\in\mathcal{Z}_{k} and degzi=μi\deg z^{\prime}_{i}=\mu_{i} for all ii, Aλ[M[μ,z]](v)=v|λ|tλ(μ)A_{\lambda}^{[M[\mu,z^{\prime}]]}(v)=v^{-|\lambda|}t^{\prime}_{\lambda}(\mu).

Corollary 8.3 ([16]).

For partitions λ,μ\lambda,\mu with |λ|=|μ||\lambda|=|\mu| and z¯=(z1,z2,,zt)\underline{z}^{\prime}=(z^{\prime}_{1},z^{\prime}_{2},\cdots,z^{\prime}_{t}) such that zi𝒵kz^{\prime}_{i}\in\mathcal{Z}_{k} and degzi=μi\deg z^{\prime}_{i}=\mu_{i} for all ii, Bλ[M[μ,z]](v)=v|λ|tλ(μ)B_{\lambda}^{[M[\mu,z^{\prime}]]}(v)=v^{-|\lambda|}t_{\lambda}(\mu).

Let 𝒢\mathcal{G} be the set of (c=(c,c0,c+),tλ)(c=(c_{-},c_{0},c_{+}),t_{\lambda}), where c:0c_{-}:\mathbb{Z}^{\leq 0}\rightarrow\mathbb{N} (resp. c+:>0c_{+}:\mathbb{Z}^{>0}\rightarrow\mathbb{N}) is function with finite support, c0Π1××Πsc_{0}\in\Pi_{1}\times\cdots\times\Pi_{s} and tλt_{\lambda} is the character of a Specht module SλS^{\lambda}. Let 𝒢a\mathcal{G}^{a} the subset of 𝒢\mathcal{G} consisting of all such (c,tλ)(c,t_{\lambda}) such that c0Π1a××Πsac_{0}\in\Pi^{a}_{1}\times\cdots\times\Pi^{a}_{s}.

For any (c,tλ)𝒢(c,t_{\lambda})\in\mathcal{G}, consider

N(c,tλ)=M(c)M(c0)SλM(c+),N(c,t_{\lambda})=\langle{M(c_{-})}\rangle\ast\langle{M(c_{0})}\rangle\ast S_{\lambda}\ast\langle{M(c_{+})}\rangle,

where c=(c,c0,c+)c=(c_{-},c_{0},c_{+}).

Proposition 8.4 ([16]).

The set {N(c,tλ)|(c,tλ)𝒢}\{N(c,t_{\lambda})\,\,|\,\,(c,t_{\lambda})\in\mathcal{G}\} is an (v)\mathbb{Q}(v)-basis of 0\mathcal{H}^{0} such that

  1. (1)

    (N(c,tλ),N(c,tλ))δ(c,tλ),(c,tλ)+v1[[v1]](v)(N(c,t_{\lambda}),N(c^{\prime},t_{\lambda}^{\prime}))\in\delta_{(c,t_{\lambda}),(c^{\prime},t_{\lambda}^{\prime})}+v^{-1}\mathbb{Q}[[v^{-1}]]\cap\mathbb{Q}(v);

  2. (2)
    N(c,tλ)N(c,tλ)=(c′′,tλ′′)𝒢P(c,tλ),(c,tλ)(c′′,tλ′′)N(c′′,tλ′′)N(c,t_{\lambda})\ast N(c^{\prime},t_{\lambda}^{\prime})=\sum_{(c^{\prime\prime},t_{\lambda}^{\prime\prime})\in\mathcal{G}}P_{(c,t_{\lambda}),(c^{\prime},t_{\lambda}^{\prime})}^{(c^{\prime\prime},t_{\lambda}^{\prime\prime})}N(c^{\prime\prime},t_{\lambda}^{\prime\prime})

    with P(c,tλ),(c,tλ)(c′′,tλ′′)[v,v1]P_{(c,t_{\lambda}),(c^{\prime},t_{\lambda}^{\prime})}^{(c^{\prime\prime},t_{\lambda}^{\prime\prime})}\in\mathbb{Z}[v,v^{-1}].

There is a ”combinatorial” order << on 𝒢\mathcal{G} defined as follows. For c,c:0c_{-},c_{-}^{\prime}:\mathbb{Z}^{\leq 0}\rightarrow\mathbb{N}, define c<cc_{-}<c_{-}^{\prime} if and only if there exists j0j\leq 0 such that c(t)=c(t)c_{-}(t)=c_{-}^{\prime}(t) for all j<t0j<t\leq 0 and c(j)>c(j)c_{-}(j)>c_{-}^{\prime}(j). For c+,c+:>0c_{+},c_{+}^{\prime}:\mathbb{Z}^{>0}\rightarrow\mathbb{N}, define c+<c+c_{+}<c_{+}^{\prime} if and only if there exists j>0j>0 such that c(t)=c(t)c_{-}(t)=c_{-}^{\prime}(t) for all j>t>0j>t>0 and c(j)>c(j)c_{-}(j)>c_{-}^{\prime}(j). The partial order on Π\Pi is given in Section 6. For tλt_{\lambda} and tλt_{\lambda}^{\prime}, tλ<tλt_{\lambda}<t_{\lambda^{\prime}} means that λ\lambda is less than λ\lambda^{\prime} under lexicographic order of partitions.

Definition 8.5.

For (c,tλ),(c,tλ)𝒢(c,t_{\lambda}),(c^{\prime},t_{\lambda^{\prime}})\in\mathcal{G}, let c=(c,c0,c+)c=(c_{-},c_{0},c_{+}), c=(c,c0,c+)c^{\prime}=(c^{\prime}_{-},c^{\prime}_{0},c^{\prime}_{+}), c0=(π1,,πs)c_{0}=(\pi_{1},\ldots,\pi_{s}) and c0=(π1,,πs)c^{\prime}_{0}=(\pi^{\prime}_{1},\ldots,\pi^{\prime}_{s}). Define (c,tλ)<(c,tλ)(c^{\prime},t_{\lambda^{\prime}})<(c,t_{\lambda}) if one of the following three conditions holds.

  1. (1)

    c=cc=c^{\prime} and tλ>tλt_{\lambda^{\prime}}>t_{\lambda};

  2. (2)

    ccc^{\prime}_{-}\leq c_{-}, c+c+c^{\prime}_{+}\leq c_{+} but not all equalities hold;

  3. (3)

    c=cc_{-}=c_{-}^{\prime}, c+=c+c_{+}=c_{+}^{\prime}, π1π1,,πsπs\pi^{\prime}_{1}\leq\pi_{1},\ldots,\pi^{\prime}_{s}\leq\pi_{s} but not all equalities hold.

Proposition 8.6 ([16]).

For each (c,tλ)𝒢a(c,t_{\lambda})\in\mathcal{G}^{a}, there exists a monomial mω(c,tλ)m^{\omega(c,t_{\lambda})} on the divided powers of uiu_{i} such that

mω(c,tλ)=N(c,tλ)+(c,tλ)<(c,tλ)ac,tλc,tλN(c,tλ)m^{\omega(c,t_{\lambda})}=N(c,t_{\lambda})+\sum_{(c^{\prime},t_{\lambda^{\prime}})<(c,t_{\lambda})}a_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}N(c^{\prime},t_{\lambda^{\prime}})

with ac,tλc,tλ[v,v1]a_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}\in\mathbb{Z}[v,v^{-1}].

Li gave the geometric construction of this monomial basis in [6].

With this partial order on 𝒢a\mathcal{G}^{a}, every nonempty subset has a minimal element. Define E(c,tλ)E{(c,t_{\lambda})} for all (c,tλ)𝒢a(c,t_{\lambda})\in\mathcal{G}^{a} inductively by the following relations. If (c,tλ)𝒢a(c,t_{\lambda})\in\mathcal{G}^{a} is minimal,

E(c,tλ)=mω(c,tλ)=N(c,tλ)+(c,tλ)<(c,tλ)(c,tλ)𝒢\𝒢aac,tλc,tλN(c,tλ).E{(c,t_{\lambda})}=m^{\omega(c,t_{\lambda})}=N(c,t_{\lambda})+\sum_{(c^{\prime},t_{\lambda^{\prime}})<(c,t_{\lambda})\atop(c^{\prime},t_{\lambda^{\prime}})\in\mathcal{G}\backslash\mathcal{G}^{a}}a_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}N(c^{\prime},t_{\lambda^{\prime}}).

If E(c,tλ)E{(c^{\prime},t_{\lambda^{\prime}})} have been defined for all (c,tλ)>(c,tλ)𝒢a(c,t_{\lambda})>(c^{\prime},t_{\lambda^{\prime}})\in\mathcal{G}^{a}, then

E(c,tλ)\displaystyle E{(c,t_{\lambda})} =\displaystyle= mω(c,tλ)(c,tλ)<(c,tλ)(c,tλ)𝒢aac,tλc,tλE(c,tλ)\displaystyle m^{\omega(c,t_{\lambda})}-\sum_{(c^{\prime},t_{\lambda^{\prime}})<(c,t_{\lambda})\atop(c^{\prime},t_{\lambda^{\prime}})\in\mathcal{G}^{a}}a_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}E(c^{\prime},t_{\lambda^{\prime}})
=\displaystyle= N(c,tλ)+(c,tλ)<(c,tλ)(c,tλ)𝒢\𝒢abc,tλc,tλN(c,tλ).\displaystyle N(c,t_{\lambda})+\sum_{(c^{\prime},t_{\lambda^{\prime}})<(c,t_{\lambda})\atop(c^{\prime},t_{\lambda^{\prime}})\in\mathcal{G}\backslash\mathcal{G}^{a}}b_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}N(c^{\prime},t_{\lambda^{\prime}}).
Proposition 8.7 ([16]).

The set {E(c,tλ)|(c,tλ)𝒢a}\{E{(c,t_{\lambda})}\,\,|\,\,(c,t_{\lambda})\in\mathcal{G}^{a}\} is a [v,v1]\mathbb{Z}[v,v^{-1}] basis of 𝒞(Q)\mathcal{C}^{\ast}(Q), such that

  1. (1)

    {E(c,tλ)|(c,tλ)𝒢a}\{E{(c,t_{\lambda})}\,\,|\,\,(c,t_{\lambda})\in\mathcal{G}^{a}\} is independent of the choice of monomials mω(c,tλ)m^{\omega(c,t_{\lambda})}.

  2. (2)
    E(c,tλ)¯=E(c,tλ)+(c,tλ)<(c,tλ)(c,tλ)𝒢aγc,tλc,tλE(c,tλ)\overline{E{(c,t_{\lambda})}}=E{(c,t_{\lambda})}+\sum_{(c^{\prime},t_{\lambda^{\prime}})<(c,t_{\lambda})\atop(c^{\prime},t_{\lambda^{\prime}})\in\mathcal{G}^{a}}\gamma_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}E(c^{\prime},t_{\lambda^{\prime}})

    with γc,tλc,tλ[v,v1]\gamma_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}\in\mathbb{Z}[v,v^{-1}].

The set {E(c,tλ)|(c,tλ)𝒢a}\{E{(c,t_{\lambda})}\,\,|\,\,(c,t_{\lambda})\in\mathcal{G}^{a}\} is called a PBW-type basis of 𝒞(Q)\mathcal{C}^{\ast}(Q).

By Lemma 3.1, there exists a unique family of elements ζc,tλc,tλ[v1]\zeta_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}\in\mathbb{Z}[v^{-1}] defined for all (c,tλ)(c,tλ)(c^{\prime},t_{\lambda^{\prime}})\leq(c,t_{\lambda}) in 𝒢a\mathcal{G}^{a} such that

  1. (1)

    ζc,tλc,tλ=1\zeta_{c,t_{\lambda}}^{c,t_{\lambda}}=1 for all (c,tλ)𝒢a(c,t_{\lambda})\in\mathcal{G}^{a};

  2. (2)

    ζc,tλc,tλv1[v1]\zeta_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}\in v^{-1}\mathbb{Z}[v^{-1}] for all (c,tλ)(c,tλ)(c^{\prime},t_{\lambda^{\prime}})\leq(c,t_{\lambda}) in 𝒢a\mathcal{G}^{a};

  3. (3)

    for all (c,tλ)(c,tλ)(c^{\prime},t_{\lambda^{\prime}})\leq(c,t_{\lambda}) in 𝒢a\mathcal{G}^{a},

    ζc,tλc,tλ=(c′′,tλ′′),(c,tλ)(c′′,tλ′′)(c,tλ)ζc,tλc′′,tλ′′¯γc,tλc′′,tλ′′.\zeta_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}=\sum_{(c^{\prime\prime},t_{\lambda^{\prime\prime}}),(c^{\prime},t_{\lambda^{\prime}})\leq(c^{\prime\prime},t_{\lambda^{\prime\prime}})\leq(c,t_{\lambda})}\overline{\zeta_{c^{\prime},t_{\lambda^{\prime}}}^{c^{\prime\prime},t_{\lambda^{\prime\prime}}}}\gamma_{c,t_{\lambda}}^{c^{\prime\prime},t_{\lambda^{\prime\prime}}}.

For any (c,tλ)𝒢a(c,t_{\lambda})\in\mathcal{G}^{a}, let

C(c,tλ)=E(c,tλ)+(c,tλ)<(c,tλ)(c,tλ)𝒢aζc,tλc,tλE(c,tλ).C{(c,t_{\lambda})}=E{(c,t_{\lambda})}+\sum_{(c^{\prime},t_{\lambda^{\prime}})<(c,t_{\lambda})\atop(c^{\prime},t_{\lambda^{\prime}})\in\mathcal{G}^{a}}\zeta_{c,t_{\lambda}}^{c^{\prime},t_{\lambda^{\prime}}}E{(c^{\prime},t_{\lambda^{\prime}})}.
Theorem 8.8 ([16]).

The set {C(c,tλ)|(c,tλ)𝒢a}\{C{(c,t_{\lambda})}\,\,|\,\,(c,t_{\lambda})\in\mathcal{G}^{a}\} is a [v,v1]\mathbb{Z}[v,v^{-1}]-basis of 𝒞(Q)[v,v1]\mathcal{C}^{\ast}(Q)_{\mathbb{Z}[v,v^{-1}]} satisfying the following conditions.

  1. (1)

    C(c,tλ)¯=C(c,tλ)\overline{C{(c,t_{\lambda})}}=C{(c,t_{\lambda})};

  2. (2)

    (C(c,tλ),C(c,tλ))δ(c,tλ),(c,tλ)+v1[[v1]](v)(C{(c,t_{\lambda})},C{(c^{\prime},t_{\lambda^{\prime}})})\in\delta_{(c,t_{\lambda}),(c^{\prime},t_{\lambda^{\prime}})}+v^{-1}\mathbb{Z}[[v^{-1}]]\cap\mathbb{Q}(v).

Corollary 8.9 ([16]).

The set {C(c,tλ)|(c,tλ)𝒢a}\{C{(c,t_{\lambda})}\,\,|\,\,(c,t_{\lambda})\in\mathcal{G}^{a}\} is the canonical basis of 𝒞(Q)\mathcal{C}^{\ast}(Q).

References

  • [1] J. Beck, V. Chari, and A. Pressley. An algebraic characterization of the affine canonical basis. Duke Math. J., 99(3):455–487, 1998.
  • [2] J. Beck and H. Nakajima. Crystal bases and two-sided cells of quantum affine algebras. Duke Math. J., 123(2):335–402, 2002.
  • [3] X. Chen. Root vectors of the composition algebra of the Kronecker algebra. Algebra Discrete Math., 2004(1):37–56, 2004.
  • [4] B. Deng, J. Du, and J. Xiao. Generic extensions and canonical bases for cyclic quivers. Canad. J. Math., 59(6):1260–1283, 2007.
  • [5] J. A. Green. Hall algebras, hereditary algebras and quantum groups. Invent. Math., 120(1):361–377, 1995.
  • [6] Y. Li. Notes on affine canonical and monomial bases. arXiv preprint arXiv:math/0610449, 2006.
  • [7] Z. Lin, J. Xiao, and G. Zhang. Representations of tame quivers and affine canonical bases. Publ. Res. Inst. Math. Sci., 47:825–885, 2011.
  • [8] G. Lusztig. Canonical bases arising from quantized enveloping algebras. J. Amer. Math. Soc., 3(2):447–498, 1990.
  • [9] G. Lusztig. Quivers, perverse sheaves, and quantized enveloping algebras. J. Amer. Math. Soc., 4(2):365–421, 1991.
  • [10] G. Lusztig. Affine quivers and canonical bases. Publications Mathématiques de l’IHÉS, 76:111–163, 1992.
  • [11] G. Lusztig. Introduction to quantum groups. Springer, 2010.
  • [12] K. Mcgerty. The Kronecker quiver and bases of quantum affine 𝔰𝔩2\mathfrak{sl}_{2}. Adv. Math., 197(2):411–429, 2005.
  • [13] H. Nakajima. Crystal, canonical and PBW bases of quantum affine algebras. In Algebraic groups and homogeneous spaces, pages 389–421. Narosa Publishing House, New Delhi, India, 2007.
  • [14] C. M. Ringel. Hall algebras and quantum groups. Invent. Math., 101(1):583–591, 1990.
  • [15] C. M. Ringel. The Hall algebra approach to quantum groups. In Proceedings E.L.A.M., pages 85–114. Aportaciones Matemáticas Comunicaciones, 1995.
  • [16] J. Xiao, H. Xu, and M. Zhao. Tame quivers and affine bases I: a Hall algebra approach to the canonical bases. arXiv preprint arXiv:2107.03095, 2021.
  • [17] P. Zhang. PBW-basis for the composition algebra of the Kronecker algebra. J. Reine Angew. Math., 2000(527):97–116, 2000.