Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A cluster algebra approach to q-characters of Kirillov-Reshetikhin modules

A cluster algebra approach to q-characters of Kirillov-Reshetikhin modules We describe a cluster algebra algorithm for calculating q-characters of Kirillov-Resheti- khin modules for any untwisted quantum affine algebra U (bg). This yields a geometric q- character formula for tensor products of Kirillov-Reshetikhin modules. When g is of type A, D, E , this formula extends Nakajima’s formula for q-characters of standard modules in terms of homology of graded quiver varieties. Contents 1 Introduction 1 2 Definitions and notation 3 3 An algorithm for the q-characters of Kirillov-Reshetikhin modules 8 4 A geometric character formula for Kirillov-Reshetikhin modules 17 5 Beyond Kirillov-Reshetikhin modules 29 6 Appendix 32 1 Introduction Let g be a simple Lie algebra over C, and let U (g) be the corresponding untwisted quantum affine algebra with quantum parameter q ∈ C not a root of unity. The finite-dimensional complex representations of U (g) have been studied by many authors during the past twenty years. We refer the reader to [CP1] for a classical introduction, and to [CH, Le2] for recent surveys on this topic. In [HL1], we started to explore some new connections between this rich representation theory and the cluster algebras of Fomin and Zelevinsky. The main result, proved in [HL1] in type A and D , and extended to any A-D-E type by Nakajima [N4], shows the existence of a tensor n 4 category C of finite-dimensional U (bg)-modules whose Grothendieck ring is a cluster algebra of 1 q the same finite Dynkin type, such that the classes of simple modules coincide with the set of cluster monomials. As a consequence, the q-characters of the simple objects of C can be computed algorithmically using the combinatorics of cluster algebras. Moreover, the Caldero-Chapoton formula for cluster expansions leads to some new geometric formulae for these characters, in terms of Euler characteristics of quiver Grassmannians. Unfortunately the category C is quite small. For instance it contains only three Kirillov- Reshetikhin modules for each node of the Dynkin diagram of g. Another limitation of the papers [HL1] and [N4] is that g is assumed to be of simply laced type. In fact, the general proof of Nakajima uses in a crucial way his geometric construction of the standard U (bg)-modules [N1], which is only available in the simply laced case. arXiv:1303.0744v1 [math.QA] 4 Mar 2013 In this paper we drop the assumption of being simply laced, and we consider a much larger tensor subcategory C which contains, up to spectral shifts, all the irreducible finite-dimensional representations of U (bg). Our first main result (Theorem 3.1) is an algorithm which calculates the q-character of an arbitrary Kirillov-Reshetikhin module in C as the result of a sequence of cluster mutations. The only input for this calculation is the initial seed of our cluster algebra A , which is encoded in a quiver obtained from the Cartan matrix of g by a simple and uniform recipe. (It may be worth noting that A is always a skew-symmetric cluster algebra, even when g is not simply laced.) The proof of this theorem is based on the fact that the q-characters of the Kirillov-Reshetikhin modules are solutions of the corresponding T -system of Kuniba, Nakanishi and Suzuki [KNS1, N2, H]. This will come as no surprise, given the many papers already devoted to the relationships between cluster algebras and T -systems (see in particular [IIKNS], [IIKKN1], [IIKKN2]; in fact our algorithm is inspired from [GLS2, §13], where similar T -system formulas are obtained for generalized minors of symmetric Kac-Moody groups). We find it nevertheless remarkable that, by interpreting the T -system equations as appropriate cluster transformations, one is able to obtain the Kirillov-Reshetikhin q-characters starting from their highest weight monomials via a procedure of successive approximations. To the best of our knowledge this simple “bootstrap” algorithm had not been noticed before, although, in retrospect, it could certainly have been formulated and proved without knowledge of the cluster algebra theory. At this stage, we should recall that Frenkel and Mukhin [FM] have described long ago a completely different algorithm, which can be used for computing the q-characters of the Kirillov- Reshetikhin modules [N2, H]. The advantage of our approach is that we are now in a position to apply deep results of the theory of cluster algebras and obtain new formulas for the Kirillov- Reshetikhin q-characters. In [DWZ1, DWZ2], Derksen, Weyman and Zelevinsky have con- structed a categorical model for a large class of cluster algebras using quivers with potentials. In particular they have proved a far-reaching generalization of the Caldero-Chapoton formula, ex- pressing any cluster variable in terms of the F -polynomial of an associated quiver representation (see also [Pl1] for a different proof of this generalized formula). Applying this formula in our con- text, we get a geometric character formula for arbitrary Kirillov-Reshetikhin modules, and also for their tensor products (Theorem 4.8). When g is simply laced, and we restrict our attention to the simplest Kirillov-Reshetikhin modules and their tensor products, namely the fundamental modules and the standard modules, the quiver Grassmannians involved in our formula are homeomorphic to the projective varieties L (V,W ) used by Nakajima [N3, §4] in his geometric construction of the standard modules. This suggests that the quiver Grassmannians we introduce, in connection with general Kirillov- Reshetikhin modules of not necessarily simply laced type, might be interesting new varieties. When g is a classical Lie algebra of type A, B, C, D, there exist tableau sum formulas for the q- characters of certain Kirillov-Reshetikhin modules (see [KNS2, §7] and references therein). From the geometric point of view of Theorem 4.8, these formulas can be explained by the fact that the corresponding quiver representations have a nice and regular “grid structure”, and in many cases their quiver Grassmannians are reduced to points (see e.g. §6.4, §6.5, §6.6). The cluster algebra approach also suggests that our results should extend far beyond the Kirillov-Reshetikhin modules. Indeed, we show (Theorem 5.1) that the cluster algebra A is iso- morphic to the Grothendieck ring of C . It is then natural to conjecture that this isomorphism maps all cluster monomials of A to the classes of certain simple objects of C (Conjecture 5.2), and to extend the above geometric character formula to all these simple objects (Conjecture 5.3). The results of [HL1, HL2] and [N4] provide some evidence supporting these conjectures in the simply laced case. 2 Here is a more precise outline of the paper. In Section 2 we associate with every simple Lie algebra g some quivers (§2.1), from which we define a cluster algebra A (§2.2). We also introduce the untwisted quantum affine algebra U (bg) (§2.3). In Section 3 we state and prove our algorithm for computing the Kirillov-Reshetikhin q-characters as special cluster variables of A . The proof uses T -systems (§3.2.1) and the notion of truncated q-characters (§3.2.2). In Section 4, we con- sider an algebra A defined by a quiver with potential, coming from our initial seed for A (§4.1). (i) We introduce certain distinguished A-modules K (§4.3), and we state our geometric formula k,m (i) for the Kirillov-Reshetikhin q-characters in terms of Grassmannians of submodules of the K k,m (Theorem 4.8). To prove it, we calculate the g-vectors of these q-characters, regarded as cluster variables of A , and we apply a result of Plamondon [Pl2] which allows to reconstruct the A- module corresponding to a given cluster variable from the knowledge of its g-vector. To be in a position to apply this result, we show that the defining potential of A is rigid, and that appropriate truncations of A are finite-dimensional (Proposition 4.17). In Section 5, we prove Theorem 5.1 and we formulate Conjecture 5.2 and Conjecture 5.3. The paper closes with an appendix illustrating our results with many examples. 2 Definitions and notation 2.1 Quivers 2.1.1 Cartan matrix Let C = (c ) be an indecomposable n× n Cartan matrix of finite type [Ka, §4.3]. There is a i j i, j∈I diagonal matrix D = diag(d | i ∈ I) with entries in Z such that the product i >0 B = DC = (b ) i j i, j∈I is symmetric. We normalize D so that min{d | i ∈ I} = 1, and we put t := max{d | i ∈ I}. Thus i i 1 if C is of type A , D , E , E or E ,  n n 6 7 8 t = 2 if C is of type B , C or F , n n 4 3 if C is of type G . It is easy to check by inspection that (d > 1 and c < 0) =⇒ (c = −1). (1) i i j i j One attaches to C a Dynkin diagram δ with vertex set I [Ka, §4.7]. Since C is assumed to be indecomposable and of finite type, δ is a tree. All the objects that we consider below depend on C, but we shall not always repeat it, neither record it explicitly in our notation. Example 2.1 The Cartan matrix C of type B in the Cartan-Killing classification is defined by   2 −1 0   C = −1 2 −1 0 −2 2 We have D = diag(2, 2, 1) and the symmetric matrix B is given by   4 −2 0   B = −2 4 −2 0 −2 2 3 . (1,2) (3,2) OO ❑ OO ❑ s ❑ s ❑ s ❑ s %% yys (2,1) OO ❑ s ❑ s ❑ s ❑ yys %% (1,0) (3,0) OO ❑ OO ❑ s ❑ s %% yys (2,0) OO (2,−1) s ❑ OO ❑ s ❑ s s ❑ s ❑ yys %% s ❑ yys %% (1,−1) (3,−1) OO ❑ OO (1,−2) (3,−2) ❑ s OO ❑ OO ❑ s s ❑ s ❑ s %% yys ❑ s ❑ s %% yys (2,−2) OO ❑ (2,−3) s ❑ s ❑ OO ❑ s ❑ s ❑ s ❑ s ❑ yys %% s ❑ s ❑ yys %% (1,−3) (3,−3) OO ❑ OO (1,−4) (3,−4) ❑ s OO ❑ OO ❑ s ❑ s ❑ s %% yys ❑ s ❑ s %% yys (2,−4) (2,−5) s ❑ s ❑ s ❑ s ❑ yys %% s ❑ s ❑ yys %% (1,−5) (3,−5) (1,−6) (3,−6) . Figure 1: The quivers Γ and G in type A . 2.1.2 Infinite quiver e e e e Put V = I ×Z. We introduce a quiver Γ with vertex set V . The arrows of Γ are given by ((i, r) → ( j, s)) ⇐⇒ (b 6= 0 and s = r + b ). i j i j Lemma 2.2 The quiver Γ has two isomorphic connected components. Proof — Let i ∈ I be such that d = 1. For every r ∈Z we have an arrow (i, r) → (i, r+ 2). Since the Dynkin diagram δ is connected, every vertex ( j, s) ∈ V is connected to a vertex of the form (i, r), so Γ has at most two connected components. Moreover, since δ is a tree, any path from (i, r) to (i, s) in Γ contains as many arrows of the form ( j, p) → (k, p+ b ) with j 6= k, as it contains arrows of the jk form (k, t) → ( j, t + b ). Since b = b , and since b ∈ 2Z for every j ∈ I, it follows that if there k j jk k j j j is a path from (i, r) to (i, s) then s− r ∈ 2Z. Therefore Γ has exactly two connected components. These two components are isomorphic via the map ( j, r) 7→ ( j, r + 1) (( j, r) ∈ V ×Z)). ✷ We pick one of the two isomorphic connected components of Γ and call it Γ. The vertex set of Γ is denoted by V . 2.1.3 Semi-infinite quiver We will have to use a second labelling of the vertices of Γ. It is deduced from the first one by means of the function ψ defined by ψ(i, r) = (i, r + d ), ((i, r) ∈ V ). (2) 4 . . . . OO . . (2,5) (1,5) OO OO s s s s s s s s yys yys (1,3) (2,3) OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% (2,1) (1,1) OO OO s s s s s s s s yys yys (2,0) OO (1,−1) (2,−1) OO ❑ OO ❑ ❑ ❑ ❑ ❑ %% ❑ ❑ ❑ ❑ %% %% (2,−2) (1,−1) OO OO (2,−3) (1,−3) s s s s OO OO s s s s s s s s yys yys s s s s yys yys (1,−3) (2,−4) OO ❑ OO ❑ (1,−5) (2,−5) ❑ ❑ ❑ ❑ OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% ❑ ❑ ❑ ❑ %% %% (2,−6) (1,−5) OO OO (2,−7) (1,−7) s s s s OO OO s s s s s s s s yys yys s s s s yys yys (1,−7) (2,−8) (1,−9) (2,−9) . . . . . . . . . . . . . . . . . . Figure 2: The quivers Γ and G in type B . Let W ⊂ I ×Z be the image of V under ψ . We shall denote by G the same quiver as Γ but with − − vertices labelled by W . Put W := W ∩ (I ×Z ). Let G be the full subquiver of G with vertex ≤0 set W . Example 2.3 The definitions of §2.1.2 and §2.1.3 are illustrated in Figure 1 and Figure 2. We find it convenient to always display the quivers Γ in the following way. We decide to draw all arrows of the form (i, r) → (i, r + b ) vertically, going upwards. Moreover, if (i, r) and (i, s) are two vertices ii with r− s 6∈ b Z, we draw them in different columns. Hence, the quivers attached to C always have ii d columns. Finally, the integer r determines the altitude of the vertex (i, r) in Γ. Therefore, i∈I since for i 6= j we have b ≤ 0, the arrows (i, r) → ( j, r + b ) are represented as oblique arrows i j i j going down. − − Figure 1 displays the quivers Γ and G for C of type A . Figure 2 shows Γ and G for C of type B . In both cases we have chosen to call Γ the connected component of Γ containing the vertex (2, 1). For another illustration, with C of type G , see Figure 3. More examples can be found in the Appendix, §6.5, §6.6, §6.7. 2.2 Cluster algebras We refer the reader to [FZ2] and [GSV] for an introduction to cluster algebras, and for any un- defined terminology. 2.2.1 Cluster algebra attached to G − − Consider an infinite set of indeterminates z = {z | (i, r) ∈ W } over Q. Let A be the cluster i,r − − algebra defined by the initial seed (z , G ). Thus, A is the Q-subalgebra of the field of rational 5 − − functions Q(z ) generated by all the elements obtained from some element of z via a finite sequence of seed mutations, see [GG, Definition 3.1]. Note that there are no frozen variables. Cluster algebras of infinite rank have not received much attention up to now. (In fact we are not aware of another paper than [GG]; in [GG], a specific example of type A is developed, in connection with a triangulated category studied by Holm and Jorgensen [HoJo].) For our purposes in this paper, it is always possible to work with sufficiently large finite sub- − − seeds of the seed (z , G ), and replace A by the genuine cluster subalgebras attached to them. On the other hand, statements become nicer if we allow ourselves to formulate them in terms of the infinite rank cluster algebra A . 2.2.2 Monomial change of variables − − Let Y = {Y | (i, r) ∈ W} be a new set of indeterminates over Q. Let Y = {Y ∈ Y | (i, r) ∈ W }. i,r i,r For (i, r) ∈ W , we perform the substitution z = Y . (3) i,r i, r+kb ∏ ii k≥0, r+kb ≤0 ii Note that all variables in the right-hand side of (3) belong to Y . Example 2.4 If G is as in Figure 2, we have z = Y , z = Y Y , z = Y Y Y , z = Y Y Y Y , 2,0 2,0 2,−2 2,−2 2,0 2,−4 2,−4 2,−2 2,0 2,−6 2,−6 2,−4 2,−2 2,0 z = Y , z = Y Y , z = Y Y Y , z = Y Y Y Y , 1,−1 1,−1 1,−5 1,−5 1,−1 1,−9 1,−9 1,−5 1,−1 1,−13 1,−13 1,−9 1,−5 1,−1 z = Y , z = Y Y , z = Y Y Y , etc. 1,−3 1,−3 1,−7 1,−7 1,−3 1,−11 1,−11 1,−7 1,−3 2.2.3 Sequence of vertices As explained in Example 2.3, the arrows of G of the form (i, r) −→ (i, r + b ) are called vertical ii and displayed in columns. To each column we attach an initial label given by the index of its top vertex (i, r), for which r is maximal among the vertices of the column. We now form a sequence of t n columns by induction as follows. At each step we pick a column whose label (i, r) has maximal r among labels of all columns. After picking a column with label (i, r), we change its label to (i, r− b ). Finally, reading column after column in this ordering, from ii top to bottom, we get an infinite sequence S of vertices of G . Example 2.5 If G is as in Figure 1, then t = 1, the sequence of columns consists of 3 columns, and we obtain the following sequence of vertices: S = ((2, 0),(2,−2),(2,−4),... ,(1,−1),(1,−5),(1,−9),... ,(3,−1),(3,−3),(3,−5),...). (Here, the columns labelled (1,-1) and (3,-1) could be interchanged.) If G is as in Figure 2, then t = 2, the sequence of columns consists of 4 columns and gives the following sequence of vertices: S = ((2, 0),(2,−2),(2,−4),... ,(1,−1),(1,−5),(1,−9),... , (2, 0),(2,−2),(2,−4),... ,(1,−3),(1,−7),(1,−11),...) Note that the column with vertices (2, r) appears two times. It appears first because its initial label is (2, 0). After picking it, its label is changed to (2,−2), so it appears again between the columns labelled (1,−1) and (1,−3). 6 g A B C D E E E F G n n n n 6 7 8 4 2 t 1 2 2 1 1 1 1 2 3 hˇ n + 1 2n− 1 n + 1 2n− 2 12 18 30 9 4 Table 1: Dual Coxeter numbers Finally, for (i, r) ∈ G , we define k to be the unique positive integer k satisfying i,r 0 < kb −|r| ≤ b . (4) ii ii In other words, (i, r) is the kth vertex in its column, counting from the top. Example 2.6 If G is as in Figure 2, then k = 2, k = 3. 2,−2 1,−9 2.3 Quantum affine algebras 2.3.1 The algebra U (bg) Let g be the simple Lie algebra over C with Cartan matrix C. We denote by α (i ∈ I) the simple roots of g, and by ϖ (i ∈ I) the fundamental weights. They are related by α = c ϖ . (5) i ∑ ji j j∈I Let hˇ be the dual Coxeter number of g, see [Ka, §6.1]. The values of hˇ are recalled in Table 1. Let g be the corresponding untwisted affine Lie algebra. Thus if g has type X in the Cartan- (1) Killing classification, bg has type X in the Kac classification [Ka, §4.8]. Let U (bg) be the n q b b Drinfeld-Jimbo quantum enveloping algebra of g, see e.g. [CP1]. We regard U (g) as a C-algebra with quantum parameter q ∈ C not a root of unity. 2.3.2 q-characters Frenkel and Reshetikin [FR] have attached to every complex finite-dimensional representation of U (bg) a q-character χ (M). If M is irreducible, it is determined up to isomorphism by its q q q-character. The irreducible finite-dimensional representations of U (g) have been classified by Chari and Pressley in terms of Drinfeld polynomials, see [CP1, Theorem 12.2.6]. Equivalently, ir- reducible finite-dimensional representations of U (g) can be parametrized by the highest dominant monomial of their q-character [FR], and this is the parametrization we shall use. By definition, the q-character χ (M) is a Laurent polynomial with positive integer coefficients in the infinite set of variables Y = {Y | i ∈ I, a ∈ C }, which should be seen as a quantum affine i,a analogue of {e | i ∈ I}. In this paper we will be concerned only with polynomials involving the subset of variables Y , ((i, r) ∈ W ). i,q For simplicity of notation, we shall therefore write Y instead of Y . Thus our q-characters will i,r i,q be Laurent polynomials in the variables of the set Y introduced in §2.2.2. 7 Let m be a dominant monomial in the variables Y ∈ Y, that is, a monomial with nonnegative i,r exponents. We denote by L(m) the corresponding irreducible representation of U (bg), and by χ (m) = χ (L(m)) its q-character. For example, if m is of the form q q k−1 m = Y , (i ∈ I, r ∈ Z, k ≥ 1), i, r+ jb ∏ ii j=0 (i) L(m) is called a Kirillov-Reshetikhin module, and usually denoted by W . In particular, if k = 1 k,r (i) (i) we get a fundamental module W = L(Y ). By convention, if k = 0 the module W is the trivial i,r 1,r 0,r one-dimensional module for every (i, r), and its q-character is equal to 1. Finally, following [FR], for (i, r) ∈ V we introduce the following quantum affine analogue of e : ! ! ! −1 −1 −1 A := Y Y Y Y Y Y Y Y (6) i,r i,r−d i,r+d j,r j,r−1 j,r+1 j,r−2 j,r j,r+2 i i ∏ ∏ ∏ j: c =−1 j: c =−2 j: c =−3 ji ji ji Note that since (i, r) ∈ V , we have (i, r± d ) ∈ W . If c < 0, we also have, because of (1), i ji ( j, r + c + 1) = ( j, r + d (c + 1)) = ( j, r + b + d ) ∈ W. ji j ji i j j It follows that A is a Laurent monomial in the variables Y with ( j, s) ∈ W . i,r j,s 3 An algorithm for the q-characters of Kirillov-Reshetikhin modules 3.1 Statement and examples − − Let A be the cluster algebra defined in §2.2.1, with initial seed Σ = (z , G ), and let S = ((i , r ),(i , r ),(i , r ),...) 1 1 2 2 3 3 be the sequence of vertices of the quiver of A defined in §2.2.3. We denote by μ (Σ) the new seed obtained after performing the sequence of mutations indexed by S , that is, by mutating first at vertex (i , r ), then at vertex (i , r ), etc. More generally, for m ≥ 1, let Σ = μ (Σ) be the 1 1 2 2 m (m) seed obtained from Σ after m repetitions of the mutation sequence μ . Let z be the cluster i,r variable of Σ sitting at vertex (i, r) ∈ W ; this is a Laurent polynomial in the initial variables (m) (m) z , ( j, s) ∈ W . Let y be the Laurent polynomial obtained from z by performing the change j,s i,r i,r of variables (3) of §2.2.2; this is a Laurent polynomial in the variables Y , ( j, s) ∈ W . j,s Theorem 3.1 (a) The quiver of μ (Σ) is equal to the quiver of Σ, that is, to G . (m) (b) Suppose that m ≥ hˇ/2. Then, the y are the q-characters of the Kirillov-Reshetikhin i,r modules. More precisely, for m ≥ hˇ/2 there holds (m) (i) y = χ W . i,r k, r−2t m where k = k is defined as in §2.2.3. i,r (i) (i) Remark 3.2 It is well known that, for p ∈ Z, the q-character χ (W ) is deduced from χ (W ) q q k,r+ p k,r by applying the ring automorphism mapping Y to Y for every ( j, s) ∈ I×Z. Therefore, mod- j,s j,s+ p ulo these straightforward automorphisms, Theorem 3.1 describes the q-characters of all Kirillov- Reshetikhin modules. 8 Remark 3.3 Although the statement of Theorem 3.1 involves an infinite seed Σ and an infinite sequence of mutations S , the calculation of the q-character of a given Kirillov-Reshetikhin mod- ule requires only a finite number of mutations on a finite initial segment of the semi-infinite (i) quiver. More precisely, the proof of Theorem 3.1 will show that all the q-characters χ (W ) k,s ′ ′ ′ with k = 1,..., l can be calculated using (h + 2l − 1)h n/2 mutations, where h = ⌈hˇ/2⌉. Example 3.4 Let g be of type A . The quiver G of the initial seed is displayed in Figure 1. The initial cluster variables are z = Y , z = Y Y , z = Y Y Y , etc. 2,0 2,0 2,−2 2,−2 2,0 2,−4 2,−4 2,−2 2,0 z = Y , z = Y Y , z = Y Y Y , etc. 1,−1 1,−1 1,−3 1,−3 1,−1 1,−5 1,−5 1,−3 1,−1 z = Y , z = Y Y , z = Y Y Y , etc. 3,−1 3,−1 3,−3 3,−3 3,−1 3,−5 3,−5 3,−3 3,−1 After the mutation sequence μ , the first new cluster variables are (1) −1 y = Y + Y Y Y , 2,−2 1,−1 3,−1 2,0 2,0 (1) −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y Y Y Y , 2,−4 2,−2 1,−1 3,−1 2,−4 1,−3 1,−1 3,−3 3,−1 2,−2 2,0 2,−2 2,0 (1) −1 −1 −1 y = Y Y Y + Y Y Y Y Y + Y Y Y Y Y Y Y , 2,−6 2,−4 2,−2 1,−1 3,−1 2,−6 2,−4 1,−3 1,−1 2,−6 3,−3 3,−1 2,−4 2,0 2,−2 2,0 −1 −1 −1 + Y Y Y Y Y Y Y Y Y , 1,−5 1,−3 1,−1 3,−5 3,−3 3,−1 2,−4 2,−2 2,0 (1) −1 −1 y = Y + Y Y + Y Y , 1,−3 2,−2 3,−1 1,−1 1,−1 2,0 (1) −1 −1 −1 −1 y = Y Y + Y Y Y + Y Y Y + Y Y Y Y 1,−5 1,−3 1,−5 2,−2 1,−5 3,−1 2,−4 2,−2 1,−3 1,−1 2,0 1,−3 1,−1 −1 −1 −1 −1 + Y Y Y Y + Y Y Y Y , 2,−4 3,−1 3,−3 3,−1 1,−3 2,0 2,−2 2,0 (1) −1 −1 −1 −1 y = Y Y Y + Y Y Y Y + Y Y Y Y + Y Y Y Y Y 1,−7 1,−5 1,−3 1,−7 1,−5 2,−2 1,−7 1,−5 3,−1 1,−7 2,−4 2,−2 1,−5 1,−1 2,0 1,−3 1,−1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y + Y Y Y Y Y + Y Y Y Y Y Y 1,−7 2,−4 3,−1 1,−7 3,−3 3,−1 2,−6 2,−4 2,−2 1,−3 2,0 2,−2 2,0 1,−5 1,−3 1,−1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y Y 2,−6 2,−4 3,−1 2,−6 3,−3 3,−1 1,−5 1,−3 2,0 1,−5 2,−2 2,0 −1 −1 −1 + Y Y Y Y Y Y , 3,−5 3,−3 3,−1 2,−4 2,−2 2,0 (1) (1) (1) (1) (1) (1) (We omit the variables y , y , y , since they are readily obtained from y , y , y 3,−1 3,−5 3,−5 1,−1 1,−5 1,−5 via the symmetry (1 ↔ 3).) After a second application of the mutation sequence μ , the first new cluster variables are (2) −1 −1 −1 −1 −1 −1 y = Y + Y Y Y + Y Y + Y Y + Y Y Y + Y , 2,−4 1,−3 3,−3 1,−3 3,−3 2,−2 2,0 2,−2 3,−1 1,−1 1,−1 3,−1 2,0 (2) −1 −1 −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y Y Y Y + Y Y Y 2,−6 2,−4 1,−3 3,−3 2,−6 1,−5 1,−3 3,−5 3,−3 1,−5 2,−2 2,−2 2,−4 2,−2 2,0 3,−3 −1 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y + Y Y Y Y + Y Y + Y Y Y Y + Y Y Y 3,−5 2,−4 2,−6 1,−5 3,−5 1,−3 2,−6 1,−3 2,0 1,−3 2,0 3,−3 2,0 2,−4 2,0 3,−1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y + Y Y Y Y Y + Y Y Y + Y Y Y Y 1,−5 1,−3 1,−5 1,−3 3,−5 2,−6 3,−3 3,−5 3,−3 3,−3 3,−1 2,−4 3,−1 1,−1 1,−3 1,−1 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y + Y Y Y Y Y Y + Y Y Y Y 1,−5 3,−5 3,−3 2,−4 2,−2 2,−6 2,−2 1,−1 2,−4 1,−3 1,−1 3,−3 3,−1 1,−1 3,−1 −1 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y + Y Y Y Y Y + Y Y Y Y Y Y 2,−2 3,−5 1,−5 2,−2 1,−5 2,−2 3,−5 1,−3 1,−1 3,−1 1,−1 3,−3 3,−1 1,−1 2,−4 3,−1 −1 −1 +Y Y , 2,−2 2,0 (2) −1 −1 −1 y = Y + Y Y + Y Y + Y , 1,−5 2,−4 3,−3 1,−1 1,−3 2,−2 3,−1 (2) −1 −1 −1 −1 y = Y Y + Y Y Y + Y Y Y + Y Y Y Y 1,−7 1,−5 1,−7 2,−4 1,−7 3,−3 2,−6 2,−4 1,−3 1,−3 2,−2 1,−5 1,−3 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y + Y Y Y Y + Y Y Y + Y Y 2,−6 3,−3 3,−5 3,−3 2,−6 1,−7 1,−5 2,−2 2,−4 2,−2 1,−5 3,−1 3,−1 −1 −1 −1 −1 + Y Y Y + Y Y , 3,−5 2,−4 3,−1 3,−3 3,−1 Here hˇ/2 = 2, so we can observe that the cluster variables obtained after performing 2 times the 9 mutation sequence μ are indeed q-characters of Kirillov-Reshetikhin modules, namely, (2) (2) y = χ (Y ), y = χ (Y Y ), etc. q 2,−4 q 2,−6 2,−4 2,0 2,−2 (2) (2) y = χ (Y ), y = χ (Y Y ), etc. q 1,−5 q 1,−7 1,−5 1,−1 1,−3 (2) (2) y = χ (Y ), y = χ (Y Y ), etc. q 3,−5 q 3,−7 3,−5 3,−1 3,−3 Example 3.5 Let g be of type B . The quiver G of the initial seed is displayed in Figure 2. The initial cluster variables are z = Y , z = Y Y , z = Y Y Y , etc. 2,0 2,0 2,−2 2,−2 2,0 2,−4 2,−4 2,−2 2,0 z = Y , z = Y Y , z = Y Y Y , etc. 1,−1 1,−1 1,−5 1,−5 1,−1 1,−9 1,−9 1,−5 1,−1 z = Y , z = Y Y , z = Y Y Y , etc. 1,−3 1,−3 1,−7 1,−7 1,−3 1,−11 1,−11 1,−7 1,−3 After the mutation sequence μ , the first new cluster variables are (1) −1 y = Y + Y Y , 2,−4 1,−3 2,0 2,−2 (1) −1 −1 −1 −1 y = Y Y + Y Y Y + Y Y Y Y + Y Y , 2,−6 2,−4 1,−3 2,−6 1,−5 1,−3 1,−3 2,−2 2,−2 2,−4 2,−2 1,−1 (1) −1 −1 −1 y = Y Y Y + Y Y Y Y + Y Y Y Y Y 2,−8 2,−6 2,−4 1,−3 2,−8 2,−6 1,−5 1,−3 2,−8 2,−4 2,−2 2,−4 2,−2 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y + Y Y Y , 1,−7 1,−5 1,−3 1,−7 1,−3 1,−3 2,−8 2,−6 2,−4 2,−2 1,−1 2,−6 1,−1 (1) −1 −1 −1 −1 y = Y + Y Y Y + Y Y + Y Y Y , 1,−5 2,−4 2,−2 2,−4 1,−3 1,−1 1,−1 2,0 2,−2 2,0 (1) −1 −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y + Y Y Y Y 1,−5 1,−9 1,−9 2,−4 2,−2 1,−9 2,−4 1,−9 1,−3 1,−5 1,−1 2,0 2,−2 2,0 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 2,−8 2,−6 2,−4 2,−2 2,−8 2,−6 2,−4 1,−5 1,−1 1,−5 2,0 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 1,−3 2,−8 2,−6 1,−3 2,−8 2,−2 2,0 2,−4 2,−2 2,0 1,−5 −1 −1 −1 −1 + Y Y Y Y Y Y , 1,−7 1,−3 2,−6 2,−4 2,−2 2,0 (1) −1 −1 −1 −1 −1 y = Y + Y Y Y + Y Y + Y Y Y + Y , 1,−7 2,−6 2,−4 2,−6 1,−5 1,−3 1,−3 2,−2 2,−4 2,−2 1,−1 (1) −1 −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y + Y Y Y Y 1,−7 1,−11 1,−11 2,−6 2,−4 1,−11 2,−6 1,−11 1,−5 1,−7 1,−3 2,−2 2,−4 2,−2 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 2,−10 2,−8 2,−6 2,−4 2,−10 2,−8 2,−6 1,−7 1,−3 1,−7 2,−2 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 1,−5 2,−10 2,−8 1,−5 2,−10 1,−7 2,−4 2,−2 2,−6 2,−4 2,−2 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y + Y Y 1,−9 1,−5 1,−9 1,−11 2,−8 2,−6 2,−4 2,−2 1,−1 2,−8 2,−6 1,−1 −1 −1 −1 −1 −1 −1 + Y Y Y + Y Y Y Y + Y Y . 2,−10 2,−10 2,−8 1,−1 2,−6 1,−7 1,−1 1,−5 1,−1 Here hˇ/2 = 3/2, and we can observe that certain cluster variables are not yet q-characters of Kirillov-Reshetikhin modules. But some already are, namely (1) (1) y = χ (Y ), y = χ (Y Y ), etc. q 1,−7 q 1,−7 1,−11 1,−3 1,−7 After a second application of the mutation sequence μ , since 2 > 3/2, all the new cluster vari- ables are q-characters of Kirillov-Reshetikhin modules. For example (2) −1 −1 −1 y = Y + Y Y + Y Y + Y = χ (Y ). 2,−8 1,−7 2,−4 q 2,−8 2,0 2,−6 1,−3 2,−2 3.2 Proof of Theorem 3.1 The proof relies on two main ingredients which we shall first review, namely, the T -systems, and the truncated q-characters. 10 3.2.1 T -systems With the quantum affine algebra U (g) is associated a system of difference equations called a T -system [KNS1]. Its unknowns are denoted by (i) T , (i ∈ I, k ∈ N, r ∈ Z). k,r We fix the initial boundary condition (i) T = 1, (i ∈ I, r ∈ Z). (7) 0,r If g is of type A , D , E , the T -system equations are n n n (i) (i) (i) (i) ( j) T T = T T + T , (i ∈ I, k ≥ 1, r ∈ Z). (8) k,r+1 k,r−1 k−1,r+1 k+1,r−1 k,r j: c =−1 i j If g is not of simply laced type, the T -system equations are more complicated. They can be written in the form (i) (i) (i) (i) (i) T T = T T + S , (i ∈ I, k ≥ 1, r ∈ Z), (9) k,r+d k,r−d k−1,r+d k+1,r−d k,r i i i i (i) where S is defined as follows. If d ≥ 2 then k,r (i) ( j) ( j) S = T T . (10) ∏ ∏ k,r k,r d k, r−d +1 i i j: c =−1 j: c ≤−2 ji ji If d = 1 and t = 2, then ( j) ( j) ( j)  T T T , if k = 2l, ∏ ∏  k,r l,r l,r+2 j: c =−1 j: c =−2 i j i j (i) S = (11) k,r ( j) ( j) ( j)  T T T if k = 2l + 1.  ∏ ∏ k,r l+1,r l,r+2 j: c =−1 j: c =−2 i j i j Finally, if d = 1 and t = 3, that is, if g is of type G , denoting by j the other vertex of δ we have i 2 d = 3 and ( j) ( j) ( j) T T T if k = 3l, l,r l,r+2 l,r+4 (i) ( j) ( j) ( j) S = (12) T T T if k = 3l + 1, k,r l+1,r l,r+2 l,r+4 ( j) ( j) ( j) T T T if k = 3l + 2. l+1,r l+1,r+2 l,r+4 Example 3.6 Let g be of type B . The Cartan matrix is 2 −1 C = −2 2 and we have d = 2 and d = 1. The T -system reads: 1 2 (1) (1) (1) (1) (2) T T = T T + T , (k ≥ 1, r ∈ Z), k,r+2 k,r−2 k−1,r+2 k+1,r−2 2k,r−1 (2) (2) (2) (2) (1) (1) T T = T T + T T , (l ≥ 1, r ∈ Z), 2l,r+1 2l,r−1 2l−1,r+1 2l+1,r−1 l,r l,r+2 (2) (2) (2) (2) (1) (1) T T = T T + T T , (l ≥ 0, r ∈ Z). 2l+1,r+1 2l+1,r−1 2l,r+1 2l+2,r−1 l+1,r l,r+2 11 It was conjectured in [KNS1], and proved in [N2] (for g of type A, D, E ) and [H] (general case), that the q-characters of the Kirillov-Reshetikhin modules of U (bg) satisfy the corresponding T -system. More precisely, we have Theorem 3.7 ([N2][H]) For i ∈ I, k ∈ N, r ∈ Z, (i) (i) T = χ W , k,r k,r h i ±1 is a solution of the T -system in the ring Z Y | (i, r) ∈ I ×Z . i,r 3.2.2 Truncated q-characters Let C be the full subcategory of the category of finite-dimensional U (bg)-modules whose ob- jects have all their composition factors of the form L(m) where m is a dominant monomial in the variables of Y . h i ±1 Lemma 3.8 The q-character of an object in C belongs to Z Y | Y ∈ Y . i,r i,r Proof — A simple object of C is a quotient of a tensor product of fundamental representations of C . But the q-character of a fundamental representation can be calculated by means of the Frenkel-Mukhin algorithm [FM]. At each step the algorithm produces monomials which involve only variables Y ∈ Y. Hence the result. ✷ i,r Note that for a dominant monomial m in the variables of Y , the q-character χ (m) may ′ − contain Laurent monomials m involving variables Y ∈ Y\ Y . Following [HL1], we define the i,r truncated q-character χ (m) to be the Laurent polynomial obtained from χ (m) by discarding all h i ±1 ′ − − these monomials m . So, by definition, χ (m) ∈ Z Y | Y ∈ Y . i,r i,r Example 3.9 Let g be of type B . We keep the notation of Example 3.6. The fundamental modules L(Y ) and L(Y ) have q-characters equal to 1,−3 2,−4 −1 −1 −1 −1 −1 χ (Y ) = Y + Y Y Y + Y Y + Y Y Y + Y , q 1,−3 1,−3 2,−2 2,0 2,−2 1,−1 1,1 2,2 2,0 2,2 1,3 −1 −1 −1 χ (Y ) = Y + Y Y + Y Y + Y . q 2,−4 2,−4 1,−3 2,0 2,−2 1,1 2,2 The corresponding truncated q-characters are χ (Y ) = Y , 1,−3 1,−3 − −1 χ (Y ) = Y + Y Y . 2,−4 2,−4 1,−3 q 2,−2 Proposition 3.10 (i) C is a tensor category. (ii) The assignment [L(m)] 7→ χ (m) extends to an injective ring homomorphism from the h i ±1 − − Grothendieck ring K (C ) to Z Y | Y ∈ Y . 0 i,r i,r Proof — The argument follows the same lines as [HL1, §5.2.4, §6.2]. Recall the Laurent monomi- als A introduced in (6). By [FR], a Laurent monomial m of the q-character of a simple object of i,r − ′ C can always be written in the form m = mM where m is a dominant monomial in the variables 12 −1 of Y , and M is a monomial in the variables A with (i, k + d ) ∈ W . Note that the Y -variable ap- i,k −1 pearing in A with the highest spectral parameter is Y . It follows that A is a right-negative i,r i,r+d i,r monomial in the sense of [FM], that is, the Y -variable with highest spectral parameter occuring in −1 A has a negative exponent. i,r ′ − ′ Let L(m) and L(m ) be simple objects of C , that is, m and m are dominant monomials in − ′′ ′ ′′ the variables of Y . If L(m ) is a composition factor of L(m)⊗ L(m ), then m is a product of ′ ′′ ′ monomials of χ (m) and χ (m ). So, we have m = mm M where M is a monomial in the variables q q −1 ′′ A . We claim that, since m is dominant, the spectral parameters r have to satisfy r + d ≤ 0. i,r ′′ ′′ Indeed otherwise m would be right-negative. Therefore, using Lemma 3.8, the monomial m − ′′ − − contains only variables of Y , hence L(m ) is in C , and C is a monoidal category. Moreover, by [CP2, Prop. 5.1], the category C is stable by duals, so it is a tensor category. This proves (i). To prove (ii) consider now an arbitrary Laurent monomial m of the q-character of an object − ′ of C . As above, it can be written in the form m = mM where m is a dominant monomial in the − −1 ′ variables of Y , and M is a monomial in the variables A with (i, k + d ) ∈ W . Now m contains i,k a variable Y 6∈ Y if and only if M contains a negative power of A for some pair (i, r) such that j,s i,r h i ±1 (i, r + d ) 6∈ W . So, if R denotes the subring of Z Y | Y ∈ Y generated by all the monomials i i,r i,r of the q-characters of the objects of C , and if I denotes the linear span of those monomials containing a variable Y ∈ Y \ Y , we see that I is an ideal of R. Hence, if π : R → R/I is the j,s natural projection, we can realize the truncated q-character map χ as χ = π ◦ χ , h i ±1 − − − which shows that χ is a ring homomorphism K (C ) → Z Y | Y ∈ Y . Finally, the fact q 0 i,r i,r that χ is injective follows from the fact that I contains only non-dominant monomials, and that two q-characters having the same dominant monomials with the same coefficients are equal. ✷ 3.2.3 Proof of the theorem We first notice that the initial cluster variables z are equal, after the change of variables (3), to i,r the truncated q-characters of certain Kirillov-Reshetikhin modules, namely, (i) z = Y = χ W , i,r+kb i,r ∏ ii q k ,r i,r k≥0, r+kb ≤0 ii where k is defined as in (4). Indeed, the level of truncation is chosen so that after truncation only i,r the highest dominant monomial of these q-characters survives. Now, the main idea of the proof is that the quiver G and the mutation sequence μ are designed in such a way that, at every step of the mutation sequence, the exchange relation is nothing else than a T -system equation. Let us first check this when g is of rank two. For g of type A , the sequence of mutated quivers obtained at each step of μ is shown in 2 S Appendix §6.1. The mutations take place at the boxed vertices. Reading the second quiver of §6.1, we see that the new cluster variable obtained after the first mutation is equal to (2) (1) − − χ W + χ W q 2,−2 q 1,−1 z + z 2,−2 1,−1 (2) =   = χ W . 1,−2 (2) 2,0 χ W 1,0 13 Here we have used Theorem 3.7 and Proposition 3.10. Similarly, reading the third quiver of §6.1, the new cluster variable obtained after the second mutation is equal to (2) (2) (1) − − − χ W χ W + χ W q q q 3,−4 1,−2 2,−3 (2) = χ W . 2,−4 (2) χ W 2,−2 An easy induction shows that, after every vertex of the second column has been mutated, each clus- (2) (2) − − ter variable of the form χ W has been replaced by the new cluster variable χ W . q q k,−2k+2 k,−2k We now continue by mutating vertices of the first column. We first get, at the top vertex (1) (2) − − χ W + χ W q q 2,−3 1,−2 (1) = χ W . q 1,−3 (1) χ W 1,−1 Then, mutating at the next vertex gives (1) (1) (2) − − − χ W χ W + χ W q q q 3,−5 1,−3 2,−4 (1) = χ W . 2,−5 (1) χ W 2,−3 By induction one sees that, after every vertex of the first column has been mutated, each cluster (1) (1) − − variable of the form χ W has been replaced by a new cluster variable χ W . q q k,−2k+1 k,−2k−1 Moreover, one sees that the new quiver obtained after μ is nothing else than G . Hence we conclude that one application of μ produces a seed with the same quiver, and in which every (i) (i) − − cluster variable χ W has been replaced by χ W . In other words, the effect of μ is q q k,r k,r−2 merely a uniform shift of the spectral parameters r by −2. The argument is similar for g of type B . The sequence of mutated quivers obtained at each step of μ is displayed in Appendix §6.2. Reading the second quiver of §6.2, we see that the new cluster variable obtained after the first mutation is equal to (2) (1) − − χ W + χ W q 2,−2 q 1,−1 z + z 2,−2 1,−1 (2) =   = χ W . 1,−2 (2) 2,0 χ W 1,0 Similarly, reading the third quiver of §6.2, the new cluster variable obtained after the second mutation is equal to (2) (2) (1) (1) − − − − χ W χ W + χ W χ W q q q q 3,−4 1,−2 1,−1 1,−3 (2) = χ W . q 2,−4 (2) χ W 2,−2 By induction, after every vertex of the second column has been mutated, each cluster variable (2) (2) − − of the form χ W has been replaced by the new cluster variable χ W . We now q k,−2k+2 q k,−2k continue by mutating vertices of the third column. We first get, at the top vertex (1) (2) − − χ W + χ W q q 2,−4 2,−5 (1) = χ W . 1,−5 (1) χ W 1,−1 14 Then, mutating at the next vertex gives (1) (1) (2) − − − χ W χ W + χ W q q q 3,−9 1,−5 4,−8 (1) = χ W . q 2,−9 (1) χ W 2,−5 By induction one sees that, after every vertex of the third column has been mutated, each cluster (1) (1) − − variable of the form χ W has been replaced by the new cluster variable χ W . q k,−4k+3 q k,−4k−1 μ , we mutate again along the second column. One checks that after that, each For the third part of (2) cluster variable of the form χ W produced after the first part of μ has been replaced by k,−2k (2) χ W . Finally, the fourth part of μ along the first column replaces each cluster variable k,−2k−2 (1) (1) − − of the form χ W by the new cluster variable χ W . Moreover, one sees that q q k,−4k+1 k,−4k−3 the new quiver obtained after μ is nothing else than G . Hence we conclude that one application (i) of μ produces a seed with the same quiver, and in which every cluster variable χ W has k,r (i) been replaced by χ W . In other words, the effect of μ is merely a uniform shift of the k,r−4 spectral parameters r by −4. The argument is similar for g of type G . The quiver G for this case is dispayed in Figure 3, and the mutation sequence is (2, 0),(2,−2),(2,−4),... ,(1,−1),(1,−7),(1,−13),... , (2, 0),(2,−2),(2,−4),... ,(1,−3),(1,−9),(1,−15),... , (2, 0),(2,−2),(2,−4),... ,(1,−5),(1,−11),(1,−17),... . The sequence of mutated quivers obtained at each step of μ is displayed in Appendix §6.3. For a general g, we use a reduction to rank two. Namely, we show that mutation sequences and T -systems equations are compatible with rank two reductions. First, by construction, the sequence of vertices S is a union of t n columns: S = (S ,S ,···,S ), 1 2 t n where each column S is a subset of i ×Z for a certain i ∈ I. As above we use μ to denote k k ≤0 k S the sequence of mutations indexed by S . So we have μ = μ ◦ μ ◦···◦ μ . S S S S t n t n−1 1 For 0 ≤ k ≤ t n, we get the mutated quiver Σ = (μ ◦ μ ◦···◦ μ )(Σ). k S S S k k−1 1 For a subset J ⊂ I, let us denote by (Σ ) the subquiver of Σ obtained by deleting the vertices k J k (i, r) such that i ∈ / J, and the edges whose tail or head is such a vertex. For any i ∈ I, the mutation sequence μ modifies (Σ ) to itself. Consequently, (Σ ) = (Σ) does not depend on k (it is a S k i k i i disjoint union of d semi-infinite linear quivers). Besides, the mutation sequence μ modifies i S only the edges whose tail (resp. head) is in i ×Z and head (resp. tail) is in j ×Z where c < 0. k i j This is because each mutation of the sequence takes place at a vertex (i , r) having two incoming arrows from vertices (i , r ± d ) and outgoing arrows to vertices of the form ( j, s) with c < 0. k i i j Consequently, for each i 6= j in I, the effect of the mutation sequence μ on (Σ) is the same as {i, j} the effect of an iteration of the mutation sequence corresponding to the rank two Lie subalgebra 15 (2,0) OO ❀ (2,−2) OO ✝ ❀ (1,−1) OO ✝ ✝ ✝ ✝ ✝ ✝ (2,−4) ✝ ✝ OO ● ✝ ✝ ✝ ✝ ✝ ✝ ✝● ✝ ● (1,−3) ✝ ● OO ✽ ● ✽ ✝ ✝ ● (2,−6) ✽ OO ✽ ● ✽ ✽ ✽ ✽ ## ✽ ✽ ✽ ✽ (1,−5) ✽ ✽ OO ✽ ✽ ✽ ✇ (2,−8) OO ✽ ✇ ✝ ✽ ✝ ✇ (1,−7) OO ✝ ✇ ✇ ✝ ✝ ✝ ✝ {{✇ ✝ ✝ ✝ (2,−10) ✝ ✝ OO ● ✝ ✝ ✝ ✝ ✝ ● ✝ ✝● (1,−9) ● OO ● ✝ ● (2,−12) ● OO ● ## (1,−11) . . . . . . . . . . . . Figure 3: The quiver G for g of type G . of g attached to {i, j} ⊂ I. But we have established the result for rank two Lie algebras, so this implies (μ (Σ)) = (Σ) . {i, j} {i, j} As this is true for any i 6= j in I, we get μ (Σ) = Σ. Secondly, a T -system equation involves only a certain index i ∈ I and the indices j ∈ I with c < 0. The T -system equations do not change by reduction, in the sense that for such a j, the i j ( j) (i) powers of the factors T in the second term S of the right-hand side of (9) are the same as for l,s k,r the T -system equation associated with the rank two Lie subalgebra of g attached to {i, j}. Com- bining with our results above for the subquivers (Σ ) , we have proved that, for a general g, k {i, j} all exchange relations of cluster variables of our mutation sequence are in fact T -system equa- tions. Moreover, the mutation sequence μ replaces the initial seed Σ by a seed with the same quiver; the cluster variables, expressed in terms of the Y via (3), are truncated q-characters of the i,r same Kirillov-Reshetikhin modules, the only difference being that their spectral parameters are uniformly shifted by −2t . 16 Hence, after m applications of μ we will get the truncated q-characters (m) (i) y = χ W . i,r q k , r−2t m i,r Now taking into account [FM, Corollary 6.14], we see that if 2t m ≥ t hˇ, then all the monomials of (i) the q-character of W are lower than the level of truncation, that is, k , r−2t m i,r (i) (i) χ W = χ W . q k , r−2t m k , r−2t m i,r i,r This finishes the proof of Theorem 3.1. 4 A geometric character formula for Kirillov-Reshetikhin modules 4.1 Semi-infinite quivers with potentials − −1 − − Recall the map ψ :V → W of §2.1.3. Put V := ψ (W ), and denote by Γ the full subquiver − − − of Γ with vertex set V . Thus Γ is the same graph as G , but with a change of labelling of its vertices. (Compare for instance Figure 3 and Figure 7.) − − For every i 6= j in I with c 6= 0, and every (i, m) in V , we have in Γ an oriented cycle: i j (i, m) (13) OO ❑ (i, m− b ) ❑ ii %% . ( j, m + b ) i j (i, m + 2b + b ) s i j ii s OO yys (i, m + 2b ) i j There are 2|b |/b = |c | consecutive vertical up arrows, hence this cycle has length 2+|c |. We i j ii i j i j define a potential S as the formal sum of all these oriented cycles up to cyclic permutations, see [DWZ1, §3]. This is an infinite sum, but note that a given arrow of Γ can only occur in a finite number of summands. Hence all the cyclic derivatives of S, defined as in [DWZ1, Definition 3.1], are finite sums of paths in Γ . Let R be the list of all cyclic derivatives of S. Let J denote the two-sided ideal of the path algebra CΓ generated by R. Following [DWZ1], we now introduce Definition 4.1 Let A be the infinite-dimensional C-algebra CΓ /J. Example 4.2 Let g be of type A . Then Γ is the first graph in Figure 4. The ideal J is generated by the following 7 families of linear combinations of paths, for every m ∈ Z , <0 ((1, 2m),(2, 2m− 1),(1, 2m− 2)), ((3, 2m),(2, 2m− 1),(3, 2m− 2)), ((1, 2m),(1, 2m + 2),(2, 2m + 1)) + ((1, 2m),(2, 2m− 1),(2, 2m + 1)), ((3, 2m),(3, 2m + 2),(2, 2m + 1)) + ((3, 2m),(2, 2m− 1),(2, 2m + 1)), ((2, 2m− 1),(1, 2m− 2),(1, 2m)) + ((2, 2m− 1),(2, 2m + 1),(1, 2m)), ((2, 2m− 1),(3, 2m− 2),(3, 2m)) + ((2, 2m− 1),(2, 2m + 1),(3, 2m)), ((2, 2m + 1),(1, 2m),(2, 2m− 1)) + ((2, 2m + 1),(3, 2m),(2, 2m− 1)). 17 (2,−1) (2,−1) OO ❑ OO ❑ s ❑ ❑ s ❑ ❑ ❑ ❑ s ❑ ❑ yys %% %% (1,−2) (3,−2) (2,−3) (1,−3) OO ❑ OO OO OO ❑ s s s ❑ s s s ❑ s s s s s s %% yys yys yys (2,−3) (1,−5) (2,−5) OO ❑ OO ❑ OO ❑ ❑ ❑ ❑ s ❑ ❑ ❑ s ❑ ❑ ❑ ❑ ❑ ❑ yys %% %% %% (1,−4) (3,−4) (2,−7) (1,−7) OO ❑ OO OO OO s s s ❑ s s s ❑ s s s s s s %% yys yys yys (2,−5) (1,−9) (2,−9) ✇ ● ✇ ● ✇ ● ✇ . . . {{✇ ## . . . . . . (1,−6) . (3,−6) Figure 4: The quivers Γ for g of type A and B . 3 2 Here, using the fact that there is at most one arrow between two vertices of Γ , we have denoted unambiguously paths by sequences of vertices. Thus ((1, 2m),(2, 2m− 1),(1, 2m− 2)) denotes the path of length 2 starting at (1, 2m), passing by (2, 2m− 1) and ending in (1, 2m− 2)). Also, for m = −1, the third and fourth linear combinations of paths reduce respectively to the single paths ((1,−2),(2,−3),(2,−1)) and ((3,−2),(2,−3),(2,−1)). Example 4.3 Let g be of type B . Then Γ is the second graph of Figure 4. The ideal J is generated by the following 4 families of linear combinations of paths, for every m ∈ Z , <0 ((1, 2m− 1),(2, 2m− 3),(1, 2m− 5)), ((1, 2m− 1),(1, 2m + 3),(2, 2m + 1)) + ((1, 2m− 1),(2, 2m− 3),(2, 2m− 1),(2, 2m + 1)), ((2, 2m− 3),(1, 2m− 5),(1, 2m− 1)) + ((2, 2m− 3),(2, 2m− 1),(2, 2m + 1),(1, 2m− 1)), ((2, 2m + 1),(2, 2m + 3),(1, 2m + 1),(2, 2m− 1)) + ((2, 2m + 1),(1, 2m− 1),(2, 2m− 3),(2, 2m− 1)). For m = −1 and m = −2 the second linear combinations of paths reduce respectively to the single paths ((1,−3),(2,−5),(2,−3),(2,−1)) and ((1,−5),(2,−7),(2,−5),(2,−3)). For m = −1 the fourth linear combination of paths reduces to the single path ((2,−1),(1,−3),(2,−5),(2,−3)). 4.2 F -polynomials of A-modules Let M be a finite-dimensional A-module, and let e ∈ N be a dimension vector. Let Gr (M) be the variety of submodules of M with dimension vector e. This is a projective complex variety, and we denote by χ(Gr (M)) its Euler characteristic. Following [DWZ2], consider the polynomial i,r F = χ(Gr (M)) v (14) M e ∑ ∏ i,r − − V (i,r)∈V e∈N 18 − in the indeterminates v ((i, r) ∈ V ), called the F -polynomial of M. Note that, for Gr (M) i,r e to be nonempty, one must take e between 0 and the dimension vector of M (componentwise). Moreover, if e = 0 or e = dim(M), the variety Gr (M) is just a point, so F is a monic polynomial e M with constant term equal to 1. In the sequel, we shall evaluate the variables of the F -polynomials at the inverses of the vari- ables A introduced in (6), namely: i,r −1 −1 −1 v := A = Y Y Y Y Y Y Y Y . (15) i,r j,r j,r−1 j,r+1 j,r−2 j,r j,r+2 i,r ∏ ∏ ∏ i,r−d i,r+d i i j: c =−1 j: c =−2 j: c =−3 ji ji ji 4.3 Generic kernels Suppose that X and Y are A-modules such that Hom (X,Y ) is finite-dimensional. Assume also that there exists f ∈ Hom (X,Y ) such that Ker( f ) is finite-dimensional. Then, there is an open dense e e subset O of Hom (X,Y ) such that the kernels of all elements of O are finite-dimensional. More- over, since the map sending a homomorphism f to the F -polynomial of Ker( f ) is constructible (see [Pa, §2]), O contains an open dense subset O of Hom (X,Y ) such that the F -polynomials of the kernels of all elements of O coincide. We shall say that an element of O is a generic homomorphism from X to Y . Let us denote by S the one-dimensional A-module supported on (i, m) ∈ V . Let I be the i,m i,m (infinite-dimensional) injective A-module with socle isomorphic to S . The C-vector space I i,m i,m has a basis indexed by classes modulo J of paths in Γ with final vertex (i, m). In particular, for every k ≥ 0 we have in Γ a path ((i, m− kb ),(i, m− (k− 1)b ),...,(i, m)) (16) ii ii of length k from (i, m − kb ) to (i, m), whose class modulo J is nonzero. Thus the (i, m − kb )- ii ii component of the dimension vector of I is nonzero, and it follows that i,m Hom (I , I ) 6= 0, ((i, m) ∈ V , k ≥ 0). (17) A i,m i,m−kb ii More precisely, Hom (I , I ) has finite dimension equal to the (i, m − kb )-component of A i,m i,m−kb ii ii the dimension vector of I . The next Lemma will be proven in §4.5.3. i,m Lemma 4.4 There exists f ∈ Hom (I , I ) with Ker( f ) finite-dimensional. A i,m i,m−kb ii Because of this lemma, the following definition makes sense. (i) Definition 4.5 Let K be the kernel of a generic A-module homomorphism from I to I . i,m i,m−kb ii k,m (i) Example 4.6 Figure 5 and Figure 6 show the structure of some modules K in type A . Our k,m convention for displaying these quiver representations is the following. We only keep the vertices of Γ whose corresponding vector space is nonzero, and the arrows whose corresponding linear map is nonzero. Moreover, in these small examples, almost all vertices carry a vector space of (2) dimension 1. The only exception is the module K in Figure 6, whose vertex (2,−3) carries a 2,−3 vector space of dimension 2. The maps associated with the arrows incident to this vertex have the following matrices α = β = γ = , δ = ε = κ = ( 0 1). 19 (3,−2) (2,−1) (1,−2) ❑ ❑ s s ❑ ❑ s s ❑ ❑ ❑ ❑ s s s s ❑ ❑ yys yys %% %% (2,−3) (1,−2) (3,−2) (2,−3) ❑ ❑ s ❑ s ❑ s ❑ s ❑ ❑ ❑ s s s ❑ s ❑ yys %% yys %% (1,−4) (2,−3) (3,−4) (1) (2) (3) Figure 5: The modules K , K , K for g of type A . 1,−4 1,−3 1,−4 (3,−2) (2,−1) (1,−2) OO OO ❑ OO ❑ s s ❑ ❑ ❑ ❑ s s s s ❑ ❑ s s ❑ ❑ yys yys %% %% (2,−3) (1,−2) (3,−2) (2,−3) OO OO ❑ OO OO ❑ s ❑α s ❑ ❑ ❑ s s s ❑ s ❑ s ❑ s ❑ yy %% yys %% (1,−4) (3,−4) (2,−3) (1,−4) (3,−4) OO OO ❑ ❑ OO s s ❑ ε ❑ s s ❑ ❑ s s ❑ ❑ s s ❑ ❑ yys yys %% %% (2,−5) (1,−4) (3,−4) (2,−5) ❑ ❑ s ❑ s ❑ s ❑ s ❑ ❑ ❑ s s s ❑ s ❑ yys %% yys %% (1,−6) (2,−5) (3,−6) (1) (2) (3) Figure 6: The modules K , K , and K for g of type A . 2,−4 2,−3 2,−4 All other arrows carry linear maps with matrix (±1), whose sign is easily deduced from the defining relations of A. It is a nice exercise to check that the modules shown in Figure 5 and Figure 6 are indeed (i) the claimed modules K (see also Example 4.7 below). For instance, one can easily see that k,m the (1,−6)-component of the dimension vector of I is equal to 1. Hence Hom (I , I ) 1,−4 A 1,−4 1,−6 (1) is of dimension 1, and K is the kernel of any nonzero homomorphism. It is also easy to 1,−4 see that the (2,−5)-component of the dimension vector of I is equal to 2. In this case we 2,−3 have a stratification of the 2-plane Hom (I , I ) with three strata of dimension 0, 1, 2. The A 2,−3 2,−5 (2) module K is the kernel of any homomorphism in the open stratum, that is, of any surjective 1,−3 homomorphism. The image of any homomorphism in the one-dimensional stratum is the unique submodule X of I with dimension vector given by 2,−5 1 if i = 2 and m = −5− 2 j for some j ∈ N, dim(X ) = i,m 0 otherwise. The kernel of such a homomorphism is infinite-dimensional. (i) Example 4.7 Let us assume that g is of type A, D, E . In this case, the modules K are closely 1,r related to the indecomposable injective modules over the preprojective algebra Λ of δ . Consider the subalgebra Λ of A generated by the images modulo J of the arrows of Γ of the form (i, m) → ( j, m − 1), for every edge between i and j in δ , and every (i, m) ∈ V . In other words, if Δ is the subquiver of Γ obtained by erasing all the vertical arrows (i, m− 2) → (i, m), then Λ is isomorphic to the quotient of CΔ by the two-sided ideal generated by the relations ((i, m),( j, m− 1),(i, m− 2)) = 0, ((i, m) ∈ V ). j: c <0 i j 20 e e Thus, Λ is a Z -graded version of Λ. We can of course regard the simple A-module S as a Λ- <0 i,r module. Let H be the injective Λ-module with socle S . Then H is finite-dimensional. More i,r i,r i,r precisely, for r ≤ 1− h, H is just a graded version of the indecomposable injective Λ-module I i,r i with socle the one-dimensional Λ-module S supported on vertex i of δ . For r > 1− h, H is a i i,r graded version of a submodule of I . Any Λ-module X can be given the structure of an A-module by letting the vertical arrows (i, m − 2) → (i, m) act by 0 on X . In particular we can regard H as a finite-dimensional A- i,r module. Then one can check that I has a unique submodule isomorphic to H , giving rise to a i,r i,r non-split short exact sequence 0 → H → I → I → 0, ((i, r) ∈ V ). i,r i,r i,r−2 (i) (i) It follows that the module K is isomorphic to H . In particular, when m ≤ 1 − h, K is a i,m 1,m 1,m graded version of the injective Λ-module I . 4.4 A geometric character formula (i) Recall the A-module K defined in §4.3. We can now state our second main result. k,r (i) Theorem 4.8 Let (i, r) ∈ V and k ∈ N. The F -polynomial of K is equal to the normalized k,r (i) truncated q-character of the Kirillov-Reshetikhin module W . More precisely, we have k, r−(2k−1)d (i) χ W = Y F , (i) ∏ i, r−(2s−1)d q i k, r−(2k−1)d i K k,r s=1 where the variables v of the F -polynomial are evaluated as in (15). i,r (i) Remark 4.9 If r ≤ d − t hˇ, then the truncated q-character of W is equal to the complete k, r−(2k−1)d q-character. Hence, Theorem 4.8 gives a geometric formula for the q-character of any Kirillov- Reshetikhin module (up to a spectral shift). Remark 4.10 If M and N are two finite-dimensional A-modules, then F = F F [DWZ2, M⊕N M N (i) Proposition 3.2]. It follows immediately that, replacing in Theorem 4.8 the module K by a k,r direct sum of such modules, we obtain a similar geometric character formula for arbitrary ten- sor products of Kirillov-Reshetikhin modules. In particular, we get a geometric formula for the standard modules, which are isomorphic to tensor products of fundamental modules. Remark 4.11 Let g be of type A, D, E . Let V and W be finite-dimensional vector spaces graded − • by V . In [N1] (see also [N4]), Nakajima has introduced a graded quiver variety L (V, W) and has endowed the sum of cohomologies ∗ • H (L (V, W)) with the structure of a standard U (bg)-module, with highest weight encoded by W. It was proved by Lusztig (in the ungraded case), and by Savage and Tingley (in the graded case), that L (V, W) is homeomorphic to a Grassmannian of submodules of an injective module over the graded pre- (i) projective algebra (see [Le2, §2.8]). Therefore, using the description of K given in Example 4.7, 1,r we see that the varieties   M ⊕a i,r (i)   Gr K 1,r (i,r) 21 (2,−1) OO ❁ (2,−3) OO ✆ ❁ (1,−4) OO ✆ ✆ ✆ ✆ ✆ ✆ ✆ ✆ (2,−5) ✆ ✆ OO ❍ ✆ ✆ ✆ ✆ ✆ ✆ ✆❍ ✆ ❍ (1,−6) ✆ ❍ OO ✾ ❍ ✾ ✆ ✆ ❍ ✾ ❍ (2,−7) ✾ OO ✾ ❍ ✾ ✾ ✾ ✾ ## ✾ ✾ ✾ ✾ (1,−8) ✾ ✾ OO ✾ ✾ ✾ ✈ ✾ ✈ (2,−9) OO ✾ ✈ ✆ ✾ ✈✾ ✆ ✈ (1,−10) OO ✆ ✈ ✈ ✆ ✆ ✆ ✆ {{✈ ✆ ✆ ✆ (2,−11) ✆ ✆ OO ❍ ✆ ✆ ✆ ✆ ✆ ❍ ✆ ✆❍ (1,−12) ❍ OO ❍ ✆ ❍ (2,−13) ❍ OO ## (1,−14) . . . . . . . . . . . . Figure 7: The quiver Γ for g of type G . involved in our geometric q-character formula for standard modules in the simply laced case are homeomorphic to certain Nakajima varieties L (V, W). Here, the multiplicities a are the di- i,r mensions of the graded components of W, and we assume that a = 0 if r > 1− h. Similarly the i,r graded dimension of V is encoded by the dimension vector e. Example 4.12 Let g be of type A . We have −1 −1 −1 −1 −1 −1 v = Y Y Y , v = Y Y Y Y , v = Y Y Y . 1,r 2,r 2,r 1,r 3,r 3,r 2,r 1,r−1 1,r+1 2,r−1 2,r+1 3,r−1 3,r+1 We continue Example 4.6. The submodule structure of the A-modules displayed in Figure 5 is very (i) simple. Indeed, in this case, all the nonempty varieties Gr (K ) are reduced to a single point, and k,r their Euler characteristics are equal to 1. Therefore the F -polynomial reduces to a generating (i) polynomial for the dimension vectors of the (finitely many) submodules of K . This yields the k,r 22 following well known formulas for the q-characters of the fundamental modules: χ (L(Y )) = Y (1 + v + v v + v v v ) q 1,−5 1,−5 1,−4 1,−4 2,−3 1,−4 2,−3 3,−2 −1 −1 −1 = Y + Y Y + Y Y + Y , 1,−5 2,−4 3,−3 1,−3 2,−2 3,−1 χ (L(Y )) = Y (1 + v + v v + v v + v v v q 2,−4 2,−4 2,−3 1,−2 2,−3 2,−3 3,−2 1,−2 2,−3 3,−2 + v v v v ) 1,−2 2,−3 3,−2 2,−1 −1 −1 −1 −1 −1 −1 = Y + Y Y Y + Y Y + Y Y + Y Y Y + Y , 2,−4 1,−3 3,−3 3,−3 1,−3 2,−2 2,−2 1,−1 3,−1 1,−1 3,−1 2,0 Similarly, the A-modules shown in Figure 6 give the following Kirillov-Reshetikhin q-characters: χ (L(Y Y )) = Y Y (1 + v (1 + v + v + v v + v v q 1,−7 1,−5 1,−7 1,−5 1,−4 1,−6 2,−3 1,−6 2,−3 2,−3 3,−2 + v v v + v v v + v v v v 1,−6 2,−3 2,−5 1,−6 2,−3 3,−2 1,−6 2,−3 2,−5 3,−2 + v v v v v )), 1,−6 2,−3 2,−5 3,−2 3,−4 χ (L(Y Y )) = Y Y (1 + v (1 + v + v + v + v v + v v q 2,−6 2,−4 2,−6 2,−4 2,−3 1,−2 2,−5 3,−2 1,−2 2,−5 1,−2 3,−2 + v v + v v v + v v v + v v v 2,−5 3,−2 1,−2 2,−5 3,−2 1,−2 2,−5 1,−4 1,−2 3,−2 2,−1 + v v v + v v v v + v v v v 2,−5 3,−2 3,−4 1,−2 2,−5 3,−2 1,−4 1,−2 2,−5 3,−2 2,−1 + v v v v + v v v v v 1,−2 2,−5 3,−2 3,−4 1,−2 2,−5 3,−2 1,−4 2,−1 + v v v v v + v v v v v 1,−2 2,−5 3,−2 1,−4 3,−4 1,−2 2,−5 3,−2 3,−4 2,−1 + v v v v v v + v v v v v v v )), 1,−2 2,−5 3,−2 1,−4 2,−1 3,−4 1,−2 2,−5 3,−2 1,−4 2,−1 3,−4 2,−3 We omit the q-characters χ (L(Y )) and χ (L(Y Y )), since they are readily obtained from q 3,−5 q 3,−5 3,−7 χ (L(Y )) and χ (L(Y Y )) via the symmetry 1 ↔ 3. q 1,−5 q 1,−5 1,−7 Example 4.13 Let g be of type G , with the long root being α . The quiver Γ is shown in 2 1 (1) (2) Figure 7. The modules K and K with r ≤ −10 and s ≤ −11 have dimension 10 and 6, 1,r 1,s (1) (2) (1) respectively. For instance, K and K are represented in Figure 8. In the module K the 1,−10 1,−11 1,−10 vector space sitting at vertex (2,−7) has dimension 2 (all other spaces have dimension 1). The maps incident to this space are given by the following matrices (see Figure 8): 0 1 α = , β = ( 1 0), γ = , γ = ( 0 1). 1 0 The corresponding fundamental modules have dimension dim L(Y ) = 15, dim L(Y ) = 7. 1,−13 2,−12 (1) (2) The Grassmannians of submodules of K and K are in this case again all reduced to points, 1,−10 1,−11 and the formula of Theorem 4.8 amounts to an enumeration of the dimension vectors of all sub- modules. This gives χ (L(Y )) = Y (1 + v (1 + v (1 + v (1 + v + v + v v q 1,−13 1,−13 1,−10 2,−7 2,−9 1,−6 2,−11 1,−6 2,−11 + v v + v v + v v v + v v v 1,−6 2,−3 2,−11 1,−8 1,−6 2,−11 2,−3 1,−6 2,−11 1,−8 + v v v v (1 + v (1 + v (1 + v ))))))), 1,−6 2,−11 2,−3 1,−8 2,−5 2,−7 1,−4 χ (L(Y )) = Y (1 + v (1 + v (1 + v (1 + v (1 + v (1 + v )))))), q 2,−12 2,−12 2,−11 1,−8 2,−5 2,−7 1,−4 2,−1 where, following (15), we have −1 −1 −1 −1 v = Y Y Y Y Y , v = Y Y Y . 1,r 2,r+2 2,r 2,r−2 2,r 1,r 1,r+3 1,r−3 2,r+1 2,r−1 23 (2,−3) (2,−1) OO ❃ ✝ ❃ (1,−4) ❃ ✝ ✆ ✝ ✆ ❃ ✝ ✆ ❃ ✆ ❃ ✝ ✆ ❃ (2,−5) ✝ ✆ OO ● ✝ ✆ ✝ (1,−4) ✝ ✆ ✝ ● ✆ ′ ✄ γ ● (1,−6) ✄ ✆ ● ✽ ● ● ✄ ✆ (2,−5) ● ✄ ✽ ✆ OO ■ ✄ ✆ ● ✽ ● ● ■✄ (2,−7) ● ✄ ■ ✽ ● ✽ OO ✾ ● ● ✄ ✽ ✾ ■ ● ✄ ■ ● ■ ✽ ✾ ## ■ ✽ ✾ ■ ✽ γ ✾ (1,−8) ✽ (2,−7) ■ ✽ ✾ ■ ✇ ■ ✽ ✾ ✽ ✾ ■ ✾ ■ ✇ $$ (2,−9) ✾ ✇ (1,−8) OO ✾ ✇ ✾ s ✇ s ✇  s ✇ s (1,−10) ✇ s ✇ s {{✇ (2,−11) s yys (2,−11) (1) (2) Figure 8: The modules K and K for g of type G . 1,−10 1,−11 Remark 4.14 Assuming Theorem 4.8, we can easily calculate the dimension vectors of the A- (i) modules K for r ≤ d − t hˇ. Indeed, by [FM, Lemma 6.8], the lowest monomial of χ (Y ) i q i,r−d 1,r i −1 is equal to Y , where ν is the involution of I defined by w (α ) = −α . Denote by 0 i ν(i) ν(i),r−d +t hˇ (i) (i) d (K ) the dimension vector of K . Then, we have j,s 1,r 1,r (i) d (K ) j,s 1,r −1 Y = Y v , i,r−d i ∏ ν(i),r−d +t hˇ j,s ( j,s)∈V (i) and using (15), this equation determines the numbers d (K ). In particular, if we introduce the j,s 1,r (i) ungraded dimension vector (d (i)) of K by 1,r (i) d (i) := d (K ), (r ≤ d − t hˇ), j ∑ j,s i 1,r we can deduce from this the nice formula d (i)α = β, (18) j j ∑ ∑ i, j∈I β∈Φ >0 where Φ is the set of positive roots of g. This can be observed in Figure 5 and Figure 8 (see >0 also §6.4, §6.5, §6.6, §6.7 below). When g is of type A, D, E , as explained in Remark 4.7 the (i) modules K are graded versions of the indecomposable injective modules over the preprojective 1,r algebra Λ, and formula (18) recovers a well known property of Λ. 24 4.5 Proof of the theorem The proof relies on Theorem 3.1, and on the categorification of cluster algebras by means of quivers with potentials, developed by Derksen, Weyman and Zelevinsky [DWZ1, DWZ2]. This categorification provides (among other things) a description of cluster variables in terms of Grass- mannians of submodules, which will be our key ingredient. An important additional result will be borrowed from Plamondon [Pl2]. 4.5.1 F -polynomials and g-vectors of cluster variables − − Recall the cluster algebra A of §2.2.1, with initial seed (z , G ). Following [FZ3, (3.7)], define −1 − yb := z z , ((i, r) ∈ W ). (19) i,r ∏ j,s ∏ j,s (i,r)→( j,s) ( j,s)→(i,r) Here the first (resp. second) product is over all outgoing (resp. incoming) arrows at the vertex (i, r) of the graph G . The following result is similar to [HL1, Lemma 7.2]. Lemma 4.15 After performing in (19) the change of variables (3), there holds −1 − yb = A , ((i, r) ∈ W ), i,r i, r−d where the Laurent monomials A are given by (6). i,r Proof — Using the definition of the quiver G , we can rewrite (19) as z z j, r+b +d −d i, r+b i j j i ii yb = , i,r ∏ z z i, r−b j, r−b +d −d ii i j j i j6=i where the product is over all j’s such that c 6= 0. Here we use the convention that z = 1 for i j i,s every (i, s) with s > 0. Using the change of variables (3), we obtain −1 −1 yb = Y Y Y Y ··· Y . i,r r, r−d +b +d r, r−d +b +3d r, r−d −b −d i,r ∏ i i j j i i j j i i j j i, r−b ii j6=i; c 6=0 i j The result then follows by comparison with (6), if we notice again that b + d = c + 1 because i j j ji of (1). ✷ In [FZ3] Fomin and Zelevinsky attach to every cluster variable x of A a polynomial F with − (W ) integer coefficients in the set of variables b y = {yb | (i, r) ∈ W }, and a vector g ∈ Z , such i,r x that [FZ3, Corollary 6.3] x = z F (b y). (20) Note that A has no frozen cluster variables, so there is no denominator in (20). The polynomial F and the integer vector g are called the F -polynomial and g-vector of the cluster variable x, x x respectively. We refer the reader to [FZ3] for their definition. On the other hand, it follows from the theory of q-characters that for every simple U (g)- − − module L(m) in the category C , the truncated q-character χ (L(m)) can be written as χ (L(m)) = mP , (21) −1 − where P is a polynomial with integer coefficients in the variables {A | (i, r) ∈ W }. More- i, r−d over, P has constant term 1. 25 Now, by the proof of Theorem 3.1, among the cluster variables of A , we find all the truncated q-characters of the Kirillov-Reshetikhin modules of C . These are of the form L(m) with k−1 (i) m = m := Y , ((i, r) ∈ W , r + (k− 1)b ≤ 0). (22) i, r+ jb ii ∏ ii k,r j=0 Proposition 4.16 The g-vector of the truncated q-character of the Kirillov-Reshetikhin module (i) (i) W = L m , considered as a cluster variable of A , is given by k,r k,r  1 if ( j, s) = (i, r), g = −1 if ( j, s) = (i, r + kb ) and r + kb ≤ 0, j,s ii ii 0 otherwise. (i) Proof — Write for short m = m , and denote by x the cluster variable χ (L(m)). Then, compar- k,r ing (20) with (21), we have −1 g P = m z F , m x where, by Lemma 4.15, P and F are polynomials in the same variables m x −1 yb = A . i,r i,r−d −g Since P has constant term 1, it follows that mz is a monomial in the variables yb which divides m i,r the F -polynomial F . But, by [FZ3, Proposition 5.2], F is not divisible by any yb . So, using (3), x x i,r i,r z = m = , i, r+kb ii where as above, we set z = 1 if s > 0. ✷ i,s 4.5.2 Truncated algebras Let ℓ ∈ Z . Let Γ be the full subquiver of Γ with set of vertices <0 − − V := {(i, m) ∈ V | m ≥ ℓ}. Let S be the corresponding truncation of the potential S, that is, S is defined as the sum of all ℓ ℓ − − cycles in S which only involve vertices of V . Let J denote the two-sided ideal of CΓ generated ℓ ℓ by all cyclic derivatives of S . Finally, define the truncated algebra at height ℓ as A := CΓ /J . ℓ ℓ Proposition 4.17 For every ℓ we have: (i) the algebra A is finite-dimensional; (ii) the quiver with potential (Γ , J ) is rigid. Proof — The proof is similar to [DWZ1, Example 8.7]. Let π : CΓ → A be the natural projec- tion. To prove (i), we show that A is spanned by the images under π of a finite number of paths. The arrows of Γ are of two types: 26 (a) the vertical arrows of the form (i, m) → (i, m + b ); ii (b) the oblique arrows of the form (i, m) → ( j, m + b ) provided c < 0. i j i j Let us say that a path from (i, m) to ( j, s) in Γ is going up (resp. down) if m < s (resp. m > s). Note that all vertical arrows go up and all oblique arrows go down. Each oblique arrow of the boundary of Γ belongs to a single cycle of the potential S , and each interior oblique arrow belongs to exactly two cycles. Therefore each interior oblique arrow gives rise to a “commutativity relation” in A : π(( j, m + b ),(i, m + 2b ),(i, m + 2b + b ),...,(i, m− b ),(i, m)) ji ji ji ii ii = −π(( j, m + b ),( j, m + b + b ),... ,( j, m− b − b ),( j, m− b ),(i, m)) ji ji j j ji j j ji The path in the left-hand side consists of an oblique arrow followed by |c | vertical arrows, while i j the right-hand side has |c | vertical arrows followed by an oblique arrow. Let p be a path in Γ ji with origin (i, m). Using only the above type of commutativity relations, we can bring a number of vertical arrows to the front of p and write π( p) = π( p )π( p ), 2 1 where p is a path with origin (i, m) consisting only of vertical arrows, and p is a path satisfying 1 2 the following property: if q is a maximal factor of p containing only vertical arrows, then q is preceded by at least one oblique arrow, say ( j, s) → (k, s + b ), and q contains less than |c | jk k j arrows. Hence q can be non trivial only if |c | > 1. k j In particular in the simply laced case, then p contains only oblique arrows. In that case, we can immediately conclude that all arrows of p go up and all arrows of p go down, so the lengths 1 2 of p and p are both bounded by ℓ, and therefore A is finite-dimensional. 1 2 ℓ Otherwise, if q is non trivial and p contains other vertical arrows after q, then q needs to be followed by at least two oblique arrows. Indeed, using the same notation as above, q consists of N vertical arrows of the form (k, r) → (k, r + b ) with 1 ≤ N < |c |. Now, by (1), the inequality kk k j |c | > 1 implies d = 1 and d = |b |. Let (k, t) → (l, t + b ) be the first arrow coming after q. k j k j k j kl Then, since d = 1 we have |c | = 1. If this oblique arrow is followed by a vertical one (l, t + k lk b ) → (l, t + b + b ), then we can use the commutativity relation and bring it, together with all kl kl ll the vertical arrows possibly following it, on top of q. In this way, we replace q by a vertical path q followed by two consecutive oblique arrows. One then easily checks by inspection that the subpath of p containing q together with the oblique arrow preceding it and the oblique arrow following it, is going down. Therefore, by induction, p can be factored into a product of paths, each of them of length less than t + 2, and all these paths go down (except possibly the last one, which might end with less than t vertical arrows). So again, the length of p is bounded above, and this proves (i) in all cases. To prove (ii), it is enough to show that every cycle of the form (13) is cyclically equivalent to an element of J . Up to cyclic equivalence, this cycle γ can be written with origin in (i, m). Then, we have: π(γ) = π((i, m),( j, m + b ),(i, m + 2b ),(i, m + 2b + b ),...,(i, m− b ),(i, m)) i j i j i j ii ii = π((i, m),( j, m + b ),( j, m + b + b ),...,( j, m− b − b ),( j, m− b ),(i, m)) i j i j j j i j j j i j = π((i, m),(i, m + b ),...,(i, m− 2b − b ),(i, m− 2b ),( j, m− b ),(i, m)), ii i j ii i j i j and the last path is cyclically equivalent to ((i, m− 2b ),( j, m− b ),(i, m),(i, m + b ),... ,(i, m− 2b − b ),(i, m− 2b )). i j i j ii i j ii i j 27 This cycle is nothing else than γ shifted vertically up by −2b . Hence, iterating this process, we i j can replace, modulo J and cyclic equivalence, any cycle γ of the form (13) by a similar cycle ′ − ′ γ sitting at the top boundary of Γ . Now the upper oblique arrow of γ does not belong to any other cycle, so it gives rise to a zero relation in A . In other words, γ is cyclically equivalent to an element of J . This proves (ii). ✷ Remark 4.18 In the simply laced case and when |ℓ| is less than the Coxeter number, the algebra A arises as the endomorphism algebra of a (finite-dimensional) rigid module over the preprojective algebra Λ associated with δ , and appears in the works of Geiss, Schro ¨ er and the second author (see [GLS1, GLS2]). This gives another proof of Proposition 4.17 (i) in this case. 4.5.3 Proof of Lemma 4.4 and Theorem 4.8 (i) − − Let (i, r) ∈ V and k ∈ N. By Theorem 3.1, the truncated q-character χ W is a k, r−(2k−1)d cluster variable x of A . By Proposition 4.16, the g-vector of x is given by 1 if ( j, s) = (i, r− 2kd + d ), i i g = −1 if ( j, s) = (i, r + d ), (23) j,s i 0 otherwise. − − − − Note that, since (i, r) ∈ V , we have (i, r + d ) ∈ W . For ℓ < 0, let W := ψ(V ), and put ℓ ℓ − − − − − z ={z | (i, r) ∈ W }. We denote by G the same quiver as Γ , but with vertices labelled by W . i,r ℓ ℓ ℓ ℓ ℓ Clearly, the cluster variable x is a Laurent polynomial in the variables of z for some ℓ ≪ 0, and − − can be regarded as a cluster variable of the cluster algebra A defined by the initial seed z , G . ℓ ℓ By Proposition 4.17 (ii), we can apply the theory of [DWZ1, DWZ2] and deduce that the F - polynomial of x coincides with the polynomial F associated with a certain A -module M. In order M ℓ to identify this module, we apply [Pl2, Remark 4.1], which states in our setting that M is the kernel of a generic element of the homomorphism space between two injective A -modules corresponding to the negative and positive components of the g-vector of x. More precisely, let us denote by S i,m − ℓ the one-dimensional A -module supported on (i, m) ∈ V . Let I be the injective A -module ℓ ℓ i,m with socle isomorphic to S . Then, using (23) and taking into account the change of labelling i,m − − ℓ ℓ ψ :V → W given by (2), we get that M is the kernel of a generic element of Hom (I , I ). ℓ i,r i,r−kb ℓ ℓ ii Finally we can identify M with the kernel of a generic homomorphism between injective A- modules. Indeed, for m < ℓ < 0 we have a natural projection A → A whose kernel is generated m ℓ − − − − by all arrows of Γ starting or ending at a vertex v ∈ V \ V . This induces for every (i, r) ∈ V m m ℓ ℓ ℓ m ℓ an embedding I → I , and we can regard the A-module I as the direct limit of I along (i,r) (i,r) (i,r) (i,r) these maps. Since F is independent of ℓ ≪ 0, we see that M is also the kernel of a generic (i) (i) element of Hom (I , I ), that is, M = K . In particular K is finite-dimensional. This A i,r i,r−kb ii k,r k,r proves Lemma 4.4 and finishes the proof of Theorem 4.8. Remark 4.19 Using the same formula as (14), we can attach to the infinite-dimensional A-module I a formal power series F in the variables v . This series also has an interpretation in terms of i,m I j,r i,m quantum affine algebras. Indeed, by [HJ], the category of finite-dimensional U (bg)-modules can be seen as a subcategory of a category O of (possibly infinite-dimensional) representations of a Borel subalgebra of U (bg). The q-character morphism can be extended to the Grothendieck ring of O (the target ring is also completed). This category contains distinguished simple representations − ∗ called negative fundamental representations L (i ∈ I, a ∈ C ) [HJ, Definition 3.7]. Denote by i,a 28 − − χ (L ) the normalized q-character of L , that is, its q-character divided by its highest weight i,a i,a −1 monomial. This normalized q-character is a formal power series in the variables A [HJ, Theo- j,b rem 6.1], and it is obtained as a limit of normalized q-characters of Kirillov-Reshetikhin modules. It is not difficult to deduce from Theorem 4.8 and Remark 4.9 that, for m ≤ d − t hˇ, χ L = F . q I m−d i,m i, q i This is the first geometric description of the q-character of these negative fundamental representa- tions. 5 Beyond Kirillov-Reshetikhin modules 5.1 Grothendieck rings − − Let us consider again the cluster algebra A , with initial seed Σ = (z , G ) whose cluster variables z are given by (3). The Laurent phenomenon for cluster algebras implies that A is a subring of i,r ±1 − Z[Y | Y ∈ Y ]. The following theorem gives the precise relationship between A and the i,r i,r Grothendieck ring of the category C . Theorem 5.1 The cluster algebra A is equal to the image of the injective ring homomorphism ±1 − − − from K (C ) to Z[Y | Y ∈ Y ] given by [L(m)] 7→ χ (m) (see Proposition 3.10). Hence A is 0 i,r i,r isomorphic to the Grothendieck ring of C . − − − Proof — Let R denote the image of the homomorphism [L(m)] 7→ χ (m). By [FR], K (C ) is − − the polynomial ring in the classes of the fundamental modules of C , hence R is the polynomial − − ring in the truncated q-characters χ (Y ) (Y ∈ Y ). By Theorem 3.1, A contains all these i,r i,r fundamental truncated q-characters, hence A contains R . To prove the reverse inclusion, we will use a description of the image of the q-character ho- momorphism as an intersection of kernels of screening operators [FR, FM]. To do this, we need to work with complete (i.e. untruncated) q-characters. So let us consider as in §3.2.2 the larger set of variables Y. Following [FR, §7.1], for every i ∈ I, we have a linear operator S from the ring ±1 Z[Y | Y ∈ Y] to a certain free module Y over this ring, which satisfies the Leibniz rule i,r i i,r ±1 S (xy) = x S (y) + y S (x), (x, y ∈ Z[Y | Y ∈ Y]). i i i i,r i,r ±1 It was conjectured in [FR] and proved in [FM] that an element of Z[Y | Y ∈ Y] is a polynomial i,r i,r χ (Y ) (Y ∈ Y) if and only if it belongs to in the q-characters q i,r i,r Ker S . i∈I Let us now introduce an auxilliary cluster algebra A . It is defined using the same initial seed − − ′ (z , G ) as A , but the initial variables of A are given by the following modification of (3) z := Y , i, r+kb +2t hˇ i,r ∏ ii k≥0, r+kb ≤0 ii in which the spectral parameters are all shifted upwards by 2t hˇ. By Theorem 3.1, if we apply to this initial seed of A the sequence of mutations μ repeated hˇ times, we will obtain a new seed 29 ′ − ′ − Σ with the same quiver G . Moreover, the cluster variable of Σ sitting at vertex (i, r) ∈ W is (i) nothing else than the complete q-character χ (W ). k ,r i,r Consider a cluster variable x of A . By definition, x is obtained from Σ by a finite sequence of mutations μ . We want to show that x belongs to R . By Theorem 3.1, all cluster variables of Σ belong to R , so by induction on the length, we may assume that the last exchange relation of μ is of the form xy = M + M , 1 2 where y is a cluster variable of A , M and M are cluster monomials of A , and y, M , M belong 1 2 1 2 − ′ ′ to R . Let us apply the same sequence of mutations μ in the cluster algebra A to the seed Σ . The last exchange relation will be of the form ′ ′ ′ ′ x y = M + M , 1 2 ′ ′ ′ − where y , M , M are polynomials in the complete fundamental q-characters χ (Y ) (Y ∈ Y ). q i,r i,r 1 2 ′ ′ ′ ′ Moreover, x , y , M , M give back x, y, M , M by application of the truncation ring homomor- 1 2 1 2 ′ ′ ′ ′ ′ phism. By the Laurent phenomenon [FZ1] in the cluster algebra A , we know that x , y , M , M 1 2 are Laurent polynomials in the variables of Y. Since S is a derivation, we have ′ ′ ′ ′ ′ ′ ′ ′ S (x y ) = x S (y ) + y S (x ) = S (M ) + S (M ), i i i i i 1 2 ′ ′ ′ ′ ′ hence S (x ) = 0 because S (y ) = S (M ) = S (M ) = 0. It follows that x is annihilated by all the i i i i 1 2 ′ − screening operators, so x is a polynomial in the q-characters χ (Y ) (Y ∈ Y ). This implies that q i,r i,r − − − x is a polynomial in the truncated q-characters χ (Y ) (Y ∈ Y ), that is, x ∈ R . ✷ i,r i,r 5.2 Conjectures 5.2.1 Cluster monomials In view of Theorem 5.1, it is natural to formulate some conjectures. Following [Le1], let us say that a simple U (g)-module S is real if S⊗ S is simple. Conjecture 5.2 In the above identification of the cluster algebra A with the ring of truncated q-characters of C , the cluster monomials get identified with the truncated q-characters of the real simple modules of C . When g is of type A, D, E , Conjecture 5.2 is essentially equivalent to [HL1, Conjecture 13.2]. But the initial seed used here is different and allows a direct connection between cluster expansions and (truncated) q-characters. 5.2.2 Geometric q-character formulas Using the methods and tools of §4, we can translate Conjecture 5.2 into a new conjectural geomet- ric formula for the (truncated) q-character of a real simple module of C . Let m be a dominant monomial in the variables Y ∈ Y . Using the change of variables (3), i,r which we can express as i,r Y = , ((i, r) ∈ W ), i,r i,r+b ii (where we understand z = 1 if s > 0), we can rewrite i,s g (m) g(m) i,r m = z := z . i,r (i,r)∈W 30 (2,−1) s ❑ s ❑ s ❑ yys %% (1,−2) (3,−2) ❑ OO ❑ s ❑ s ❑ s %% yys (2,−3) OO (3,−4) yys (2,−5) yys (1,−6) Figure 9: The A-module K(m) for m = Y Y in type A . 1,−7 2,−4 3 (W ) Let us call the integer vector g(m) ∈ Z the g-vector of L(m). Following §4.3, let us attach to m the A-module K(m) defined as the kernel of a generic A-module homomorphism from the injective − + A-module I(m) to the injective A-module I(m) , where M M ⊕g (m) ⊕|g (m)| + i,r − i,r I(m) = I , I(m) = I . i,r−d i,r−d i i g (m)>0 g (m)<0 i,r i,r Finally define the F -polynomial F of K(m) as in §4.2. We can now state the following conjec- K(m) tural generalization of Theorem 4.8. Conjecture 5.3 Suppose that L(m) is an irreducible real U (bg)-module in C . Then the truncated q-character of L(m) is equal to χ (L(m)) = mF , K(m) where the variables v of the F -polynomial are evaluated as in (15). i,r Example 5.4 Let g be of type A . Take m = Y Y . We have 3 1,−7 2,−4 + − I(m) = I ⊕ I , I(m) = I ⊕ I . 1,−8 2,−5 1,−6 2,−3 The module K(m) has dimension 7 and is displayed in Figure 9. Using for instance the fact that L(m) is a minimal affinization (in the sense of [C]), we can compute its q-character. We find: χ (L(Y Y )) = Y Y (1 + v + v + v v + v v + v v q 1,−7 2,−4 1,−7 2,−4 1,−6 2,−3 1,−6 2,−3 1,−2 2,−3 2,−3 3,−2 + v v v + v v v + v v v + v v v 1,−6 1,−2 2,−3 1,−6 2,−3 3,−2 1,−6 2,−3 2,−5 1,−2 2,−3 3,−2 + v v v v + v v v v + v v v v 1,−6 1,−2 2,−5 2,−3 1,−6 1,−2 2,−3 3,−2 1,−6 2,−5 2,−3 3,−2 + v v v v + v v v v v 1,−2 2,−5 2,−3 3,−2 1,−6 1,−2 2,−5 2,−3 3,−2 + v v v v v + v v v v v 1,−6 1,−2 2,−3 2,−1 3,−2 1,−6 2,−5 2,−3 3,−4 3,−2 + v v v v v v + v v v v v v 1,−6 1,−2 2,−5 2,−3 3,−4 3,−2 1,−6 1,−2 2,−5 2,−3 2,−1 3,−2 + v v v v v v v ), 1,−6 1,−2 2,−5 2,−3 2,−1 3,−4 3,−2 in agreement with Conjecture 5.3. 31 6 Appendix 6.1 Mutation sequence in type A We display the sequence of mutated quivers obtained from G at each step of the mutation se- quence μ . The first quiver is G , and in the next quivers the box indicates at which vertex a mutation has been performed. (2,0) (2,0) (2,0) (2,0) 99 99 OO OO OO ✉ s s ;; s s ✉ ✎ ✍ ✇ s s ✇ s s ✎ ✍ ✉ s s zz✉ s s ✇ ✍ ✎ ✍ (1,−1) (1,−1) (1,−1) (1,−1) ✎ OO ■ OO OO ■ ✎ OO $$ ✎ (2,−2) (2,−2) (2,−2) (2,−2) ✍ OO 99 OO ✉ s OO ✍ ✉ s ✎ t ✎ :: ✉ ✍ s t ✎ ✉ ✎ ✈ s zz✉ ✈ ✍ s ✈ ✎ zzt ✎ ✈ (1,−3) (1,−3) (1,−3) ✎ OO ■ OO ■ (1,−3) OO ❏ ■ OO ✎ $$ ✎ $$ ✎ (2,−4) (2,−4) (2,−4) OO (2,−4) OO t ✎ ✉ OO :: ✉ s ✎ ✈ zz✉ s ✈ s ✎ ✈ zzt s ✈ (1,−5) yys ✎ ✈ (1,−5) OO ■ ■ (1,−5) (1,−5) OO ❏ ■ OO ❑ OO ■ ❑ $$ $$ %% (2,−6) (2,−6) (2,−6) (2,−6) . t s s ✉ . . zz✉ s s t . . . . s s zzt s s . . (1,−7) yys yys . . (1,−7) (1,−7) (1,−7) (2,0) (2,0) (2,0) (2,0) (2,0) OO OO OO OO OO s s s s s s s s ✍ ✈ s s s s s ✈ s s s s ✈ s s s s zz✈ yys yys yys (1,−1) (1,−1) (1,−1) (1,−1) (1,−1) OO OO ❑ OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ✍ ❑ ❑ ❑ ❑ ❑ ❑ %% %% %% (2,−2) (2,−2) (2,−2) (2,−2) ✍ 99 OO OO OO OO s (2,−2) ✈ s s s ✍ s s 99 OO ✍ s ✈ s s s s ✈ s s ✍ s s ✍ zz✈ yys yys (1,−3) ✍ (1,−3) (1,−3) (1,−3) OO ✍ OO ❑ OO ❑ (1,−3) ❑ ❑ ✍ ❑ ❑ OO ❑ ❑ ❑ ❑ %% %% (2,−4) (2,−4) (2,−4) 99 OO OO OO ✍ (2,−4) (2,−4) s ✈ s s ✎ ✈ s s 99 OO 99 OO s ✍ ✈ s ✍ s s ✈ s ✍ ✍ ✍ s s s zz✈ yys s s ✍ ✍ s s ✍ s  s (1,−5) ✍ ✍ (1,−5) (1,−5) OO ✍ ✍ OO ❑ (1,−5) (1,−5) ❑ ✍ ✍ ✎ OO OO ❑ ✍ ✍ ❑ ✎ %% ✍ ✍ ✍ ✍ (2,−6) ✎ ✍ ✍ (2,−6) ✍ (2,−6) ✍ (2,−6) (2,−6) ✎ ✈ ✍ ✍ :: 99 99 99 s s s ✎ ✈ ✍ ✍ ✈ . ✈ s s s zz✈ . . . . ✎ ✈ ✍ s ✍ s s . ✈ s s s . . . . ✎ ✈ ✍ s ✍ s  s . . . . (1,−7) (1,−7) (1,−7) (1,−7) (1,−7) 6.2 Mutation sequence in type B We display the sequence of mutated quivers obtained from G at each step of the mutation se- quence μ . 32 (2,0) (2,0) ❏ (2,0) OO OO ❏ ✟ cc● ● ✟ $$ ✟ (2,−2) (1,−1)  ✟ oo // ✟ (2,−2) (1,−1) OO OO s t (2,−2) (1,−1) ✟ OO s t :: s t OO OO ✟ s t s t ✉ s ✟ ✉ yys zzt t s t ✉ s t ✉ yys zzt ✟✉ (1,−3) (2,−4) oo OO ❑ OO ❏ ❑ ❏ (1,−3) (2,−4) (1,−3) (2,−4) ❑ ❏ OO ❑ OO ❏ OO ▲ OO ❑ ❑ ❏ ❑ ❏ ❑ ❏ ❑ ❑ ❏ ▲ ❑ %% $$ ❑ ❏ ▲ ❑ ❑ ❏ ▲ ❑ %% $$ %% %% (2,−6) (1,−5) OO OO s t (2,−6) (1,−5) (2,−6) (1,−5) s t OO OO OO OO s t s t s t r s yys zzt s t r s s t s s t r s yys zzt yyr yys (1,−7) (2,−8) OO ❑ OO ❏ ❑ ❏ (1,−7) (2,−8) (1,−7) (2,−8) ❑ ❏ ❑ ❏ OO ❑ OO ❏ OO ▲ OO ❑ ▲ ❑ ❑ ❏ ❑ ❏ %% $$ ❑ ❏ ▲ ❑ ❑ ❏ ▲ ❑ ❑ ▲ ❑ %% $$ %% %% (2,−10) (1,−9) s (2,−10) (1,−9) (2,−10) (1,−9) s s r yys s r s r . . s r yys yyr (1,−11) . . . . . . . . (1,−11) (1,−11) . . . . . . . . (2,0) (2,0) (2,0) OO OO OO ✝ ☎ ☎ ☎ ☎ ✝ ☎ ☎ ✝ ☎ ☎ oo oo oo ✝ ☎ ☎ (1,−1) (2,−2) (1,−1) (2,−2) (2,−2) (1,−1) ✝ ☎ ☎ OO ✺ OO OO ✼ OO OO ✾ OO ✝ ☎ ☎ ✺ ✼ ✾ ✝ ☎ ☎ ✝ ✺ ☎ ✼ ☎ ✝ ✺ ☎ ✼ ☎ ✝ ✺ // // ✺ (1,−3) (2,−4) (1,−3) (2,−4) // ✼ (1,−3) (2,−4) ✺ OO OO ✼ OO OO OO ✺ ✼ ✟ ✝ dd❍ ✺ ✼ ✝ ✟ ✾ ❍ ✺ ✼ ✝ ❍ ✟ oo ✟ (2,−6) (1,−5) oo // ✟ (2,−6) (1,−5) OO ✺ OO (2,−6) (1,−5) ✟ OO :: OO OO ✟ ✝ ✺ r s ✉ r ✟ ✉ ✺ s ✝ r s ✉ ✟ ✝ ✺ r s ✉ yyr yys ✟✉  ✺ // oo (1,−7) (2,−8) ✺ (1,−7) (2,−8) (1,−7) (2,−8) OO ✺ ▲ ❑ ▲ ❑ OO OO OO OO dd❍ ✺ ▲ ❑ ▲ ❑ ❍ ▲ ❑ ▲ ❑ ❍ ✺ ▲ ❑ ▲ ❑ ✺ ▲ ❑ ▲ ❑ ❍ %% %% %% %% // (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) r r r r r r r r r r r r yyr yyr yyr . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) OO OO OO ✄ ✠ ✝ ✄ ✝ ✄ ✝ oo // ✄ ✝ (2,−2) (1,−1) (2,−2) (1,−1) ✠ // ✄ ✝ ✠ (2,−2) (1,−1) OO ✿ OO OO OO ✄ ✝ ✿ ✠ OO ✄ ✝ ✿ ✠ ✝ ✄ ✝ ✿ ✠ ✝ ✄ ✝ ✿ ✝ // // ✿ ✝ (1,−3) (2,−4) ✠ (1,−3) (2,−4) ✿ ✝ // OO OO OO OO ✿ ✝ (1,−3) (2,−4) ✄ ✿ ✠ ✄ OO OO ✿ ✠ ✝ ✄ ✝ ✠ ✝ ✄ ✝ ✠ ✝ oo ✄ ✝ (2,−6) (1,−5) ✠ // ✄ ✝ oo ✠ (2,−6) (1,−5) OO ✽ OO ✄ ✝ (2,−6) (1,−5) ✽ ✠ OO OO ✿ OO ✽ ✠ ✄ ✝ ✽ ✿ ✝ ✠ ✽ ✿ ✝ ✠ // ✽ ✿ ✝ ✠ (1,−7) (2,−8) ✽ ✿ // // OO OO ✽ ✿ (1,−7) (2,−8) (1,−7) (2,−8) ✞ ✽ ✿ OO OO OO OO ✞ ✿ ✝ ✝ ✞ ✽ ✿ ✝ ✝ ✞  ✿ ✝ ✝ oo ✝ ✝ (2,−10) (1,−9) ✞ oo oo ✝ ✝ (2,−10) (1,−9) (2,−10) (1,−9) ✝ ✝ :: ✝ ✝ ✞ t ✝ ✝ ✞ t ✝ ✝ ✞t ✝ ✝ . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . 33 (2,0) (2,0) (2,0) ❍ OO OO ✝ ❍ ✝ CC ❍ ✝ ✞ ✞ $$ // (2,−2) (1,−1) ✞ oo ✞ // (2,−2) (1,−1) OO OO ✞ (2,−2) (1,−1) :: OO :: ✈ ☎ ✞ OO OO ✞ ✝ ✈ s ✉ ☎ ✞ s ✉ ✈ ✞ ☎ ✞ s ✞ ✉ ✈ // s ✉ ✈ ☎ ✞yys ✞ ✉  ✈ (1,−3) (2,−4) ✞ ☎ ✞ // ✞ OO OO ☎ ✞ (1,−3) (2,−4) (1,−3) (2,−4) ✝ ☎ ✞ OO OO OO OO ✝ ✞ ☎ ✞ ✝ ☎ ✞ ✝ ✞ ✝ ✝ ☎ ✞ // ✝ ☎ ✝ ✞ ✞ (2,−6) (1,−5) ✝ ✝ ☎ // // OO OO ☎ ✝ ✝ (2,−6) (1,−5) (2,−6) (1,−5) ✝ ✝ ☎ OO OO OO OO ✝ ✝ ☎ ✞ ✆ ✝ ✝ ✝ ☎ ✝ ✞ ✆ ✝ ☎ // ✞ ✆ ✝ ✝ ☎ (1,−7) (2,−8) ✝ ✞ ✆ OO OO // // ✞ ✆ (1,−7) (2,−8) (1,−7) (2,−8) ✠ ✝ ✞ ✆ OO OO OO OO ✠ ✞ ✆ ✝ ☎ ✠ ✝ ✞ ✆ ✝ ☎ ✠ ✝ ✞ ✆ ✝ ☎ ✞ ✆ ✠ // ✝ ☎ (2,−10) (1,−9) ✠ // // ✝ ☎ (2,−10) (1,−9) (2,−10) (1,−9) ✝ ☎ ✠ ✝ ☎ ✝ ☎ ✝ ☎ ✠ ✝ ☎ . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) ❑ ❑ ❑ OO OO OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% %% (2,−2) (2,−2) (1,−1) (2,−2) (1,−1) (1,−1) :: OO OO 99 OO OO 99 OO OO r r ✉ ✈ s s ✉ ✎ ✈ r ✎ s r s r s r s ✉ ✈ ✎ ✎ ✍ ✉ ✈ r s r s zz✈ r yys r yys ✉ ✎ ✎ ✍ ✎ ✎ (1,−3) (2,−4) (1,−3) (2,−4) ✎ ✎ (1,−3) (2,−4) OO OO ❍ OO OO ❑ ❍ ❑ OO ✎ ✎ ❍ ❑ ✎ CC ✎ ❍ ❑ ❍ ✍ ✎ ✎ ❍ %% ✝ ✍ $$ ✎ ✎ ✝ ✍ (2,−6) (1,−5) ✎ ✎ oo // (2,−6) (1,−5) :: OO OO ✎ ✝ ✎ (2,−6) (1,−5) :: OO ✍ ✉ ✈ ✝ ✎ ✎ ✎ ✉ ✈ :: ✈ OO OO ✞ ✍ ✈ ✝ ✉ r ✉ ✈ ✎ ✎ ✎ ✉ ✈ ✝ r ✆ ✉ ✈ ✞ ✍ zz✈ ✉ ✎ ✎ r ✆ ✎ ✉ ✈ ✝ ✞ ✍ ✉ r ✉ ✈ ✆ ✎ ✎✝yyr ✎ ✉  ✈ ✞ // ✞ (1,−7) (2,−8) (1,−7) (2,−8) (1,−7) (2,−8) ✎ ✞ OO OO OO OO OO ✞ ✎ CC ☎ ☎ ✆ ✞ ✎ ✝ ☎ ☎ ✝ ✆ ✞ ☎ ☎ ✆ ✞ ☎ ☎ ✝ // // // ☎ ☎ ✎ ✝ (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) ☎ ☎ ✝ ☎ ☎ ✝ ☎ ☎ ✝ r ✎ r ☎ ☎ ✝ ☎ ☎ ✎✝yyr . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) ❏ ❏ OO ▲ OO OO ❏ ❏ ❏ ❏ ▲ ❏ ❏ ▲ ❏ ❏ %% $$ $$ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) OO OO OO OO OO OO r r r ✉ ✇ t r ✌ r ✉ ✇ r t r ✉ r t r ✇ ✌ r r r ✉ ✇ t r yyr zz✉ yyr zzt ✌ ✇ {{✇ (1,−3) (2,−4) (1,−3) (2,−4) (1,−3) (2,−4) OO OO ▲ OO ▲ OO ❏ ✌ ▲ ▲ ❏ OO ● ▲ ▲ ❏ ✌ ● ▲ ▲ ❏ ▲ ▲ ❏ %% %% $$ ## ✌ (2,−6) (1,−5) (2,−6) (1,−5) ✌ 88 OO OO OO OO r r (2,−6) (1,−5) ✌ ✉ ✇ r ✍ r ✉ ✇ r 99 OO OO ✌ r ✉ r ✍ ✇ r r t ✉ r ✇ r ✌ ✌ r ✍ yyr t zz✉ r t ✌ {{✇ r t r zzt (1,−7) (2,−8) ✌ (1,−7) (2,−8) OO OO ■ ✌ ■ (1,−7) (2,−8) OO ● ✍ ■ OO OO ❏ ● ✌ ● ✍ ❏ ■ ● ■ ✌ ✍ ❏ ● $$ ● ✍ ✌ $$ ## oo ✍ (2,−10) (1,−9) oo oo ✍ (2,−10) (1,−9) (2,−10) (1,−9) ✍ :: ✌ 99 99 t r r ✍ ✌ t r r ✍ t ✌ r r t r r ✍ t ✌ r  r . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . 6.3 Mutation sequence in type G We display the sequence of mutated quivers obtained from G at each step of the mutation se- quence μ . 34 (2,0) (2,0) (2,0) OO OO cc● ❏ ✠ ❏ ● $$ ● ✠ (2,−2) (1,−1) ✠ // oo (2,−2) (1,−1) ✠ (2,−2) (1,−1) OO OO t OO OO ✠ :: OO t ✈ t ✞ ✞ ✠ :: ✞ t ✈ ✈ ✞ t ✞ ✠ ✈ ✞ zzt ✈ t ✈ ✞ ✞ ✈ zzt ✠ ✞ ✈ ✈ ✞ ✞ (1,−3) (2,−4) ✠✈  ✈ ✞ ✞❚ ❚ (1,−3) (2,−4) oo ✞ OO ✼ OO ❚ ✞ ❚ ✞❚ ❚ (1,−3) (2,−4) ✼ OO ✼ OO ❚ ❯✞ ❚ ✞ ❚ ❯ ✞ ❚ ❚ OO ✾ OO ❯ ✼ ❚ ✼ ❚ ✞ ✞ ❚ ✞ ❯ ❚ ❯ ✼ ❚ ✼ ❚ ✾ ❚ ✞ ❚ ✞ ❯ ✞ ❯ ** ❚ ✼ ✼ ❚ ✾ ✞ ❯ ✞ ❯ ** ✼ ✼ ✾ ❯ ✞ ❯ (2,−6) (1,−5) ** ✼ ✼ ✾ (2,−6) (1,−5) OO ❏ OO ✼ ✾ ❏ t (2,−6) (1,−5) OO ❏ OO ✼ ❏ ✼ ✾ t ❏ t OO ❑ OO ❏ t ✼ ✾ ✼ ❏ t ❏ ❑ t t ❏ ✼ t ✾ ❑ $$ ❏ t t t ❑ ✾ t t  $$ t ❑ t t  %% (2,−8) (1,−7) t (2,−8) (1,−7) t OO OO t t t t (2,−8) (1,−7) OO OO t t ✞ t t t t OO t OO t ✞ t ✞ t t t t s t t ✞ ✆ zzt zzt ✞ s t t t ✞ ✞ s ✆ zzt zzt s t ✞ ✆ ✞ yys yyt (1,−9) (2,−10) ✞ ✞ ✆ ❚ (1,−9) (2,−10) OO OO ❚ ❚ ✆ ✞ ❚ ✞❚ (1,−9) (2,−10) ❚ OO OO ❚ ❯ ✞ ❚ ✞ ❚ ✆❯ ❚ ❚ OO OO ❯ ❚ ❚ ✆ ❯ ✞ ✞ ❚ ❚ ❯ ❚ ❯ ✞ ✞ ❚ ✆ ** ❚ ❯ ❚ ❯ ✞ ** ✆ . . ❯ (2,−12) (1,−11) . . ** . . (2,−12) (1,−11) . . . . . . . (2,−12) . (1,−11) . . . . (2,0) (2,0) (2,0) OO OO OO ✞ ✆ ✆ ✞ ✆ ✆ ✆ ✆ ✆ ✆ ✆ ✆ ✞ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) ✆ ✆ ✞ OO ❍ OO OO ■ OO OO ❏ OO ✆ ✆ ❍ ✈ ■ s ❏ s ✈ ✆ s ✆ s ❏ ❍ ■ ✞ ✈ s s ❏ ❍ ✆ ■ ✆ ✈ s s ❏ ✞ zz✈ ❍ ✆ yys ■ ✆ yys ❍ ■ ✞ ❏ ❍ ■ // ❍ (1,−3) (2,−4) ■ (1,−3) (2,−4) (1,−3) (2,−4) ❍ ■ ❏ OO ❍ OO ✺ OO ❑ OO ✼ ii❙ ❍ ❍ ■ ❑ ❏ OO ✼dd❍ ✺ ■ ✼ ❏ ❙ ❍ ❍ ❑ ✼ ❙ ❍ ■ ❑ ❏ ❙ ❍ ✺ ✼ ❍ ❍ ❙ ❍ ■ ❑ ❏ ✼ ❍ ❙ $$ ✺ %% ✼ %% ❙ ❍ ■ ✼ ❙ ❍ ❙ ❍ ✺ $$ ✼ ✼  ## ✺ oo (2,−6) ✼ (1,−5) ✼ (2,−6) ✺ (1,−5) ❦ // OO ✼ ❦ OO (2,−6) (1,−5) ❦ ❦ ✼ ✺ ❦ OO ✠ ✼ ❦ ❦ ❦ ✼ OO ❑ OO dd❍ ✺ ❦ ❦ ❦ ✠ ✼❦ ❑ ❍ ✉ t ❦ ❦ ✼ ❍ ✺ ✼ ❑ t ❦ ✉ ✠ ❦ ❦ ❦ ❑ ❍ ✼ t ❦ ✺ ✉ uu❦ ❑ ❦ ❍ ✠ ❦ ✉ %%  ❦ t ✉ ✠ oo ✉ ✠ (2,−8) (1,−7) // (2,−8) (1,−7) (2,−8) (✉ 1,−7) ✠ :: OO t ✈ OO OO OO OO ✠ :: ✞ s t s ✈ ✉ ✈ t ✆ ✆ ✠ ✈ ✞ s s ✈ s t ✆ s ✆ ✈ ✈ ✠ ✞ s t s ✉ ✈ ✈ ✆ ✆ yys yyt yys zz✉ ✠✈  ✈ ✞ ✆ ✆ oo ✞ ✆ ✆ (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) ❯✆ ❯ ❯ ❯ ✆❯ ❯ ❯ ❯ ❯ OO OO OO OO OO OO ✞ ✆ ❯ ✆ ❯ ❯ ❯ ❯ ❯ ❯ ❯ ✞ ❯ ✆ ❯ ✆ ❯ ❯ ❯ ❯ ❯ ✆ ❯ ✆ ❯ ✞ ❯ ❯ ❯ ❯ ❯ ❯ ❯ ✆ ** ✆ ** ✞ ** . . . . . . (2,−12) (1,−11) (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) OO OO OO ☎ ☎ ☎ ☎ ✡ ☎ ☎ ☎ ☎ ☎ ☎ (2,−2) (1,−1) (2,−2) (1,−1) ✡ ☎ ☎ (2,−2) (1,−1) OO ❑ OO OO ❑ OO ✡ ☎ ☎ ❑ r ❑ r ✡ OO ❏ ☎ r ☎ r ❑ ❑ ❏ r r :: ❑ ❑ ✡ ☎ ☎ ❏ r r ✉ ❑ ❑ ✡ ☎ yyr ☎ yyr ✉❏ ❑ ❑ ✉ ❏ ❑ ❑ ✉ ❏ ❑ ❑ ✡ ✉ ❏ (1,−3) (2,−4) (1,−3) (2,−4) ❑ ❑ ❑ ❑ OO ▲ OO ✿ OO ▲ OO ✿ ▲ ❑ ▲ ❑ (1,−3) (2,−4) ✿ ❑ ✿ ❑ ▲ ▲ ▲ ✿ ❑ ▲ ✿ ❑ OO ❏ OO ❏ ❏ ▲ ❑ ▲ ❑ ✿ ✿ ❏ %% %% %% %% ❏ ✿ ✿ ❏ ❏ ❏ ❏ ✿ ✿ $$ %% (2,−6) (1,−5) (2,−6) (1,−5) ✿ ✐ ✿ ✐ ✐ ✐ OO OO OO OO ✿ ✐ ✿ ✐ ✐ ✐ (2,−6) (1,−5) ✝ ✐ ☎ ✐ ✿ ✐ ✿ ✐ ✐ ✐ ✐ ✐ OO OO ✝ ✐✿ ☎ ✐✿ ✐ ✐ ✐ ✐ ✝ ✐ ☎ ✐ ✞ ✐  ✐  ✐ tt✐ tt✐ ✐ ✝ ☎ ✞ ✐ ✝ ☎ ✞ (2,−8) (1,−7) (2,−8) (1,−7) tt✐ ✝ ☎ ✞ OO ❍ OO OO ■ OO ✝ ☎ ✞ r (2,−8) (1,−7) ❍ ✉ ■ ✝ ✉ ☎ ■r ✞ ✉ r OO ▲ OO ✝ ☎ ■ ✞ ❍ ▲ r ✉ r ❍ ■ ▲ ✝ zz✉ ☎ yyr ✞ r ❍ r ▲ ✝ ✞ ■ r ▲ ✞ yyr ▲ // ❍ (1,−9) (2,−10) ■ ▲ (1,−9) (2,−10) ❍ ▲ OO ■ OO ■ ii❚ ❍ ■ ■ (1,−9) (2,−10) ▲ OO dd■ ❚ ❍ ▲ ■ ■ ■ ❚ ❚ ❍ ■ ■ OO ❏ OO ▲ ❚ ■ ❏ ▲ ❍ ■ ■ ❚ ❏ ▲ ❚ $$ ■ ❍ ■ ❚ ❏ ▲ $$ ■ ❚ ❍ ❏ ▲ $$ $$ %% . oo . . . (2,−12) (1,−11) . . // . . (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . 35 (2,0) (2,0) (2,0) OO OO ❂ DD ❂ ✞ ✟ ❂ ✞ ✟ ✞ ❂ (2,−2) (1,−1) ✟ ▲ 99 ✟ (2,−2) (1,−1) (2,−2) (1,−1) OO OO ❂ ▲ r ✟ OO ❏ :: OO :: OO ▲ r ✓ ❂ t ❏ t dd■ ✈ r ▲ ✟ :: ❂ ✞ t t❏ ✓ ✓ r ▲✓ ✈ ■✈ ▲ ✟ t t ❏ ✈ ❂ ✞ r ✈ ■ ✓ ✓ t t ❏✓ ▲ ✈ ✈ ■ ❂ ✟zzt t ▲ ❏ ✈ ✈ ✓ ✓■ ✓ ❏ ❂ (1,−3) (2,−4) ▲ ✈  ✈ ▲ ❏ ❂ ✓ ✓ ✓ ▲ (1,−3) (2,−4) oo ■ OO ❏ OO ❏ ❏ ▲ ✓ ❏ (1,−3) (2,−4) ✓ ■ ✓ OO ❏ OO ❯ ❏ ▲ ❯ ■ ❂ ❏ ❏ OO ❑ OO ❯ ❏ ✓ ▲ ✓ ❏ ✓❯ ■ ❏ ❂ ❏ ▲ ❑ ❯ ❏ ❏ ❯ ■ ❑ ✓ ❂ $$ ✓ %% ❏ ✓ ❏ ❯ ❯ ■ $$ $$ ❯ ❂ ✓ ❑ ✓ ❯ ■ %% ❯ (2,−6) (1,−5) ** ✓ ✓ ❥ (2,−6) ❥ (1,−5) OO OO ❥ ❥ ✓ ✓ ❥ ✓ (2,−6) (1,−5) OO ❥ OO ✡ ❥ ❥ ✐ ✓ ✓ ❥ OO ✓ ✐ OO ❥ ✞ ❥ ✐ ✡ ❥ ❥ ✐ ❥ ❥ ✆ ✓ ✐ ✡ ✓ ❥ ✞ ✓ ❥ ✐ ❥ ❥ ✐ ❥ ✞ ❥ ✆ ✡ ✓ ✓ ✐ uu❥ ✓tt❥ ✐ ✞ ✐ ✡ ✓tt✐ ✓ ✓ ✞ ✆ (2,−8) (1,−7) (2,−8) (1,−7) ✓ ✆ ✓ ✞ ✓ (2,−8) (1,−7) ✡ OO ❏ OO ❏ :: OO ✞ ✆ ✓ ❏ ✓ ❏ :: ✓ t OO ❏ 99 OO ❏ ✞ t❏ ✆ ✓ ✉ ✓ ❏ s ✡ ✓ t ✉❏ ✞ ❏ ✆ s ❏ ❏ t ❏ ✡ ✓ ✉ ✓ s ❏ ✞ ✓ t ✆ ✉ ❏ ❏ s ❏ ✡ ✓ ✉ ❏ ✆ ✓ s ❏ ❏ ❏ ❏ (1,−9) (2,−10) ❏ ❏ ❏ (1,−9) (2,−10) ❏ ❏ (1,−9) (2,−10) OO ❏ OO ❏ ❏ ❏ ❏ OO ❏ OO OO ❑ OO ❏ ❏ ❏ ❏ ❑ ❏ ❏ ❏ ❏ ❏ ❏ ❑ ❏ ❏ ❏ ❏ ❑ ❏ ❏ ❏ $$ $$ ❑ ❏ $$ %% %% %% . . . . (2,−12) (1,−11) . . . . . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . (2,0) (2,0) (2,0) OO ❂ OO ❄ OO ❅ ❂ ❄ ❅ ❂ ❄ ❅ ❂ ❄ ❂ ❄ ❅ // // ❅// ❂ ❄ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) ❅ (1,−1) OO DD OO OO GG OO OO ❅OO ❂ ❄ ✈ ❅ ✠❂ ✞ ❄ ✓ ✎✓ ✎ ✈ ❅ ✈ ✠ ❂ ✞ ❄ ✓ ✎✓ ✎ ✈ ❅ ❂ ✞ zz✈ ✠ ❄ ✎ ❅ ✓ ✎✓ ❂ ✞ ❄ ✠ ❅ // // ❂ ✎ ✎ ✓ ✞ ✓ ❄ // ✠ (1,−3) (2,−4) (1,−3) (2,−4) ❅ ❂ ✞ ❄ ✎ (1,−3) (2,−4) ✓ ✎✓ ❅ ✠ OO OO OO OO ii❙ ✞ ❄ ✎ ❅ ❙ ✎ OO ✠ ✓ ✓ ❙ ❂ ✎✎ ✞ ❄ ✎ ❅ ✠ ❙ ✎ ✓ ❙ ❂ ✓ ✎✎ ✞ ❄ ✎ ❅ ✠ ❙ ✎ ❂ ✞ ✎ ❄ ✓ ❙ ✓ ✎✎ ✎ ✠ ❙ ❂ ✎ ❙ ✎ ✓ ✓ oo ❙ ✎✎ ✎ ✠  ✎ oo (2,−6) (1,−5) ✓ ✎✎ ✓ ✎ (2,−6) (1,−5) ✎ // OO ❂ OO (2,−6) ✓ (1,−5) ✓ ✎✎ ✎ OO ✎❂ ✐ ✎ OO ✓ OO ✎✎ CC ✓ ✎ ❂ ✐ ✎ ✆ ✓ ✐ ✞ ✓ ❂ ✐ ✎✎ ✎ ✎ ✆ ✐ ✎ ❂ ✐ ✞ ✓ ✎✎ ✓ ✎ ✆ ✐ ❂ ✞  ✎ ✓tt✐ ✓ ✆ ✎✎ ✎ ❂ ✞ (2,−8) (1,−7) ✆ ✓ ✎✎ ✓ ✎ ❂ (2,−8) (1,−7) (2,−8) (1,−7) ✞ :: ❂OO ✆ ✓ ✓ ✎✎ ✎ dd■ ✈ OO ❏ 99 OO ✞ OO ❏ 99 OO :: ✆ ❂ ❏ s ❏ ■✈ ✓ ✎✎ s ✓ s ✎ ✈ ✞ ✈ ❂ ✆ s ❏ s s ❏ ✈ ■ ✓ s ✎✎ s ✓ s ✎ ✈ ✈ ✆ ❏ ✞ ❏ ■ ❂ s s s ✈ ❏ ❏ ✈ ■ ❂ ✆ ✓ s ✎✎✞yys ✓ s ✎ ✈  ✈ ❏ ❏ ■ ❏ ❏ ❂ oo ❏ ❏ ❂ (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) ■ ❏ ❏ ❯ ■ ❂ ❏ ❏ OO ❑ OO OO ❑ OO OO ❑ OO ❯ ❏ ❏ ❯ ■ ❂ ❑ ❑ ❑ ❯ ❑ ❏ ❑ ❏ ❑ ❯ ❯ ❂ ❏ ❏ ■ ❑ ❑ ❑ ❯ ❯ ❂ ❑ ❏ ❑ ❏ ❑ ❯ ■ %% %% %% %% %% ** . . . . . . (2,−12) (1,−11) (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) OO ❆ OO ❆ OO ❂ ❆ ❆ ❆ ❆ ❂ ❆ ❆ ❂ ❆ ❆ ❆// ❆// ❂// (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) ❆ ❆ ❆ ❆ OO OO OO OO OO ❂OO ❆ ❆ ✌ ✌ ✏❂ ❆ ❆ ✌ ❆ ✌ ❆ ✏ ❆ ❆ ✌ ✌ ✏ ❆ ❆ // ✌ // ✌ ✏ ❂ (1,−3) (2,−4) ❆ (1,−3) (2,−4) ❆ oo ✌ ✌ ❂ ❆ ❆ ✏ (1,−3) (2,−4) OO OO OO OO ✌ ❆ ✌ ❆ OO ✏ ✍ ✌ ❂ ❆ ❆ ✌ ✌ ✍ ❆ ✌ ❆ ❂ ✌ ✌ ✏ ❂ ✍ ✌ ✌ ✌ oo ✏ ✍ ✌ oo (2,−6) ✌ (1,−5) (2,−6) ✌ (1,−5) ✍ ✌ ✌ ✌ oo OO ❃ OO OO ❄ OO ✍ ✌ ✏ (2,−6) (1,−5) ✌❃ ✌❄ ✍ ✏ ✌ ❄ ✌ ❃ ✌ OO ❄ OO ✍ ❃ ✌ ❄ ✏ ❄ ✌ ✌ ❄ ❄ ✍ ❃ ✌ ✏ // ❄// ❃ ❄ ✍ ✏ (2,−8) (1,−7) ✌ (2,−8) (1,−7) ❄ ❄ ✌ ❄ ❄// OO ❃CC OO OO OO ✍ ✉ ❄ (2,−8) ❄ (1,−7) ✟❃ ✝ ❄ OO ❄OO ✍ ✉ ✟ ❃ ✌ ✝ ✉ ❄ ❄ ✍ ✟ ❃ ✝ zz✉ ✌ ❄ ❄ ❃ ✝ ✍ ✟ // ❄ ❄ ❃ ✝ ✟ (1,−9) (2,−10) ❄ ❄ // ❃ ✝ (1,−9) (2,−10) ❄ ❄ ✟ // OO ■ OO ❃ ✝ ii❚ ■ ❄ (1,−9) (2,−10) ❄ OO ✟ ❚ ❃ ■ ✝ ❚ ❄ ❄ ■ OO OO ✟ ❚ ❚ ❃ ✝ ❄ ❄ ✟ ❃ $$ ✝ ❄ ❄ ❚ ❃  ❄ . . oo . . (2,−12) (1,−11) . . // . . oo . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . 36 (2,0) (2,0) (2,0) OO ❄ OO ❄ ^^❃ ❄ ❃ ❍ ❄ ❃ ❍ $$ // ❃ (2,−2) (1,−1) ❄ // oo ❄ (2,−2) (1,−1) (2,−2) (1,−1) OO OO ❃ OO ❏ OO :: OO ✎ ❃ ❄ ❏ dd■ ✈ ✎ ❏ ✎ ✎ ❄ ■✈ ❏ ❃ ✈ ❄ ✎ ■ ✎ ✎ ❏ ✈ ❃ ■ oo ❄ ✎ ❏ ✈ ✎ ✎ ■ ❏ ❃ (1,−3) (2,−4) ❄  ✈ oo ✎ ✎ ■ ✎ ❏ ❃ (1,−3) (2,−4) oo ■ OO ✵ OO ❏ ✎ ✎ ❄ ✎ ❃ ❏ (1,−3) (2,−4) ■ OO ✴ OO ❯ ✵ ❃ ❯ ■ ❄ ✎ ❏ ✎ ✎ OO ✵ OO ❯ ✴ ❏ ❯ ■ ✵ ❄ ❃ ✎ ✎ ❯ ✎ ❏ ❯ ■ ❃ ✵ ✵  ✴ ❏ ❯ ✎ ✎ ❯ ■ $$ ✵ ❯ ✵ oo ✴ ❯ ■ ✎ ❯ (2,−6) ✎ (1,−5) ✵ ** ✴ oo ✵ ✎ (2,−6) ✎ (1,−5) ✎ ✵ oo OO ❂ OO ✴ ✎ (2,−6) ✎ (1,−5) ✎❂ OO ❄ OO ✵ ✴ ✎ ❄ OO ❅✎ OO ✎ ❂ ✵ ✴ ❂ ❄ ✎ ❅ ✎ ✵ ✵ ✴ ❄ ❅ ❂ ✎ // ❄ ✵ ❅ ❂ ✴ ✵ ✎ (2,−8) (1,−7) ❄ // ❅ ❂ ✵ ✵ ✴(2,−8) (1,−7) ❄ ❅// OO ❂OO ✵ ✴ ❄ (2,−8) ❅ (1,−7) OO OO ❄ ✵ ✵ ✴ OO ❅OO ❄ ❅ ✵ ✴ ✵ ❄ ❅ ❂ ✵ ❄ ❅ oo oo (1,−9) (2,−10) (1,−9) (2,−10) ❄ ❅ oo ❂ ❄ ❅ (1,−9) (2,−10) OO OO OO ❄ ❅ OO OO ❄ ❅ ❂ ❄ ❅ ❂  ❅ . oo . . . (2,−12) (1,−11) . . oo . . oo . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . (2,0) (2,0) (2,0) ❑ ❑ ❑ OO ❑ OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% %% (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) OO DD OO OO OO OO OO ✈ ✈ s s ✠ ✞ ✆ ✎ s s ✈ ✈ ✈ ✈ ✠ s ✞ s ✆ ✈ s s ✈ ✞ ✆ zz✈ ✠ yys yys ✞ ✆ zz✈ ✠ ✞ ✆ // ✠ (1,−3) (2,−4) (1,−3) (2,−4) ii❙ jj❯ ✞ ✆❯ (1,−3) (2,−4) ✎ ❙ ✠ OO ✴ZZ✺ ❍ OO ✻ ❙ OO ✼ OO ✻ ❯ ii❙ ✞ ❙ ✆ ❙✎ ❍ ❙ ❯ OO ✴ dd❍ ✠ ✺ ✻ ✼ ✻ ❯ ❙ ✴ ❍ ✞ ❙ ✆ ❍ ❯ ✠ ✎ ❙ ❍ ❙ ✼ ❯ ✴ ❙ ✺ ✻ ❙ ✻ ❍ ✴ ❍ ✞ ✆ ❯ ❙ ❙ ❯ ❍ ✠ ❙ ❙ ✼ ✴ ✎ ✺ $$ ✞ ✻ ✆ ✻ ❯ ❙ ❙ ❍ ✴ ❙ ✼ ✠ ❙ ✴ ❍ ✎ ❙ ✺ ✻ ✻ ❙ ✴ ✼ ✴ ✺ ✻ (2,−6) ✻ (1,−5) ✴ ✼ ✺ (2,−6) (1,−5) ✻ OO ❍ ✻ OO ✴ ✎ ✼ (2,−6) (1,−5) ✴ ✺ ❍ ✻ OO ✻ ✴ ✼ ❍ OO ❅ OO ✴ ✺ __❄ ✻ ✼ ❍ ✻ ✎ ❅ ❄ ❍ ✴ ✻  ✻ ✴ ❍ ✎ ❅ ❄ ✺ ✻ $$ ✻ ✴ ❅ ✎  ✻ ✻ ❅ oo ✴ ❄ ✻ (2,−8) (1,−✻ 7) ❅// //❄ ✴(2,−8) (1,−7) (2,−8) (1,−7) ❅ ✻ :: ✻ ❄ OO ✴ ❅ ✴ dd■ ✈ OO OO OO ❏ OO ✻ ✻ ❏ ■✈ ✴ ❅ ✴ ❄ ✻ ✈ ✻ ❏ ■ ✴ ✴ ✈ ❏ ❄ ✻ ■ ✻ ❅ ✈ ❄ ■ ❅  ✻  ✈ ✻ ❏ ■ ❏ ❄ ❅ ✻ ■ ✻ oo oo oo ❏ ❄ (1,−9) (2,−10) ❅ (1,−9) (2,−10) (1,−9) (2,−10) ■ ❏ ✻ ✻ ❅ ❄ ❯ ■ OO OO OO OO ✻ OO OO ❯ ✻ ❏ ❄ ❯ ■ ❅ ❯ ❏ ✻ ❯ ✻ ❅ ❄ ❯ ❏ ■ ❅ ❄✻ ❯ ✻ ❏ ❯ ■ %%  ** . oo . . oo . . oo . (2,−12) (1,−11) (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) ▲ ▲ OO OO OO ▲ ▲ ❏ ▲ ▲ ▲ ▲ ▲ ▲ ❏ ▲ ▲ ❏ %% %% $$ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) OO OO OO OO OO OO r r ☎ ☎ ✟ r r t r ☎ r ☎ t r r ☎ ☎ yyr yyr zzt ✟ ☎ ☎ ☎ ☎ (1,−3) (2,−4) (1,−3) (2,−4) (1,−3) (2,−4) jj❯ jj❯ ✟ ☎❯ ☎❯ ❙ ❯ ❯ OO ✿ OO ✼ ❯ OO ✿ OO ✽ ❯ OO ✹ OO ✟ ❙ ☎ ☎ ❯ ❯ ❙ ✿ ❯ ✿ ❯ ✟ ❙ ✼ ✽ ✹ ☎ ❯ ☎ ❯ ❙ ❯ ❯ ❙ ✿ ❯ ✿ ✽ ❯ ✟ ☎✼ ☎ ✹ ❙ ❯ ❯ ✿ ❯ ✿ ❯ ❙ ❯ ✽ ❯ ✹ ✟ ❙ ☎ ✼ ☎ ✿ ✿ )) ✽ ✹ ✟ ✿ ✿ (2,−6) ✼ (1,−5) (2,−6) ✽ (1,−5) ✹ ✿ ✿ ▲ ✼ ▲ ✽ ✹ (2,−6) (1,−5) OO OO OO OO ✿ ✿ ▲ ▲ ✽ ✼ ✹ OO ● ✿ ▲ ✿ ▲ ✽ ● ▲✼ ▲ ✹ ✿ ✿ ● ▲ ▲ ✼ ✹ ● %%  %% ✽ ✹ ✼ ● ## (2,−8) (1,✼−7) (2,−8) (1,−7) OO CC OO OO OO ✉ ✉ r (2,−8) (1,−7) ✟ r ✝ ✉ ✉ ✽ OO OO ✉ ✼ r ✝ ✉ ✟ ✽ ✉ r t ✉ ✞ ✟ ✼ ✝ t zz✉ yyr ✽ ✝ t ✞ zz✉ ✟ ✼ ✽ t ✝ zzt ✟ ✼ (1,−9) (2,−10) // ✽ ii❚✝ ✞ (1,−9) (2,−10) ✟ ✼ ❚ OO ■ OO ❚ ✽ ✝ ✞ ii❚ ✼ ■ ❚ (1,−9) (2,−10) OO dd■ ✟ ❚ ✽ jj❯ ■ ✝ ❚ ✞ ■ ❚ ❚ ❯ ✼ ■ OO OO ❯ ✟ ❚ ❚ ✽ ■ ❚ ✝ ❚ ✞ ❯ ■ ❯ ❚ ✼ ❚ ❯ ■ ✟ $$ ✝ ✽ ✞ ❚ ❚ ❯ ■ ❚ ❚ ❯ ✟ ✼ ■ ❚  ✞ ❯ ❚ ❯ ✟  ✞ . . . (2,−12) (1,−11) . . . . . . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . 37 6.4 Examples of A-modules for g of type B (i) We describe some A-modules K for g of type B . The quiver Γ is: k,m (2,−1) OO ❑ %% (2,−3) (1,−3) OO OO s s s s s s s s yys yys (1,−5) (2,−5) OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% (2,−7) (1,−7) OO OO s s s s s s s s yys yys (1,−9) (2,−9) . . . . . . . . . Following the convention of Example 4.6, unless otherwise specified, in the following figures the vertices carry one-dimensional spaces, and the arrows carry linear maps with matrix (±1). (1) (1) The modules K and K are: 1,−5 1,−7 (2,−3) (1,−3) (1,−5) (2,−5) OO ❑ OO ❑ ❑ ❑ s s s s ❑ ❑ s s ❑ ❑ ❑ ❑ s s yys yys %% %% (1,−5) (2,−5) (2,−7) (1,−7) (2) (2) The modules K and K are: 1,−5 1,−7 (2,−1) (2,−3) ❑ s ❑ s ❑ s yys %% (1,−5) (1,−3) s ❑ s ❑ s ❑ yys %% (2,−5) (2,−7) Applying Theorem 4.8, we recover the following well known formulas for the q-characters of the fundamental U (bg)-modules: χ (L(Y )) = Y (1 + v (1 + v (1 + v (1 + v )))), q 1,−7 1,−7 1,−5 2,−3 2,−5 1,−3 χ (L(Y )) = Y (1 + v (1 + v (1 + v ))). q 2,−6 2,−6 2,−5 1,−3 2,−1 (1) (2) The modules K and K are: 2,−5 2,−7 (2,−3) (1,−3) (2,−3) OO OO OO s s s s s s s s s s s s yys yys yys (1,−5) (2,−5) (1,−5) (2,−5) OO OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% (2,−7) (1,−7) (2,−7) (1,−7) OO OO s s s s s s s s s s s s yys yys yys (1,−9) (2,−9) (2,−9) They correspond under Theorem 4.8 to the Kirillov-Reshetikhin modules (1) (2) W = L(Y Y ) and W = L(Y Y ). 1,−11 1,−7 2,−10 2,−8 2,−11 2,−10 38 (1) (2) The modules K and K are: 3,−5 3,−7 (2,−3) (1,−3) (2,−3) OO OO OO r r s r r s r r r r s xxr xxr yys (1,−5) (2,−5) (1,−5) (2,−5) OO OO OO ▲ OO ▲ ▲α ▲ ▲ ′ ▲ ▲ γ ▲ ▲ ▲ %% %% (2,−7) (1,−7) (2,−7) (1,−7) OO OO OO r r s s r r s s r r s s r r s s xxr xxr yys yys (1,−9) (2,−9) (1,−9) (2,−9) OO OO ▲ OO %% (2,−11) (1,−11) (2,−11) OO r r r r r r r r xxr xxr (1,−13) (2,−13) (2) In K , the vertex (2,−7) carries a two-dimensional vector space. The linear maps carried by 3,−7 the adjacent arrows have the following matrices: α = γ = , β = γ = ( 0 1). They correspond under Theorem 4.8 to the Kirillov-Reshetikhin modules: (1) (2) W = L(Y Y Y ) and W = L(Y Y Y ). 1,−15 1,−11 1,−7 2,−12 2,−10 2,−8 3,−15 3,−12 6.5 Examples of A-modules for g of type B Let g be of type B , with the short root being α . The quiver Γ is: 3 3 (3,−1) OO xxr (2,−3) (3,−3) (1,−3) OO ▲ OO ▲ OO r ▲ ▲ r ▲ ▲ r r r ▲ ▲ r r ▲ ▲ r xxr && && xxr (1,−5) (3,−5) (2,−5) OO ▲ OO OO ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ && xxr xxr && (2,−7) (3,−7) (1,−7) OO ▲ OO ▲ OO r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r xxr && && xxr (1,−9) (3,−9) (2,−9) OO ▲ OO OO ▲ ▲ r r ▲ ▲ r r ▲ ▲ ▲ r r ▲ r r ▲ && xxr xxr && (2,−11) (3,−11) (1,−11) OO ▲ OO ▲ OO r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r xxr && && xxr (1,−13) (3,−13) (2,−13) OO OO OO . . . . . . . . . . . . . . . (1) The module K is: 1,−9 (1,−3) yys (3,−5) (2,−5) OO s s s s s s s s yys yys (2,−7) (3,−7) yys (1,−9) 39 (2) (3) The modules K and K are: 1,−11 1,−11 (3,−5) (2,−5) (3,−3) OO ❑ ▲ r s ❑ ▲ r s ❑ ▲ ❑ ▲ r s r s ❑ ▲ xxr yys %% %% (2,−7) (3,−7) (1,−7) (2,−5) ▲ ▲ ❑ s ▲ ▲ s s ❑ s ▲ ▲ s s ❑ ▲ ▲ ❑ s s s s ▲ ▲ s s ❑ yys && %% yys yys %% (1,−9) (3,−9) (2,−9) (3,−7) (1,−7) ▲ OO ▲ ▲ r s ▲ s ▲ r s ▲ s ▲ r ▲ s ▲ r s ▲ s %% xxr yys %% yys (2,−11) (3,−11) (2,−9) yys (3,−11) The corresponding fundamental U (bg)-modules are L(Y ), L(Y ), and L(Y ), of respec- q 1,−11 2,−13 3,−12 tive dimensions 7, 22, and 8. 6.6 Examples of A-modules for g of type C Let g is of type C , with the long root being α . The quiver Γ is: 3 3 (1,−1) OO ▲ %% (2,−2) (3,−2) OO OO s ③ s ❖ s ❖③ yys ❖ ③ ❖ (1,−3) ❖ ③ ❖ OO ▲ ③ ❖ ③ ❖ ▲ ③ %% ||③ '' (2,−4) (3,−4) OO ❉ OO s ❉ ❉ ♦ s ♦ yys ❉ ♦ ❉ ♦ ❉ ♦ (1,−5) ♦ ♦❉ OO ▲ ▲ ❉ ▲ ♦ ❉ ▲ ♦ ▲ ❉ %% ww♦ "" (2,−6) (3,−6) OO OO ❖ ③ ❖ ③ s ③ yys ③❖ ③ ❖ (1,−7) ③ ❖ OO ▲ ▲ ③ ③ ❖ ▲ ❖ ▲ ③ ❖ %% ||③ '' (2,−8) (3,−8) OO ❉ OO ❉ ♦ s ♦ s ♦ s ❉ yy s ❉ ♦ ❉ ♦ (1,−9) ♦ ❉ ♦ OO ▲ ♦❉ ▲ ❉ ♦ ❉ ▲ ♦ %% ww♦ "" (2,−10) (3,−10) OO OO . . . . . . . . . . . . 40 (1) (2) The modules K and K are: 1,−7 1,−8 (1,−1) (2,−2) ❑ ✽ ❑ s ❑ s s ✽ ❑ s %% yys ✽ (2,−2) (1,−3) ❑ ❑ ❑ ❑ ❑ ❑ ✽ ❑ ❑ ❑ ❑ %% %% (3,−4) (2,−4) (3,−4) OO ❑ s ❑ s s ❑s s ❑ s s ❑ yys yys %% (2,−6) (2,−6) (3,−6) s s s s s s s s yys yys ✝ (1,−7) (1,−7) %% ✝ (2,−8) (3) The module K is: 1,−8 (2,−4) (3,−4) OO ❈ ♠ ♣ ♠ ♣ ♠ ❈ ♣ ♠ ♣ ♠ ❈ vv♠ ❈ ♣ (1,−5) κ ◗ ♣ OO ❈ ◗ ♣ β ◗ ♣ ❈ ♣ ❈ ◗ ♣ ◗ ♣ (( ww♣ !! (2,−6) (3,−6) OO ◆ ♠ ④ ◆ ④ ♠ ④ vv♠ ε ④◆ ④ ◆ γ ◆ (1,−7) ④ ◆ ◗ ◆ ◗ ◆ ◗ ④ ◆ ◗ ④ ◗ ◆ ◗ ◆ (( }}④ '' (2,−8) (3,−8) Here, the vector space sitting at vertex (2,−6) has dimension 2. The maps incident to this space are given by the following matrices: 1 0 1 α = , β = , γ = , ε = ( 0 1), κ = ( 0 1), ι = ( 1 0). 0 1 0 The corresponding fundamental U (bg)-modules are L(Y ), L(Y ), and L(Y ), of respec- q 1,−8 2,−10 3,−10 tive dimensions 6, 14, and 14. 41 6.7 Examples of A-modules for g of type F Let g be of type F . We label the simple roots α , α , α , α , so that the short simple roots are α 4 1 2 3 4 1 and α . The quiver Γ is: (1,−1) OO ▲ %% (2,−2) (3,−2) (4,−2) OO ❖ OO ❘ OO ❖ ❘ ❖ ③ ❘ ✆ ❖ ❘ ③ ❘ s ❖ ✆ s ❖③ yys ✆ ③❖ ❘ ❖ ❘ ③ ✆ ❖ ❘ ③ ❖ ✆❘ (1,−3) ❖ ③ ❘ ❖ ✆ OO ▲ ③ ❘ ❖ ✆ ▲ ③ ❘ ▲ ✆ ❘ ③ ❖ ▲ ❘ ❖ ✆ ③ ❘ ▲ ❖ %% ||③ '' ✆ )) (2,−4) (3,−4) (4,−4) OO ❉ OO ✾ ❧ OO ♦ ❧ s ❉ ♦ ✾ ❧ ♦ ❧ s ❉ ❧ ♦ ✾ s ❉ ♦ ✾ ❧ yys ♦ ❉ ❧ ♦ ✾ ❧ ♦ ❧ ❉ ♦ ✾ (1,−5) ♦ ❉ ❧ ♦ ✾ ♦❉ OO ▲ ❧ ♦ ❧ ✾ ❉ ❧ ♦ ✾ ▲ ❧ ♦ ❉ ▲ ❧ ♦ ✾ ▲ ❉ ❧ %% ww♦ "" vv❧ (2,−6) (3,−6) (4,−6) OO OO ❘ OO s ③ ❘ ❖ ✆ s ❘ ❖ ③ s ❘ ❖ ✆ ③ ❘ s ❖ ❘ ✆ yys ❖ ③ ❘ ❖ ❘ ③ ❖ ❘ ❖ ❘ ③ ✆ (1,−7) ❖ ③ ❘ ❖ ✆ ③ ❖ ❘ OO ▲ ✆ ❖ ❘ ▲ ③ ❖ ❘ ▲ ❖ ❘ ▲ ❖ ❘ ▲ ③ ❘ ❖ ❘ %% ||③ '' ✆ )) (2,−8) (3,−8) (4,−8) OO ❉ OO ✾ ❧ OO s ❧ ❉ ♦ ❧ s ❧ ❉ ♦ s ✾ ❧ s ❉ ❧ ✾ ❧ yys ❉ ♦ ❧ ♦ ❧ ❉ ♦ ❧ ♦ ❧ ❉ ✾ (1,−9) ♦ ❧ ❉ ✾ ♦ ❧ ♦❉ ❧ OO ▲ ✾ ♦ ❧ ▲ ❉ ♦ ❧ ▲ ❧ ✾ ♦ ❉ ▲ ♦ ❧ ✾ ▲ ♦ ❉ %% ww♦ "" vv❧ (2,−10) (3,−10) (4,−10) OO OO OO . . . . . . . . . . . . . . . . . . (1) The module K is: 1,−17 (1,−1) && (2,−2) && (3,−4) r ▲ r ▲ r ▲ r ▲ xxr && (2,−6) (4,−6) ✿ ☎ ✿ ☎ ✿ ☎ xxr ✿ ☎ ✿ ☎ (1,−7) ✿ ☎ ✿ ☎ ✿ ☎ ✿ ☎ &&  ☎ (2,−8) (3,−8) OO ▲ ▲ r ▲r r ▲ xxr && (2,−10) (3,−10) ☎ ✿ r ☎ ✿ ☎ ✿ xxr ☎ ✿ ☎ ✿ (1,−11) ☎ ✿ ▲ ☎ ✿ ▲ ☎ ✿ ☎ ✿ && ☎ (2,−12) (4,−12) ▲ r ▲ r ▲ r && xxr (3,−14) xxr (2,−16) xxr (1,−17) 42 (4) The module K is: 1,−16 (4,−2) tt❤ (2,−4) (3,−4) OO P ❧ P ❧ P ③ ❧ P ❧ ③ P③ vv❧ ③ ′ ③ P (1,−5) P ③ P ❘ ′ P OO ❘ ′ ③ ❘ α β ❘ P ③ P ❘ ③ ||③ '' (( (2,−6) (3,−6) ❧ OO ❉ ✿ ❧ ♦ ❧ ♦ ✿ ❧ ♦ ❧ ❉ ♦ ✿ ❉ ♦ vv❧ ′ ♦ ✿ ❉ ′ ε κ ♦ ✿ ❉ ♦ ✿ (1,−7) γ ♦ ♦ ✿ ♦ ❉ ♦ ✿ ❘ ♦ ✿ ♦ ❉ ♦ ✿ ❘ ❉ ❘ ♦ (( ww♦ "" (2,−8) (3,−8) (4,−8) ❘ ❱ ❤ ❱ ❤ ❘ ❧ ❱ ❤ ❘ ❧ ❱ ❤ ❘ ❱ ❤ ❧ ❱ ❤ ❘❧ ❤❱ ❧ ❘ ❤ ❱ ❘ ❤ ❱ ❧ ❘ ❤ ❱ ❤ ❱ vv❧ (( ❤ ❱ tt❤ ** (2,−10) (3,−10) (4,−10) OO P ❧ ③ P ☎ ❧ P ③ ❧ P ☎ ❧ P vv❧ ③P P ☎ ③ P ☎ (1,−11) P P ☎ ❘ ③ P OO ☎ ❘ β P α ③ ❘ P ❘ ☎ ❘ ③ ❘ P ☎ ❘ P (( ||③ '' ☎ (2,−12) (3,−12) OO ❉ ❧ ♦ ❧ ♦ ❧ ❉ ♦ ❧ ♦ ❧ ♦ vv❧ ε ❉ κ ♦ γ ❉ ♦ (1,−13) ♦ ❘ ♦❉ ❘ ♦ ❘ ❉ ♦ ❉ ❘ ♦ ❘ ♦ (( ww♦ "" (2,−14) (3,−14) ** (4,−16) Here, the vector spaces sitting at vertex (2,−6) and (2,−12) have dimension 2. The maps incident to these spaces are given by the following matrices: 1 0 1 α = , κ = ( 1 0), β = , ε = ( 0 1), γ = , ι = ( 0 1), 0 1 0 1 0 1 ′ ′ ′ ′ ′ ′ α = , κ = ( 1 0), β = , ε = ( 0 1), γ = , ι = ( 0 1). 0 1 0 The corresponding fundamental U (g)-modules are L(Y ) and L(Y ), of respective dimen- q 1,−18 4,−18 sions 26, and 53. Acknowledgements We are grateful to P.-G. Plamondon for very helpful comments and discussions. References [C] V. C H A R I, Minimal affinizations of representations of quantum groups: the rank 2 case, Publ. Res. Inst. Math. Sci. 31 (1995), 873–911. [CH] V. C H A R I, D. H E R NA N D E Z, Beyond Kirillov-Reshetikhin modules, In Quantum affine algebras, extended affine Lie algebras, and their applications, Contemp. Math., 506, AMS Providence, 2010, 49–81. [CP1] V. C H A R I, A. P R E S S L E Y, A guide to quantum groups. Cambridge University Press 1994. 43 [CP2] V. C H A R I, A. P R E S S L E Y, Minimal affinizations of representations of quantum groups: the simply laced case, J. Algebra 184 (1996), 1–30. [DWZ1] H. D E R K S E N, J. W E Y M A N, A. Z E L E V I N S K Y, Quivers with potential and their representations I: Muta- tions, Selecta Math., 14 (2008), 59–119. [DWZ2] H. D E R K S E N, J. W E Y M A N, A. Z E L E V I N S K Y, Quivers with potential and their representations II: Appli- cations to cluster algebras, J. Amer. Math. Soc. 23 (2010), 749–790. [FM] E. F R E N K E L, E. MU K H I N, Combinatorics of q-characters of finite-dimensional representations of quan- tum affine algebras, Comm. Math. Phys. 216 (2001), 23–57. [FR] E. F R E N K E L, N. R E S H E T I K H I N, The q-characters of representations of quantum affine algebras, Recent developments in quantum affine algebras and related topics, Contemp. Math. 248 (1999), 163–205. [FZ1] S. F O M I N, A. Z E L E V I N S K Y, Cluster algebras I: Foundations, J. Amer. Math. Soc. 15 (2002), 497–529. [FZ2] S. F O M I N, A. Z E L E V I N S K Y, Cluster algebras: notes for the CDM-03 conference, in Current developments in mathematics, 2003, 1–34, Int. Press, Somerville, MA, 2003. [FZ3] S. F O M I N, A. Z E L E V I N S K Y, Cluster algebras IV: coefficients, Compos. Math. 143 (2007), 112–164. [GG] J. G R A B OWS K I, S. G R AT Z Cluster algebras of infinite rank, (2012) arxiv 1212.3528. [GLS1] C. G E I S S, B. L E C L E R C, J. S C H R O E R, Auslander algebras and initial seeds for cluster algebras, J. London Math. Soc. 75 (2007), 718–740. [GLS2] C. G E I S S, B. L E C L E R C, J. S C H R O E R, Kac-Moody groups and cluster algebras, Advances Math. 228 (2011), 329–443. [GSV] M. G E K H T M A N, M. S H A P I RO, A. VA I N S H T E I N, Cluster algebras and Poisson geometry, AMS Math. Survey and Monographs 167, AMS 2010. [H] D. H E R NA N D E Z, The Kirillov-Reshetikhin conjecture and solutions of T -systems, J. Reine Angew. Math. 596 (2006), 63–87. [HJ] D. H E R NA N D E Z, M. J I M B O, Asymptotic representations and Drinfeld rational fractions, Compositio Math. 148 (2012), 1593–1623. [HL1] D. H E R NA N D E Z, B. L E C L E R C, Cluster algebras and quantum affine algebras, Duke Math. J. 154 (2010), 265–341. [HL2] D. H E R NA N D E Z, B. L E C L E R C, Monoidal categorifications of cluster algebras of type A and D, in Sym- metries, integrable systems and representations, (K. Iohara, S. Morier-Genoud, B. Re ´my, eds.), Springer proceedings in mathematics and statistics 40, 2013, 175–193. [HoJo] T. H O L M, P. J O R G E N S E N, On a cluster category of infinite Dynkin type, and the relation to triangulations of the infinitygon, Math. Z. 270 (2012), 277–295. [IIKKN1] R. I N O U E, O. I YA M A, B. K E L L E R, A. K U N I BA, T. NA K A N I S H I, Periodicities of T and Y -systems, dilog- arithm identities, and cluster algebras I, Type B , Publ. RIMS, (to appear). [IIKKN2] R. I N O U E, O. I YA M A, B. K E L L E R, A. K U N I BA, T. NA K A N I S H I, Periodicities of T and Y -systems, dilog- arithm identities, and cluster algebras II, Type C , F and G , Publ. RIMS, (to appear). r 4 2 [IIKNS] R. I N O U E, O. I YA M A, A. K U N I BA, T. NA K A N I S H I, J. S U Z U K I, Periodicities of T and Y -systems, Nagoya Math. J. 197 (2010), 59–174. [Ka] V. G. K AC, Infinite dimensional Lie algebras, Cambridge University Press 1990. [KNS1] A. K U N I BA, T. NA K A N I S H I, J. S U Z U K I, Functional relations in solvable lattice models: I. Functional relations and representation theory, Int. J. Mod. Phys. A9 (1994), 5215–5266. 44 [KNS2] A. K U N I BA, T. NA K A N I S H I, J. S U Z U K I, T -systems and Y -systems in integrable systems, J. Phys. A 44 (2011), no. 10, 103001, 146 pp. [Le1] B. L E C L E R C, Imaginary vectors in the dual canonical basis of U (n), Transform. Groups 8 (2003), 95–104. [Le2] B. L E C L E R C, Quantum loop algebras, quiver varieties, and cluster algebras, in Representations of Alge- bras and Related Topics, (A. Skowron ´ ski and K. Yamagata, eds.), European Math. Soc. Series of Congress Reports, 2011, 117–152. [Lu] G. L U S Z T I G, On quiver varieties, Adv. Math. 136 (1998), 141–182. [N1] H. NA K A J I M A, Quiver varieties and finite-dimensional representations of quantum affine algebras. J. Amer. Math. Soc. 14 (2001), 145–238. [N2] H. NA K A J I M A, t -analogs of q-characters of Kirillov-Reshetikhin modules of quantum affine algebras, Rep- resent. Theory 7 (2003), 259–274. [N3] H. NA K A J I M A, Quiver varieties and t -analogs of q-characters of quantum affine algebras. Annals of Math. 160 (2004), 1057–1097. [N4] H. NA K A J I M A, Quiver varieties and cluster algebras, Kyoto J. Math. 51 (2011), 71–126. [Pa] Y. PA L U, Cluster characters II: A multiplication formula, Proceedings London Math. Soc. (to appear). [Pl1] P.-G. P L A M O N D O N, Cluster characters for cluster categories with infinite-dimensional morphism spaces, Adv. Math. 227 (2011), 1–39. [Pl2] P.-G. P L A M O N D O N, Generic bases for cluster algebras from the cluster category, Int. Math. Research. Notices, (to appear). David HE RNANDE Z : Universite ´ Paris Diderot-Paris 7, Institut de Mathe ´matiques de Jussieu - Paris Rive Gauche CNRS UMR 7586, Ba ˆtiment Sophie Germain, Case 7012, 75205 Paris Cedex 13, France. email : hernandez@math.jussieu.fr Bernard L E CL E RC : Normandie Univ, France UNICAEN, LMNO F-14032 Caen, France CNRS UMR 6139, F-14032 Caen, France Institut Universitaire de France. email : bernard.leclerc@unicaen.fr http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Mathematics arXiv (Cornell University)

A cluster algebra approach to q-characters of Kirillov-Reshetikhin modules

Mathematics , Volume 2020 (1303) – Mar 4, 2013

Loading next page...
 
/lp/arxiv-cornell-university/a-cluster-algebra-approach-to-q-characters-of-kirillov-reshetikhin-8WUBtt159w
ISSN
1435-9855
eISSN
ARCH-3343
DOI
10.4171/JEMS/609
Publisher site
See Article on Publisher Site

Abstract

We describe a cluster algebra algorithm for calculating q-characters of Kirillov-Resheti- khin modules for any untwisted quantum affine algebra U (bg). This yields a geometric q- character formula for tensor products of Kirillov-Reshetikhin modules. When g is of type A, D, E , this formula extends Nakajima’s formula for q-characters of standard modules in terms of homology of graded quiver varieties. Contents 1 Introduction 1 2 Definitions and notation 3 3 An algorithm for the q-characters of Kirillov-Reshetikhin modules 8 4 A geometric character formula for Kirillov-Reshetikhin modules 17 5 Beyond Kirillov-Reshetikhin modules 29 6 Appendix 32 1 Introduction Let g be a simple Lie algebra over C, and let U (g) be the corresponding untwisted quantum affine algebra with quantum parameter q ∈ C not a root of unity. The finite-dimensional complex representations of U (g) have been studied by many authors during the past twenty years. We refer the reader to [CP1] for a classical introduction, and to [CH, Le2] for recent surveys on this topic. In [HL1], we started to explore some new connections between this rich representation theory and the cluster algebras of Fomin and Zelevinsky. The main result, proved in [HL1] in type A and D , and extended to any A-D-E type by Nakajima [N4], shows the existence of a tensor n 4 category C of finite-dimensional U (bg)-modules whose Grothendieck ring is a cluster algebra of 1 q the same finite Dynkin type, such that the classes of simple modules coincide with the set of cluster monomials. As a consequence, the q-characters of the simple objects of C can be computed algorithmically using the combinatorics of cluster algebras. Moreover, the Caldero-Chapoton formula for cluster expansions leads to some new geometric formulae for these characters, in terms of Euler characteristics of quiver Grassmannians. Unfortunately the category C is quite small. For instance it contains only three Kirillov- Reshetikhin modules for each node of the Dynkin diagram of g. Another limitation of the papers [HL1] and [N4] is that g is assumed to be of simply laced type. In fact, the general proof of Nakajima uses in a crucial way his geometric construction of the standard U (bg)-modules [N1], which is only available in the simply laced case. arXiv:1303.0744v1 [math.QA] 4 Mar 2013 In this paper we drop the assumption of being simply laced, and we consider a much larger tensor subcategory C which contains, up to spectral shifts, all the irreducible finite-dimensional representations of U (bg). Our first main result (Theorem 3.1) is an algorithm which calculates the q-character of an arbitrary Kirillov-Reshetikhin module in C as the result of a sequence of cluster mutations. The only input for this calculation is the initial seed of our cluster algebra A , which is encoded in a quiver obtained from the Cartan matrix of g by a simple and uniform recipe. (It may be worth noting that A is always a skew-symmetric cluster algebra, even when g is not simply laced.) The proof of this theorem is based on the fact that the q-characters of the Kirillov-Reshetikhin modules are solutions of the corresponding T -system of Kuniba, Nakanishi and Suzuki [KNS1, N2, H]. This will come as no surprise, given the many papers already devoted to the relationships between cluster algebras and T -systems (see in particular [IIKNS], [IIKKN1], [IIKKN2]; in fact our algorithm is inspired from [GLS2, §13], where similar T -system formulas are obtained for generalized minors of symmetric Kac-Moody groups). We find it nevertheless remarkable that, by interpreting the T -system equations as appropriate cluster transformations, one is able to obtain the Kirillov-Reshetikhin q-characters starting from their highest weight monomials via a procedure of successive approximations. To the best of our knowledge this simple “bootstrap” algorithm had not been noticed before, although, in retrospect, it could certainly have been formulated and proved without knowledge of the cluster algebra theory. At this stage, we should recall that Frenkel and Mukhin [FM] have described long ago a completely different algorithm, which can be used for computing the q-characters of the Kirillov- Reshetikhin modules [N2, H]. The advantage of our approach is that we are now in a position to apply deep results of the theory of cluster algebras and obtain new formulas for the Kirillov- Reshetikhin q-characters. In [DWZ1, DWZ2], Derksen, Weyman and Zelevinsky have con- structed a categorical model for a large class of cluster algebras using quivers with potentials. In particular they have proved a far-reaching generalization of the Caldero-Chapoton formula, ex- pressing any cluster variable in terms of the F -polynomial of an associated quiver representation (see also [Pl1] for a different proof of this generalized formula). Applying this formula in our con- text, we get a geometric character formula for arbitrary Kirillov-Reshetikhin modules, and also for their tensor products (Theorem 4.8). When g is simply laced, and we restrict our attention to the simplest Kirillov-Reshetikhin modules and their tensor products, namely the fundamental modules and the standard modules, the quiver Grassmannians involved in our formula are homeomorphic to the projective varieties L (V,W ) used by Nakajima [N3, §4] in his geometric construction of the standard modules. This suggests that the quiver Grassmannians we introduce, in connection with general Kirillov- Reshetikhin modules of not necessarily simply laced type, might be interesting new varieties. When g is a classical Lie algebra of type A, B, C, D, there exist tableau sum formulas for the q- characters of certain Kirillov-Reshetikhin modules (see [KNS2, §7] and references therein). From the geometric point of view of Theorem 4.8, these formulas can be explained by the fact that the corresponding quiver representations have a nice and regular “grid structure”, and in many cases their quiver Grassmannians are reduced to points (see e.g. §6.4, §6.5, §6.6). The cluster algebra approach also suggests that our results should extend far beyond the Kirillov-Reshetikhin modules. Indeed, we show (Theorem 5.1) that the cluster algebra A is iso- morphic to the Grothendieck ring of C . It is then natural to conjecture that this isomorphism maps all cluster monomials of A to the classes of certain simple objects of C (Conjecture 5.2), and to extend the above geometric character formula to all these simple objects (Conjecture 5.3). The results of [HL1, HL2] and [N4] provide some evidence supporting these conjectures in the simply laced case. 2 Here is a more precise outline of the paper. In Section 2 we associate with every simple Lie algebra g some quivers (§2.1), from which we define a cluster algebra A (§2.2). We also introduce the untwisted quantum affine algebra U (bg) (§2.3). In Section 3 we state and prove our algorithm for computing the Kirillov-Reshetikhin q-characters as special cluster variables of A . The proof uses T -systems (§3.2.1) and the notion of truncated q-characters (§3.2.2). In Section 4, we con- sider an algebra A defined by a quiver with potential, coming from our initial seed for A (§4.1). (i) We introduce certain distinguished A-modules K (§4.3), and we state our geometric formula k,m (i) for the Kirillov-Reshetikhin q-characters in terms of Grassmannians of submodules of the K k,m (Theorem 4.8). To prove it, we calculate the g-vectors of these q-characters, regarded as cluster variables of A , and we apply a result of Plamondon [Pl2] which allows to reconstruct the A- module corresponding to a given cluster variable from the knowledge of its g-vector. To be in a position to apply this result, we show that the defining potential of A is rigid, and that appropriate truncations of A are finite-dimensional (Proposition 4.17). In Section 5, we prove Theorem 5.1 and we formulate Conjecture 5.2 and Conjecture 5.3. The paper closes with an appendix illustrating our results with many examples. 2 Definitions and notation 2.1 Quivers 2.1.1 Cartan matrix Let C = (c ) be an indecomposable n× n Cartan matrix of finite type [Ka, §4.3]. There is a i j i, j∈I diagonal matrix D = diag(d | i ∈ I) with entries in Z such that the product i >0 B = DC = (b ) i j i, j∈I is symmetric. We normalize D so that min{d | i ∈ I} = 1, and we put t := max{d | i ∈ I}. Thus i i 1 if C is of type A , D , E , E or E ,  n n 6 7 8 t = 2 if C is of type B , C or F , n n 4 3 if C is of type G . It is easy to check by inspection that (d > 1 and c < 0) =⇒ (c = −1). (1) i i j i j One attaches to C a Dynkin diagram δ with vertex set I [Ka, §4.7]. Since C is assumed to be indecomposable and of finite type, δ is a tree. All the objects that we consider below depend on C, but we shall not always repeat it, neither record it explicitly in our notation. Example 2.1 The Cartan matrix C of type B in the Cartan-Killing classification is defined by   2 −1 0   C = −1 2 −1 0 −2 2 We have D = diag(2, 2, 1) and the symmetric matrix B is given by   4 −2 0   B = −2 4 −2 0 −2 2 3 . (1,2) (3,2) OO ❑ OO ❑ s ❑ s ❑ s ❑ s %% yys (2,1) OO ❑ s ❑ s ❑ s ❑ yys %% (1,0) (3,0) OO ❑ OO ❑ s ❑ s %% yys (2,0) OO (2,−1) s ❑ OO ❑ s ❑ s s ❑ s ❑ yys %% s ❑ yys %% (1,−1) (3,−1) OO ❑ OO (1,−2) (3,−2) ❑ s OO ❑ OO ❑ s s ❑ s ❑ s %% yys ❑ s ❑ s %% yys (2,−2) OO ❑ (2,−3) s ❑ s ❑ OO ❑ s ❑ s ❑ s ❑ s ❑ yys %% s ❑ s ❑ yys %% (1,−3) (3,−3) OO ❑ OO (1,−4) (3,−4) ❑ s OO ❑ OO ❑ s ❑ s ❑ s %% yys ❑ s ❑ s %% yys (2,−4) (2,−5) s ❑ s ❑ s ❑ s ❑ yys %% s ❑ s ❑ yys %% (1,−5) (3,−5) (1,−6) (3,−6) . Figure 1: The quivers Γ and G in type A . 2.1.2 Infinite quiver e e e e Put V = I ×Z. We introduce a quiver Γ with vertex set V . The arrows of Γ are given by ((i, r) → ( j, s)) ⇐⇒ (b 6= 0 and s = r + b ). i j i j Lemma 2.2 The quiver Γ has two isomorphic connected components. Proof — Let i ∈ I be such that d = 1. For every r ∈Z we have an arrow (i, r) → (i, r+ 2). Since the Dynkin diagram δ is connected, every vertex ( j, s) ∈ V is connected to a vertex of the form (i, r), so Γ has at most two connected components. Moreover, since δ is a tree, any path from (i, r) to (i, s) in Γ contains as many arrows of the form ( j, p) → (k, p+ b ) with j 6= k, as it contains arrows of the jk form (k, t) → ( j, t + b ). Since b = b , and since b ∈ 2Z for every j ∈ I, it follows that if there k j jk k j j j is a path from (i, r) to (i, s) then s− r ∈ 2Z. Therefore Γ has exactly two connected components. These two components are isomorphic via the map ( j, r) 7→ ( j, r + 1) (( j, r) ∈ V ×Z)). ✷ We pick one of the two isomorphic connected components of Γ and call it Γ. The vertex set of Γ is denoted by V . 2.1.3 Semi-infinite quiver We will have to use a second labelling of the vertices of Γ. It is deduced from the first one by means of the function ψ defined by ψ(i, r) = (i, r + d ), ((i, r) ∈ V ). (2) 4 . . . . OO . . (2,5) (1,5) OO OO s s s s s s s s yys yys (1,3) (2,3) OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% (2,1) (1,1) OO OO s s s s s s s s yys yys (2,0) OO (1,−1) (2,−1) OO ❑ OO ❑ ❑ ❑ ❑ ❑ %% ❑ ❑ ❑ ❑ %% %% (2,−2) (1,−1) OO OO (2,−3) (1,−3) s s s s OO OO s s s s s s s s yys yys s s s s yys yys (1,−3) (2,−4) OO ❑ OO ❑ (1,−5) (2,−5) ❑ ❑ ❑ ❑ OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% ❑ ❑ ❑ ❑ %% %% (2,−6) (1,−5) OO OO (2,−7) (1,−7) s s s s OO OO s s s s s s s s yys yys s s s s yys yys (1,−7) (2,−8) (1,−9) (2,−9) . . . . . . . . . . . . . . . . . . Figure 2: The quivers Γ and G in type B . Let W ⊂ I ×Z be the image of V under ψ . We shall denote by G the same quiver as Γ but with − − vertices labelled by W . Put W := W ∩ (I ×Z ). Let G be the full subquiver of G with vertex ≤0 set W . Example 2.3 The definitions of §2.1.2 and §2.1.3 are illustrated in Figure 1 and Figure 2. We find it convenient to always display the quivers Γ in the following way. We decide to draw all arrows of the form (i, r) → (i, r + b ) vertically, going upwards. Moreover, if (i, r) and (i, s) are two vertices ii with r− s 6∈ b Z, we draw them in different columns. Hence, the quivers attached to C always have ii d columns. Finally, the integer r determines the altitude of the vertex (i, r) in Γ. Therefore, i∈I since for i 6= j we have b ≤ 0, the arrows (i, r) → ( j, r + b ) are represented as oblique arrows i j i j going down. − − Figure 1 displays the quivers Γ and G for C of type A . Figure 2 shows Γ and G for C of type B . In both cases we have chosen to call Γ the connected component of Γ containing the vertex (2, 1). For another illustration, with C of type G , see Figure 3. More examples can be found in the Appendix, §6.5, §6.6, §6.7. 2.2 Cluster algebras We refer the reader to [FZ2] and [GSV] for an introduction to cluster algebras, and for any un- defined terminology. 2.2.1 Cluster algebra attached to G − − Consider an infinite set of indeterminates z = {z | (i, r) ∈ W } over Q. Let A be the cluster i,r − − algebra defined by the initial seed (z , G ). Thus, A is the Q-subalgebra of the field of rational 5 − − functions Q(z ) generated by all the elements obtained from some element of z via a finite sequence of seed mutations, see [GG, Definition 3.1]. Note that there are no frozen variables. Cluster algebras of infinite rank have not received much attention up to now. (In fact we are not aware of another paper than [GG]; in [GG], a specific example of type A is developed, in connection with a triangulated category studied by Holm and Jorgensen [HoJo].) For our purposes in this paper, it is always possible to work with sufficiently large finite sub- − − seeds of the seed (z , G ), and replace A by the genuine cluster subalgebras attached to them. On the other hand, statements become nicer if we allow ourselves to formulate them in terms of the infinite rank cluster algebra A . 2.2.2 Monomial change of variables − − Let Y = {Y | (i, r) ∈ W} be a new set of indeterminates over Q. Let Y = {Y ∈ Y | (i, r) ∈ W }. i,r i,r For (i, r) ∈ W , we perform the substitution z = Y . (3) i,r i, r+kb ∏ ii k≥0, r+kb ≤0 ii Note that all variables in the right-hand side of (3) belong to Y . Example 2.4 If G is as in Figure 2, we have z = Y , z = Y Y , z = Y Y Y , z = Y Y Y Y , 2,0 2,0 2,−2 2,−2 2,0 2,−4 2,−4 2,−2 2,0 2,−6 2,−6 2,−4 2,−2 2,0 z = Y , z = Y Y , z = Y Y Y , z = Y Y Y Y , 1,−1 1,−1 1,−5 1,−5 1,−1 1,−9 1,−9 1,−5 1,−1 1,−13 1,−13 1,−9 1,−5 1,−1 z = Y , z = Y Y , z = Y Y Y , etc. 1,−3 1,−3 1,−7 1,−7 1,−3 1,−11 1,−11 1,−7 1,−3 2.2.3 Sequence of vertices As explained in Example 2.3, the arrows of G of the form (i, r) −→ (i, r + b ) are called vertical ii and displayed in columns. To each column we attach an initial label given by the index of its top vertex (i, r), for which r is maximal among the vertices of the column. We now form a sequence of t n columns by induction as follows. At each step we pick a column whose label (i, r) has maximal r among labels of all columns. After picking a column with label (i, r), we change its label to (i, r− b ). Finally, reading column after column in this ordering, from ii top to bottom, we get an infinite sequence S of vertices of G . Example 2.5 If G is as in Figure 1, then t = 1, the sequence of columns consists of 3 columns, and we obtain the following sequence of vertices: S = ((2, 0),(2,−2),(2,−4),... ,(1,−1),(1,−5),(1,−9),... ,(3,−1),(3,−3),(3,−5),...). (Here, the columns labelled (1,-1) and (3,-1) could be interchanged.) If G is as in Figure 2, then t = 2, the sequence of columns consists of 4 columns and gives the following sequence of vertices: S = ((2, 0),(2,−2),(2,−4),... ,(1,−1),(1,−5),(1,−9),... , (2, 0),(2,−2),(2,−4),... ,(1,−3),(1,−7),(1,−11),...) Note that the column with vertices (2, r) appears two times. It appears first because its initial label is (2, 0). After picking it, its label is changed to (2,−2), so it appears again between the columns labelled (1,−1) and (1,−3). 6 g A B C D E E E F G n n n n 6 7 8 4 2 t 1 2 2 1 1 1 1 2 3 hˇ n + 1 2n− 1 n + 1 2n− 2 12 18 30 9 4 Table 1: Dual Coxeter numbers Finally, for (i, r) ∈ G , we define k to be the unique positive integer k satisfying i,r 0 < kb −|r| ≤ b . (4) ii ii In other words, (i, r) is the kth vertex in its column, counting from the top. Example 2.6 If G is as in Figure 2, then k = 2, k = 3. 2,−2 1,−9 2.3 Quantum affine algebras 2.3.1 The algebra U (bg) Let g be the simple Lie algebra over C with Cartan matrix C. We denote by α (i ∈ I) the simple roots of g, and by ϖ (i ∈ I) the fundamental weights. They are related by α = c ϖ . (5) i ∑ ji j j∈I Let hˇ be the dual Coxeter number of g, see [Ka, §6.1]. The values of hˇ are recalled in Table 1. Let g be the corresponding untwisted affine Lie algebra. Thus if g has type X in the Cartan- (1) Killing classification, bg has type X in the Kac classification [Ka, §4.8]. Let U (bg) be the n q b b Drinfeld-Jimbo quantum enveloping algebra of g, see e.g. [CP1]. We regard U (g) as a C-algebra with quantum parameter q ∈ C not a root of unity. 2.3.2 q-characters Frenkel and Reshetikin [FR] have attached to every complex finite-dimensional representation of U (bg) a q-character χ (M). If M is irreducible, it is determined up to isomorphism by its q q q-character. The irreducible finite-dimensional representations of U (g) have been classified by Chari and Pressley in terms of Drinfeld polynomials, see [CP1, Theorem 12.2.6]. Equivalently, ir- reducible finite-dimensional representations of U (g) can be parametrized by the highest dominant monomial of their q-character [FR], and this is the parametrization we shall use. By definition, the q-character χ (M) is a Laurent polynomial with positive integer coefficients in the infinite set of variables Y = {Y | i ∈ I, a ∈ C }, which should be seen as a quantum affine i,a analogue of {e | i ∈ I}. In this paper we will be concerned only with polynomials involving the subset of variables Y , ((i, r) ∈ W ). i,q For simplicity of notation, we shall therefore write Y instead of Y . Thus our q-characters will i,r i,q be Laurent polynomials in the variables of the set Y introduced in §2.2.2. 7 Let m be a dominant monomial in the variables Y ∈ Y, that is, a monomial with nonnegative i,r exponents. We denote by L(m) the corresponding irreducible representation of U (bg), and by χ (m) = χ (L(m)) its q-character. For example, if m is of the form q q k−1 m = Y , (i ∈ I, r ∈ Z, k ≥ 1), i, r+ jb ∏ ii j=0 (i) L(m) is called a Kirillov-Reshetikhin module, and usually denoted by W . In particular, if k = 1 k,r (i) (i) we get a fundamental module W = L(Y ). By convention, if k = 0 the module W is the trivial i,r 1,r 0,r one-dimensional module for every (i, r), and its q-character is equal to 1. Finally, following [FR], for (i, r) ∈ V we introduce the following quantum affine analogue of e : ! ! ! −1 −1 −1 A := Y Y Y Y Y Y Y Y (6) i,r i,r−d i,r+d j,r j,r−1 j,r+1 j,r−2 j,r j,r+2 i i ∏ ∏ ∏ j: c =−1 j: c =−2 j: c =−3 ji ji ji Note that since (i, r) ∈ V , we have (i, r± d ) ∈ W . If c < 0, we also have, because of (1), i ji ( j, r + c + 1) = ( j, r + d (c + 1)) = ( j, r + b + d ) ∈ W. ji j ji i j j It follows that A is a Laurent monomial in the variables Y with ( j, s) ∈ W . i,r j,s 3 An algorithm for the q-characters of Kirillov-Reshetikhin modules 3.1 Statement and examples − − Let A be the cluster algebra defined in §2.2.1, with initial seed Σ = (z , G ), and let S = ((i , r ),(i , r ),(i , r ),...) 1 1 2 2 3 3 be the sequence of vertices of the quiver of A defined in §2.2.3. We denote by μ (Σ) the new seed obtained after performing the sequence of mutations indexed by S , that is, by mutating first at vertex (i , r ), then at vertex (i , r ), etc. More generally, for m ≥ 1, let Σ = μ (Σ) be the 1 1 2 2 m (m) seed obtained from Σ after m repetitions of the mutation sequence μ . Let z be the cluster i,r variable of Σ sitting at vertex (i, r) ∈ W ; this is a Laurent polynomial in the initial variables (m) (m) z , ( j, s) ∈ W . Let y be the Laurent polynomial obtained from z by performing the change j,s i,r i,r of variables (3) of §2.2.2; this is a Laurent polynomial in the variables Y , ( j, s) ∈ W . j,s Theorem 3.1 (a) The quiver of μ (Σ) is equal to the quiver of Σ, that is, to G . (m) (b) Suppose that m ≥ hˇ/2. Then, the y are the q-characters of the Kirillov-Reshetikhin i,r modules. More precisely, for m ≥ hˇ/2 there holds (m) (i) y = χ W . i,r k, r−2t m where k = k is defined as in §2.2.3. i,r (i) (i) Remark 3.2 It is well known that, for p ∈ Z, the q-character χ (W ) is deduced from χ (W ) q q k,r+ p k,r by applying the ring automorphism mapping Y to Y for every ( j, s) ∈ I×Z. Therefore, mod- j,s j,s+ p ulo these straightforward automorphisms, Theorem 3.1 describes the q-characters of all Kirillov- Reshetikhin modules. 8 Remark 3.3 Although the statement of Theorem 3.1 involves an infinite seed Σ and an infinite sequence of mutations S , the calculation of the q-character of a given Kirillov-Reshetikhin mod- ule requires only a finite number of mutations on a finite initial segment of the semi-infinite (i) quiver. More precisely, the proof of Theorem 3.1 will show that all the q-characters χ (W ) k,s ′ ′ ′ with k = 1,..., l can be calculated using (h + 2l − 1)h n/2 mutations, where h = ⌈hˇ/2⌉. Example 3.4 Let g be of type A . The quiver G of the initial seed is displayed in Figure 1. The initial cluster variables are z = Y , z = Y Y , z = Y Y Y , etc. 2,0 2,0 2,−2 2,−2 2,0 2,−4 2,−4 2,−2 2,0 z = Y , z = Y Y , z = Y Y Y , etc. 1,−1 1,−1 1,−3 1,−3 1,−1 1,−5 1,−5 1,−3 1,−1 z = Y , z = Y Y , z = Y Y Y , etc. 3,−1 3,−1 3,−3 3,−3 3,−1 3,−5 3,−5 3,−3 3,−1 After the mutation sequence μ , the first new cluster variables are (1) −1 y = Y + Y Y Y , 2,−2 1,−1 3,−1 2,0 2,0 (1) −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y Y Y Y , 2,−4 2,−2 1,−1 3,−1 2,−4 1,−3 1,−1 3,−3 3,−1 2,−2 2,0 2,−2 2,0 (1) −1 −1 −1 y = Y Y Y + Y Y Y Y Y + Y Y Y Y Y Y Y , 2,−6 2,−4 2,−2 1,−1 3,−1 2,−6 2,−4 1,−3 1,−1 2,−6 3,−3 3,−1 2,−4 2,0 2,−2 2,0 −1 −1 −1 + Y Y Y Y Y Y Y Y Y , 1,−5 1,−3 1,−1 3,−5 3,−3 3,−1 2,−4 2,−2 2,0 (1) −1 −1 y = Y + Y Y + Y Y , 1,−3 2,−2 3,−1 1,−1 1,−1 2,0 (1) −1 −1 −1 −1 y = Y Y + Y Y Y + Y Y Y + Y Y Y Y 1,−5 1,−3 1,−5 2,−2 1,−5 3,−1 2,−4 2,−2 1,−3 1,−1 2,0 1,−3 1,−1 −1 −1 −1 −1 + Y Y Y Y + Y Y Y Y , 2,−4 3,−1 3,−3 3,−1 1,−3 2,0 2,−2 2,0 (1) −1 −1 −1 −1 y = Y Y Y + Y Y Y Y + Y Y Y Y + Y Y Y Y Y 1,−7 1,−5 1,−3 1,−7 1,−5 2,−2 1,−7 1,−5 3,−1 1,−7 2,−4 2,−2 1,−5 1,−1 2,0 1,−3 1,−1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y + Y Y Y Y Y + Y Y Y Y Y Y 1,−7 2,−4 3,−1 1,−7 3,−3 3,−1 2,−6 2,−4 2,−2 1,−3 2,0 2,−2 2,0 1,−5 1,−3 1,−1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y Y 2,−6 2,−4 3,−1 2,−6 3,−3 3,−1 1,−5 1,−3 2,0 1,−5 2,−2 2,0 −1 −1 −1 + Y Y Y Y Y Y , 3,−5 3,−3 3,−1 2,−4 2,−2 2,0 (1) (1) (1) (1) (1) (1) (We omit the variables y , y , y , since they are readily obtained from y , y , y 3,−1 3,−5 3,−5 1,−1 1,−5 1,−5 via the symmetry (1 ↔ 3).) After a second application of the mutation sequence μ , the first new cluster variables are (2) −1 −1 −1 −1 −1 −1 y = Y + Y Y Y + Y Y + Y Y + Y Y Y + Y , 2,−4 1,−3 3,−3 1,−3 3,−3 2,−2 2,0 2,−2 3,−1 1,−1 1,−1 3,−1 2,0 (2) −1 −1 −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y Y Y Y + Y Y Y 2,−6 2,−4 1,−3 3,−3 2,−6 1,−5 1,−3 3,−5 3,−3 1,−5 2,−2 2,−2 2,−4 2,−2 2,0 3,−3 −1 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y + Y Y Y Y + Y Y + Y Y Y Y + Y Y Y 3,−5 2,−4 2,−6 1,−5 3,−5 1,−3 2,−6 1,−3 2,0 1,−3 2,0 3,−3 2,0 2,−4 2,0 3,−1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y + Y Y Y Y Y + Y Y Y + Y Y Y Y 1,−5 1,−3 1,−5 1,−3 3,−5 2,−6 3,−3 3,−5 3,−3 3,−3 3,−1 2,−4 3,−1 1,−1 1,−3 1,−1 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y + Y Y Y Y Y Y + Y Y Y Y 1,−5 3,−5 3,−3 2,−4 2,−2 2,−6 2,−2 1,−1 2,−4 1,−3 1,−1 3,−3 3,−1 1,−1 3,−1 −1 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y + Y Y Y Y Y + Y Y Y Y Y Y 2,−2 3,−5 1,−5 2,−2 1,−5 2,−2 3,−5 1,−3 1,−1 3,−1 1,−1 3,−3 3,−1 1,−1 2,−4 3,−1 −1 −1 +Y Y , 2,−2 2,0 (2) −1 −1 −1 y = Y + Y Y + Y Y + Y , 1,−5 2,−4 3,−3 1,−1 1,−3 2,−2 3,−1 (2) −1 −1 −1 −1 y = Y Y + Y Y Y + Y Y Y + Y Y Y Y 1,−7 1,−5 1,−7 2,−4 1,−7 3,−3 2,−6 2,−4 1,−3 1,−3 2,−2 1,−5 1,−3 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y + Y Y Y Y + Y Y Y + Y Y 2,−6 3,−3 3,−5 3,−3 2,−6 1,−7 1,−5 2,−2 2,−4 2,−2 1,−5 3,−1 3,−1 −1 −1 −1 −1 + Y Y Y + Y Y , 3,−5 2,−4 3,−1 3,−3 3,−1 Here hˇ/2 = 2, so we can observe that the cluster variables obtained after performing 2 times the 9 mutation sequence μ are indeed q-characters of Kirillov-Reshetikhin modules, namely, (2) (2) y = χ (Y ), y = χ (Y Y ), etc. q 2,−4 q 2,−6 2,−4 2,0 2,−2 (2) (2) y = χ (Y ), y = χ (Y Y ), etc. q 1,−5 q 1,−7 1,−5 1,−1 1,−3 (2) (2) y = χ (Y ), y = χ (Y Y ), etc. q 3,−5 q 3,−7 3,−5 3,−1 3,−3 Example 3.5 Let g be of type B . The quiver G of the initial seed is displayed in Figure 2. The initial cluster variables are z = Y , z = Y Y , z = Y Y Y , etc. 2,0 2,0 2,−2 2,−2 2,0 2,−4 2,−4 2,−2 2,0 z = Y , z = Y Y , z = Y Y Y , etc. 1,−1 1,−1 1,−5 1,−5 1,−1 1,−9 1,−9 1,−5 1,−1 z = Y , z = Y Y , z = Y Y Y , etc. 1,−3 1,−3 1,−7 1,−7 1,−3 1,−11 1,−11 1,−7 1,−3 After the mutation sequence μ , the first new cluster variables are (1) −1 y = Y + Y Y , 2,−4 1,−3 2,0 2,−2 (1) −1 −1 −1 −1 y = Y Y + Y Y Y + Y Y Y Y + Y Y , 2,−6 2,−4 1,−3 2,−6 1,−5 1,−3 1,−3 2,−2 2,−2 2,−4 2,−2 1,−1 (1) −1 −1 −1 y = Y Y Y + Y Y Y Y + Y Y Y Y Y 2,−8 2,−6 2,−4 1,−3 2,−8 2,−6 1,−5 1,−3 2,−8 2,−4 2,−2 2,−4 2,−2 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y + Y Y Y , 1,−7 1,−5 1,−3 1,−7 1,−3 1,−3 2,−8 2,−6 2,−4 2,−2 1,−1 2,−6 1,−1 (1) −1 −1 −1 −1 y = Y + Y Y Y + Y Y + Y Y Y , 1,−5 2,−4 2,−2 2,−4 1,−3 1,−1 1,−1 2,0 2,−2 2,0 (1) −1 −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y + Y Y Y Y 1,−5 1,−9 1,−9 2,−4 2,−2 1,−9 2,−4 1,−9 1,−3 1,−5 1,−1 2,0 2,−2 2,0 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 2,−8 2,−6 2,−4 2,−2 2,−8 2,−6 2,−4 1,−5 1,−1 1,−5 2,0 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 1,−3 2,−8 2,−6 1,−3 2,−8 2,−2 2,0 2,−4 2,−2 2,0 1,−5 −1 −1 −1 −1 + Y Y Y Y Y Y , 1,−7 1,−3 2,−6 2,−4 2,−2 2,0 (1) −1 −1 −1 −1 −1 y = Y + Y Y Y + Y Y + Y Y Y + Y , 1,−7 2,−6 2,−4 2,−6 1,−5 1,−3 1,−3 2,−2 2,−4 2,−2 1,−1 (1) −1 −1 −1 −1 y = Y Y + Y Y Y Y + Y Y Y + Y Y Y Y 1,−7 1,−11 1,−11 2,−6 2,−4 1,−11 2,−6 1,−11 1,−5 1,−7 1,−3 2,−2 2,−4 2,−2 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 2,−10 2,−8 2,−6 2,−4 2,−10 2,−8 2,−6 1,−7 1,−3 1,−7 2,−2 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y Y 1,−5 2,−10 2,−8 1,−5 2,−10 1,−7 2,−4 2,−2 2,−6 2,−4 2,−2 −1 −1 −1 −1 −1 −1 −1 −1 + Y Y Y Y Y Y + Y Y Y Y + Y Y 1,−9 1,−5 1,−9 1,−11 2,−8 2,−6 2,−4 2,−2 1,−1 2,−8 2,−6 1,−1 −1 −1 −1 −1 −1 −1 + Y Y Y + Y Y Y Y + Y Y . 2,−10 2,−10 2,−8 1,−1 2,−6 1,−7 1,−1 1,−5 1,−1 Here hˇ/2 = 3/2, and we can observe that certain cluster variables are not yet q-characters of Kirillov-Reshetikhin modules. But some already are, namely (1) (1) y = χ (Y ), y = χ (Y Y ), etc. q 1,−7 q 1,−7 1,−11 1,−3 1,−7 After a second application of the mutation sequence μ , since 2 > 3/2, all the new cluster vari- ables are q-characters of Kirillov-Reshetikhin modules. For example (2) −1 −1 −1 y = Y + Y Y + Y Y + Y = χ (Y ). 2,−8 1,−7 2,−4 q 2,−8 2,0 2,−6 1,−3 2,−2 3.2 Proof of Theorem 3.1 The proof relies on two main ingredients which we shall first review, namely, the T -systems, and the truncated q-characters. 10 3.2.1 T -systems With the quantum affine algebra U (g) is associated a system of difference equations called a T -system [KNS1]. Its unknowns are denoted by (i) T , (i ∈ I, k ∈ N, r ∈ Z). k,r We fix the initial boundary condition (i) T = 1, (i ∈ I, r ∈ Z). (7) 0,r If g is of type A , D , E , the T -system equations are n n n (i) (i) (i) (i) ( j) T T = T T + T , (i ∈ I, k ≥ 1, r ∈ Z). (8) k,r+1 k,r−1 k−1,r+1 k+1,r−1 k,r j: c =−1 i j If g is not of simply laced type, the T -system equations are more complicated. They can be written in the form (i) (i) (i) (i) (i) T T = T T + S , (i ∈ I, k ≥ 1, r ∈ Z), (9) k,r+d k,r−d k−1,r+d k+1,r−d k,r i i i i (i) where S is defined as follows. If d ≥ 2 then k,r (i) ( j) ( j) S = T T . (10) ∏ ∏ k,r k,r d k, r−d +1 i i j: c =−1 j: c ≤−2 ji ji If d = 1 and t = 2, then ( j) ( j) ( j)  T T T , if k = 2l, ∏ ∏  k,r l,r l,r+2 j: c =−1 j: c =−2 i j i j (i) S = (11) k,r ( j) ( j) ( j)  T T T if k = 2l + 1.  ∏ ∏ k,r l+1,r l,r+2 j: c =−1 j: c =−2 i j i j Finally, if d = 1 and t = 3, that is, if g is of type G , denoting by j the other vertex of δ we have i 2 d = 3 and ( j) ( j) ( j) T T T if k = 3l, l,r l,r+2 l,r+4 (i) ( j) ( j) ( j) S = (12) T T T if k = 3l + 1, k,r l+1,r l,r+2 l,r+4 ( j) ( j) ( j) T T T if k = 3l + 2. l+1,r l+1,r+2 l,r+4 Example 3.6 Let g be of type B . The Cartan matrix is 2 −1 C = −2 2 and we have d = 2 and d = 1. The T -system reads: 1 2 (1) (1) (1) (1) (2) T T = T T + T , (k ≥ 1, r ∈ Z), k,r+2 k,r−2 k−1,r+2 k+1,r−2 2k,r−1 (2) (2) (2) (2) (1) (1) T T = T T + T T , (l ≥ 1, r ∈ Z), 2l,r+1 2l,r−1 2l−1,r+1 2l+1,r−1 l,r l,r+2 (2) (2) (2) (2) (1) (1) T T = T T + T T , (l ≥ 0, r ∈ Z). 2l+1,r+1 2l+1,r−1 2l,r+1 2l+2,r−1 l+1,r l,r+2 11 It was conjectured in [KNS1], and proved in [N2] (for g of type A, D, E ) and [H] (general case), that the q-characters of the Kirillov-Reshetikhin modules of U (bg) satisfy the corresponding T -system. More precisely, we have Theorem 3.7 ([N2][H]) For i ∈ I, k ∈ N, r ∈ Z, (i) (i) T = χ W , k,r k,r h i ±1 is a solution of the T -system in the ring Z Y | (i, r) ∈ I ×Z . i,r 3.2.2 Truncated q-characters Let C be the full subcategory of the category of finite-dimensional U (bg)-modules whose ob- jects have all their composition factors of the form L(m) where m is a dominant monomial in the variables of Y . h i ±1 Lemma 3.8 The q-character of an object in C belongs to Z Y | Y ∈ Y . i,r i,r Proof — A simple object of C is a quotient of a tensor product of fundamental representations of C . But the q-character of a fundamental representation can be calculated by means of the Frenkel-Mukhin algorithm [FM]. At each step the algorithm produces monomials which involve only variables Y ∈ Y. Hence the result. ✷ i,r Note that for a dominant monomial m in the variables of Y , the q-character χ (m) may ′ − contain Laurent monomials m involving variables Y ∈ Y\ Y . Following [HL1], we define the i,r truncated q-character χ (m) to be the Laurent polynomial obtained from χ (m) by discarding all h i ±1 ′ − − these monomials m . So, by definition, χ (m) ∈ Z Y | Y ∈ Y . i,r i,r Example 3.9 Let g be of type B . We keep the notation of Example 3.6. The fundamental modules L(Y ) and L(Y ) have q-characters equal to 1,−3 2,−4 −1 −1 −1 −1 −1 χ (Y ) = Y + Y Y Y + Y Y + Y Y Y + Y , q 1,−3 1,−3 2,−2 2,0 2,−2 1,−1 1,1 2,2 2,0 2,2 1,3 −1 −1 −1 χ (Y ) = Y + Y Y + Y Y + Y . q 2,−4 2,−4 1,−3 2,0 2,−2 1,1 2,2 The corresponding truncated q-characters are χ (Y ) = Y , 1,−3 1,−3 − −1 χ (Y ) = Y + Y Y . 2,−4 2,−4 1,−3 q 2,−2 Proposition 3.10 (i) C is a tensor category. (ii) The assignment [L(m)] 7→ χ (m) extends to an injective ring homomorphism from the h i ±1 − − Grothendieck ring K (C ) to Z Y | Y ∈ Y . 0 i,r i,r Proof — The argument follows the same lines as [HL1, §5.2.4, §6.2]. Recall the Laurent monomi- als A introduced in (6). By [FR], a Laurent monomial m of the q-character of a simple object of i,r − ′ C can always be written in the form m = mM where m is a dominant monomial in the variables 12 −1 of Y , and M is a monomial in the variables A with (i, k + d ) ∈ W . Note that the Y -variable ap- i,k −1 pearing in A with the highest spectral parameter is Y . It follows that A is a right-negative i,r i,r+d i,r monomial in the sense of [FM], that is, the Y -variable with highest spectral parameter occuring in −1 A has a negative exponent. i,r ′ − ′ Let L(m) and L(m ) be simple objects of C , that is, m and m are dominant monomials in − ′′ ′ ′′ the variables of Y . If L(m ) is a composition factor of L(m)⊗ L(m ), then m is a product of ′ ′′ ′ monomials of χ (m) and χ (m ). So, we have m = mm M where M is a monomial in the variables q q −1 ′′ A . We claim that, since m is dominant, the spectral parameters r have to satisfy r + d ≤ 0. i,r ′′ ′′ Indeed otherwise m would be right-negative. Therefore, using Lemma 3.8, the monomial m − ′′ − − contains only variables of Y , hence L(m ) is in C , and C is a monoidal category. Moreover, by [CP2, Prop. 5.1], the category C is stable by duals, so it is a tensor category. This proves (i). To prove (ii) consider now an arbitrary Laurent monomial m of the q-character of an object − ′ of C . As above, it can be written in the form m = mM where m is a dominant monomial in the − −1 ′ variables of Y , and M is a monomial in the variables A with (i, k + d ) ∈ W . Now m contains i,k a variable Y 6∈ Y if and only if M contains a negative power of A for some pair (i, r) such that j,s i,r h i ±1 (i, r + d ) 6∈ W . So, if R denotes the subring of Z Y | Y ∈ Y generated by all the monomials i i,r i,r of the q-characters of the objects of C , and if I denotes the linear span of those monomials containing a variable Y ∈ Y \ Y , we see that I is an ideal of R. Hence, if π : R → R/I is the j,s natural projection, we can realize the truncated q-character map χ as χ = π ◦ χ , h i ±1 − − − which shows that χ is a ring homomorphism K (C ) → Z Y | Y ∈ Y . Finally, the fact q 0 i,r i,r that χ is injective follows from the fact that I contains only non-dominant monomials, and that two q-characters having the same dominant monomials with the same coefficients are equal. ✷ 3.2.3 Proof of the theorem We first notice that the initial cluster variables z are equal, after the change of variables (3), to i,r the truncated q-characters of certain Kirillov-Reshetikhin modules, namely, (i) z = Y = χ W , i,r+kb i,r ∏ ii q k ,r i,r k≥0, r+kb ≤0 ii where k is defined as in (4). Indeed, the level of truncation is chosen so that after truncation only i,r the highest dominant monomial of these q-characters survives. Now, the main idea of the proof is that the quiver G and the mutation sequence μ are designed in such a way that, at every step of the mutation sequence, the exchange relation is nothing else than a T -system equation. Let us first check this when g is of rank two. For g of type A , the sequence of mutated quivers obtained at each step of μ is shown in 2 S Appendix §6.1. The mutations take place at the boxed vertices. Reading the second quiver of §6.1, we see that the new cluster variable obtained after the first mutation is equal to (2) (1) − − χ W + χ W q 2,−2 q 1,−1 z + z 2,−2 1,−1 (2) =   = χ W . 1,−2 (2) 2,0 χ W 1,0 13 Here we have used Theorem 3.7 and Proposition 3.10. Similarly, reading the third quiver of §6.1, the new cluster variable obtained after the second mutation is equal to (2) (2) (1) − − − χ W χ W + χ W q q q 3,−4 1,−2 2,−3 (2) = χ W . 2,−4 (2) χ W 2,−2 An easy induction shows that, after every vertex of the second column has been mutated, each clus- (2) (2) − − ter variable of the form χ W has been replaced by the new cluster variable χ W . q q k,−2k+2 k,−2k We now continue by mutating vertices of the first column. We first get, at the top vertex (1) (2) − − χ W + χ W q q 2,−3 1,−2 (1) = χ W . q 1,−3 (1) χ W 1,−1 Then, mutating at the next vertex gives (1) (1) (2) − − − χ W χ W + χ W q q q 3,−5 1,−3 2,−4 (1) = χ W . 2,−5 (1) χ W 2,−3 By induction one sees that, after every vertex of the first column has been mutated, each cluster (1) (1) − − variable of the form χ W has been replaced by a new cluster variable χ W . q q k,−2k+1 k,−2k−1 Moreover, one sees that the new quiver obtained after μ is nothing else than G . Hence we conclude that one application of μ produces a seed with the same quiver, and in which every (i) (i) − − cluster variable χ W has been replaced by χ W . In other words, the effect of μ is q q k,r k,r−2 merely a uniform shift of the spectral parameters r by −2. The argument is similar for g of type B . The sequence of mutated quivers obtained at each step of μ is displayed in Appendix §6.2. Reading the second quiver of §6.2, we see that the new cluster variable obtained after the first mutation is equal to (2) (1) − − χ W + χ W q 2,−2 q 1,−1 z + z 2,−2 1,−1 (2) =   = χ W . 1,−2 (2) 2,0 χ W 1,0 Similarly, reading the third quiver of §6.2, the new cluster variable obtained after the second mutation is equal to (2) (2) (1) (1) − − − − χ W χ W + χ W χ W q q q q 3,−4 1,−2 1,−1 1,−3 (2) = χ W . q 2,−4 (2) χ W 2,−2 By induction, after every vertex of the second column has been mutated, each cluster variable (2) (2) − − of the form χ W has been replaced by the new cluster variable χ W . We now q k,−2k+2 q k,−2k continue by mutating vertices of the third column. We first get, at the top vertex (1) (2) − − χ W + χ W q q 2,−4 2,−5 (1) = χ W . 1,−5 (1) χ W 1,−1 14 Then, mutating at the next vertex gives (1) (1) (2) − − − χ W χ W + χ W q q q 3,−9 1,−5 4,−8 (1) = χ W . q 2,−9 (1) χ W 2,−5 By induction one sees that, after every vertex of the third column has been mutated, each cluster (1) (1) − − variable of the form χ W has been replaced by the new cluster variable χ W . q k,−4k+3 q k,−4k−1 μ , we mutate again along the second column. One checks that after that, each For the third part of (2) cluster variable of the form χ W produced after the first part of μ has been replaced by k,−2k (2) χ W . Finally, the fourth part of μ along the first column replaces each cluster variable k,−2k−2 (1) (1) − − of the form χ W by the new cluster variable χ W . Moreover, one sees that q q k,−4k+1 k,−4k−3 the new quiver obtained after μ is nothing else than G . Hence we conclude that one application (i) of μ produces a seed with the same quiver, and in which every cluster variable χ W has k,r (i) been replaced by χ W . In other words, the effect of μ is merely a uniform shift of the k,r−4 spectral parameters r by −4. The argument is similar for g of type G . The quiver G for this case is dispayed in Figure 3, and the mutation sequence is (2, 0),(2,−2),(2,−4),... ,(1,−1),(1,−7),(1,−13),... , (2, 0),(2,−2),(2,−4),... ,(1,−3),(1,−9),(1,−15),... , (2, 0),(2,−2),(2,−4),... ,(1,−5),(1,−11),(1,−17),... . The sequence of mutated quivers obtained at each step of μ is displayed in Appendix §6.3. For a general g, we use a reduction to rank two. Namely, we show that mutation sequences and T -systems equations are compatible with rank two reductions. First, by construction, the sequence of vertices S is a union of t n columns: S = (S ,S ,···,S ), 1 2 t n where each column S is a subset of i ×Z for a certain i ∈ I. As above we use μ to denote k k ≤0 k S the sequence of mutations indexed by S . So we have μ = μ ◦ μ ◦···◦ μ . S S S S t n t n−1 1 For 0 ≤ k ≤ t n, we get the mutated quiver Σ = (μ ◦ μ ◦···◦ μ )(Σ). k S S S k k−1 1 For a subset J ⊂ I, let us denote by (Σ ) the subquiver of Σ obtained by deleting the vertices k J k (i, r) such that i ∈ / J, and the edges whose tail or head is such a vertex. For any i ∈ I, the mutation sequence μ modifies (Σ ) to itself. Consequently, (Σ ) = (Σ) does not depend on k (it is a S k i k i i disjoint union of d semi-infinite linear quivers). Besides, the mutation sequence μ modifies i S only the edges whose tail (resp. head) is in i ×Z and head (resp. tail) is in j ×Z where c < 0. k i j This is because each mutation of the sequence takes place at a vertex (i , r) having two incoming arrows from vertices (i , r ± d ) and outgoing arrows to vertices of the form ( j, s) with c < 0. k i i j Consequently, for each i 6= j in I, the effect of the mutation sequence μ on (Σ) is the same as {i, j} the effect of an iteration of the mutation sequence corresponding to the rank two Lie subalgebra 15 (2,0) OO ❀ (2,−2) OO ✝ ❀ (1,−1) OO ✝ ✝ ✝ ✝ ✝ ✝ (2,−4) ✝ ✝ OO ● ✝ ✝ ✝ ✝ ✝ ✝ ✝● ✝ ● (1,−3) ✝ ● OO ✽ ● ✽ ✝ ✝ ● (2,−6) ✽ OO ✽ ● ✽ ✽ ✽ ✽ ## ✽ ✽ ✽ ✽ (1,−5) ✽ ✽ OO ✽ ✽ ✽ ✇ (2,−8) OO ✽ ✇ ✝ ✽ ✝ ✇ (1,−7) OO ✝ ✇ ✇ ✝ ✝ ✝ ✝ {{✇ ✝ ✝ ✝ (2,−10) ✝ ✝ OO ● ✝ ✝ ✝ ✝ ✝ ● ✝ ✝● (1,−9) ● OO ● ✝ ● (2,−12) ● OO ● ## (1,−11) . . . . . . . . . . . . Figure 3: The quiver G for g of type G . of g attached to {i, j} ⊂ I. But we have established the result for rank two Lie algebras, so this implies (μ (Σ)) = (Σ) . {i, j} {i, j} As this is true for any i 6= j in I, we get μ (Σ) = Σ. Secondly, a T -system equation involves only a certain index i ∈ I and the indices j ∈ I with c < 0. The T -system equations do not change by reduction, in the sense that for such a j, the i j ( j) (i) powers of the factors T in the second term S of the right-hand side of (9) are the same as for l,s k,r the T -system equation associated with the rank two Lie subalgebra of g attached to {i, j}. Com- bining with our results above for the subquivers (Σ ) , we have proved that, for a general g, k {i, j} all exchange relations of cluster variables of our mutation sequence are in fact T -system equa- tions. Moreover, the mutation sequence μ replaces the initial seed Σ by a seed with the same quiver; the cluster variables, expressed in terms of the Y via (3), are truncated q-characters of the i,r same Kirillov-Reshetikhin modules, the only difference being that their spectral parameters are uniformly shifted by −2t . 16 Hence, after m applications of μ we will get the truncated q-characters (m) (i) y = χ W . i,r q k , r−2t m i,r Now taking into account [FM, Corollary 6.14], we see that if 2t m ≥ t hˇ, then all the monomials of (i) the q-character of W are lower than the level of truncation, that is, k , r−2t m i,r (i) (i) χ W = χ W . q k , r−2t m k , r−2t m i,r i,r This finishes the proof of Theorem 3.1. 4 A geometric character formula for Kirillov-Reshetikhin modules 4.1 Semi-infinite quivers with potentials − −1 − − Recall the map ψ :V → W of §2.1.3. Put V := ψ (W ), and denote by Γ the full subquiver − − − of Γ with vertex set V . Thus Γ is the same graph as G , but with a change of labelling of its vertices. (Compare for instance Figure 3 and Figure 7.) − − For every i 6= j in I with c 6= 0, and every (i, m) in V , we have in Γ an oriented cycle: i j (i, m) (13) OO ❑ (i, m− b ) ❑ ii %% . ( j, m + b ) i j (i, m + 2b + b ) s i j ii s OO yys (i, m + 2b ) i j There are 2|b |/b = |c | consecutive vertical up arrows, hence this cycle has length 2+|c |. We i j ii i j i j define a potential S as the formal sum of all these oriented cycles up to cyclic permutations, see [DWZ1, §3]. This is an infinite sum, but note that a given arrow of Γ can only occur in a finite number of summands. Hence all the cyclic derivatives of S, defined as in [DWZ1, Definition 3.1], are finite sums of paths in Γ . Let R be the list of all cyclic derivatives of S. Let J denote the two-sided ideal of the path algebra CΓ generated by R. Following [DWZ1], we now introduce Definition 4.1 Let A be the infinite-dimensional C-algebra CΓ /J. Example 4.2 Let g be of type A . Then Γ is the first graph in Figure 4. The ideal J is generated by the following 7 families of linear combinations of paths, for every m ∈ Z , <0 ((1, 2m),(2, 2m− 1),(1, 2m− 2)), ((3, 2m),(2, 2m− 1),(3, 2m− 2)), ((1, 2m),(1, 2m + 2),(2, 2m + 1)) + ((1, 2m),(2, 2m− 1),(2, 2m + 1)), ((3, 2m),(3, 2m + 2),(2, 2m + 1)) + ((3, 2m),(2, 2m− 1),(2, 2m + 1)), ((2, 2m− 1),(1, 2m− 2),(1, 2m)) + ((2, 2m− 1),(2, 2m + 1),(1, 2m)), ((2, 2m− 1),(3, 2m− 2),(3, 2m)) + ((2, 2m− 1),(2, 2m + 1),(3, 2m)), ((2, 2m + 1),(1, 2m),(2, 2m− 1)) + ((2, 2m + 1),(3, 2m),(2, 2m− 1)). 17 (2,−1) (2,−1) OO ❑ OO ❑ s ❑ ❑ s ❑ ❑ ❑ ❑ s ❑ ❑ yys %% %% (1,−2) (3,−2) (2,−3) (1,−3) OO ❑ OO OO OO ❑ s s s ❑ s s s ❑ s s s s s s %% yys yys yys (2,−3) (1,−5) (2,−5) OO ❑ OO ❑ OO ❑ ❑ ❑ ❑ s ❑ ❑ ❑ s ❑ ❑ ❑ ❑ ❑ ❑ yys %% %% %% (1,−4) (3,−4) (2,−7) (1,−7) OO ❑ OO OO OO s s s ❑ s s s ❑ s s s s s s %% yys yys yys (2,−5) (1,−9) (2,−9) ✇ ● ✇ ● ✇ ● ✇ . . . {{✇ ## . . . . . . (1,−6) . (3,−6) Figure 4: The quivers Γ for g of type A and B . 3 2 Here, using the fact that there is at most one arrow between two vertices of Γ , we have denoted unambiguously paths by sequences of vertices. Thus ((1, 2m),(2, 2m− 1),(1, 2m− 2)) denotes the path of length 2 starting at (1, 2m), passing by (2, 2m− 1) and ending in (1, 2m− 2)). Also, for m = −1, the third and fourth linear combinations of paths reduce respectively to the single paths ((1,−2),(2,−3),(2,−1)) and ((3,−2),(2,−3),(2,−1)). Example 4.3 Let g be of type B . Then Γ is the second graph of Figure 4. The ideal J is generated by the following 4 families of linear combinations of paths, for every m ∈ Z , <0 ((1, 2m− 1),(2, 2m− 3),(1, 2m− 5)), ((1, 2m− 1),(1, 2m + 3),(2, 2m + 1)) + ((1, 2m− 1),(2, 2m− 3),(2, 2m− 1),(2, 2m + 1)), ((2, 2m− 3),(1, 2m− 5),(1, 2m− 1)) + ((2, 2m− 3),(2, 2m− 1),(2, 2m + 1),(1, 2m− 1)), ((2, 2m + 1),(2, 2m + 3),(1, 2m + 1),(2, 2m− 1)) + ((2, 2m + 1),(1, 2m− 1),(2, 2m− 3),(2, 2m− 1)). For m = −1 and m = −2 the second linear combinations of paths reduce respectively to the single paths ((1,−3),(2,−5),(2,−3),(2,−1)) and ((1,−5),(2,−7),(2,−5),(2,−3)). For m = −1 the fourth linear combination of paths reduces to the single path ((2,−1),(1,−3),(2,−5),(2,−3)). 4.2 F -polynomials of A-modules Let M be a finite-dimensional A-module, and let e ∈ N be a dimension vector. Let Gr (M) be the variety of submodules of M with dimension vector e. This is a projective complex variety, and we denote by χ(Gr (M)) its Euler characteristic. Following [DWZ2], consider the polynomial i,r F = χ(Gr (M)) v (14) M e ∑ ∏ i,r − − V (i,r)∈V e∈N 18 − in the indeterminates v ((i, r) ∈ V ), called the F -polynomial of M. Note that, for Gr (M) i,r e to be nonempty, one must take e between 0 and the dimension vector of M (componentwise). Moreover, if e = 0 or e = dim(M), the variety Gr (M) is just a point, so F is a monic polynomial e M with constant term equal to 1. In the sequel, we shall evaluate the variables of the F -polynomials at the inverses of the vari- ables A introduced in (6), namely: i,r −1 −1 −1 v := A = Y Y Y Y Y Y Y Y . (15) i,r j,r j,r−1 j,r+1 j,r−2 j,r j,r+2 i,r ∏ ∏ ∏ i,r−d i,r+d i i j: c =−1 j: c =−2 j: c =−3 ji ji ji 4.3 Generic kernels Suppose that X and Y are A-modules such that Hom (X,Y ) is finite-dimensional. Assume also that there exists f ∈ Hom (X,Y ) such that Ker( f ) is finite-dimensional. Then, there is an open dense e e subset O of Hom (X,Y ) such that the kernels of all elements of O are finite-dimensional. More- over, since the map sending a homomorphism f to the F -polynomial of Ker( f ) is constructible (see [Pa, §2]), O contains an open dense subset O of Hom (X,Y ) such that the F -polynomials of the kernels of all elements of O coincide. We shall say that an element of O is a generic homomorphism from X to Y . Let us denote by S the one-dimensional A-module supported on (i, m) ∈ V . Let I be the i,m i,m (infinite-dimensional) injective A-module with socle isomorphic to S . The C-vector space I i,m i,m has a basis indexed by classes modulo J of paths in Γ with final vertex (i, m). In particular, for every k ≥ 0 we have in Γ a path ((i, m− kb ),(i, m− (k− 1)b ),...,(i, m)) (16) ii ii of length k from (i, m − kb ) to (i, m), whose class modulo J is nonzero. Thus the (i, m − kb )- ii ii component of the dimension vector of I is nonzero, and it follows that i,m Hom (I , I ) 6= 0, ((i, m) ∈ V , k ≥ 0). (17) A i,m i,m−kb ii More precisely, Hom (I , I ) has finite dimension equal to the (i, m − kb )-component of A i,m i,m−kb ii ii the dimension vector of I . The next Lemma will be proven in §4.5.3. i,m Lemma 4.4 There exists f ∈ Hom (I , I ) with Ker( f ) finite-dimensional. A i,m i,m−kb ii Because of this lemma, the following definition makes sense. (i) Definition 4.5 Let K be the kernel of a generic A-module homomorphism from I to I . i,m i,m−kb ii k,m (i) Example 4.6 Figure 5 and Figure 6 show the structure of some modules K in type A . Our k,m convention for displaying these quiver representations is the following. We only keep the vertices of Γ whose corresponding vector space is nonzero, and the arrows whose corresponding linear map is nonzero. Moreover, in these small examples, almost all vertices carry a vector space of (2) dimension 1. The only exception is the module K in Figure 6, whose vertex (2,−3) carries a 2,−3 vector space of dimension 2. The maps associated with the arrows incident to this vertex have the following matrices α = β = γ = , δ = ε = κ = ( 0 1). 19 (3,−2) (2,−1) (1,−2) ❑ ❑ s s ❑ ❑ s s ❑ ❑ ❑ ❑ s s s s ❑ ❑ yys yys %% %% (2,−3) (1,−2) (3,−2) (2,−3) ❑ ❑ s ❑ s ❑ s ❑ s ❑ ❑ ❑ s s s ❑ s ❑ yys %% yys %% (1,−4) (2,−3) (3,−4) (1) (2) (3) Figure 5: The modules K , K , K for g of type A . 1,−4 1,−3 1,−4 (3,−2) (2,−1) (1,−2) OO OO ❑ OO ❑ s s ❑ ❑ ❑ ❑ s s s s ❑ ❑ s s ❑ ❑ yys yys %% %% (2,−3) (1,−2) (3,−2) (2,−3) OO OO ❑ OO OO ❑ s ❑α s ❑ ❑ ❑ s s s ❑ s ❑ s ❑ s ❑ yy %% yys %% (1,−4) (3,−4) (2,−3) (1,−4) (3,−4) OO OO ❑ ❑ OO s s ❑ ε ❑ s s ❑ ❑ s s ❑ ❑ s s ❑ ❑ yys yys %% %% (2,−5) (1,−4) (3,−4) (2,−5) ❑ ❑ s ❑ s ❑ s ❑ s ❑ ❑ ❑ s s s ❑ s ❑ yys %% yys %% (1,−6) (2,−5) (3,−6) (1) (2) (3) Figure 6: The modules K , K , and K for g of type A . 2,−4 2,−3 2,−4 All other arrows carry linear maps with matrix (±1), whose sign is easily deduced from the defining relations of A. It is a nice exercise to check that the modules shown in Figure 5 and Figure 6 are indeed (i) the claimed modules K (see also Example 4.7 below). For instance, one can easily see that k,m the (1,−6)-component of the dimension vector of I is equal to 1. Hence Hom (I , I ) 1,−4 A 1,−4 1,−6 (1) is of dimension 1, and K is the kernel of any nonzero homomorphism. It is also easy to 1,−4 see that the (2,−5)-component of the dimension vector of I is equal to 2. In this case we 2,−3 have a stratification of the 2-plane Hom (I , I ) with three strata of dimension 0, 1, 2. The A 2,−3 2,−5 (2) module K is the kernel of any homomorphism in the open stratum, that is, of any surjective 1,−3 homomorphism. The image of any homomorphism in the one-dimensional stratum is the unique submodule X of I with dimension vector given by 2,−5 1 if i = 2 and m = −5− 2 j for some j ∈ N, dim(X ) = i,m 0 otherwise. The kernel of such a homomorphism is infinite-dimensional. (i) Example 4.7 Let us assume that g is of type A, D, E . In this case, the modules K are closely 1,r related to the indecomposable injective modules over the preprojective algebra Λ of δ . Consider the subalgebra Λ of A generated by the images modulo J of the arrows of Γ of the form (i, m) → ( j, m − 1), for every edge between i and j in δ , and every (i, m) ∈ V . In other words, if Δ is the subquiver of Γ obtained by erasing all the vertical arrows (i, m− 2) → (i, m), then Λ is isomorphic to the quotient of CΔ by the two-sided ideal generated by the relations ((i, m),( j, m− 1),(i, m− 2)) = 0, ((i, m) ∈ V ). j: c <0 i j 20 e e Thus, Λ is a Z -graded version of Λ. We can of course regard the simple A-module S as a Λ- <0 i,r module. Let H be the injective Λ-module with socle S . Then H is finite-dimensional. More i,r i,r i,r precisely, for r ≤ 1− h, H is just a graded version of the indecomposable injective Λ-module I i,r i with socle the one-dimensional Λ-module S supported on vertex i of δ . For r > 1− h, H is a i i,r graded version of a submodule of I . Any Λ-module X can be given the structure of an A-module by letting the vertical arrows (i, m − 2) → (i, m) act by 0 on X . In particular we can regard H as a finite-dimensional A- i,r module. Then one can check that I has a unique submodule isomorphic to H , giving rise to a i,r i,r non-split short exact sequence 0 → H → I → I → 0, ((i, r) ∈ V ). i,r i,r i,r−2 (i) (i) It follows that the module K is isomorphic to H . In particular, when m ≤ 1 − h, K is a i,m 1,m 1,m graded version of the injective Λ-module I . 4.4 A geometric character formula (i) Recall the A-module K defined in §4.3. We can now state our second main result. k,r (i) Theorem 4.8 Let (i, r) ∈ V and k ∈ N. The F -polynomial of K is equal to the normalized k,r (i) truncated q-character of the Kirillov-Reshetikhin module W . More precisely, we have k, r−(2k−1)d (i) χ W = Y F , (i) ∏ i, r−(2s−1)d q i k, r−(2k−1)d i K k,r s=1 where the variables v of the F -polynomial are evaluated as in (15). i,r (i) Remark 4.9 If r ≤ d − t hˇ, then the truncated q-character of W is equal to the complete k, r−(2k−1)d q-character. Hence, Theorem 4.8 gives a geometric formula for the q-character of any Kirillov- Reshetikhin module (up to a spectral shift). Remark 4.10 If M and N are two finite-dimensional A-modules, then F = F F [DWZ2, M⊕N M N (i) Proposition 3.2]. It follows immediately that, replacing in Theorem 4.8 the module K by a k,r direct sum of such modules, we obtain a similar geometric character formula for arbitrary ten- sor products of Kirillov-Reshetikhin modules. In particular, we get a geometric formula for the standard modules, which are isomorphic to tensor products of fundamental modules. Remark 4.11 Let g be of type A, D, E . Let V and W be finite-dimensional vector spaces graded − • by V . In [N1] (see also [N4]), Nakajima has introduced a graded quiver variety L (V, W) and has endowed the sum of cohomologies ∗ • H (L (V, W)) with the structure of a standard U (bg)-module, with highest weight encoded by W. It was proved by Lusztig (in the ungraded case), and by Savage and Tingley (in the graded case), that L (V, W) is homeomorphic to a Grassmannian of submodules of an injective module over the graded pre- (i) projective algebra (see [Le2, §2.8]). Therefore, using the description of K given in Example 4.7, 1,r we see that the varieties   M ⊕a i,r (i)   Gr K 1,r (i,r) 21 (2,−1) OO ❁ (2,−3) OO ✆ ❁ (1,−4) OO ✆ ✆ ✆ ✆ ✆ ✆ ✆ ✆ (2,−5) ✆ ✆ OO ❍ ✆ ✆ ✆ ✆ ✆ ✆ ✆❍ ✆ ❍ (1,−6) ✆ ❍ OO ✾ ❍ ✾ ✆ ✆ ❍ ✾ ❍ (2,−7) ✾ OO ✾ ❍ ✾ ✾ ✾ ✾ ## ✾ ✾ ✾ ✾ (1,−8) ✾ ✾ OO ✾ ✾ ✾ ✈ ✾ ✈ (2,−9) OO ✾ ✈ ✆ ✾ ✈✾ ✆ ✈ (1,−10) OO ✆ ✈ ✈ ✆ ✆ ✆ ✆ {{✈ ✆ ✆ ✆ (2,−11) ✆ ✆ OO ❍ ✆ ✆ ✆ ✆ ✆ ❍ ✆ ✆❍ (1,−12) ❍ OO ❍ ✆ ❍ (2,−13) ❍ OO ## (1,−14) . . . . . . . . . . . . Figure 7: The quiver Γ for g of type G . involved in our geometric q-character formula for standard modules in the simply laced case are homeomorphic to certain Nakajima varieties L (V, W). Here, the multiplicities a are the di- i,r mensions of the graded components of W, and we assume that a = 0 if r > 1− h. Similarly the i,r graded dimension of V is encoded by the dimension vector e. Example 4.12 Let g be of type A . We have −1 −1 −1 −1 −1 −1 v = Y Y Y , v = Y Y Y Y , v = Y Y Y . 1,r 2,r 2,r 1,r 3,r 3,r 2,r 1,r−1 1,r+1 2,r−1 2,r+1 3,r−1 3,r+1 We continue Example 4.6. The submodule structure of the A-modules displayed in Figure 5 is very (i) simple. Indeed, in this case, all the nonempty varieties Gr (K ) are reduced to a single point, and k,r their Euler characteristics are equal to 1. Therefore the F -polynomial reduces to a generating (i) polynomial for the dimension vectors of the (finitely many) submodules of K . This yields the k,r 22 following well known formulas for the q-characters of the fundamental modules: χ (L(Y )) = Y (1 + v + v v + v v v ) q 1,−5 1,−5 1,−4 1,−4 2,−3 1,−4 2,−3 3,−2 −1 −1 −1 = Y + Y Y + Y Y + Y , 1,−5 2,−4 3,−3 1,−3 2,−2 3,−1 χ (L(Y )) = Y (1 + v + v v + v v + v v v q 2,−4 2,−4 2,−3 1,−2 2,−3 2,−3 3,−2 1,−2 2,−3 3,−2 + v v v v ) 1,−2 2,−3 3,−2 2,−1 −1 −1 −1 −1 −1 −1 = Y + Y Y Y + Y Y + Y Y + Y Y Y + Y , 2,−4 1,−3 3,−3 3,−3 1,−3 2,−2 2,−2 1,−1 3,−1 1,−1 3,−1 2,0 Similarly, the A-modules shown in Figure 6 give the following Kirillov-Reshetikhin q-characters: χ (L(Y Y )) = Y Y (1 + v (1 + v + v + v v + v v q 1,−7 1,−5 1,−7 1,−5 1,−4 1,−6 2,−3 1,−6 2,−3 2,−3 3,−2 + v v v + v v v + v v v v 1,−6 2,−3 2,−5 1,−6 2,−3 3,−2 1,−6 2,−3 2,−5 3,−2 + v v v v v )), 1,−6 2,−3 2,−5 3,−2 3,−4 χ (L(Y Y )) = Y Y (1 + v (1 + v + v + v + v v + v v q 2,−6 2,−4 2,−6 2,−4 2,−3 1,−2 2,−5 3,−2 1,−2 2,−5 1,−2 3,−2 + v v + v v v + v v v + v v v 2,−5 3,−2 1,−2 2,−5 3,−2 1,−2 2,−5 1,−4 1,−2 3,−2 2,−1 + v v v + v v v v + v v v v 2,−5 3,−2 3,−4 1,−2 2,−5 3,−2 1,−4 1,−2 2,−5 3,−2 2,−1 + v v v v + v v v v v 1,−2 2,−5 3,−2 3,−4 1,−2 2,−5 3,−2 1,−4 2,−1 + v v v v v + v v v v v 1,−2 2,−5 3,−2 1,−4 3,−4 1,−2 2,−5 3,−2 3,−4 2,−1 + v v v v v v + v v v v v v v )), 1,−2 2,−5 3,−2 1,−4 2,−1 3,−4 1,−2 2,−5 3,−2 1,−4 2,−1 3,−4 2,−3 We omit the q-characters χ (L(Y )) and χ (L(Y Y )), since they are readily obtained from q 3,−5 q 3,−5 3,−7 χ (L(Y )) and χ (L(Y Y )) via the symmetry 1 ↔ 3. q 1,−5 q 1,−5 1,−7 Example 4.13 Let g be of type G , with the long root being α . The quiver Γ is shown in 2 1 (1) (2) Figure 7. The modules K and K with r ≤ −10 and s ≤ −11 have dimension 10 and 6, 1,r 1,s (1) (2) (1) respectively. For instance, K and K are represented in Figure 8. In the module K the 1,−10 1,−11 1,−10 vector space sitting at vertex (2,−7) has dimension 2 (all other spaces have dimension 1). The maps incident to this space are given by the following matrices (see Figure 8): 0 1 α = , β = ( 1 0), γ = , γ = ( 0 1). 1 0 The corresponding fundamental modules have dimension dim L(Y ) = 15, dim L(Y ) = 7. 1,−13 2,−12 (1) (2) The Grassmannians of submodules of K and K are in this case again all reduced to points, 1,−10 1,−11 and the formula of Theorem 4.8 amounts to an enumeration of the dimension vectors of all sub- modules. This gives χ (L(Y )) = Y (1 + v (1 + v (1 + v (1 + v + v + v v q 1,−13 1,−13 1,−10 2,−7 2,−9 1,−6 2,−11 1,−6 2,−11 + v v + v v + v v v + v v v 1,−6 2,−3 2,−11 1,−8 1,−6 2,−11 2,−3 1,−6 2,−11 1,−8 + v v v v (1 + v (1 + v (1 + v ))))))), 1,−6 2,−11 2,−3 1,−8 2,−5 2,−7 1,−4 χ (L(Y )) = Y (1 + v (1 + v (1 + v (1 + v (1 + v (1 + v )))))), q 2,−12 2,−12 2,−11 1,−8 2,−5 2,−7 1,−4 2,−1 where, following (15), we have −1 −1 −1 −1 v = Y Y Y Y Y , v = Y Y Y . 1,r 2,r+2 2,r 2,r−2 2,r 1,r 1,r+3 1,r−3 2,r+1 2,r−1 23 (2,−3) (2,−1) OO ❃ ✝ ❃ (1,−4) ❃ ✝ ✆ ✝ ✆ ❃ ✝ ✆ ❃ ✆ ❃ ✝ ✆ ❃ (2,−5) ✝ ✆ OO ● ✝ ✆ ✝ (1,−4) ✝ ✆ ✝ ● ✆ ′ ✄ γ ● (1,−6) ✄ ✆ ● ✽ ● ● ✄ ✆ (2,−5) ● ✄ ✽ ✆ OO ■ ✄ ✆ ● ✽ ● ● ■✄ (2,−7) ● ✄ ■ ✽ ● ✽ OO ✾ ● ● ✄ ✽ ✾ ■ ● ✄ ■ ● ■ ✽ ✾ ## ■ ✽ ✾ ■ ✽ γ ✾ (1,−8) ✽ (2,−7) ■ ✽ ✾ ■ ✇ ■ ✽ ✾ ✽ ✾ ■ ✾ ■ ✇ $$ (2,−9) ✾ ✇ (1,−8) OO ✾ ✇ ✾ s ✇ s ✇  s ✇ s (1,−10) ✇ s ✇ s {{✇ (2,−11) s yys (2,−11) (1) (2) Figure 8: The modules K and K for g of type G . 1,−10 1,−11 Remark 4.14 Assuming Theorem 4.8, we can easily calculate the dimension vectors of the A- (i) modules K for r ≤ d − t hˇ. Indeed, by [FM, Lemma 6.8], the lowest monomial of χ (Y ) i q i,r−d 1,r i −1 is equal to Y , where ν is the involution of I defined by w (α ) = −α . Denote by 0 i ν(i) ν(i),r−d +t hˇ (i) (i) d (K ) the dimension vector of K . Then, we have j,s 1,r 1,r (i) d (K ) j,s 1,r −1 Y = Y v , i,r−d i ∏ ν(i),r−d +t hˇ j,s ( j,s)∈V (i) and using (15), this equation determines the numbers d (K ). In particular, if we introduce the j,s 1,r (i) ungraded dimension vector (d (i)) of K by 1,r (i) d (i) := d (K ), (r ≤ d − t hˇ), j ∑ j,s i 1,r we can deduce from this the nice formula d (i)α = β, (18) j j ∑ ∑ i, j∈I β∈Φ >0 where Φ is the set of positive roots of g. This can be observed in Figure 5 and Figure 8 (see >0 also §6.4, §6.5, §6.6, §6.7 below). When g is of type A, D, E , as explained in Remark 4.7 the (i) modules K are graded versions of the indecomposable injective modules over the preprojective 1,r algebra Λ, and formula (18) recovers a well known property of Λ. 24 4.5 Proof of the theorem The proof relies on Theorem 3.1, and on the categorification of cluster algebras by means of quivers with potentials, developed by Derksen, Weyman and Zelevinsky [DWZ1, DWZ2]. This categorification provides (among other things) a description of cluster variables in terms of Grass- mannians of submodules, which will be our key ingredient. An important additional result will be borrowed from Plamondon [Pl2]. 4.5.1 F -polynomials and g-vectors of cluster variables − − Recall the cluster algebra A of §2.2.1, with initial seed (z , G ). Following [FZ3, (3.7)], define −1 − yb := z z , ((i, r) ∈ W ). (19) i,r ∏ j,s ∏ j,s (i,r)→( j,s) ( j,s)→(i,r) Here the first (resp. second) product is over all outgoing (resp. incoming) arrows at the vertex (i, r) of the graph G . The following result is similar to [HL1, Lemma 7.2]. Lemma 4.15 After performing in (19) the change of variables (3), there holds −1 − yb = A , ((i, r) ∈ W ), i,r i, r−d where the Laurent monomials A are given by (6). i,r Proof — Using the definition of the quiver G , we can rewrite (19) as z z j, r+b +d −d i, r+b i j j i ii yb = , i,r ∏ z z i, r−b j, r−b +d −d ii i j j i j6=i where the product is over all j’s such that c 6= 0. Here we use the convention that z = 1 for i j i,s every (i, s) with s > 0. Using the change of variables (3), we obtain −1 −1 yb = Y Y Y Y ··· Y . i,r r, r−d +b +d r, r−d +b +3d r, r−d −b −d i,r ∏ i i j j i i j j i i j j i, r−b ii j6=i; c 6=0 i j The result then follows by comparison with (6), if we notice again that b + d = c + 1 because i j j ji of (1). ✷ In [FZ3] Fomin and Zelevinsky attach to every cluster variable x of A a polynomial F with − (W ) integer coefficients in the set of variables b y = {yb | (i, r) ∈ W }, and a vector g ∈ Z , such i,r x that [FZ3, Corollary 6.3] x = z F (b y). (20) Note that A has no frozen cluster variables, so there is no denominator in (20). The polynomial F and the integer vector g are called the F -polynomial and g-vector of the cluster variable x, x x respectively. We refer the reader to [FZ3] for their definition. On the other hand, it follows from the theory of q-characters that for every simple U (g)- − − module L(m) in the category C , the truncated q-character χ (L(m)) can be written as χ (L(m)) = mP , (21) −1 − where P is a polynomial with integer coefficients in the variables {A | (i, r) ∈ W }. More- i, r−d over, P has constant term 1. 25 Now, by the proof of Theorem 3.1, among the cluster variables of A , we find all the truncated q-characters of the Kirillov-Reshetikhin modules of C . These are of the form L(m) with k−1 (i) m = m := Y , ((i, r) ∈ W , r + (k− 1)b ≤ 0). (22) i, r+ jb ii ∏ ii k,r j=0 Proposition 4.16 The g-vector of the truncated q-character of the Kirillov-Reshetikhin module (i) (i) W = L m , considered as a cluster variable of A , is given by k,r k,r  1 if ( j, s) = (i, r), g = −1 if ( j, s) = (i, r + kb ) and r + kb ≤ 0, j,s ii ii 0 otherwise. (i) Proof — Write for short m = m , and denote by x the cluster variable χ (L(m)). Then, compar- k,r ing (20) with (21), we have −1 g P = m z F , m x where, by Lemma 4.15, P and F are polynomials in the same variables m x −1 yb = A . i,r i,r−d −g Since P has constant term 1, it follows that mz is a monomial in the variables yb which divides m i,r the F -polynomial F . But, by [FZ3, Proposition 5.2], F is not divisible by any yb . So, using (3), x x i,r i,r z = m = , i, r+kb ii where as above, we set z = 1 if s > 0. ✷ i,s 4.5.2 Truncated algebras Let ℓ ∈ Z . Let Γ be the full subquiver of Γ with set of vertices <0 − − V := {(i, m) ∈ V | m ≥ ℓ}. Let S be the corresponding truncation of the potential S, that is, S is defined as the sum of all ℓ ℓ − − cycles in S which only involve vertices of V . Let J denote the two-sided ideal of CΓ generated ℓ ℓ by all cyclic derivatives of S . Finally, define the truncated algebra at height ℓ as A := CΓ /J . ℓ ℓ Proposition 4.17 For every ℓ we have: (i) the algebra A is finite-dimensional; (ii) the quiver with potential (Γ , J ) is rigid. Proof — The proof is similar to [DWZ1, Example 8.7]. Let π : CΓ → A be the natural projec- tion. To prove (i), we show that A is spanned by the images under π of a finite number of paths. The arrows of Γ are of two types: 26 (a) the vertical arrows of the form (i, m) → (i, m + b ); ii (b) the oblique arrows of the form (i, m) → ( j, m + b ) provided c < 0. i j i j Let us say that a path from (i, m) to ( j, s) in Γ is going up (resp. down) if m < s (resp. m > s). Note that all vertical arrows go up and all oblique arrows go down. Each oblique arrow of the boundary of Γ belongs to a single cycle of the potential S , and each interior oblique arrow belongs to exactly two cycles. Therefore each interior oblique arrow gives rise to a “commutativity relation” in A : π(( j, m + b ),(i, m + 2b ),(i, m + 2b + b ),...,(i, m− b ),(i, m)) ji ji ji ii ii = −π(( j, m + b ),( j, m + b + b ),... ,( j, m− b − b ),( j, m− b ),(i, m)) ji ji j j ji j j ji The path in the left-hand side consists of an oblique arrow followed by |c | vertical arrows, while i j the right-hand side has |c | vertical arrows followed by an oblique arrow. Let p be a path in Γ ji with origin (i, m). Using only the above type of commutativity relations, we can bring a number of vertical arrows to the front of p and write π( p) = π( p )π( p ), 2 1 where p is a path with origin (i, m) consisting only of vertical arrows, and p is a path satisfying 1 2 the following property: if q is a maximal factor of p containing only vertical arrows, then q is preceded by at least one oblique arrow, say ( j, s) → (k, s + b ), and q contains less than |c | jk k j arrows. Hence q can be non trivial only if |c | > 1. k j In particular in the simply laced case, then p contains only oblique arrows. In that case, we can immediately conclude that all arrows of p go up and all arrows of p go down, so the lengths 1 2 of p and p are both bounded by ℓ, and therefore A is finite-dimensional. 1 2 ℓ Otherwise, if q is non trivial and p contains other vertical arrows after q, then q needs to be followed by at least two oblique arrows. Indeed, using the same notation as above, q consists of N vertical arrows of the form (k, r) → (k, r + b ) with 1 ≤ N < |c |. Now, by (1), the inequality kk k j |c | > 1 implies d = 1 and d = |b |. Let (k, t) → (l, t + b ) be the first arrow coming after q. k j k j k j kl Then, since d = 1 we have |c | = 1. If this oblique arrow is followed by a vertical one (l, t + k lk b ) → (l, t + b + b ), then we can use the commutativity relation and bring it, together with all kl kl ll the vertical arrows possibly following it, on top of q. In this way, we replace q by a vertical path q followed by two consecutive oblique arrows. One then easily checks by inspection that the subpath of p containing q together with the oblique arrow preceding it and the oblique arrow following it, is going down. Therefore, by induction, p can be factored into a product of paths, each of them of length less than t + 2, and all these paths go down (except possibly the last one, which might end with less than t vertical arrows). So again, the length of p is bounded above, and this proves (i) in all cases. To prove (ii), it is enough to show that every cycle of the form (13) is cyclically equivalent to an element of J . Up to cyclic equivalence, this cycle γ can be written with origin in (i, m). Then, we have: π(γ) = π((i, m),( j, m + b ),(i, m + 2b ),(i, m + 2b + b ),...,(i, m− b ),(i, m)) i j i j i j ii ii = π((i, m),( j, m + b ),( j, m + b + b ),...,( j, m− b − b ),( j, m− b ),(i, m)) i j i j j j i j j j i j = π((i, m),(i, m + b ),...,(i, m− 2b − b ),(i, m− 2b ),( j, m− b ),(i, m)), ii i j ii i j i j and the last path is cyclically equivalent to ((i, m− 2b ),( j, m− b ),(i, m),(i, m + b ),... ,(i, m− 2b − b ),(i, m− 2b )). i j i j ii i j ii i j 27 This cycle is nothing else than γ shifted vertically up by −2b . Hence, iterating this process, we i j can replace, modulo J and cyclic equivalence, any cycle γ of the form (13) by a similar cycle ′ − ′ γ sitting at the top boundary of Γ . Now the upper oblique arrow of γ does not belong to any other cycle, so it gives rise to a zero relation in A . In other words, γ is cyclically equivalent to an element of J . This proves (ii). ✷ Remark 4.18 In the simply laced case and when |ℓ| is less than the Coxeter number, the algebra A arises as the endomorphism algebra of a (finite-dimensional) rigid module over the preprojective algebra Λ associated with δ , and appears in the works of Geiss, Schro ¨ er and the second author (see [GLS1, GLS2]). This gives another proof of Proposition 4.17 (i) in this case. 4.5.3 Proof of Lemma 4.4 and Theorem 4.8 (i) − − Let (i, r) ∈ V and k ∈ N. By Theorem 3.1, the truncated q-character χ W is a k, r−(2k−1)d cluster variable x of A . By Proposition 4.16, the g-vector of x is given by 1 if ( j, s) = (i, r− 2kd + d ), i i g = −1 if ( j, s) = (i, r + d ), (23) j,s i 0 otherwise. − − − − Note that, since (i, r) ∈ V , we have (i, r + d ) ∈ W . For ℓ < 0, let W := ψ(V ), and put ℓ ℓ − − − − − z ={z | (i, r) ∈ W }. We denote by G the same quiver as Γ , but with vertices labelled by W . i,r ℓ ℓ ℓ ℓ ℓ Clearly, the cluster variable x is a Laurent polynomial in the variables of z for some ℓ ≪ 0, and − − can be regarded as a cluster variable of the cluster algebra A defined by the initial seed z , G . ℓ ℓ By Proposition 4.17 (ii), we can apply the theory of [DWZ1, DWZ2] and deduce that the F - polynomial of x coincides with the polynomial F associated with a certain A -module M. In order M ℓ to identify this module, we apply [Pl2, Remark 4.1], which states in our setting that M is the kernel of a generic element of the homomorphism space between two injective A -modules corresponding to the negative and positive components of the g-vector of x. More precisely, let us denote by S i,m − ℓ the one-dimensional A -module supported on (i, m) ∈ V . Let I be the injective A -module ℓ ℓ i,m with socle isomorphic to S . Then, using (23) and taking into account the change of labelling i,m − − ℓ ℓ ψ :V → W given by (2), we get that M is the kernel of a generic element of Hom (I , I ). ℓ i,r i,r−kb ℓ ℓ ii Finally we can identify M with the kernel of a generic homomorphism between injective A- modules. Indeed, for m < ℓ < 0 we have a natural projection A → A whose kernel is generated m ℓ − − − − by all arrows of Γ starting or ending at a vertex v ∈ V \ V . This induces for every (i, r) ∈ V m m ℓ ℓ ℓ m ℓ an embedding I → I , and we can regard the A-module I as the direct limit of I along (i,r) (i,r) (i,r) (i,r) these maps. Since F is independent of ℓ ≪ 0, we see that M is also the kernel of a generic (i) (i) element of Hom (I , I ), that is, M = K . In particular K is finite-dimensional. This A i,r i,r−kb ii k,r k,r proves Lemma 4.4 and finishes the proof of Theorem 4.8. Remark 4.19 Using the same formula as (14), we can attach to the infinite-dimensional A-module I a formal power series F in the variables v . This series also has an interpretation in terms of i,m I j,r i,m quantum affine algebras. Indeed, by [HJ], the category of finite-dimensional U (bg)-modules can be seen as a subcategory of a category O of (possibly infinite-dimensional) representations of a Borel subalgebra of U (bg). The q-character morphism can be extended to the Grothendieck ring of O (the target ring is also completed). This category contains distinguished simple representations − ∗ called negative fundamental representations L (i ∈ I, a ∈ C ) [HJ, Definition 3.7]. Denote by i,a 28 − − χ (L ) the normalized q-character of L , that is, its q-character divided by its highest weight i,a i,a −1 monomial. This normalized q-character is a formal power series in the variables A [HJ, Theo- j,b rem 6.1], and it is obtained as a limit of normalized q-characters of Kirillov-Reshetikhin modules. It is not difficult to deduce from Theorem 4.8 and Remark 4.9 that, for m ≤ d − t hˇ, χ L = F . q I m−d i,m i, q i This is the first geometric description of the q-character of these negative fundamental representa- tions. 5 Beyond Kirillov-Reshetikhin modules 5.1 Grothendieck rings − − Let us consider again the cluster algebra A , with initial seed Σ = (z , G ) whose cluster variables z are given by (3). The Laurent phenomenon for cluster algebras implies that A is a subring of i,r ±1 − Z[Y | Y ∈ Y ]. The following theorem gives the precise relationship between A and the i,r i,r Grothendieck ring of the category C . Theorem 5.1 The cluster algebra A is equal to the image of the injective ring homomorphism ±1 − − − from K (C ) to Z[Y | Y ∈ Y ] given by [L(m)] 7→ χ (m) (see Proposition 3.10). Hence A is 0 i,r i,r isomorphic to the Grothendieck ring of C . − − − Proof — Let R denote the image of the homomorphism [L(m)] 7→ χ (m). By [FR], K (C ) is − − the polynomial ring in the classes of the fundamental modules of C , hence R is the polynomial − − ring in the truncated q-characters χ (Y ) (Y ∈ Y ). By Theorem 3.1, A contains all these i,r i,r fundamental truncated q-characters, hence A contains R . To prove the reverse inclusion, we will use a description of the image of the q-character ho- momorphism as an intersection of kernels of screening operators [FR, FM]. To do this, we need to work with complete (i.e. untruncated) q-characters. So let us consider as in §3.2.2 the larger set of variables Y. Following [FR, §7.1], for every i ∈ I, we have a linear operator S from the ring ±1 Z[Y | Y ∈ Y] to a certain free module Y over this ring, which satisfies the Leibniz rule i,r i i,r ±1 S (xy) = x S (y) + y S (x), (x, y ∈ Z[Y | Y ∈ Y]). i i i i,r i,r ±1 It was conjectured in [FR] and proved in [FM] that an element of Z[Y | Y ∈ Y] is a polynomial i,r i,r χ (Y ) (Y ∈ Y) if and only if it belongs to in the q-characters q i,r i,r Ker S . i∈I Let us now introduce an auxilliary cluster algebra A . It is defined using the same initial seed − − ′ (z , G ) as A , but the initial variables of A are given by the following modification of (3) z := Y , i, r+kb +2t hˇ i,r ∏ ii k≥0, r+kb ≤0 ii in which the spectral parameters are all shifted upwards by 2t hˇ. By Theorem 3.1, if we apply to this initial seed of A the sequence of mutations μ repeated hˇ times, we will obtain a new seed 29 ′ − ′ − Σ with the same quiver G . Moreover, the cluster variable of Σ sitting at vertex (i, r) ∈ W is (i) nothing else than the complete q-character χ (W ). k ,r i,r Consider a cluster variable x of A . By definition, x is obtained from Σ by a finite sequence of mutations μ . We want to show that x belongs to R . By Theorem 3.1, all cluster variables of Σ belong to R , so by induction on the length, we may assume that the last exchange relation of μ is of the form xy = M + M , 1 2 where y is a cluster variable of A , M and M are cluster monomials of A , and y, M , M belong 1 2 1 2 − ′ ′ to R . Let us apply the same sequence of mutations μ in the cluster algebra A to the seed Σ . The last exchange relation will be of the form ′ ′ ′ ′ x y = M + M , 1 2 ′ ′ ′ − where y , M , M are polynomials in the complete fundamental q-characters χ (Y ) (Y ∈ Y ). q i,r i,r 1 2 ′ ′ ′ ′ Moreover, x , y , M , M give back x, y, M , M by application of the truncation ring homomor- 1 2 1 2 ′ ′ ′ ′ ′ phism. By the Laurent phenomenon [FZ1] in the cluster algebra A , we know that x , y , M , M 1 2 are Laurent polynomials in the variables of Y. Since S is a derivation, we have ′ ′ ′ ′ ′ ′ ′ ′ S (x y ) = x S (y ) + y S (x ) = S (M ) + S (M ), i i i i i 1 2 ′ ′ ′ ′ ′ hence S (x ) = 0 because S (y ) = S (M ) = S (M ) = 0. It follows that x is annihilated by all the i i i i 1 2 ′ − screening operators, so x is a polynomial in the q-characters χ (Y ) (Y ∈ Y ). This implies that q i,r i,r − − − x is a polynomial in the truncated q-characters χ (Y ) (Y ∈ Y ), that is, x ∈ R . ✷ i,r i,r 5.2 Conjectures 5.2.1 Cluster monomials In view of Theorem 5.1, it is natural to formulate some conjectures. Following [Le1], let us say that a simple U (g)-module S is real if S⊗ S is simple. Conjecture 5.2 In the above identification of the cluster algebra A with the ring of truncated q-characters of C , the cluster monomials get identified with the truncated q-characters of the real simple modules of C . When g is of type A, D, E , Conjecture 5.2 is essentially equivalent to [HL1, Conjecture 13.2]. But the initial seed used here is different and allows a direct connection between cluster expansions and (truncated) q-characters. 5.2.2 Geometric q-character formulas Using the methods and tools of §4, we can translate Conjecture 5.2 into a new conjectural geomet- ric formula for the (truncated) q-character of a real simple module of C . Let m be a dominant monomial in the variables Y ∈ Y . Using the change of variables (3), i,r which we can express as i,r Y = , ((i, r) ∈ W ), i,r i,r+b ii (where we understand z = 1 if s > 0), we can rewrite i,s g (m) g(m) i,r m = z := z . i,r (i,r)∈W 30 (2,−1) s ❑ s ❑ s ❑ yys %% (1,−2) (3,−2) ❑ OO ❑ s ❑ s ❑ s %% yys (2,−3) OO (3,−4) yys (2,−5) yys (1,−6) Figure 9: The A-module K(m) for m = Y Y in type A . 1,−7 2,−4 3 (W ) Let us call the integer vector g(m) ∈ Z the g-vector of L(m). Following §4.3, let us attach to m the A-module K(m) defined as the kernel of a generic A-module homomorphism from the injective − + A-module I(m) to the injective A-module I(m) , where M M ⊕g (m) ⊕|g (m)| + i,r − i,r I(m) = I , I(m) = I . i,r−d i,r−d i i g (m)>0 g (m)<0 i,r i,r Finally define the F -polynomial F of K(m) as in §4.2. We can now state the following conjec- K(m) tural generalization of Theorem 4.8. Conjecture 5.3 Suppose that L(m) is an irreducible real U (bg)-module in C . Then the truncated q-character of L(m) is equal to χ (L(m)) = mF , K(m) where the variables v of the F -polynomial are evaluated as in (15). i,r Example 5.4 Let g be of type A . Take m = Y Y . We have 3 1,−7 2,−4 + − I(m) = I ⊕ I , I(m) = I ⊕ I . 1,−8 2,−5 1,−6 2,−3 The module K(m) has dimension 7 and is displayed in Figure 9. Using for instance the fact that L(m) is a minimal affinization (in the sense of [C]), we can compute its q-character. We find: χ (L(Y Y )) = Y Y (1 + v + v + v v + v v + v v q 1,−7 2,−4 1,−7 2,−4 1,−6 2,−3 1,−6 2,−3 1,−2 2,−3 2,−3 3,−2 + v v v + v v v + v v v + v v v 1,−6 1,−2 2,−3 1,−6 2,−3 3,−2 1,−6 2,−3 2,−5 1,−2 2,−3 3,−2 + v v v v + v v v v + v v v v 1,−6 1,−2 2,−5 2,−3 1,−6 1,−2 2,−3 3,−2 1,−6 2,−5 2,−3 3,−2 + v v v v + v v v v v 1,−2 2,−5 2,−3 3,−2 1,−6 1,−2 2,−5 2,−3 3,−2 + v v v v v + v v v v v 1,−6 1,−2 2,−3 2,−1 3,−2 1,−6 2,−5 2,−3 3,−4 3,−2 + v v v v v v + v v v v v v 1,−6 1,−2 2,−5 2,−3 3,−4 3,−2 1,−6 1,−2 2,−5 2,−3 2,−1 3,−2 + v v v v v v v ), 1,−6 1,−2 2,−5 2,−3 2,−1 3,−4 3,−2 in agreement with Conjecture 5.3. 31 6 Appendix 6.1 Mutation sequence in type A We display the sequence of mutated quivers obtained from G at each step of the mutation se- quence μ . The first quiver is G , and in the next quivers the box indicates at which vertex a mutation has been performed. (2,0) (2,0) (2,0) (2,0) 99 99 OO OO OO ✉ s s ;; s s ✉ ✎ ✍ ✇ s s ✇ s s ✎ ✍ ✉ s s zz✉ s s ✇ ✍ ✎ ✍ (1,−1) (1,−1) (1,−1) (1,−1) ✎ OO ■ OO OO ■ ✎ OO $$ ✎ (2,−2) (2,−2) (2,−2) (2,−2) ✍ OO 99 OO ✉ s OO ✍ ✉ s ✎ t ✎ :: ✉ ✍ s t ✎ ✉ ✎ ✈ s zz✉ ✈ ✍ s ✈ ✎ zzt ✎ ✈ (1,−3) (1,−3) (1,−3) ✎ OO ■ OO ■ (1,−3) OO ❏ ■ OO ✎ $$ ✎ $$ ✎ (2,−4) (2,−4) (2,−4) OO (2,−4) OO t ✎ ✉ OO :: ✉ s ✎ ✈ zz✉ s ✈ s ✎ ✈ zzt s ✈ (1,−5) yys ✎ ✈ (1,−5) OO ■ ■ (1,−5) (1,−5) OO ❏ ■ OO ❑ OO ■ ❑ $$ $$ %% (2,−6) (2,−6) (2,−6) (2,−6) . t s s ✉ . . zz✉ s s t . . . . s s zzt s s . . (1,−7) yys yys . . (1,−7) (1,−7) (1,−7) (2,0) (2,0) (2,0) (2,0) (2,0) OO OO OO OO OO s s s s s s s s ✍ ✈ s s s s s ✈ s s s s ✈ s s s s zz✈ yys yys yys (1,−1) (1,−1) (1,−1) (1,−1) (1,−1) OO OO ❑ OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ✍ ❑ ❑ ❑ ❑ ❑ ❑ %% %% %% (2,−2) (2,−2) (2,−2) (2,−2) ✍ 99 OO OO OO OO s (2,−2) ✈ s s s ✍ s s 99 OO ✍ s ✈ s s s s ✈ s s ✍ s s ✍ zz✈ yys yys (1,−3) ✍ (1,−3) (1,−3) (1,−3) OO ✍ OO ❑ OO ❑ (1,−3) ❑ ❑ ✍ ❑ ❑ OO ❑ ❑ ❑ ❑ %% %% (2,−4) (2,−4) (2,−4) 99 OO OO OO ✍ (2,−4) (2,−4) s ✈ s s ✎ ✈ s s 99 OO 99 OO s ✍ ✈ s ✍ s s ✈ s ✍ ✍ ✍ s s s zz✈ yys s s ✍ ✍ s s ✍ s  s (1,−5) ✍ ✍ (1,−5) (1,−5) OO ✍ ✍ OO ❑ (1,−5) (1,−5) ❑ ✍ ✍ ✎ OO OO ❑ ✍ ✍ ❑ ✎ %% ✍ ✍ ✍ ✍ (2,−6) ✎ ✍ ✍ (2,−6) ✍ (2,−6) ✍ (2,−6) (2,−6) ✎ ✈ ✍ ✍ :: 99 99 99 s s s ✎ ✈ ✍ ✍ ✈ . ✈ s s s zz✈ . . . . ✎ ✈ ✍ s ✍ s s . ✈ s s s . . . . ✎ ✈ ✍ s ✍ s  s . . . . (1,−7) (1,−7) (1,−7) (1,−7) (1,−7) 6.2 Mutation sequence in type B We display the sequence of mutated quivers obtained from G at each step of the mutation se- quence μ . 32 (2,0) (2,0) ❏ (2,0) OO OO ❏ ✟ cc● ● ✟ $$ ✟ (2,−2) (1,−1)  ✟ oo // ✟ (2,−2) (1,−1) OO OO s t (2,−2) (1,−1) ✟ OO s t :: s t OO OO ✟ s t s t ✉ s ✟ ✉ yys zzt t s t ✉ s t ✉ yys zzt ✟✉ (1,−3) (2,−4) oo OO ❑ OO ❏ ❑ ❏ (1,−3) (2,−4) (1,−3) (2,−4) ❑ ❏ OO ❑ OO ❏ OO ▲ OO ❑ ❑ ❏ ❑ ❏ ❑ ❏ ❑ ❑ ❏ ▲ ❑ %% $$ ❑ ❏ ▲ ❑ ❑ ❏ ▲ ❑ %% $$ %% %% (2,−6) (1,−5) OO OO s t (2,−6) (1,−5) (2,−6) (1,−5) s t OO OO OO OO s t s t s t r s yys zzt s t r s s t s s t r s yys zzt yyr yys (1,−7) (2,−8) OO ❑ OO ❏ ❑ ❏ (1,−7) (2,−8) (1,−7) (2,−8) ❑ ❏ ❑ ❏ OO ❑ OO ❏ OO ▲ OO ❑ ▲ ❑ ❑ ❏ ❑ ❏ %% $$ ❑ ❏ ▲ ❑ ❑ ❏ ▲ ❑ ❑ ▲ ❑ %% $$ %% %% (2,−10) (1,−9) s (2,−10) (1,−9) (2,−10) (1,−9) s s r yys s r s r . . s r yys yyr (1,−11) . . . . . . . . (1,−11) (1,−11) . . . . . . . . (2,0) (2,0) (2,0) OO OO OO ✝ ☎ ☎ ☎ ☎ ✝ ☎ ☎ ✝ ☎ ☎ oo oo oo ✝ ☎ ☎ (1,−1) (2,−2) (1,−1) (2,−2) (2,−2) (1,−1) ✝ ☎ ☎ OO ✺ OO OO ✼ OO OO ✾ OO ✝ ☎ ☎ ✺ ✼ ✾ ✝ ☎ ☎ ✝ ✺ ☎ ✼ ☎ ✝ ✺ ☎ ✼ ☎ ✝ ✺ // // ✺ (1,−3) (2,−4) (1,−3) (2,−4) // ✼ (1,−3) (2,−4) ✺ OO OO ✼ OO OO OO ✺ ✼ ✟ ✝ dd❍ ✺ ✼ ✝ ✟ ✾ ❍ ✺ ✼ ✝ ❍ ✟ oo ✟ (2,−6) (1,−5) oo // ✟ (2,−6) (1,−5) OO ✺ OO (2,−6) (1,−5) ✟ OO :: OO OO ✟ ✝ ✺ r s ✉ r ✟ ✉ ✺ s ✝ r s ✉ ✟ ✝ ✺ r s ✉ yyr yys ✟✉  ✺ // oo (1,−7) (2,−8) ✺ (1,−7) (2,−8) (1,−7) (2,−8) OO ✺ ▲ ❑ ▲ ❑ OO OO OO OO dd❍ ✺ ▲ ❑ ▲ ❑ ❍ ▲ ❑ ▲ ❑ ❍ ✺ ▲ ❑ ▲ ❑ ✺ ▲ ❑ ▲ ❑ ❍ %% %% %% %% // (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) r r r r r r r r r r r r yyr yyr yyr . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) OO OO OO ✄ ✠ ✝ ✄ ✝ ✄ ✝ oo // ✄ ✝ (2,−2) (1,−1) (2,−2) (1,−1) ✠ // ✄ ✝ ✠ (2,−2) (1,−1) OO ✿ OO OO OO ✄ ✝ ✿ ✠ OO ✄ ✝ ✿ ✠ ✝ ✄ ✝ ✿ ✠ ✝ ✄ ✝ ✿ ✝ // // ✿ ✝ (1,−3) (2,−4) ✠ (1,−3) (2,−4) ✿ ✝ // OO OO OO OO ✿ ✝ (1,−3) (2,−4) ✄ ✿ ✠ ✄ OO OO ✿ ✠ ✝ ✄ ✝ ✠ ✝ ✄ ✝ ✠ ✝ oo ✄ ✝ (2,−6) (1,−5) ✠ // ✄ ✝ oo ✠ (2,−6) (1,−5) OO ✽ OO ✄ ✝ (2,−6) (1,−5) ✽ ✠ OO OO ✿ OO ✽ ✠ ✄ ✝ ✽ ✿ ✝ ✠ ✽ ✿ ✝ ✠ // ✽ ✿ ✝ ✠ (1,−7) (2,−8) ✽ ✿ // // OO OO ✽ ✿ (1,−7) (2,−8) (1,−7) (2,−8) ✞ ✽ ✿ OO OO OO OO ✞ ✿ ✝ ✝ ✞ ✽ ✿ ✝ ✝ ✞  ✿ ✝ ✝ oo ✝ ✝ (2,−10) (1,−9) ✞ oo oo ✝ ✝ (2,−10) (1,−9) (2,−10) (1,−9) ✝ ✝ :: ✝ ✝ ✞ t ✝ ✝ ✞ t ✝ ✝ ✞t ✝ ✝ . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . 33 (2,0) (2,0) (2,0) ❍ OO OO ✝ ❍ ✝ CC ❍ ✝ ✞ ✞ $$ // (2,−2) (1,−1) ✞ oo ✞ // (2,−2) (1,−1) OO OO ✞ (2,−2) (1,−1) :: OO :: ✈ ☎ ✞ OO OO ✞ ✝ ✈ s ✉ ☎ ✞ s ✉ ✈ ✞ ☎ ✞ s ✞ ✉ ✈ // s ✉ ✈ ☎ ✞yys ✞ ✉  ✈ (1,−3) (2,−4) ✞ ☎ ✞ // ✞ OO OO ☎ ✞ (1,−3) (2,−4) (1,−3) (2,−4) ✝ ☎ ✞ OO OO OO OO ✝ ✞ ☎ ✞ ✝ ☎ ✞ ✝ ✞ ✝ ✝ ☎ ✞ // ✝ ☎ ✝ ✞ ✞ (2,−6) (1,−5) ✝ ✝ ☎ // // OO OO ☎ ✝ ✝ (2,−6) (1,−5) (2,−6) (1,−5) ✝ ✝ ☎ OO OO OO OO ✝ ✝ ☎ ✞ ✆ ✝ ✝ ✝ ☎ ✝ ✞ ✆ ✝ ☎ // ✞ ✆ ✝ ✝ ☎ (1,−7) (2,−8) ✝ ✞ ✆ OO OO // // ✞ ✆ (1,−7) (2,−8) (1,−7) (2,−8) ✠ ✝ ✞ ✆ OO OO OO OO ✠ ✞ ✆ ✝ ☎ ✠ ✝ ✞ ✆ ✝ ☎ ✠ ✝ ✞ ✆ ✝ ☎ ✞ ✆ ✠ // ✝ ☎ (2,−10) (1,−9) ✠ // // ✝ ☎ (2,−10) (1,−9) (2,−10) (1,−9) ✝ ☎ ✠ ✝ ☎ ✝ ☎ ✝ ☎ ✠ ✝ ☎ . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) ❑ ❑ ❑ OO OO OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% %% (2,−2) (2,−2) (1,−1) (2,−2) (1,−1) (1,−1) :: OO OO 99 OO OO 99 OO OO r r ✉ ✈ s s ✉ ✎ ✈ r ✎ s r s r s r s ✉ ✈ ✎ ✎ ✍ ✉ ✈ r s r s zz✈ r yys r yys ✉ ✎ ✎ ✍ ✎ ✎ (1,−3) (2,−4) (1,−3) (2,−4) ✎ ✎ (1,−3) (2,−4) OO OO ❍ OO OO ❑ ❍ ❑ OO ✎ ✎ ❍ ❑ ✎ CC ✎ ❍ ❑ ❍ ✍ ✎ ✎ ❍ %% ✝ ✍ $$ ✎ ✎ ✝ ✍ (2,−6) (1,−5) ✎ ✎ oo // (2,−6) (1,−5) :: OO OO ✎ ✝ ✎ (2,−6) (1,−5) :: OO ✍ ✉ ✈ ✝ ✎ ✎ ✎ ✉ ✈ :: ✈ OO OO ✞ ✍ ✈ ✝ ✉ r ✉ ✈ ✎ ✎ ✎ ✉ ✈ ✝ r ✆ ✉ ✈ ✞ ✍ zz✈ ✉ ✎ ✎ r ✆ ✎ ✉ ✈ ✝ ✞ ✍ ✉ r ✉ ✈ ✆ ✎ ✎✝yyr ✎ ✉  ✈ ✞ // ✞ (1,−7) (2,−8) (1,−7) (2,−8) (1,−7) (2,−8) ✎ ✞ OO OO OO OO OO ✞ ✎ CC ☎ ☎ ✆ ✞ ✎ ✝ ☎ ☎ ✝ ✆ ✞ ☎ ☎ ✆ ✞ ☎ ☎ ✝ // // // ☎ ☎ ✎ ✝ (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) ☎ ☎ ✝ ☎ ☎ ✝ ☎ ☎ ✝ r ✎ r ☎ ☎ ✝ ☎ ☎ ✎✝yyr . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) ❏ ❏ OO ▲ OO OO ❏ ❏ ❏ ❏ ▲ ❏ ❏ ▲ ❏ ❏ %% $$ $$ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) OO OO OO OO OO OO r r r ✉ ✇ t r ✌ r ✉ ✇ r t r ✉ r t r ✇ ✌ r r r ✉ ✇ t r yyr zz✉ yyr zzt ✌ ✇ {{✇ (1,−3) (2,−4) (1,−3) (2,−4) (1,−3) (2,−4) OO OO ▲ OO ▲ OO ❏ ✌ ▲ ▲ ❏ OO ● ▲ ▲ ❏ ✌ ● ▲ ▲ ❏ ▲ ▲ ❏ %% %% $$ ## ✌ (2,−6) (1,−5) (2,−6) (1,−5) ✌ 88 OO OO OO OO r r (2,−6) (1,−5) ✌ ✉ ✇ r ✍ r ✉ ✇ r 99 OO OO ✌ r ✉ r ✍ ✇ r r t ✉ r ✇ r ✌ ✌ r ✍ yyr t zz✉ r t ✌ {{✇ r t r zzt (1,−7) (2,−8) ✌ (1,−7) (2,−8) OO OO ■ ✌ ■ (1,−7) (2,−8) OO ● ✍ ■ OO OO ❏ ● ✌ ● ✍ ❏ ■ ● ■ ✌ ✍ ❏ ● $$ ● ✍ ✌ $$ ## oo ✍ (2,−10) (1,−9) oo oo ✍ (2,−10) (1,−9) (2,−10) (1,−9) ✍ :: ✌ 99 99 t r r ✍ ✌ t r r ✍ t ✌ r r t r r ✍ t ✌ r  r . . . . . . (1,−11) (1,−11) (1,−11) . . . . . . . . . . . . 6.3 Mutation sequence in type G We display the sequence of mutated quivers obtained from G at each step of the mutation se- quence μ . 34 (2,0) (2,0) (2,0) OO OO cc● ❏ ✠ ❏ ● $$ ● ✠ (2,−2) (1,−1) ✠ // oo (2,−2) (1,−1) ✠ (2,−2) (1,−1) OO OO t OO OO ✠ :: OO t ✈ t ✞ ✞ ✠ :: ✞ t ✈ ✈ ✞ t ✞ ✠ ✈ ✞ zzt ✈ t ✈ ✞ ✞ ✈ zzt ✠ ✞ ✈ ✈ ✞ ✞ (1,−3) (2,−4) ✠✈  ✈ ✞ ✞❚ ❚ (1,−3) (2,−4) oo ✞ OO ✼ OO ❚ ✞ ❚ ✞❚ ❚ (1,−3) (2,−4) ✼ OO ✼ OO ❚ ❯✞ ❚ ✞ ❚ ❯ ✞ ❚ ❚ OO ✾ OO ❯ ✼ ❚ ✼ ❚ ✞ ✞ ❚ ✞ ❯ ❚ ❯ ✼ ❚ ✼ ❚ ✾ ❚ ✞ ❚ ✞ ❯ ✞ ❯ ** ❚ ✼ ✼ ❚ ✾ ✞ ❯ ✞ ❯ ** ✼ ✼ ✾ ❯ ✞ ❯ (2,−6) (1,−5) ** ✼ ✼ ✾ (2,−6) (1,−5) OO ❏ OO ✼ ✾ ❏ t (2,−6) (1,−5) OO ❏ OO ✼ ❏ ✼ ✾ t ❏ t OO ❑ OO ❏ t ✼ ✾ ✼ ❏ t ❏ ❑ t t ❏ ✼ t ✾ ❑ $$ ❏ t t t ❑ ✾ t t  $$ t ❑ t t  %% (2,−8) (1,−7) t (2,−8) (1,−7) t OO OO t t t t (2,−8) (1,−7) OO OO t t ✞ t t t t OO t OO t ✞ t ✞ t t t t s t t ✞ ✆ zzt zzt ✞ s t t t ✞ ✞ s ✆ zzt zzt s t ✞ ✆ ✞ yys yyt (1,−9) (2,−10) ✞ ✞ ✆ ❚ (1,−9) (2,−10) OO OO ❚ ❚ ✆ ✞ ❚ ✞❚ (1,−9) (2,−10) ❚ OO OO ❚ ❯ ✞ ❚ ✞ ❚ ✆❯ ❚ ❚ OO OO ❯ ❚ ❚ ✆ ❯ ✞ ✞ ❚ ❚ ❯ ❚ ❯ ✞ ✞ ❚ ✆ ** ❚ ❯ ❚ ❯ ✞ ** ✆ . . ❯ (2,−12) (1,−11) . . ** . . (2,−12) (1,−11) . . . . . . . (2,−12) . (1,−11) . . . . (2,0) (2,0) (2,0) OO OO OO ✞ ✆ ✆ ✞ ✆ ✆ ✆ ✆ ✆ ✆ ✆ ✆ ✞ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) ✆ ✆ ✞ OO ❍ OO OO ■ OO OO ❏ OO ✆ ✆ ❍ ✈ ■ s ❏ s ✈ ✆ s ✆ s ❏ ❍ ■ ✞ ✈ s s ❏ ❍ ✆ ■ ✆ ✈ s s ❏ ✞ zz✈ ❍ ✆ yys ■ ✆ yys ❍ ■ ✞ ❏ ❍ ■ // ❍ (1,−3) (2,−4) ■ (1,−3) (2,−4) (1,−3) (2,−4) ❍ ■ ❏ OO ❍ OO ✺ OO ❑ OO ✼ ii❙ ❍ ❍ ■ ❑ ❏ OO ✼dd❍ ✺ ■ ✼ ❏ ❙ ❍ ❍ ❑ ✼ ❙ ❍ ■ ❑ ❏ ❙ ❍ ✺ ✼ ❍ ❍ ❙ ❍ ■ ❑ ❏ ✼ ❍ ❙ $$ ✺ %% ✼ %% ❙ ❍ ■ ✼ ❙ ❍ ❙ ❍ ✺ $$ ✼ ✼  ## ✺ oo (2,−6) ✼ (1,−5) ✼ (2,−6) ✺ (1,−5) ❦ // OO ✼ ❦ OO (2,−6) (1,−5) ❦ ❦ ✼ ✺ ❦ OO ✠ ✼ ❦ ❦ ❦ ✼ OO ❑ OO dd❍ ✺ ❦ ❦ ❦ ✠ ✼❦ ❑ ❍ ✉ t ❦ ❦ ✼ ❍ ✺ ✼ ❑ t ❦ ✉ ✠ ❦ ❦ ❦ ❑ ❍ ✼ t ❦ ✺ ✉ uu❦ ❑ ❦ ❍ ✠ ❦ ✉ %%  ❦ t ✉ ✠ oo ✉ ✠ (2,−8) (1,−7) // (2,−8) (1,−7) (2,−8) (✉ 1,−7) ✠ :: OO t ✈ OO OO OO OO ✠ :: ✞ s t s ✈ ✉ ✈ t ✆ ✆ ✠ ✈ ✞ s s ✈ s t ✆ s ✆ ✈ ✈ ✠ ✞ s t s ✉ ✈ ✈ ✆ ✆ yys yyt yys zz✉ ✠✈  ✈ ✞ ✆ ✆ oo ✞ ✆ ✆ (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) ❯✆ ❯ ❯ ❯ ✆❯ ❯ ❯ ❯ ❯ OO OO OO OO OO OO ✞ ✆ ❯ ✆ ❯ ❯ ❯ ❯ ❯ ❯ ❯ ✞ ❯ ✆ ❯ ✆ ❯ ❯ ❯ ❯ ❯ ✆ ❯ ✆ ❯ ✞ ❯ ❯ ❯ ❯ ❯ ❯ ❯ ✆ ** ✆ ** ✞ ** . . . . . . (2,−12) (1,−11) (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) OO OO OO ☎ ☎ ☎ ☎ ✡ ☎ ☎ ☎ ☎ ☎ ☎ (2,−2) (1,−1) (2,−2) (1,−1) ✡ ☎ ☎ (2,−2) (1,−1) OO ❑ OO OO ❑ OO ✡ ☎ ☎ ❑ r ❑ r ✡ OO ❏ ☎ r ☎ r ❑ ❑ ❏ r r :: ❑ ❑ ✡ ☎ ☎ ❏ r r ✉ ❑ ❑ ✡ ☎ yyr ☎ yyr ✉❏ ❑ ❑ ✉ ❏ ❑ ❑ ✉ ❏ ❑ ❑ ✡ ✉ ❏ (1,−3) (2,−4) (1,−3) (2,−4) ❑ ❑ ❑ ❑ OO ▲ OO ✿ OO ▲ OO ✿ ▲ ❑ ▲ ❑ (1,−3) (2,−4) ✿ ❑ ✿ ❑ ▲ ▲ ▲ ✿ ❑ ▲ ✿ ❑ OO ❏ OO ❏ ❏ ▲ ❑ ▲ ❑ ✿ ✿ ❏ %% %% %% %% ❏ ✿ ✿ ❏ ❏ ❏ ❏ ✿ ✿ $$ %% (2,−6) (1,−5) (2,−6) (1,−5) ✿ ✐ ✿ ✐ ✐ ✐ OO OO OO OO ✿ ✐ ✿ ✐ ✐ ✐ (2,−6) (1,−5) ✝ ✐ ☎ ✐ ✿ ✐ ✿ ✐ ✐ ✐ ✐ ✐ OO OO ✝ ✐✿ ☎ ✐✿ ✐ ✐ ✐ ✐ ✝ ✐ ☎ ✐ ✞ ✐  ✐  ✐ tt✐ tt✐ ✐ ✝ ☎ ✞ ✐ ✝ ☎ ✞ (2,−8) (1,−7) (2,−8) (1,−7) tt✐ ✝ ☎ ✞ OO ❍ OO OO ■ OO ✝ ☎ ✞ r (2,−8) (1,−7) ❍ ✉ ■ ✝ ✉ ☎ ■r ✞ ✉ r OO ▲ OO ✝ ☎ ■ ✞ ❍ ▲ r ✉ r ❍ ■ ▲ ✝ zz✉ ☎ yyr ✞ r ❍ r ▲ ✝ ✞ ■ r ▲ ✞ yyr ▲ // ❍ (1,−9) (2,−10) ■ ▲ (1,−9) (2,−10) ❍ ▲ OO ■ OO ■ ii❚ ❍ ■ ■ (1,−9) (2,−10) ▲ OO dd■ ❚ ❍ ▲ ■ ■ ■ ❚ ❚ ❍ ■ ■ OO ❏ OO ▲ ❚ ■ ❏ ▲ ❍ ■ ■ ❚ ❏ ▲ ❚ $$ ■ ❍ ■ ❚ ❏ ▲ $$ ■ ❚ ❍ ❏ ▲ $$ $$ %% . oo . . . (2,−12) (1,−11) . . // . . (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . 35 (2,0) (2,0) (2,0) OO OO ❂ DD ❂ ✞ ✟ ❂ ✞ ✟ ✞ ❂ (2,−2) (1,−1) ✟ ▲ 99 ✟ (2,−2) (1,−1) (2,−2) (1,−1) OO OO ❂ ▲ r ✟ OO ❏ :: OO :: OO ▲ r ✓ ❂ t ❏ t dd■ ✈ r ▲ ✟ :: ❂ ✞ t t❏ ✓ ✓ r ▲✓ ✈ ■✈ ▲ ✟ t t ❏ ✈ ❂ ✞ r ✈ ■ ✓ ✓ t t ❏✓ ▲ ✈ ✈ ■ ❂ ✟zzt t ▲ ❏ ✈ ✈ ✓ ✓■ ✓ ❏ ❂ (1,−3) (2,−4) ▲ ✈  ✈ ▲ ❏ ❂ ✓ ✓ ✓ ▲ (1,−3) (2,−4) oo ■ OO ❏ OO ❏ ❏ ▲ ✓ ❏ (1,−3) (2,−4) ✓ ■ ✓ OO ❏ OO ❯ ❏ ▲ ❯ ■ ❂ ❏ ❏ OO ❑ OO ❯ ❏ ✓ ▲ ✓ ❏ ✓❯ ■ ❏ ❂ ❏ ▲ ❑ ❯ ❏ ❏ ❯ ■ ❑ ✓ ❂ $$ ✓ %% ❏ ✓ ❏ ❯ ❯ ■ $$ $$ ❯ ❂ ✓ ❑ ✓ ❯ ■ %% ❯ (2,−6) (1,−5) ** ✓ ✓ ❥ (2,−6) ❥ (1,−5) OO OO ❥ ❥ ✓ ✓ ❥ ✓ (2,−6) (1,−5) OO ❥ OO ✡ ❥ ❥ ✐ ✓ ✓ ❥ OO ✓ ✐ OO ❥ ✞ ❥ ✐ ✡ ❥ ❥ ✐ ❥ ❥ ✆ ✓ ✐ ✡ ✓ ❥ ✞ ✓ ❥ ✐ ❥ ❥ ✐ ❥ ✞ ❥ ✆ ✡ ✓ ✓ ✐ uu❥ ✓tt❥ ✐ ✞ ✐ ✡ ✓tt✐ ✓ ✓ ✞ ✆ (2,−8) (1,−7) (2,−8) (1,−7) ✓ ✆ ✓ ✞ ✓ (2,−8) (1,−7) ✡ OO ❏ OO ❏ :: OO ✞ ✆ ✓ ❏ ✓ ❏ :: ✓ t OO ❏ 99 OO ❏ ✞ t❏ ✆ ✓ ✉ ✓ ❏ s ✡ ✓ t ✉❏ ✞ ❏ ✆ s ❏ ❏ t ❏ ✡ ✓ ✉ ✓ s ❏ ✞ ✓ t ✆ ✉ ❏ ❏ s ❏ ✡ ✓ ✉ ❏ ✆ ✓ s ❏ ❏ ❏ ❏ (1,−9) (2,−10) ❏ ❏ ❏ (1,−9) (2,−10) ❏ ❏ (1,−9) (2,−10) OO ❏ OO ❏ ❏ ❏ ❏ OO ❏ OO OO ❑ OO ❏ ❏ ❏ ❏ ❑ ❏ ❏ ❏ ❏ ❏ ❏ ❑ ❏ ❏ ❏ ❏ ❑ ❏ ❏ ❏ $$ $$ ❑ ❏ $$ %% %% %% . . . . (2,−12) (1,−11) . . . . . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . (2,0) (2,0) (2,0) OO ❂ OO ❄ OO ❅ ❂ ❄ ❅ ❂ ❄ ❅ ❂ ❄ ❂ ❄ ❅ // // ❅// ❂ ❄ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) ❅ (1,−1) OO DD OO OO GG OO OO ❅OO ❂ ❄ ✈ ❅ ✠❂ ✞ ❄ ✓ ✎✓ ✎ ✈ ❅ ✈ ✠ ❂ ✞ ❄ ✓ ✎✓ ✎ ✈ ❅ ❂ ✞ zz✈ ✠ ❄ ✎ ❅ ✓ ✎✓ ❂ ✞ ❄ ✠ ❅ // // ❂ ✎ ✎ ✓ ✞ ✓ ❄ // ✠ (1,−3) (2,−4) (1,−3) (2,−4) ❅ ❂ ✞ ❄ ✎ (1,−3) (2,−4) ✓ ✎✓ ❅ ✠ OO OO OO OO ii❙ ✞ ❄ ✎ ❅ ❙ ✎ OO ✠ ✓ ✓ ❙ ❂ ✎✎ ✞ ❄ ✎ ❅ ✠ ❙ ✎ ✓ ❙ ❂ ✓ ✎✎ ✞ ❄ ✎ ❅ ✠ ❙ ✎ ❂ ✞ ✎ ❄ ✓ ❙ ✓ ✎✎ ✎ ✠ ❙ ❂ ✎ ❙ ✎ ✓ ✓ oo ❙ ✎✎ ✎ ✠  ✎ oo (2,−6) (1,−5) ✓ ✎✎ ✓ ✎ (2,−6) (1,−5) ✎ // OO ❂ OO (2,−6) ✓ (1,−5) ✓ ✎✎ ✎ OO ✎❂ ✐ ✎ OO ✓ OO ✎✎ CC ✓ ✎ ❂ ✐ ✎ ✆ ✓ ✐ ✞ ✓ ❂ ✐ ✎✎ ✎ ✎ ✆ ✐ ✎ ❂ ✐ ✞ ✓ ✎✎ ✓ ✎ ✆ ✐ ❂ ✞  ✎ ✓tt✐ ✓ ✆ ✎✎ ✎ ❂ ✞ (2,−8) (1,−7) ✆ ✓ ✎✎ ✓ ✎ ❂ (2,−8) (1,−7) (2,−8) (1,−7) ✞ :: ❂OO ✆ ✓ ✓ ✎✎ ✎ dd■ ✈ OO ❏ 99 OO ✞ OO ❏ 99 OO :: ✆ ❂ ❏ s ❏ ■✈ ✓ ✎✎ s ✓ s ✎ ✈ ✞ ✈ ❂ ✆ s ❏ s s ❏ ✈ ■ ✓ s ✎✎ s ✓ s ✎ ✈ ✈ ✆ ❏ ✞ ❏ ■ ❂ s s s ✈ ❏ ❏ ✈ ■ ❂ ✆ ✓ s ✎✎✞yys ✓ s ✎ ✈  ✈ ❏ ❏ ■ ❏ ❏ ❂ oo ❏ ❏ ❂ (1,−9) (2,−10) (1,−9) (2,−10) (1,−9) (2,−10) ■ ❏ ❏ ❯ ■ ❂ ❏ ❏ OO ❑ OO OO ❑ OO OO ❑ OO ❯ ❏ ❏ ❯ ■ ❂ ❑ ❑ ❑ ❯ ❑ ❏ ❑ ❏ ❑ ❯ ❯ ❂ ❏ ❏ ■ ❑ ❑ ❑ ❯ ❯ ❂ ❑ ❏ ❑ ❏ ❑ ❯ ■ %% %% %% %% %% ** . . . . . . (2,−12) (1,−11) (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) OO ❆ OO ❆ OO ❂ ❆ ❆ ❆ ❆ ❂ ❆ ❆ ❂ ❆ ❆ ❆// ❆// ❂// (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) ❆ ❆ ❆ ❆ OO OO OO OO OO ❂OO ❆ ❆ ✌ ✌ ✏❂ ❆ ❆ ✌ ❆ ✌ ❆ ✏ ❆ ❆ ✌ ✌ ✏ ❆ ❆ // ✌ // ✌ ✏ ❂ (1,−3) (2,−4) ❆ (1,−3) (2,−4) ❆ oo ✌ ✌ ❂ ❆ ❆ ✏ (1,−3) (2,−4) OO OO OO OO ✌ ❆ ✌ ❆ OO ✏ ✍ ✌ ❂ ❆ ❆ ✌ ✌ ✍ ❆ ✌ ❆ ❂ ✌ ✌ ✏ ❂ ✍ ✌ ✌ ✌ oo ✏ ✍ ✌ oo (2,−6) ✌ (1,−5) (2,−6) ✌ (1,−5) ✍ ✌ ✌ ✌ oo OO ❃ OO OO ❄ OO ✍ ✌ ✏ (2,−6) (1,−5) ✌❃ ✌❄ ✍ ✏ ✌ ❄ ✌ ❃ ✌ OO ❄ OO ✍ ❃ ✌ ❄ ✏ ❄ ✌ ✌ ❄ ❄ ✍ ❃ ✌ ✏ // ❄// ❃ ❄ ✍ ✏ (2,−8) (1,−7) ✌ (2,−8) (1,−7) ❄ ❄ ✌ ❄ ❄// OO ❃CC OO OO OO ✍ ✉ ❄ (2,−8) ❄ (1,−7) ✟❃ ✝ ❄ OO ❄OO ✍ ✉ ✟ ❃ ✌ ✝ ✉ ❄ ❄ ✍ ✟ ❃ ✝ zz✉ ✌ ❄ ❄ ❃ ✝ ✍ ✟ // ❄ ❄ ❃ ✝ ✟ (1,−9) (2,−10) ❄ ❄ // ❃ ✝ (1,−9) (2,−10) ❄ ❄ ✟ // OO ■ OO ❃ ✝ ii❚ ■ ❄ (1,−9) (2,−10) ❄ OO ✟ ❚ ❃ ■ ✝ ❚ ❄ ❄ ■ OO OO ✟ ❚ ❚ ❃ ✝ ❄ ❄ ✟ ❃ $$ ✝ ❄ ❄ ❚ ❃  ❄ . . oo . . (2,−12) (1,−11) . . // . . oo . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . 36 (2,0) (2,0) (2,0) OO ❄ OO ❄ ^^❃ ❄ ❃ ❍ ❄ ❃ ❍ $$ // ❃ (2,−2) (1,−1) ❄ // oo ❄ (2,−2) (1,−1) (2,−2) (1,−1) OO OO ❃ OO ❏ OO :: OO ✎ ❃ ❄ ❏ dd■ ✈ ✎ ❏ ✎ ✎ ❄ ■✈ ❏ ❃ ✈ ❄ ✎ ■ ✎ ✎ ❏ ✈ ❃ ■ oo ❄ ✎ ❏ ✈ ✎ ✎ ■ ❏ ❃ (1,−3) (2,−4) ❄  ✈ oo ✎ ✎ ■ ✎ ❏ ❃ (1,−3) (2,−4) oo ■ OO ✵ OO ❏ ✎ ✎ ❄ ✎ ❃ ❏ (1,−3) (2,−4) ■ OO ✴ OO ❯ ✵ ❃ ❯ ■ ❄ ✎ ❏ ✎ ✎ OO ✵ OO ❯ ✴ ❏ ❯ ■ ✵ ❄ ❃ ✎ ✎ ❯ ✎ ❏ ❯ ■ ❃ ✵ ✵  ✴ ❏ ❯ ✎ ✎ ❯ ■ $$ ✵ ❯ ✵ oo ✴ ❯ ■ ✎ ❯ (2,−6) ✎ (1,−5) ✵ ** ✴ oo ✵ ✎ (2,−6) ✎ (1,−5) ✎ ✵ oo OO ❂ OO ✴ ✎ (2,−6) ✎ (1,−5) ✎❂ OO ❄ OO ✵ ✴ ✎ ❄ OO ❅✎ OO ✎ ❂ ✵ ✴ ❂ ❄ ✎ ❅ ✎ ✵ ✵ ✴ ❄ ❅ ❂ ✎ // ❄ ✵ ❅ ❂ ✴ ✵ ✎ (2,−8) (1,−7) ❄ // ❅ ❂ ✵ ✵ ✴(2,−8) (1,−7) ❄ ❅// OO ❂OO ✵ ✴ ❄ (2,−8) ❅ (1,−7) OO OO ❄ ✵ ✵ ✴ OO ❅OO ❄ ❅ ✵ ✴ ✵ ❄ ❅ ❂ ✵ ❄ ❅ oo oo (1,−9) (2,−10) (1,−9) (2,−10) ❄ ❅ oo ❂ ❄ ❅ (1,−9) (2,−10) OO OO OO ❄ ❅ OO OO ❄ ❅ ❂ ❄ ❅ ❂  ❅ . oo . . . (2,−12) (1,−11) . . oo . . oo . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . (2,0) (2,0) (2,0) ❑ ❑ ❑ OO ❑ OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% %% (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) OO DD OO OO OO OO OO ✈ ✈ s s ✠ ✞ ✆ ✎ s s ✈ ✈ ✈ ✈ ✠ s ✞ s ✆ ✈ s s ✈ ✞ ✆ zz✈ ✠ yys yys ✞ ✆ zz✈ ✠ ✞ ✆ // ✠ (1,−3) (2,−4) (1,−3) (2,−4) ii❙ jj❯ ✞ ✆❯ (1,−3) (2,−4) ✎ ❙ ✠ OO ✴ZZ✺ ❍ OO ✻ ❙ OO ✼ OO ✻ ❯ ii❙ ✞ ❙ ✆ ❙✎ ❍ ❙ ❯ OO ✴ dd❍ ✠ ✺ ✻ ✼ ✻ ❯ ❙ ✴ ❍ ✞ ❙ ✆ ❍ ❯ ✠ ✎ ❙ ❍ ❙ ✼ ❯ ✴ ❙ ✺ ✻ ❙ ✻ ❍ ✴ ❍ ✞ ✆ ❯ ❙ ❙ ❯ ❍ ✠ ❙ ❙ ✼ ✴ ✎ ✺ $$ ✞ ✻ ✆ ✻ ❯ ❙ ❙ ❍ ✴ ❙ ✼ ✠ ❙ ✴ ❍ ✎ ❙ ✺ ✻ ✻ ❙ ✴ ✼ ✴ ✺ ✻ (2,−6) ✻ (1,−5) ✴ ✼ ✺ (2,−6) (1,−5) ✻ OO ❍ ✻ OO ✴ ✎ ✼ (2,−6) (1,−5) ✴ ✺ ❍ ✻ OO ✻ ✴ ✼ ❍ OO ❅ OO ✴ ✺ __❄ ✻ ✼ ❍ ✻ ✎ ❅ ❄ ❍ ✴ ✻  ✻ ✴ ❍ ✎ ❅ ❄ ✺ ✻ $$ ✻ ✴ ❅ ✎  ✻ ✻ ❅ oo ✴ ❄ ✻ (2,−8) (1,−✻ 7) ❅// //❄ ✴(2,−8) (1,−7) (2,−8) (1,−7) ❅ ✻ :: ✻ ❄ OO ✴ ❅ ✴ dd■ ✈ OO OO OO ❏ OO ✻ ✻ ❏ ■✈ ✴ ❅ ✴ ❄ ✻ ✈ ✻ ❏ ■ ✴ ✴ ✈ ❏ ❄ ✻ ■ ✻ ❅ ✈ ❄ ■ ❅  ✻  ✈ ✻ ❏ ■ ❏ ❄ ❅ ✻ ■ ✻ oo oo oo ❏ ❄ (1,−9) (2,−10) ❅ (1,−9) (2,−10) (1,−9) (2,−10) ■ ❏ ✻ ✻ ❅ ❄ ❯ ■ OO OO OO OO ✻ OO OO ❯ ✻ ❏ ❄ ❯ ■ ❅ ❯ ❏ ✻ ❯ ✻ ❅ ❄ ❯ ❏ ■ ❅ ❄✻ ❯ ✻ ❏ ❯ ■ %%  ** . oo . . oo . . oo . (2,−12) (1,−11) (2,−12) (1,−11) (2,−12) (1,−11) . . . . . . . . . . . . (2,0) (2,0) (2,0) ▲ ▲ OO OO OO ▲ ▲ ❏ ▲ ▲ ▲ ▲ ▲ ▲ ❏ ▲ ▲ ❏ %% %% $$ (2,−2) (1,−1) (2,−2) (1,−1) (2,−2) (1,−1) OO OO OO OO OO OO r r ☎ ☎ ✟ r r t r ☎ r ☎ t r r ☎ ☎ yyr yyr zzt ✟ ☎ ☎ ☎ ☎ (1,−3) (2,−4) (1,−3) (2,−4) (1,−3) (2,−4) jj❯ jj❯ ✟ ☎❯ ☎❯ ❙ ❯ ❯ OO ✿ OO ✼ ❯ OO ✿ OO ✽ ❯ OO ✹ OO ✟ ❙ ☎ ☎ ❯ ❯ ❙ ✿ ❯ ✿ ❯ ✟ ❙ ✼ ✽ ✹ ☎ ❯ ☎ ❯ ❙ ❯ ❯ ❙ ✿ ❯ ✿ ✽ ❯ ✟ ☎✼ ☎ ✹ ❙ ❯ ❯ ✿ ❯ ✿ ❯ ❙ ❯ ✽ ❯ ✹ ✟ ❙ ☎ ✼ ☎ ✿ ✿ )) ✽ ✹ ✟ ✿ ✿ (2,−6) ✼ (1,−5) (2,−6) ✽ (1,−5) ✹ ✿ ✿ ▲ ✼ ▲ ✽ ✹ (2,−6) (1,−5) OO OO OO OO ✿ ✿ ▲ ▲ ✽ ✼ ✹ OO ● ✿ ▲ ✿ ▲ ✽ ● ▲✼ ▲ ✹ ✿ ✿ ● ▲ ▲ ✼ ✹ ● %%  %% ✽ ✹ ✼ ● ## (2,−8) (1,✼−7) (2,−8) (1,−7) OO CC OO OO OO ✉ ✉ r (2,−8) (1,−7) ✟ r ✝ ✉ ✉ ✽ OO OO ✉ ✼ r ✝ ✉ ✟ ✽ ✉ r t ✉ ✞ ✟ ✼ ✝ t zz✉ yyr ✽ ✝ t ✞ zz✉ ✟ ✼ ✽ t ✝ zzt ✟ ✼ (1,−9) (2,−10) // ✽ ii❚✝ ✞ (1,−9) (2,−10) ✟ ✼ ❚ OO ■ OO ❚ ✽ ✝ ✞ ii❚ ✼ ■ ❚ (1,−9) (2,−10) OO dd■ ✟ ❚ ✽ jj❯ ■ ✝ ❚ ✞ ■ ❚ ❚ ❯ ✼ ■ OO OO ❯ ✟ ❚ ❚ ✽ ■ ❚ ✝ ❚ ✞ ❯ ■ ❯ ❚ ✼ ❚ ❯ ■ ✟ $$ ✝ ✽ ✞ ❚ ❚ ❯ ■ ❚ ❚ ❯ ✟ ✼ ■ ❚  ✞ ❯ ❚ ❯ ✟  ✞ . . . (2,−12) (1,−11) . . . . . . (2,−12) . (1,−11) . (2,−12) . (1,−11) . . . . . . 37 6.4 Examples of A-modules for g of type B (i) We describe some A-modules K for g of type B . The quiver Γ is: k,m (2,−1) OO ❑ %% (2,−3) (1,−3) OO OO s s s s s s s s yys yys (1,−5) (2,−5) OO ❑ OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% (2,−7) (1,−7) OO OO s s s s s s s s yys yys (1,−9) (2,−9) . . . . . . . . . Following the convention of Example 4.6, unless otherwise specified, in the following figures the vertices carry one-dimensional spaces, and the arrows carry linear maps with matrix (±1). (1) (1) The modules K and K are: 1,−5 1,−7 (2,−3) (1,−3) (1,−5) (2,−5) OO ❑ OO ❑ ❑ ❑ s s s s ❑ ❑ s s ❑ ❑ ❑ ❑ s s yys yys %% %% (1,−5) (2,−5) (2,−7) (1,−7) (2) (2) The modules K and K are: 1,−5 1,−7 (2,−1) (2,−3) ❑ s ❑ s ❑ s yys %% (1,−5) (1,−3) s ❑ s ❑ s ❑ yys %% (2,−5) (2,−7) Applying Theorem 4.8, we recover the following well known formulas for the q-characters of the fundamental U (bg)-modules: χ (L(Y )) = Y (1 + v (1 + v (1 + v (1 + v )))), q 1,−7 1,−7 1,−5 2,−3 2,−5 1,−3 χ (L(Y )) = Y (1 + v (1 + v (1 + v ))). q 2,−6 2,−6 2,−5 1,−3 2,−1 (1) (2) The modules K and K are: 2,−5 2,−7 (2,−3) (1,−3) (2,−3) OO OO OO s s s s s s s s s s s s yys yys yys (1,−5) (2,−5) (1,−5) (2,−5) OO OO ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑ %% %% (2,−7) (1,−7) (2,−7) (1,−7) OO OO s s s s s s s s s s s s yys yys yys (1,−9) (2,−9) (2,−9) They correspond under Theorem 4.8 to the Kirillov-Reshetikhin modules (1) (2) W = L(Y Y ) and W = L(Y Y ). 1,−11 1,−7 2,−10 2,−8 2,−11 2,−10 38 (1) (2) The modules K and K are: 3,−5 3,−7 (2,−3) (1,−3) (2,−3) OO OO OO r r s r r s r r r r s xxr xxr yys (1,−5) (2,−5) (1,−5) (2,−5) OO OO OO ▲ OO ▲ ▲α ▲ ▲ ′ ▲ ▲ γ ▲ ▲ ▲ %% %% (2,−7) (1,−7) (2,−7) (1,−7) OO OO OO r r s s r r s s r r s s r r s s xxr xxr yys yys (1,−9) (2,−9) (1,−9) (2,−9) OO OO ▲ OO %% (2,−11) (1,−11) (2,−11) OO r r r r r r r r xxr xxr (1,−13) (2,−13) (2) In K , the vertex (2,−7) carries a two-dimensional vector space. The linear maps carried by 3,−7 the adjacent arrows have the following matrices: α = γ = , β = γ = ( 0 1). They correspond under Theorem 4.8 to the Kirillov-Reshetikhin modules: (1) (2) W = L(Y Y Y ) and W = L(Y Y Y ). 1,−15 1,−11 1,−7 2,−12 2,−10 2,−8 3,−15 3,−12 6.5 Examples of A-modules for g of type B Let g be of type B , with the short root being α . The quiver Γ is: 3 3 (3,−1) OO xxr (2,−3) (3,−3) (1,−3) OO ▲ OO ▲ OO r ▲ ▲ r ▲ ▲ r r r ▲ ▲ r r ▲ ▲ r xxr && && xxr (1,−5) (3,−5) (2,−5) OO ▲ OO OO ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ && xxr xxr && (2,−7) (3,−7) (1,−7) OO ▲ OO ▲ OO r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r xxr && && xxr (1,−9) (3,−9) (2,−9) OO ▲ OO OO ▲ ▲ r r ▲ ▲ r r ▲ ▲ ▲ r r ▲ r r ▲ && xxr xxr && (2,−11) (3,−11) (1,−11) OO ▲ OO ▲ OO r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r r ▲ ▲ r xxr && && xxr (1,−13) (3,−13) (2,−13) OO OO OO . . . . . . . . . . . . . . . (1) The module K is: 1,−9 (1,−3) yys (3,−5) (2,−5) OO s s s s s s s s yys yys (2,−7) (3,−7) yys (1,−9) 39 (2) (3) The modules K and K are: 1,−11 1,−11 (3,−5) (2,−5) (3,−3) OO ❑ ▲ r s ❑ ▲ r s ❑ ▲ ❑ ▲ r s r s ❑ ▲ xxr yys %% %% (2,−7) (3,−7) (1,−7) (2,−5) ▲ ▲ ❑ s ▲ ▲ s s ❑ s ▲ ▲ s s ❑ ▲ ▲ ❑ s s s s ▲ ▲ s s ❑ yys && %% yys yys %% (1,−9) (3,−9) (2,−9) (3,−7) (1,−7) ▲ OO ▲ ▲ r s ▲ s ▲ r s ▲ s ▲ r ▲ s ▲ r s ▲ s %% xxr yys %% yys (2,−11) (3,−11) (2,−9) yys (3,−11) The corresponding fundamental U (bg)-modules are L(Y ), L(Y ), and L(Y ), of respec- q 1,−11 2,−13 3,−12 tive dimensions 7, 22, and 8. 6.6 Examples of A-modules for g of type C Let g is of type C , with the long root being α . The quiver Γ is: 3 3 (1,−1) OO ▲ %% (2,−2) (3,−2) OO OO s ③ s ❖ s ❖③ yys ❖ ③ ❖ (1,−3) ❖ ③ ❖ OO ▲ ③ ❖ ③ ❖ ▲ ③ %% ||③ '' (2,−4) (3,−4) OO ❉ OO s ❉ ❉ ♦ s ♦ yys ❉ ♦ ❉ ♦ ❉ ♦ (1,−5) ♦ ♦❉ OO ▲ ▲ ❉ ▲ ♦ ❉ ▲ ♦ ▲ ❉ %% ww♦ "" (2,−6) (3,−6) OO OO ❖ ③ ❖ ③ s ③ yys ③❖ ③ ❖ (1,−7) ③ ❖ OO ▲ ▲ ③ ③ ❖ ▲ ❖ ▲ ③ ❖ %% ||③ '' (2,−8) (3,−8) OO ❉ OO ❉ ♦ s ♦ s ♦ s ❉ yy s ❉ ♦ ❉ ♦ (1,−9) ♦ ❉ ♦ OO ▲ ♦❉ ▲ ❉ ♦ ❉ ▲ ♦ %% ww♦ "" (2,−10) (3,−10) OO OO . . . . . . . . . . . . 40 (1) (2) The modules K and K are: 1,−7 1,−8 (1,−1) (2,−2) ❑ ✽ ❑ s ❑ s s ✽ ❑ s %% yys ✽ (2,−2) (1,−3) ❑ ❑ ❑ ❑ ❑ ❑ ✽ ❑ ❑ ❑ ❑ %% %% (3,−4) (2,−4) (3,−4) OO ❑ s ❑ s s ❑s s ❑ s s ❑ yys yys %% (2,−6) (2,−6) (3,−6) s s s s s s s s yys yys ✝ (1,−7) (1,−7) %% ✝ (2,−8) (3) The module K is: 1,−8 (2,−4) (3,−4) OO ❈ ♠ ♣ ♠ ♣ ♠ ❈ ♣ ♠ ♣ ♠ ❈ vv♠ ❈ ♣ (1,−5) κ ◗ ♣ OO ❈ ◗ ♣ β ◗ ♣ ❈ ♣ ❈ ◗ ♣ ◗ ♣ (( ww♣ !! (2,−6) (3,−6) OO ◆ ♠ ④ ◆ ④ ♠ ④ vv♠ ε ④◆ ④ ◆ γ ◆ (1,−7) ④ ◆ ◗ ◆ ◗ ◆ ◗ ④ ◆ ◗ ④ ◗ ◆ ◗ ◆ (( }}④ '' (2,−8) (3,−8) Here, the vector space sitting at vertex (2,−6) has dimension 2. The maps incident to this space are given by the following matrices: 1 0 1 α = , β = , γ = , ε = ( 0 1), κ = ( 0 1), ι = ( 1 0). 0 1 0 The corresponding fundamental U (bg)-modules are L(Y ), L(Y ), and L(Y ), of respec- q 1,−8 2,−10 3,−10 tive dimensions 6, 14, and 14. 41 6.7 Examples of A-modules for g of type F Let g be of type F . We label the simple roots α , α , α , α , so that the short simple roots are α 4 1 2 3 4 1 and α . The quiver Γ is: (1,−1) OO ▲ %% (2,−2) (3,−2) (4,−2) OO ❖ OO ❘ OO ❖ ❘ ❖ ③ ❘ ✆ ❖ ❘ ③ ❘ s ❖ ✆ s ❖③ yys ✆ ③❖ ❘ ❖ ❘ ③ ✆ ❖ ❘ ③ ❖ ✆❘ (1,−3) ❖ ③ ❘ ❖ ✆ OO ▲ ③ ❘ ❖ ✆ ▲ ③ ❘ ▲ ✆ ❘ ③ ❖ ▲ ❘ ❖ ✆ ③ ❘ ▲ ❖ %% ||③ '' ✆ )) (2,−4) (3,−4) (4,−4) OO ❉ OO ✾ ❧ OO ♦ ❧ s ❉ ♦ ✾ ❧ ♦ ❧ s ❉ ❧ ♦ ✾ s ❉ ♦ ✾ ❧ yys ♦ ❉ ❧ ♦ ✾ ❧ ♦ ❧ ❉ ♦ ✾ (1,−5) ♦ ❉ ❧ ♦ ✾ ♦❉ OO ▲ ❧ ♦ ❧ ✾ ❉ ❧ ♦ ✾ ▲ ❧ ♦ ❉ ▲ ❧ ♦ ✾ ▲ ❉ ❧ %% ww♦ "" vv❧ (2,−6) (3,−6) (4,−6) OO OO ❘ OO s ③ ❘ ❖ ✆ s ❘ ❖ ③ s ❘ ❖ ✆ ③ ❘ s ❖ ❘ ✆ yys ❖ ③ ❘ ❖ ❘ ③ ❖ ❘ ❖ ❘ ③ ✆ (1,−7) ❖ ③ ❘ ❖ ✆ ③ ❖ ❘ OO ▲ ✆ ❖ ❘ ▲ ③ ❖ ❘ ▲ ❖ ❘ ▲ ❖ ❘ ▲ ③ ❘ ❖ ❘ %% ||③ '' ✆ )) (2,−8) (3,−8) (4,−8) OO ❉ OO ✾ ❧ OO s ❧ ❉ ♦ ❧ s ❧ ❉ ♦ s ✾ ❧ s ❉ ❧ ✾ ❧ yys ❉ ♦ ❧ ♦ ❧ ❉ ♦ ❧ ♦ ❧ ❉ ✾ (1,−9) ♦ ❧ ❉ ✾ ♦ ❧ ♦❉ ❧ OO ▲ ✾ ♦ ❧ ▲ ❉ ♦ ❧ ▲ ❧ ✾ ♦ ❉ ▲ ♦ ❧ ✾ ▲ ♦ ❉ %% ww♦ "" vv❧ (2,−10) (3,−10) (4,−10) OO OO OO . . . . . . . . . . . . . . . . . . (1) The module K is: 1,−17 (1,−1) && (2,−2) && (3,−4) r ▲ r ▲ r ▲ r ▲ xxr && (2,−6) (4,−6) ✿ ☎ ✿ ☎ ✿ ☎ xxr ✿ ☎ ✿ ☎ (1,−7) ✿ ☎ ✿ ☎ ✿ ☎ ✿ ☎ &&  ☎ (2,−8) (3,−8) OO ▲ ▲ r ▲r r ▲ xxr && (2,−10) (3,−10) ☎ ✿ r ☎ ✿ ☎ ✿ xxr ☎ ✿ ☎ ✿ (1,−11) ☎ ✿ ▲ ☎ ✿ ▲ ☎ ✿ ☎ ✿ && ☎ (2,−12) (4,−12) ▲ r ▲ r ▲ r && xxr (3,−14) xxr (2,−16) xxr (1,−17) 42 (4) The module K is: 1,−16 (4,−2) tt❤ (2,−4) (3,−4) OO P ❧ P ❧ P ③ ❧ P ❧ ③ P③ vv❧ ③ ′ ③ P (1,−5) P ③ P ❘ ′ P OO ❘ ′ ③ ❘ α β ❘ P ③ P ❘ ③ ||③ '' (( (2,−6) (3,−6) ❧ OO ❉ ✿ ❧ ♦ ❧ ♦ ✿ ❧ ♦ ❧ ❉ ♦ ✿ ❉ ♦ vv❧ ′ ♦ ✿ ❉ ′ ε κ ♦ ✿ ❉ ♦ ✿ (1,−7) γ ♦ ♦ ✿ ♦ ❉ ♦ ✿ ❘ ♦ ✿ ♦ ❉ ♦ ✿ ❘ ❉ ❘ ♦ (( ww♦ "" (2,−8) (3,−8) (4,−8) ❘ ❱ ❤ ❱ ❤ ❘ ❧ ❱ ❤ ❘ ❧ ❱ ❤ ❘ ❱ ❤ ❧ ❱ ❤ ❘❧ ❤❱ ❧ ❘ ❤ ❱ ❘ ❤ ❱ ❧ ❘ ❤ ❱ ❤ ❱ vv❧ (( ❤ ❱ tt❤ ** (2,−10) (3,−10) (4,−10) OO P ❧ ③ P ☎ ❧ P ③ ❧ P ☎ ❧ P vv❧ ③P P ☎ ③ P ☎ (1,−11) P P ☎ ❘ ③ P OO ☎ ❘ β P α ③ ❘ P ❘ ☎ ❘ ③ ❘ P ☎ ❘ P (( ||③ '' ☎ (2,−12) (3,−12) OO ❉ ❧ ♦ ❧ ♦ ❧ ❉ ♦ ❧ ♦ ❧ ♦ vv❧ ε ❉ κ ♦ γ ❉ ♦ (1,−13) ♦ ❘ ♦❉ ❘ ♦ ❘ ❉ ♦ ❉ ❘ ♦ ❘ ♦ (( ww♦ "" (2,−14) (3,−14) ** (4,−16) Here, the vector spaces sitting at vertex (2,−6) and (2,−12) have dimension 2. The maps incident to these spaces are given by the following matrices: 1 0 1 α = , κ = ( 1 0), β = , ε = ( 0 1), γ = , ι = ( 0 1), 0 1 0 1 0 1 ′ ′ ′ ′ ′ ′ α = , κ = ( 1 0), β = , ε = ( 0 1), γ = , ι = ( 0 1). 0 1 0 The corresponding fundamental U (g)-modules are L(Y ) and L(Y ), of respective dimen- q 1,−18 4,−18 sions 26, and 53. Acknowledgements We are grateful to P.-G. Plamondon for very helpful comments and discussions. References [C] V. C H A R I, Minimal affinizations of representations of quantum groups: the rank 2 case, Publ. Res. Inst. Math. Sci. 31 (1995), 873–911. [CH] V. C H A R I, D. H E R NA N D E Z, Beyond Kirillov-Reshetikhin modules, In Quantum affine algebras, extended affine Lie algebras, and their applications, Contemp. Math., 506, AMS Providence, 2010, 49–81. [CP1] V. C H A R I, A. P R E S S L E Y, A guide to quantum groups. Cambridge University Press 1994. 43 [CP2] V. C H A R I, A. P R E S S L E Y, Minimal affinizations of representations of quantum groups: the simply laced case, J. Algebra 184 (1996), 1–30. [DWZ1] H. D E R K S E N, J. W E Y M A N, A. Z E L E V I N S K Y, Quivers with potential and their representations I: Muta- tions, Selecta Math., 14 (2008), 59–119. [DWZ2] H. D E R K S E N, J. W E Y M A N, A. Z E L E V I N S K Y, Quivers with potential and their representations II: Appli- cations to cluster algebras, J. Amer. Math. Soc. 23 (2010), 749–790. [FM] E. F R E N K E L, E. MU K H I N, Combinatorics of q-characters of finite-dimensional representations of quan- tum affine algebras, Comm. Math. Phys. 216 (2001), 23–57. [FR] E. F R E N K E L, N. R E S H E T I K H I N, The q-characters of representations of quantum affine algebras, Recent developments in quantum affine algebras and related topics, Contemp. Math. 248 (1999), 163–205. [FZ1] S. F O M I N, A. Z E L E V I N S K Y, Cluster algebras I: Foundations, J. Amer. Math. Soc. 15 (2002), 497–529. [FZ2] S. F O M I N, A. Z E L E V I N S K Y, Cluster algebras: notes for the CDM-03 conference, in Current developments in mathematics, 2003, 1–34, Int. Press, Somerville, MA, 2003. [FZ3] S. F O M I N, A. Z E L E V I N S K Y, Cluster algebras IV: coefficients, Compos. Math. 143 (2007), 112–164. [GG] J. G R A B OWS K I, S. G R AT Z Cluster algebras of infinite rank, (2012) arxiv 1212.3528. [GLS1] C. G E I S S, B. L E C L E R C, J. S C H R O E R, Auslander algebras and initial seeds for cluster algebras, J. London Math. Soc. 75 (2007), 718–740. [GLS2] C. G E I S S, B. L E C L E R C, J. S C H R O E R, Kac-Moody groups and cluster algebras, Advances Math. 228 (2011), 329–443. [GSV] M. G E K H T M A N, M. S H A P I RO, A. VA I N S H T E I N, Cluster algebras and Poisson geometry, AMS Math. Survey and Monographs 167, AMS 2010. [H] D. H E R NA N D E Z, The Kirillov-Reshetikhin conjecture and solutions of T -systems, J. Reine Angew. Math. 596 (2006), 63–87. [HJ] D. H E R NA N D E Z, M. J I M B O, Asymptotic representations and Drinfeld rational fractions, Compositio Math. 148 (2012), 1593–1623. [HL1] D. H E R NA N D E Z, B. L E C L E R C, Cluster algebras and quantum affine algebras, Duke Math. J. 154 (2010), 265–341. [HL2] D. H E R NA N D E Z, B. L E C L E R C, Monoidal categorifications of cluster algebras of type A and D, in Sym- metries, integrable systems and representations, (K. Iohara, S. Morier-Genoud, B. Re ´my, eds.), Springer proceedings in mathematics and statistics 40, 2013, 175–193. [HoJo] T. H O L M, P. J O R G E N S E N, On a cluster category of infinite Dynkin type, and the relation to triangulations of the infinitygon, Math. Z. 270 (2012), 277–295. [IIKKN1] R. I N O U E, O. I YA M A, B. K E L L E R, A. K U N I BA, T. NA K A N I S H I, Periodicities of T and Y -systems, dilog- arithm identities, and cluster algebras I, Type B , Publ. RIMS, (to appear). [IIKKN2] R. I N O U E, O. I YA M A, B. K E L L E R, A. K U N I BA, T. NA K A N I S H I, Periodicities of T and Y -systems, dilog- arithm identities, and cluster algebras II, Type C , F and G , Publ. RIMS, (to appear). r 4 2 [IIKNS] R. I N O U E, O. I YA M A, A. K U N I BA, T. NA K A N I S H I, J. S U Z U K I, Periodicities of T and Y -systems, Nagoya Math. J. 197 (2010), 59–174. [Ka] V. G. K AC, Infinite dimensional Lie algebras, Cambridge University Press 1990. [KNS1] A. K U N I BA, T. NA K A N I S H I, J. S U Z U K I, Functional relations in solvable lattice models: I. Functional relations and representation theory, Int. J. Mod. Phys. A9 (1994), 5215–5266. 44 [KNS2] A. K U N I BA, T. NA K A N I S H I, J. S U Z U K I, T -systems and Y -systems in integrable systems, J. Phys. A 44 (2011), no. 10, 103001, 146 pp. [Le1] B. L E C L E R C, Imaginary vectors in the dual canonical basis of U (n), Transform. Groups 8 (2003), 95–104. [Le2] B. L E C L E R C, Quantum loop algebras, quiver varieties, and cluster algebras, in Representations of Alge- bras and Related Topics, (A. Skowron ´ ski and K. Yamagata, eds.), European Math. Soc. Series of Congress Reports, 2011, 117–152. [Lu] G. L U S Z T I G, On quiver varieties, Adv. Math. 136 (1998), 141–182. [N1] H. NA K A J I M A, Quiver varieties and finite-dimensional representations of quantum affine algebras. J. Amer. Math. Soc. 14 (2001), 145–238. [N2] H. NA K A J I M A, t -analogs of q-characters of Kirillov-Reshetikhin modules of quantum affine algebras, Rep- resent. Theory 7 (2003), 259–274. [N3] H. NA K A J I M A, Quiver varieties and t -analogs of q-characters of quantum affine algebras. Annals of Math. 160 (2004), 1057–1097. [N4] H. NA K A J I M A, Quiver varieties and cluster algebras, Kyoto J. Math. 51 (2011), 71–126. [Pa] Y. PA L U, Cluster characters II: A multiplication formula, Proceedings London Math. Soc. (to appear). [Pl1] P.-G. P L A M O N D O N, Cluster characters for cluster categories with infinite-dimensional morphism spaces, Adv. Math. 227 (2011), 1–39. [Pl2] P.-G. P L A M O N D O N, Generic bases for cluster algebras from the cluster category, Int. Math. Research. Notices, (to appear). David HE RNANDE Z : Universite ´ Paris Diderot-Paris 7, Institut de Mathe ´matiques de Jussieu - Paris Rive Gauche CNRS UMR 7586, Ba ˆtiment Sophie Germain, Case 7012, 75205 Paris Cedex 13, France. email : hernandez@math.jussieu.fr Bernard L E CL E RC : Normandie Univ, France UNICAEN, LMNO F-14032 Caen, France CNRS UMR 6139, F-14032 Caen, France Institut Universitaire de France. email : bernard.leclerc@unicaen.fr

Journal

MathematicsarXiv (Cornell University)

Published: Mar 4, 2013

There are no references for this article.