Factorizable R-Matrices for Small Quantum Groups

Representations of small quantum groups uq(g) at a root of unity and their extensions provide interesting tensor categories, that appear in different areas of algebra and mathematical physics. There is an ansatz by Lusztig to endow these categories with the structure of a braided tensor category. In...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Symmetry, Integrability and Geometry: Methods and Applications
Datum:2017
Hauptverfasser: Lentner, S., Ohrmann, T.
Format: Artikel
Sprache:Englisch
Veröffentlicht: Інститут математики НАН України 2017
Online Zugang:https://nasplib.isofts.kiev.ua/handle/123456789/148764
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:Factorizable R-Matrices for Small Quantum Groups / S. Lentner, T. Ohrmann // Symmetry, Integrability and Geometry: Methods and Applications. — 2017. — Т. 13. — Бібліогр.: 19 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1860026623733006336
author Lentner, S.
Ohrmann, T.
author_facet Lentner, S.
Ohrmann, T.
citation_txt Factorizable R-Matrices for Small Quantum Groups / S. Lentner, T. Ohrmann // Symmetry, Integrability and Geometry: Methods and Applications. — 2017. — Т. 13. — Бібліогр.: 19 назв. — англ.
collection DSpace DC
container_title Symmetry, Integrability and Geometry: Methods and Applications
description Representations of small quantum groups uq(g) at a root of unity and their extensions provide interesting tensor categories, that appear in different areas of algebra and mathematical physics. There is an ansatz by Lusztig to endow these categories with the structure of a braided tensor category. In this article we determine all solutions to this ansatz that lead to a non-degenerate braiding. Particularly interesting are cases where the order of q has common divisors with root lengths. In this way we produce familiar and unfamiliar series of (non-semisimple) modular tensor categories. In the degenerate cases we determine the group of so-called transparent objects for further use.
first_indexed 2025-12-07T16:50:27Z
format Article
fulltext Symmetry, Integrability and Geometry: Methods and Applications SIGMA 13 (2017), 076, 25 pages Factorizable R-Matrices for Small Quantum Groups Simon LENTNER and Tobias OHRMANN Fachbereich Mathematik, University of Hamburg, Bundesstraße 55, 20146 Hamburg, Germany E-mail: simon.lentner@uni-hamburg.de, tobias.ohrmann@uni-hamburg.de Received January 16, 2017, in final form September 15, 2017; Published online September 25, 2017 https://doi.org/10.3842/SIGMA.2017.076 Abstract. Representations of small quantum groups uq(g) at a root of unity and their extensions provide interesting tensor categories, that appear in different areas of algebra and mathematical physics. There is an ansatz by Lusztig to endow these categories with the structure of a braided tensor category. In this article we determine all solutions to this ansatz that lead to a non-degenerate braiding. Particularly interesting are cases where the order of q has common divisors with root lengths. In this way we produce familiar and unfamiliar series of (non-semisimple) modular tensor categories. In the degenerate cases we determine the group of so-called transparent objects for further use. Key words: factorizable; R-matrix; quantum group; modular tensor category; transparent object 2010 Mathematics Subject Classification: 17B37; 20G42; 81R50; 18D10 1 Introduction Hopf algebras with R-matrices, so called quasitriangular Hopf algebras, give rise to tensor cate- gories with a braiding c : V ⊗W ∼−→W ⊗V . Of particular interest are braided tensor categories where the braiding fulfills a certain non-degeneracy condition, see Definition 5.1, which is equiva- lent to the fact that there are no transparent objects V , i.e., no objects where the double-braiding c2 : V ⊗W ∼−→ V ⊗W is the identity for all W . A C-linear tensor category with a nondegenerate braiding, as well as finiteness conditions and another natural transformation θ : V ∼−→ V (twist), is called a modular tensor category. Note that we do not require the category to be semisimple. Modular tensor categories have many interesting applications: They give rise to topological invariants and mapping class group actions [7, 19]. For example, the standard generators T , S of the mapping class group of the torus SL2(Z) are constructed from θ and c2, respectively. A different source for modular tensor categories in mathematical physics are vertex algebras. There are only few example classes of modular tensor categories, in particular non-semisimple ones. The aim of the present article is to provide modular tensor categories from small quantum groups uq(g) at a primitive `-th root of unity q for a finite-dimensional simple complex Lie algebra g. Lusztig [12] has constructed these finite-dimensional Hopf algebras and provided an ansatz for an R-matrix R0Θ̄, where the fixed element Θ̄ ∈ uq(g)− ⊗ uq(g)+ is constructed from a dual basis of PBW generators, while R0 ∈ uq(g)0⊗ uq(g)0 is a free parameter subject to some constraints. He gives one canonical solution for R0 whenever ` has no common divisors with root lengths, otherwise there are cases where no R-matrix exists [8] and the quantum group becomes more interesting [9], involving, e.g., the dual root system. Of particular interest in conformal field theory [4, 5, 6] is the most extreme case where all root lengths (α, α) divide `. In particular our article addresses the question, which modular tensor category appear in these cases. We find indeed, e.g., in Lemma 4.7 that these extremal cases give especially nice R-matrices; although in general they are not factorizable and will require modularization (see for example [2]) to match the CFT side. mailto:simon.lentner@uni-hamburg.de mailto:tobias.ohrmann@uni-hamburg.de https://doi.org/10.3842/SIGMA.2017.076 2 S. Lentner and T. Ohrmann But even if there are no common divisors with the root length, the resulting braided tensor categories may not fulfill the non-degeneracy condition and hence provides no modular tensor category. Both obstacles (existence and non-degeneracy) can be be resolved by extending the Car- tan part of the quantum group by a choice of a lattice ΛR ⊆ Λ ⊆ ΛW between root- and weight-lattice, respectively a choice of a subgroup of the fundamental group π1 := ΛW /ΛR, corresponding to a choice of a Lie group between adjoint and simply-connected form. These ex- tensions are already present in [12] as the choice of two lattices X, Y with pairing X ×Y → C× (root datum). In this way the number of possible R matrices increases and the purpose of our paper is to study them all. In a previous article [11] we have already constructed some solutions R0 in this spirit (under some additional assumptions). As it turns out, the solutions can be parametrized by subgroups H1, H2 ⊂ π1 and group pairings between H1, H2, and the set of solutions depends on the common divisors of ` not only with root lengths, but also divisors of the Cartan matrix. Some cases admit no braided structure, while others have multiple in-equivalent solutions. An interesting occurrence was for example that Bn behaves differently for n odd or even, and that D2n with non-cyclic fundamental group allows several more solutions with non-symmetric R0. In the present article we conclude this effort: First we introduce more systematical techniques that allow us to compute a list of all quasitriangular structures (without additional assumptions, so we find more solutions). Then our new techniques allow us to determine, which of these choices fulfill the non-degeneracy condition. We also determine which cases have a ribbon structure. A main role in the first part is played by a natural pairing a` on the fundamental group π1 which depends only on the common divisors of ` with the fundamental group and encapsulates the essential `-dependence. Then the non-degeneracy of the braiding turns out to depend only on the 2-torsion of the abelian group in question. Our result produces a list of modular tensor categories for representations of quantum groups. Moreover we use our methods to explicitly describe the group of transparent objects if the category is not modular, which is for example a prerequisite for modularization. We now discuss our methods and results in more detail: In Section 2 we briefly recall the Lie theory and Hopf algebra preliminaries: For every finite- dimensional (semi-)simple complex Lie algebra g and a primitive `-th root of unity q Lusztig has introduced in [12] the small quantum group uq(g) which has a triangular decomposition u+ q u 0 qu − q where the (exponentiated) Cartan algebra u0 q is the groupring of the root lattice ΛR modulo some suitable sublattice and u±q are generated by simple root vectors Eαi , Fαi fulfilling q-deformed Serre relations. In [13, Section 32] he gives an ansatz for an R-matrix in the form R0Θ̄, where Θ̄ consists of dual PBW basis’ and R0 ∈ u0 q ⊗ u0 q is an arbitrary element in the Cartan part that has to fulfill certain relations. Our goal is to study the existence and non-degeneracy of R-matrices of this form for the quan- tum group uq(g,Λ,Λ ′) with any choice of lattice between root- and weight-lattice ΛR ⊆ Λ ⊆ ΛW and any possible choice of quotient by a subgroup Λ′ ⊆ Λ in the Cartan part u0 = C[Λ/Λ′]. Later, we prove that Λ′ is in fact unique if we want a quasitriagular structure (Corollary 3.6). The R0-matrix has the following interpretation: It is an R-matrix for the groupring C[Λ/Λ′] and it appears as the braiding between highest-weight vectors in our uq(g)-modules. Thus the previous theorem clarifies which choices for an R-matrix for the group ring lift to the quantum group. In Section 3 we address the question of constructing quasi-triangular R-matrices. First we briefly recall the following general combinatorial result in [11]: Theorem 3.3. The R0-matrix is necessarily of the form f(µ, ν) = 1 d q−(µ,ν)g(µ̄, ν̄)δµ̄∈H1δν̄∈H2 , Factorizable R-Matrices for Small Quantum Groups 3 where H1, H2 are subgroups of Λ/ΛR ⊆ π1 with |H1| = |H2| =: d (not necessarily isomorphic!) and g : H1 ×H2 → C× is a pairing of groups. Then we proceed differently than in the previous article: Using the previous result, we prove in Lemma 3.5 that the quasitriangularity of R is equivalent to the assertion that the group pairing f̂ := |Λ/Λ′| · f between the preimages Gi := Λi/Λ ′ of the groups Hi is non-degenerate (which is no surprise). In particular we show that this condition fixes Λ′ uniquely. In later applications we often encounter f̂ as a natural identification of G1 and the dual Ĝ2, e.g., when studying representation theory. To find all solutions f with this property we develop a machinery to push f̂ into the funda- mental group π1, which encapsulates all the `-dependence: In Definition 3.8 we give an abstract characterization of a centralizer transfer map A` : Λ/ΛR ∼−→ Cent`Λ(ΛR)/Cent`ΛR(Λ) (without proving that it always exists). In a generic case this is just multiplication by `, but it se- verely depends on common divisors of ` with root length and divisors of the Cartan matrix. With this matrix we can transfer q−(µ,ν) to a natural form a`g on the fundamental group. We prove that f̂ is non-degenerate iff a`g(µ, ν) = q−(µ,A`(ν)) · g(µ,A`(ν)) is non-degenerate. This explains why the set of solutions, say for fundamental group Zn always looks like the subset of invertible elements Z×n but it is shifted (namely by A`) depending on ` and the root system in question. In Section 4 the remaining computational work is done for quasitriagularity: We calculate a list containing a`g for all simple g, depending on common divisors of ` with root length and divisors. We thus write down all solutions for f and hence R-matrices. The calculation starts with the Smith normal form for the Cartan matrix in question and uses three cases: For Λ = ΛW we have a generic construction, the cases An with their large fundamental group Zn+1 is treated by hand, as isD2n with non-cyclic fundamental group, which has the only cases allowing Λ1 6= Λ2. In Section 5 we address our main issue of factorizability with our new tools: In Section 5.1 we introduce factorizability. Then we calculate the monodromy matrixR21R for an arbitrary choice of R-matrix in terms the R0-part. This gives a purely lattice theoretic prob- lem equivalent to the factorizability of such an R-matrix. Then we prove in the main Theorem 5.9 that factorizability is equivalent to the non-degeneracy of a symmetrization SymG ( f̂ ) of f̂ . As will turn out later, the radical of this form is isomorphic to the group of transparent objects. In Section 5.2 we restrict ourselves to the symmetric case where H1 = H2 and f , g are symmetric. Other cases appear only in some of the non-cyclic Z2×Z2-extension for type g = D2n and are dealt with in Section 5.3 and give surprising new solutions. The main result for the symmetric case is that the radical of the form SymG ( f̂ ) is in this case simply the 2-torsion of Λ/Λ′ (Example 5.11) and that this is non-degenerate precisely for odd ` and odd Λ/ΛR as well as for g = Bn, Λ = ΛR, ` ≡ 2 mod 4 including A1. In Section 5.4 we prove the following result: Lemma 5.19. The transparent objects in the category of representations of the Hopf algebra uq(g,Λ) with R-matrix given by Lusztig’s ansatz are 1-dimensional objects Cχ and are the f - transformed of the radical of SymG ( f̂ ) : χ(µ) = f(µ, ξ), ξ ∈ Rad ( SymG ( f̂ )) . In the following we summarize our results by a table containing all quasitriangular quantum groups uq(g,Λ) with their group of transparent objects. In Section 6 we show that all our quasitriangular quantum groups admit a ribbon element. The factorizable solutions and thus modular tensor categories are ` odd, Λ = ΛR and the following new factorizable cases: 4 S. Lentner and T. Ohrmann (` odd, E6, Λ = ΛW ) and (` ≡ 2 mod 4, g = Bn, Λ = ΛR) (including A1) and (` odd, g = D2n, Λ1 6= Λ2). All other cases can be modularized as discussed in Question 7.3. The columns of the following table are labeled by 1) the finite-dimensional simple complex Lie algebra g, 2) the natural number `, determining the root of unity q = exp ( 2πi ` ) , 3) the number of possible R-matrices for the Lusztig ansatz, 4) the subgroups Hi ⊆ H = Λ/ΛR introduced in Theorem 3.3, 5) the subgroups Hi in terms of generators given by multiples of fundamental dominant weights λi ∈ ΛW , 6) the group pairing g : H1 ×H2 → C× determined by its values on generators, 7) the group of transparent objects T ⊆ Λ/Λ′ introduced in Lemma 5.19. g ` # Hi ∼= Hi (i=1,2) g T ⊆ Λ/Λ′ all ` odd 1 Z1 〈0〉 g = 1 0 ` ≡ 0 mod 4 1 Zn2 ∞ Zn−1 2 , 2 - x Zd 〈d̂λn〉 g(d̂λn, d̂λn) = exp ( 2πik d ) An≥1 Zn2 , 2 |x π1 = Zn+1 d |n+ 1 d̂ = n+1 d gcd ( d, k`−d̂n gcd(`,d̂) ) = 1 x = d` gcd(`,d̂) ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 0 ` ≡ 2 mod 4 2 Z2 〈λn〉 g(λn, λn) = ±1 Z2 Bn≥2 ` ≡ 0 mod 4 2 Z2 〈λn〉 g(λn, λn) = ±1 Zn2π1 = Z2 ` odd 1 Z2 〈λn〉 g(λn, λn) = (−1)n+1 Z2 ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Zn−2 2 ` ≡ 2 mod 4 1 Z2 〈λn〉 g(λn, λn) = 1 Zn−1 2 Cn≥3 ` ≡ 0 mod 4 2 g(λn, λn) = ±1 Zn2π1 = Z2 ` odd 1 g(λn, λn) = −1 Z2 ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Z2(n−1) 2 ` ≡ 2 mod 4 1 Z2 H1 ∼= 〈λ2n−1〉 g(λ2n−1, λ2n) = (−1)n Z2n 2 ` ≡ 0 mod 4 2δ2 |n g(λ2n−1, λ2n) = ±1, n even H2 ∼= 〈λ2n〉` odd 1 g(λ2n−1, λ2n) = −1 0 D2n≥4 π1 = Z2 × Z2 ` ≡ 2 mod 4 1 Z2 〈λ2n〉 g(λ2n, λ2n) = (−1)n+1 Z2n−1 2 ` ≡ 0 mod 4 2δ2-n g(λ2n, λ2n) = ±1, n odd Z2n 2 ` odd 1 g(λ2n, λ2n) = −1 Z2 ` even 16 Z2 × Z2 〈λ2n, λ2n+1〉 g(λ2(n−1)+i, λ2(n−1)+j) = ±1 Z2n 2 ` odd 6 det(K) = K12 +K21 = 0 mod 2 Z2 det(K) = K12 +K21 = 1 mod 2 Z2 2 Factorizable R-Matrices for Small Quantum Groups 5 ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Z2n 2 ` ≡ 2 mod 4 1 Z2 〈2λ2n+1〉 g(2λ2n+1, 2λ2n+1) = 1 Z2n+1 2 ` ≡ 0 mod 4 2 g(2λ2n+1, 2λ2n+1) = ±1 D2n+1≥5 ` odd 1 g(2λ2n+1, 2λ2n+1) = −1 Z2 π1 = Z4 ` even 4 Z4 〈λ2n+1〉 g(λ2n+1, λ2n+1) = c, c4 = 1 Z2n+1 2 ` odd 2 g(λ2n+1, λ2n+1) = ±1 Z2 ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Z6 2 ` ≡ 0 mod 3 3 Z3 〈λn〉 g(λn, λn) = c, c3 = 1 Z6 2, 2 | ` E6 ` ≡ 1 mod 3 2 g(λn, λn) = 1, exp ( 2πi2 3 ) π1 = Z3 0, 2 - ` ` ≡ 2 mod 3 2 g(λn, λn) = 1, exp ( 2πi 3 ) ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Z6 2 ` even 2 Z2 〈λn〉 g(λn, λn) = ±1 Zn2E7 π1 = Z2 ` odd 1 g(λn, λn) = 1 Z2 E8 ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Z8 2 F4 ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Z2 2 G2 ` ≡ 2 mod 4 1 Z1 〈0〉 g = 1 Z2 2 Table 1: Solutions for R0-matrices. In the case D2n, Λ = ΛW , g is uniquely defined by a (2 × 2)-matrix K ∈ gl(2,F2), s.t. g(λ2(n−1)+i, λ2(n−1)+j) = exp (2πiKg ij 2 ) for i, j ∈ {1, 2}. 2 Preliminaries 2.1 Lie theory Throughout this article, g denotes a finite-dimensional simple complex Lie algebra. We fix a choice of simple roots ∆ = {αi | i ∈ I}, so that the Cartan matrix C is given by Cij = 2 (αi,αj) (αi,αi) , where ( , ) denotes the normalized Killing form. For a root α, we define dα := (α,α) 2 and set di = dαi . By ΛR := Z[∆] and Λ∨R := Z[∆∨] we denote the (co)root lattice of g. By ΛW , we denote the weight lattice spanned by fundamental dominant weights λi, which are defined by the equation (λi, αj) = δi,jdi. Finally, we define the co-weight lattice Λ∨W as the Z- span of the elements λ∨i := λi di . The quotient π1 := ΛW /ΛR is called the fundamental group of g. One can easily see that the Killing form restricts to a perfect pairing ( , ) : Λ∨W ×ΛR → Z and that we get a string of inclusions ΛR ⊆ Λ∨R ⊆ ΛW ⊆ Λ∨W . 2.2 Lusztig’s ansatz for R-matrices The starting point for our work [11] was Lusztig’s ansatz in [13, Section 32.1] for a universal R- matrix of Uq(g). Namely, for a specific element Θ̄ ∈ U≥0 q ⊗U≤0 q from a dual basis and a suitable (not further specified) element in the coradical R0 ∈ U0 q ⊗ U0 q we are looking for R-matrices of 6 S. Lentner and T. Ohrmann the form R = R0Θ̄. We remark that there is no claim that all possible R-matrices are of this form. However they are an interesting source of examples, motivated by the interpretation of uq(g) as a quotient of a Drinfeld double and thus well-behaved with respect to the triangular decomposition. This ansatz has been successfully generalized to general diagonal Nichols algebras in [1]. In our more general setting Uq(g,Λ,Λ ′), we have R0 ∈ C[Λ/Λ′]⊗ C[Λ/Λ′]. This ansatz has been worked out by Müller in his dissertation [14, 15, 16] for small quantum groups uq(g) which we will use in the following, leading to a system of quadratic equation on R0 that are equivalent to R being an R-matrix: Theorem 2.1 (cf. [16, Theorem 8.2]). Let u := uq(g). (a) There is a unique family of elements Θβ ∈ u−β ⊗ u + β , β ∈ ΛR, such that Θ0 = 1 ⊗ 1 and Θ = ∑ β Θβ ∈ u⊗ u satisfies ∆(x)Θ = Θ∆̄(x) for all x ∈ u. (b) Let B be a vector space-basis of u−, such that Bβ = B ∩ u−β is a basis of u−β for all β. Here, u−β refers to the natural ΛR-grading of u−. Let {b∗ | b ∈ Bβ} be the basis of u+ β dual to Bβ under the non-degenerate bilinear form ( · , · ) : u− ⊗ u+ → C. We have Θβ = (−1)trβqβ ∑ b∈Bβ b− ⊗ b∗+ ∈ u−β ⊗ u + β . Theorem 2.2 (cf. [16, Theorem 8.11]). Let Λ′ ⊂ {µ ∈ Λ |Kµ central in uq(g,Λ)} a subgroup of Λ, and G1, G2 subgroups of G := Λ/Λ′, containing ΛR/Λ ′. In the following, µ, µ1, µ2 ∈ G1 and ν, ν1, ν2 ∈ G2. The element R = R0Θ̄ with an arbitrary R0 = ∑ µ,ν f(µ, ν)Kµ⊗Kν is a R-matrix for uq(g,Λ,Λ ′), if and only if for all α ∈ ΛR and µ, ν the following holds: f(µ+ α, ν) = q−(ν,α)f(µ, ν), f(µ, ν + α) = q−(µ,α)f(µ, ν), (2.1)∑ ν1+ν2=ν f(µ1, ν1)f(µ2, ν2) = δµ1,µ2f(µ1, ν), ∑ µ1+µ2=µ f(µ1, ν1)f(µ2, ν2) = δν1,ν2f(µ, ν1),∑ µ f(µ, ν) = δν,0, ∑ ν f(µ, ν) = δµ,0. 3 Conditions for the existence of R-matrices 3.1 A first set of conditions on Λ/Λ′ The target of our efforts is a Hopf algebra called small quantum group uq(g,Λ,Λ ′) with Cartan part u0 q = C[Λ/Λ′]. It is defined, e.g., in [11] and depends on lattices Λ, Λ′ defined below. For Λ = ΛR the root lattice and this is the usual small quantum group; the choice of Λ′ differs in literature. In the previous section we have discussed an R = R0Θ̄-matrix for the quantum group uq(g,Λ,Λ ′) can be obtained from an R0-matrix of the form R0 = ∑ µ,ν∈Λ f(µ, ν)Kµ ⊗Kν ∈ C[Λ/Λ′]⊗ C[Λ/Λ′]. In the following we collect necessary and sufficient conditions for R = R0Θ̄ to be an R-matrix. Factorizable R-Matrices for Small Quantum Groups 7 Definition 3.1. We fix once-and-for-all a finite-dimensional simple complex Lie algebra g and a lattice Λ between root- and weight-lattice ΛR ⊆ Λ ⊆ ΛW . These choices have a nice geometric interpretation as quantum groups associated to different Lie groups associated to the Lie algebra g. Another interesting choice is ΛR ⊆ Λ ⊆ Λ∨W ∼= Λ∗R, which would below pose no additional complications and may produce further interesting factorizable R-matrices. Definition 3.2. We fix once-and-for-all a primitive `-th root of unity q. For Λ1,Λ2 ⊆ Λ∨W we define the sublattice CentΛ1(Λ2) := { ν ∈ Λ1 | (ν, µ) ∈ ` · Z ∀µ ∈ Λ2}. Informally, this is the centralizer with respect to the braiding q−(ν,µ). Contrary to [11] we do not fix Λ′ but we prove later Corollary 3.6 that there is a necessary choice for Λ′. In this way, we get more solutions than in [11]. The only condition necessary to ensure that the Hopf algebra uq(g,Λ,Λ ′) is well-defined is Λ′ ⊆ CentΛR(ΛR). Theorem 3.3 (cf. [11, Theorem 3.4]). The R0-matrix is necessarily of the form f(µ, ν) = 1 d|ΛR/Λ′| · q−(µ,ν)g(µ̄, ν̄)δµ̄∈H1δν̄∈H2 , where H1, H2 are subgroups of H := Λ/ΛR ⊆ π1 with equal cardinality |H1| = |H2| =: d (not necessarily isomorphic!) and g : H1 ×H2 → C× is a pairing of groups. The necessity of this form (in particular that the support of f is indeed a subgroup!) amounts to a combinatorial problem of its own interest, which we solved for π1 cyclic in [10] and for Z2×Z2 by hand; a closed proof for all abelian groups would be interesting. Definition 3.4. Let g : G×H → C× be a finite group pairing, then the left radical is defined as RadL(g) := {λ ∈ G | g(λ, η) = 1 ∀ η ∈ H}. Similarly, the right radical is defined as RadR(g) := {η ∈ H | g(λ, η) = 1 ∀λ ∈ G}. The pairing g is called non-degenerate if RadL(g) = 0. If in addition RadR(g) = 0, g is called perfect. For an R0-matrix of this form, a sufficient condition is that they fulfill the so-called diamond- equations (see [11, Definition 2.7]) for each element 0 6= ζ ∈ (Cent(ΛR) ∩ Λ)/Λ′. However, we will now go into a different, more systematic direction that makes use of the following observation: Lemma 3.5. An R0-matrix of the form given in Theorem 3.3 is a solution to the equations in Theorem 2.2, and hence produces an R-matrix R0Θ̄ iff the restriction to the support f̂ := d|ΛR/Λ′| · f : G1 ×G2 → C× is a perfect group pairing, where Gi := Λi/Λ ′ ⊆ Λ/Λ′ =: G. 8 S. Lentner and T. Ohrmann Proof. We first show that a solution with restriction to the support a nondegenerate pairing solves the equation. The first equations are obviously fulfilled for the form assumed f(µ+ α, ν) = q−(ν,α)f(µ, ν), f(µ, ν + α) = q−(µ,α)f(µ, ν). For the other equations the sums get only contributions in the support Λ1/Λ ′×Λ2/Λ ′. The quan- tities f(µ, ν) ·d|ΛR/Λ′| for fixed ν (or µ) are characters on the respective support, and by the as- sumed non-degeneracy all ν 6= 0 give rise to different nontrivial characters. Then the second and third relations follows from orthogonality of characters. Note that since d|ΛR/Λ′| = |G1| = |G2| (equality of the latter was an assumption!) we were able to chose the right normalization. For the other direction assume a solution of the given form to the equations. Then already the third equation shows that no f(−, ν) may be the trivial character and hence the form on the support is nondegenerate and hence perfect by |G1| = |G2|. � Corollary 3.6. A first consequence of the perfectness of f̂ (i.e., a necessary condition for quasi- triangularity) is CentΛR(Λ1) = CentΛR(Λ2) = Λ′. This fixes Λ′ uniquely. Moreover in cases Λ1 6= Λ2, which can only happen for g = D2n, where π1 is noncyclic, we get an additional constraint relating Λ1, Λ2. In our case, the only possibility for Λ1 6= Λ2, s.t. G1 ∼= G2 is g = D2n. In this case, we have CentΛR(ΛW ) = CentΛR(ΛR) and thus the above condition is always fulfilled. Our main goal for the new approach on quasitriangularity as well as the later modularity is to reduce this non-degeneracy condition for f̂ to a non-degeneracy condition for g on H1, H2 ⊂ π1 that can be checked explicitly. 3.2 A natural form on the fundamental group We now define for each triple (Λ,Λ1,Λ2) and each `th root of unity q a natural pairing a` on the subgroups Hi := Λi/ΛR of the fundamental group π1 := ΛW /ΛR. The simplest example is a` = e−2πi(µ,ν). In general it is a transportation of the natural form q−(µ,ν) (which does not factorize over ΛR) to Hi by a suitable isomorphism A`. This isomorphism A` will encapsulate the crucial dependence on the common divisors of `, |H| and the root lengths di; moreover, for different H these forms are not simply restrictions of one another. Then, we can moreover transport any given pairing g together with q−(µ,ν) along the iso- morphism A` to the Hi and thus define forms ag` on H. The main result of this section is in Theorem 3.13 that the non-degeneracy condition in Lemma 3.5 for R0(f) depending on Hi, g is equivalent to ag` being non-degenerate. Definition 3.7. Let Λ ⊆ Λ∨W be a sublattice, s.t. ΛR ⊆ Λ. By Λ̂ ⊂ Λ∨W we denote the unique sublattice, s.t. the symmetric bilinear form ( · , · ) : Λ∨W ×Λ∨W → Q induces a commuting diagram ΛR Λ̂ Λ∨W Λ∨W ∗ Λ∗ Λ∗R, ∼= ∼= ∼= where Λ∗ := HomZ(Λ,Z). In particular, we have Λ̂R = Λ∨W and Λ̂∨W = ΛR. Factorizable R-Matrices for Small Quantum Groups 9 Definition 3.8. A centralizer transfer map is an group endomorphism A` ∈ EndZ(Λ), s.t. 1) A`(Λ) ! = Λ ∩ ` · Λ̂R = Cent`Λ(ΛR), 2) A`(ΛR) ! = ΛR ∩ ` · Λ̂ = Cent`ΛR(Λ). Such a A` induces a group isomorphism Λ/ΛR ∼−→ Cent`Λ(ΛR)/Cent`ΛR(Λ). Of course A` is not unique. Question 3.9. Are there abstract arguments for the existence of these isomorphism and for its explicit form? We will calculate explicit expressions for A` depending on the cases in the next section. At this point we give the generic answers: Example 3.10. For Λ = Λ∨W we have A` = ` · id. For Λ = ΛR the two conditions are equivalent, so existence is trivial (resp. obviously the two trivial groups are isomorphic) and we may simply take for A` any base change between left and right side. The expression may however be nontrivial. Lemma 3.11. Assume gcd(`, |Λ∨W /Λ|) = 1, then A` = ` · id. In particular this is the case if ` is prime to all root lengths and all divisors of the Cartan matrix. Moreover if ` = `1`2 with gcd(`1, |Λ∨W /Λ|) = 1, then A` = `1 ·A`2. This means we only have to calculate A` for all divisors ` of |Λ∨W /Λ|, which is a subset of all divisors of root lengths times divisors of the Cartan matrix. Proof. For the first condition we need to show for any λ ∈ Λ∨W that `λ ∈ Λ already implies λ ∈ Λ. But if by assumption the order of the quotient group Λ∨W /Λ is prime to `, then `· is an isomorphism on this abelian group, hence follows the assertion. For the second condition applies the same argument noting that |Λ̂/ΛR| = |Λ∨W /Λ|. For the second claim we simply consider the inclusion chains A`(Λ) ⊂ Λ ∩ `2 · Λ̂R ⊂ Λ ∩ ` · Λ̂R, A`(ΛR) ⊂ Λ ∩ `2 · Λ̂ ⊂ ΛR ∩ ` · Λ̂, where a first isomorphism is given by A`2 and again `1· is a second isomorphism because it is prime to the index. � Our main result of this chapter is the following: Theorem 3.12. Let ΛR ⊆ Λ1, Λ2 ⊆ ΛW be intermediate lattices, s.t. the condition in Corol- lary 3.6 is fulfilled, i.e., CentΛR(Λ1) = CentΛR(Λ2) = Λ′. Assume we have a centralizer transfer map A`. 1. The following form is well defined on the quotients: a`g : Λ1/ΛR × Λ2/ΛR −→ C×, (λ̄, µ̄) 7−→ q−(λ,A`(µ)) · g(λ,A`(µ)). 2. Let CentgΛ1 (Λ2) := { λ ∈ Λ1 | q(λ,µ) = g(λ, µ) ∀µ ∈ Λ2 } . Then the inclusion CentgΛ1 (Λ2) ↪→ Λ1 induces an isomorphism CentgΛ1 (Λ2)/Λ′ ∼= Rad ( a`g ) . 10 S. Lentner and T. Ohrmann Corollary 3.13. The quasitriangularity conditions for a choice R0 are by Lemma 3.5 equivalent to the non-degeneracy of the group pairing on Λ1/Λ ′ × Λ2/Λ ′: f̂(λ, µ) = q−(λ,µ)g(λ, µ). By the previous theorem this condition is now equivalent to the nondegeneracy of a`g. This condition on the fundamental group, which is a finite abelian group and mostly cyclic, can be checked explicitly once a`g has been calculated. Proof of Theorem 3.12. The first part of the theorem is a direct consequence of the definition of the centralizer transfer matrix A`. For the second part, we first notice that by assumption we have a commutative diagram of finite abelian groups ΛR/Λ ′ Λ1/Λ ′ Λ1/ΛR (Λ2/CentΛ2(ΛR))∧ (Λ2/Λ ′)∧ (CentΛ2(ΛR)/Λ′)∧ , q−(·,·) f̂ f̂ ′ where G∧ denotes the dual group of a group G. Now, by the five lemma we know that f̂ is an isomorphism if and only if the induced map f̂ ′ is an isomorphism. Post-composing this map with the dualized centralizer transfer matrix A∧` : (CentΛ2(ΛR)/Λ′)∧ ∼= (Λ2/ΛR)∧ gives a`g. � 4 Explicit calculation for every g In the following, we want to compute the endomorphism A` ∈ EndZ(Λ) and the pairing a` on the fundamental group explicitly in terms of the Cartan matrices and the common divisors of ` with root lengths and divisors of the Cartan matrix. We will finally give a list for all g. 4.1 Technical tools We choose the basis of simple roots αi for ΛR and the dual basis of fundamental coweights λ∨i for the dual lattice Λ∨W with (αi, λ ∨ j ) = δi,j . For any choice Λ ⊂ ΛW ⊂ Λ∨W , let AΛ be a basis matrix, i.e., any Z-linear isomorphism Λ∨W → Λ sending the basis λ∨i of Λ∨W to some basis µi of Λ. It is unique up to pre-composition of a unimodular matrix U ∈ SLn(Z). The dual basis AΛ̂ of Λ̂ is defined by( AΛ̂ ( λ∨i ) , AΛ ( λ∨j )) = δij . Explicitly, AΛ̂ is given by AΛ̂ = ( A−1 Λ AR )T , where (AR)ij = (αi, αj). Now, let AΛ = PΛSΛQΛ be the unique Smith decomposition of AΛ, which means: PΛ, QΛ are unimodular and SΛ is diagonal with diagonal entries (SΛ)ii =: dΛ i , such that dΛ i | dΛ j for i < j. Example 4.1. For the root lattice the dΛR i are the divisors of scalar product matrix (αi, αj). Their product is∏ i dΛR i = ∣∣Λ∨W /ΛR∣∣ = (∏ i di ) · |π1|, di = (αi, αi) 2 . For the coweight lattice all d Λ∨W i = 1. For the weight lattice we recover the familiar dΛW i = di. Factorizable R-Matrices for Small Quantum Groups 11 Without loss of generality, we will assume the basis matrices AΛ to be symmetric, i.e., QΛ = P TΛ . We then have the following lemma: Lemma 4.2. Let ΛR ⊆ Λ ⊆ Λ∨W be a lattice. We define lattices ACent := ( P TΛ )−1 D`P −1 Λ , D` := Diag ( ` gcd ( `, dΛ i )) . Then, CentΛR(Λ) = ARACentΛ ∨ W , CentΛ(ΛR) = AΛACentΛ ∨ W . Proof. We compute explicitly, CentΛR(Λ) = ΛR ∩ ` · Λ̂ = ARΛ∨W ∩ ( A−1 Λ AR )T `Λ∨W = ( A−1 Λ AR )T ((( A−1 Λ AR )T )−1 ARΛ∨W ∩ `Λ∨W ) = ARA −1 Λ ( AΛ ∩ `Λ∨W ) = AR ( PΛSΛP T Λ )−1( PΛSΛP T Λ Λ∨W ∩ `Λ∨W ) = AR ( P TΛ )−1 S−1 Λ ( SΛΛ∨W ∩ `Λ∨W ) = AR ( P TΛ )−1 S−1 Λ Diag(lcm(SΛii , `))Λ ∨ W = AR ( P TΛ )−1 D`Λ ∨ W = ARACentΛ ∨ W . On the other hand, CentΛ(ΛR) = Λ ∪ `Λ̂R = Λ ∪ `Λ∨W = AΛΛ∨W ∪ `Λ∨W = PΛSΛP T Λ Λ∨W ∪ `Λ∨W = PΛ ( SΛΛ∨W ∪ `Λ∨W ) = PΛSΛD`Λ ∨ W = AΛ ( P TΛ )−1 D`Λ ∨ W = AΛACentΛ ∨ W . In particular, this means AΛ̂ CentΛ(ΛR) = CentΛR(Λ). � 4.2 Case Λ = ΛW In order to exhaust all cases that appear in our setting, we continue with Λ = ΛW : Lemma 4.3. In the case Λ = ΛW , the centralizer transfer matrix A` is of the following form: A` = { AΛWACentQ T CP −1 C A−1 ΛW , gcd(`, |π1|) 6= 1, ` · id, else. Here, C = PCSCQC denotes the Smith decomposition of the Cartan matrix of g. Proof. As we noted in Example 4.1, we have AΛW = Diag(di), for di being the ith root length. Since di ∈ {1, p} for some prime number p, up to a permutation AΛW is already in Smith normal form: this means that PΛW is a permutation matrix of the form (PΛW )ij = δj,σ(i) for some σ ∈ Sn, s.t. dσ(1) ≤ · · · ≤ dσ(n). It follows that ACent = Diag ( ` gcd(`,di) ) . Using the definition Cij = (αi,αj) di , in the case gcd(`, |π1|) 6= 1 we obtain ACentC T = CACent. Thus, A`AR = AΛWACentQ T CP −1 C A−1 ΛW AR = ARC −1ACentQ T CP −1 C C = ARACent ( CT )−1 QTCP −1 C C = ARACent. By the previous lemma, this proves the first condition for A`. The second condition follows immediately from the previous lemma. The case gcd(`, |π1|) = 1 follows from Lemma 3.11 and the fact that |π1| = |Λ∨W /Λ∨R|. � 12 S. Lentner and T. Ohrmann 4.3 Case An In the following example, we treat the case g = An with fundamental group ΛW /ΛR = Zn+1 for general intermediate lattices ΛR ⊆ Λ ⊆ ΛW . Example 4.4. In order to compute the centralizer transfer map A`, we first compute the Smith decomposition of AR: AR =  2 −1 0 . . . 0 −1 2 −1 0 0 0 −1 2 . . . . . . ... 0 0 . . . . . . −1 0 ... . . . −1 2 −1 0 . . . 0 −1 2  =  −1 0 0 . . . 0 2 −1 0 0 0 2 −1 . . . ... 0 0 . . . . . . 0 ... . . . 2 −1 0 0 . . . 0 2 1   1 0 0 . . . 0 0 1 0 0 0 0 1 . . . ... ... . . . . . . . . . 1 0 0 . . . 0 n+ 1   −2 1 0 . . . 0 −3 0 1 . . . 0 −4 0 0 . . . ... ... ... . . . 1 0 −n 0 1 1 0 . . . 0 0  . A sublattice ΛR ( Λ ( ΛW is uniquely determined by a divisor d |n + 1, so that Λ/ΛR ∼= Zd and is generated by the multiple d̂λn, where d̂ := n+1 d . Then dΛ i = { 1, i < n, d, i = n. Since An is simply laced with cyclic fundamental group, the formula AΛ = PRSΛP T R gives us symmetric basis matrices of sublattices ΛR ⊆ Λ ⊆ ΛW . We also substitute the above basis matrix of the root lattice AR by AR(QR)−1P TR . It is then easy to see that the definition A` := PRD`P T R gives a centralizer transfer matrix. We calculate it explicitly (A`)ij = ( PRD`P −1 R ) ij =  δij , i < n, (n+ 1− j) ( ` gcd(`, d) − 1 ) , i = n and j < n, ` gcd(`, d) , i = j = n. Now a form g is uniquely determined by a dth root of unity g(χ, χ) = exp ( 2πi·k d ) = ζkd with some k. Then we calculate the form a`g on the generator a`g(χ, χ) = q−(χ,A`(χ))g(χ,A`(χ)) = q − (n+1)2·` d2 gcd(`,d̂) (λ∨n ,λ ∨ n) · g(χ, χ) ` gcd(`,d̂) = exp ( 2πi · (k`− d̂n) d · gcd(`, d̂) ) . For example the trivial g (i.e., k = 0) gives an R-matrix for all lattices Λ (defined by d̂d = n+1) iff d̂ gcd(`,d̂) is coprime to d. For ` coprime to the divisor n+1 this amounts to all lattices associated to decompositions of n+ 1 into two coprime factors. Factorizable R-Matrices for Small Quantum Groups 13 4.4 Case Dn Finally, we consider the root lattice Dn. Since we have π1(D2n≥4) ∼= Z2×Z2 and π1(D2n+1≥5) ∼= Z4, it is appropriate to split this investigation in two steps. We start with D2n≥4. In order to compute the respective Smith decompositions, we used the software Wolfram Mathematica. Example 4.5. In the case D2n≥4, we have three different possibilities for the lattices ΛR ⊆ Λ1,Λ2 ⊆ ΛW : 1. Λ1 6= Λ2, H1 ∼= H2 ∼= Z2: In this case, the subgroups Λi/ΛR ⊆ ΛR are spanned by the fundamental weights λ2(n−1)+i. As in the case An, we define the centralizer transfer map A` := PRD`P −1 R on H2. This is possible since the symmetric basis matrix AΛ2 = PRSΛ2P T R of Λ2 is already in Smith normal form. Using the software Wolfram Mathematica in order to compute PR, we obtain A`(λ2n) = ` gcd(2,`) . Combining this with (λ2n−1, λ2n) = n−1 2 , we get a`g(λ2n−1, λ2n) = exp ( 2πi · (kl − 2(n− 1)) 2 · gcd(2, `) ) for g(λ2n−1, λ2n) = exp ( 2πik 2 ) . 2. Λ1 = Λ2, Hi ∼= Z2: Without restrictions and in order to use the same definition for A` as above, we choose Λi, s.t. the group Λi/ΛR is spanned by λ2n. Combining the above result A`(λ2n) = ` gcd(2,`) with (λ2n, λ2n) = n 2 , we obtain a`g(λ2n, λ2n) = exp ( 2πi · (kl − 2n) 2 · gcd(2, `) ) for g(λ2n, λ2n) = exp ( 2πik 2 ) . 3. Λ1 = Λ2 = ΛW , H ∼= Z2×Z2: A group pairing g : (Z2×Z2)× (Z2×Z2)→ C× is uniquely defined by a matrix K ∈ gl(2,F2), so that g(λ2(n−1)+i, λ2(n−1)+j) = exp ( 2πiKij 2 ) . Since Dn is simply-laced, we have A` = ` · id. Using (λ2(n−1)+i, λ2(n−1)+j) mod 2 = δi+jodd, we obtain ag` (λ2(n−1)+i, λ2(n−1)+j) = exp ( 2πi ·Kij` 2 ) (−1)i+j . The last step is the case D2n+1≥5: Example 4.6. Since it it is simply-laced and its fundamental group is cyclic, the case D2n+1≥5 can be treated very similar to An. We distinguish two cases: 1. Λ1 = Λ2, Hi = 〈2λ2n+1〉 ∼= Z2. As in the case An, we define the centralizer transfer map A` := PRD`P −1 R on H2. Using (λ2n+1, λ2n+1) = 2n+1 4 , we obtain a`g(2λ2n+1, 2λ2n+1) = exp ( 2πi · (k`− 2(2n+ 1)) 2 · gcd(2, `) ) . for g(2λ2n+1, 2λ2n+1) = exp ( 2πik 2 ) . 2. Λ1 = Λ2 = ΛW , H = 〈λ2n+1〉 ∼= Z4. By an analogous argument as above, we obtain a`g(λ2n+1, λ2n+1) = exp ( 2πi · (k`− (2n+ 1)) 4 ) for g(λ2n+1, λ2n+1) = exp ( 2πik 4 ) . 14 S. Lentner and T. Ohrmann 4.5 Table of all quasitriangular quantum groups In the following table, we list all simple Lie algebras and check for which non-trivial choices of Λ, Λi, ` and g the element R0Θ̄ is an R-matrix. As before, we define Hi := Λi/ΛR and H := Λ/ΛR. In the cyclic case, if xi are generators of the Hi, then the pairing is uniquely defined by an element 1 ≤ k ≤ |Hi|, s.t. g(x1, x2) = exp ( 2πik |Hi| ) . In the case D2n, Λ = ΛW , g is uniquely defined by a 2× 2-matrix K ∈ gl(2,F2), s.t. g(λ2(n−1)+i, λ2(n−1)+j) = exp (2πiKg ij 2 ) for i, j ∈ {1, 2}. The columns of the following table are labeled by 1) the finite-dimensional simple complex Lie algebra g, 2) the natural number `, determining the root of unity q = exp ( 2πi ` ) , 3) the number of possible R-matrices for the Lusztig ansatz, 4) the subgroups Hi ⊆ H = Λ/ΛR introduced in Theorem 3.3, 5) the subgroups Hi in terms of generators given by multiples of fundamental dominant weights λi ∈ ΛW , 6) the group pairing g : H1 ×H2 → C× determined by its values on generators, 7) the group pairing a`g ⊆ Λ/Λ′ introduced in Theorem 3.12 determined by its values on generators. g ` # Hi ∼= Hi (i=1,2) g a`g all 1 Z1 〈0〉 g = 1 1 ∞ Zd 〈d̂λn〉 g(d̂λn, d̂λn) = exp ( 2πik d ) exp ( 2πi·(k`−d̂n) d·gcd(`,d̂) )An≥1 π1 = Zn+1 d |n+ 1 d̂ = n+1 d gcd ( d, k`−d̂n gcd(`,d̂) ) = 1 ` even 2 Z2 〈λn〉 g(λn, λn) = ±1 −1 Bn≥2 π1 = Z2 ` odd 1 g(λn, λn) = (−1)n+1 exp ( 2πi·(k`−n) 2 ) ` ≡ 2 mod 4 1 Z2 〈λn〉 g(λn, λn) = 1 exp ( 2πi·(k ` 2 +1) 2 ) Cn≥3 ` ≡ 0 mod 4 2 g(λn, λn) = ±1 −1 π1 = Z2 ` odd 1 g(λn, λn) = −1 exp ( 2πi·(k`−2n) 2 ) ` ≡ 2 mod 4 1 Z2 H1 ∼= 〈λ2n−1〉 g(λ2n−1, λ2n) = (−1)n exp ( 2πi·(k ` 2 −n+1)) 2 ) ` ≡ 0 mod 4 2δ2 |n g(λ2n−1, λ2n) = ±1, n even H2 ∼= 〈λ2n〉` odd 1 g(λ2n−1, λ2n) = −1 exp ( 2πi·(k`−2(n−1)) 2 ) D2n≥4 π1 = Z2 × Z2 ` ≡ 2 mod 4 1 Z2 〈λ2n〉 g(λ2n, λ2n) = (−1)n+1 exp ( 2πi(k ` 2 −n) 2 ) ` ≡ 0 mod 4 2δ2-n g(λ2n, λ2n) = ±1, n odd ` odd 1 g(λ2n, λ2n) = −1 exp ( 2πi(k`−2n) 2 ) ` even 16 Z2 × Z2 〈λ2n, λ2n+1〉 g(λ2(n−1)+i, λ2(n−1)+j) = ±1 exp ( 2πi·Kij` 2 ) (−1)i+j ` odd 6 det(K) = K12 +K12 mod 2 Factorizable R-Matrices for Small Quantum Groups 15 ` ≡ 2 mod 4 1 Z2 〈2λ2n+1〉 g(2λ2n+1, 2λ2n+1) = 1 exp ( 2πi·(k ` 2 −2n−1) 2 ) ` ≡ 0 mod 4 2 g(2λ2n+1, 2λ2n+1) = ±1 D2n+1≥5 ` odd 1 g(2λ2n+1, 2λ2n+1) = −1 exp ( 2πi·(k`−2(2n+1)) 2 ) π1 = Z4 ` even 4 Z4 〈λ2n+1〉 g(λ2n+1, λ2n+1) = c, c4 = 1 exp ( 2πi·(k`−(2n+1)) 4 ) ` odd 2 g(λ2n+1, λ2n+1) = ±1 ` ≡ 0 mod 3 3 Z3 〈λn〉 g(λn, λn) = c, c3 = 1 exp ( 2πi·(k`−1) 3 )E6 ` ≡ 1 mod 3 2 g(λn, λn) = 1, exp ( 2πi2 3 ) π1 = Z3 ` ≡ 2 mod 3 2 g(λn, λn) = 1, exp ( 2πi 3 ) ` even 2 Z2 〈λn〉 g(λn, λn) = ±1 exp ( 2πi·(k`−1) 2 )E7 π1 = Z2 ` odd 1 g(λn, λn) = 1 Table 2: Solutions for R0-matrices. The Lie algebras E8, F4 and G2 have trivial fundamental groups and thus have no non- trivial solution. We want to emphasize once more that the choice Λi = ΛR always leads to a quasitriangular quantum group. The following lemma connects our results with Lusztig’s original result: Lemma 4.7. In Lusztig’s definition of a quantum group he uses the quotient Λ′Lusz = 2 CentΛR(2ΛW ). This coincide with our choice Λ′ = CentΛR(Λ1 + Λ2), if and only if 2 gcd ( `, dΛ i ) = gcd ( `, 2dWi ) , (4.1) where the dΛ i denote the invariant factors of Λ∨W /Λ and the dWi denote the invariant factors of Λ∨W /ΛW (i.e., ordered root lengths). In particular, for ` odd these choices never coincide. For Λ = ΛW , Λ′ = Λ′Lusz holds if and only if 2di | `. This is the most extreme case of divisibility and it is precisely the case appearing in logarithmic conformal field theories. Proof. We first note that in our cases, Λ′ = CentΛR(Λ1 + Λ2) = CentΛR(Λ). We have 2 CentΛR(2ΛW ) = 2 ( ΛR ∩ 2̂ΛW ) = AR2 ( Λ∨W ∩A−1 W ` 2 Λ∨W ) = AR Diag ( 2` gcd(`, 2dWi ) ) Λ∨W . By Lemma 4.2, this coincides with Λ′ if and only if equation (4.1) holds. � 5 Factorizability of quantum group R-matrices We first recall the definition of factorizable braided tensor categories and factorizable Hopf algebras, respectively. 16 S. Lentner and T. Ohrmann Definition 5.1 ([3]). A braided tensor category C is factorizable if the canonical braided tensor functor G : C � Cop → Z(C) is an equivalence of categories. In [17], Schneider gave a different characterization of factorizable Hopf algebras in terms of its Drinfeld double, leading to the following theorem: Definition 5.2. A finite-dimensional quasitriangular Hopf algebra (H,R) is called factorizable if its monodromy matrix M := R21 ·R ∈ H ⊗H is non-degenerate, i.e., the following linear map is bijective H∗ → H, φ 7→ (id⊗φ)(M). Equivalently, this means we can write M = ∑ iR i 1 ⊗Ri2 for two basis’ Ri1, R j 2 ∈ H. Theorem 5.3. Let (H,R) be a finite-dimensional quasitriangular Hopf algebra. Then the cate- gory of finite-dimensional H-modules H − modfd is factorizable if and only if (H,R) is a fac- torizable Hopf algebra. Shimizu [18] has recently proven a number of equivalent characterizations of factorizability for arbitrary (in particular non-semisimple) braided tensor categories. Besides the two previous characterizations (equivalence to Drinfeld center and nondegeneracy of the monodromy matrix), factorizability is equivalent to the fact that the so-called transparent objects are all trivial, see Theorem 5.16 below, which will become visible during our analysis later. 5.1 Monodromy matrix in terms of R0 In order to obtain conditions for the factorizability of the quasitriangular small quantum groups (uq(g,Λ,Λ ′), R0(f)Θ̄) as in Theorem 2.2 in terms of g, q, Λ and f , we start by calculating the monodromy matrix M := R21 ·R ∈ uq(g,Λ,Λ′)⊗ uq(g,Λ,Λ′) in general as far as possible: Lemma 5.4. For R = R0(f)Θ̄ as in Theorem 2.2, the factorizability of R is equivalent to the invertibility of the following complex-valued matrix m with entries indexed by elements in µ, ν ∈ Λ/Λ′: mµ,ν := ∑ µ′,ν′∈Λ/Λ′ f(µ− µ′, ν − ν ′)f(ν ′, µ′). Proof. We first plug in the expressions for R0 from Theorem 3.3 and Θ̄ from Theorem 2.2 and simplify: M := R21 ·R = (R0)21 · Θ̄21 ·R0 · Θ̄ =  ∑ µ1,ν1∈Λ f(µ1, ν1)Kν1 ⊗Kµ1   ∑ β1∈Λ+ R (−1)trβ1qβ1 ∑ b1∈Bβ2 b∗+1 ⊗ b − 1  ×  ∑ µ2,ν2∈Λ f(µ2, ν2)Kµ2 ⊗Kν2   ∑ β2∈Λ+ R (−1)trβ2qβ2 ∑ b2∈Bβ2 b−2 ⊗ b ∗+ 2  = ∑ β1,β2∈Λ+ R (−1)trβ1+β2qβ1qβ2  ∑ µ1,µ2,ν1,ν2∈Λ f(µ1, ν1)f(µ2, ν2)qβ1(ν2−µ2)Kν1+µ2 ⊗Kµ1+ν2  ×  ∑ b1∈Bβ1 ,b2∈Bβ2 b∗+1 b−2 ⊗ b − 1 b ∗+ 2  , Factorizable R-Matrices for Small Quantum Groups 17 where Λ+ R = N0[∆]. The last equation holds since b−1 ∈ u−β1 and hence fulfills Kν2b − 1 = q−β1ν2b−1 Kν2 and similarly for b∗+1 . We have two triangular decompositions uq = u0 qu − q u + q , uq = u0 qu + q u − q , and the Λ+ R-gradation on u±q induces a gradation uq ⊗ uq ∼= ⊕ β1,β2 ( u0 ⊗ u0 )( u+ q β1 u−q β2 ⊗ u−q β1 u+ q β2 ) . The factorizability of R is equivalent to the invertibility of M interpreted as a matrix indexed by the PBW basis. The grading implies a block matrix form of M , so the invertibility M is equivalent to the invertibility of Mβ1,β2 ∈ (uq ⊗ uq)(β1,β2) for every β1, β2 ∈ Λ+ R as follows Mβ1,β2 :=  ∑ µ1,µ2,ν1,ν2∈Λ f(µ1, ν1)f(µ2, ν2)qβ1(ν2−µ2)Kν1+µ2 ⊗Kµ1+ν2  ×  ∑ b1∈Bβ1 ,b2∈Bβ2 b∗+1 b−2 ⊗ b − 1 b ∗+ 2  . Since the second sum in Mβ1,β2 runs over a basis in u+ q β1 u−q β2 ⊗u−q β1 u+ q β2 , the invertibility of M is equivalent to the invertibility for all β1 ∈ Λ+ R the following element: Mβ1 0 := ∑ µ1,µ2,ν1,ν2∈Λ/Λ′ qβ1(ν2−µ2)f(µ1, ν1)f(µ2, ν2)Kν1+µ2 ⊗Kµ1+ν2 = ∑ µ,ν∈Λ/Λ′ Kν ⊗Kµ ·  ∑ µ′,ν′∈Λ/Λ′ qβ1(µ′−ν′)f(µ− µ′, ν − ν ′)f(ν ′, µ′)  . Since Kν ⊗Kµ is a vector space basis of u0 q ⊗ u0 q = C[Λ/Λ′]⊗C[Λ/Λ′], this in turn is equivalent to the invertibility of the following family of matrices mβ1 for all β1 ∈ Λ+ R with rows/columns indexed by elements in µ, ν ∈ Λ/Λ′: mβ1 µ,ν := ∑ µ′,ν′∈Λ/Λ′ f(µ− µ′, ν − ν ′)f(ν ′, µ′)qβ1(µ′−ν′). We now use the fact that R was indeed an R-matrix. By property (2.1) in Theorem 2.2 we have mβ1 µ,ν = ∑ µ′,ν′∈Λ/Λ′ f(µ− µ′, ν − ν ′)f(ν ′ + β1, µ ′)q−β1ν′ . Since the invertibility of a matrix mµ,ν is equivalent to the invertibility of any matrix mµ,ν+β1 , we may substitute ν ′ 7→ ν ′ + β1, ν 7→ ν + β1, pull the constant factor q−β 2 1 in front (which also does not affect invertibility) and hence eliminate the first β1 from the condition. Hence the invertibility of R is equivalent to the invertibility of the following family of matrices m̃β1 for all β1 ∈ Λ+ R: m̃β1 µ,ν := ∑ µ′,ν′∈Λ/Λ′ f(µ− µ′, ν − ν ′)f(ν ′, µ′)q−β1ν′ . 18 S. Lentner and T. Ohrmann We may now use the same procedure to eliminate the second β1, hence the invertibility of R is equivalent to the invertibility of the following matrix with rows/columns induced by elements in µ, ν ∈ Λ/Λ′: mµ,ν := ∑ µ′,ν′∈Λ/Λ′ f(µ− µ′, ν − ν ′)f(ν ′, µ′). This was the assertion we wanted to prove. � Definition 5.5. Let g : G1 ×G2 → C× be a group pairing. It induces a symmetric form on the product G1 ×G2 we denote by Sym(g): Sym(g) : (G1 ×G2)×2 −→ C×, ((µ1, µ2), (ν1, ν2)) 7−→ g(µ1, ν2)g(ν1, µ2). Lemma 5.6. If g : G1 × G2 → C× is a perfect pairing of abelian groups, then the symmetric form Sym(g) is perfect. Proof. By assumption, g × g defines an isomorphism between G1 × G2 to Ĝ2 × Ĝ1. The symmetric form Sym(g) is given by the composition of this isomorphism with the canonical isomorphism Ĝ2 × Ĝ1 ∼= Ĝ1 ×G2. This proves the claim. � Consider for a finite abelian group G and subgroups G1, G2 ≤ G the canonical exact sequence 0→ G1 ∩G2 → G1 ×G2 → G1 +G2 → 0. (5.1) For µ ∈ G1 +G2, we denote its fiber by (G1 ×G2)µ := {(µ1, µ2) ∈ G1 ×G2 |µ1 + µ2 = µ}. Moreover, we define Rad := { (µ1, µ2) ∈ G1 ×G2 | Sym ( f̂ ) ((µ1, µ2), x) = 1 ∀x ∈ (G1 ×G2)0 } , Radµ := Rad∩(G1 ×G2)µ, Rad⊥0 := {µ1 + µ2 ∈ G | (µ1, µ2) ∈ Rad}. Lemma 5.7. We have two split exact sequences: 0→ Rad0 → Rad→ Rad⊥0 → 0, 0→ Rad⊥0 → G→ Rad0 → 0. Proof. The first sequence is exact by definition of the three groups. Moreover, we know Rad = ker ( ι̂ ◦ Sym ( f̂ )) ∼= ker(ι̂) = im(π̂) ∼= Ĝ ∼= G, where ι̂, π̂ denote the duals of the inclusion and projection in (5.1). In Example 5.11 we will see that in the case G1 = G2 = G, f̂ symmetric, Rad0 is the 2-torsion subgroup of G, and the second map in the second exact sequence is just the projection, hence both diagrams split in this case. If f̂ is asymmetric, we will see in Section 5.3 that Rad0 is isomorphic to Zk2 for some k ≥ 2, thus Rad⊥0 −→ Rad, x 7−→ ∑ x̃∈Radx x̃ is a section of the first exact sequence. Here we used that the sum over all elements in Zk2 vanishes. Again, it follows that both diagrams split. Finally, if G1 6= G2 (i.e., in the case D2n), then f̂ = q−(·,·) on G1 ∩ G2. By the same argument as in Example 5.11, Rad0 is the 2-torsion subgroup of G1∩G2. But we have G ∼= G1∩G2×π1 in this case, hence both sequences split. � Factorizable R-Matrices for Small Quantum Groups 19 Corollary 5.8. Using the projection α : G → Rad⊥0 and the inclusion β : Rad⊥0 → Rad from the above lemma, we can define a symmetric form on G: SymG ( f̂ ) : G×G −→ C×, (µ, ν) 7−→ Sym ( f̂ ) (β ◦ α(µ), β ◦ α(ν)). Moreover, we have Rad ( SymG ( f̂ )) ∼= Rad0. Theorem 5.9. We have shown in Theorem 2.2 and Lemma 3.5 that the assumption that R = R0(f)Θ̄ is an R-matrix is equivalent to the existence of subgroups G1, G2 ⊂ Λ/Λ′ of same order some d|ΛR/Λ′| and f restricting up to a scalar to a non-degenerate pairing f̂ : G1 × G2 → C× and f vanishes otherwise. In this notation the matrix m as defined in the previous lemma can be rewritten as mµ,ν = 1 d2|ΛR/Λ′|2 ∑ µ̃∈(G1×G2)µ ν̃∈(G1×G2)ν Sym ( f̂ ) (µ̃, ν̃). It is invertible if and only if Rad0 = 0. In this case, mµ,ν = |G1 ∩G2| d2|ΛR/Λ′|2 SymG ( f̂ ) . We first note that Rad0 = 0 implies Rad⊥0 = G and thus G = G1 + G2. Together with Corollary 3.6 this implies Corollary 5.10. Λ′ = CentΛR(Λ). Before we proof the theorem, we first give a simple example: Example 5.11. Let G1 = G2 = G (correspondingly Λ1 = Λ2 = Λ) and assume f̂ is symmetric non-degenerate, then the radical measures 2-torsion: Rad ( SymG ( f̂ )) ∼= Rad0 = {µ ∈ G | 2µ = 0}. Again, this is the only case appearing for cyclic fundamental groups. Hence in all cases except g = D2n factorizability is equivalent to |Λ/Λ′| being odd. Proof of Theorem 5.9. The first part of the theorem follows by applying Lemma 3.5 to the matrix m as given in the previous lemma. Now, assume that m is invertible. We must have G = G1 + G2, otherwise the matrix has zero-columns and rows, differently formulated: the fibers (G1 ×G2)µ in the short exact sequence must be non-empty for all µ ∈ G. If on the other hand, Rad0 = 0, then Rad⊥0 = G and thus G1 + G2 = G must also hold, thus we assume this from now on. By the short exact sequence the fiber (G1 ×G2)0 ∼= G1 ∩G2, other fibers are of the explicit form µ̃+ (G1 ×G2)0 for some choice of representative µ̃. Therefore, mµ,ν = 1 d2|ΛR/Λ′|2 ∑ µ̃∈(G1×G2)µ ν̃∈(G1×G2)ν Sym ( f̂ ) (µ̃, ν̃) = 1 d2|ΛR/Λ′|2 ∑ ν̃∈(G1×G2)ν Sym ( f̂ ) (µ̃, ν̃) ∑ η̃∈(G1×G2)0 Sym ( f̂ ) (η̃, ν̃) 20 S. Lentner and T. Ohrmann = |G1 ∩G2| d2|ΛR/Λ′|2 ∑ ν̃∈(G1×G2)ν Sym ( f̂ ) (µ̃, ν̃) · δSym(f)(ν̃, )|G1∩G2 =1 = (∗). Fix as above a representative ν̃ of the fiber of ν, i.e., ν̃ ∈ (G1×G2)ν such that Sym(f)(ν̃, )|G1∩G2 = 1 holds. Two elements fulfilling this property differ by an element in the subgroup Rad0 ≤ G1 ∩G2, thus (∗) = |G1 ∩G2| d2|ΛR/Λ′|2 Sym ( f̂ ) (µ̃, ν̃) ∑ ξ̃∈Rad0 Sym ( f̂ ) (ξ̃, ν̃) · δSym(f)(ν̃, )|G1∩G2 =1 = |G1 ∩G2||Rad0 | d2|ΛR/Λ′|2 Sym ( f̂ ) (µ̃, ν̃) · δSym(f̂)(ν̃, )|G1∩G2 =1 δSym(f̂)(µ̃, )|Rad0 =1. Since m is symmetric, we have mµ,ν = |G1 ∩G2||Rad0 | d2|ΛR/Λ′|2 Sym ( f̂ ) (µ̃, ν̃) · δSym(f̂)(ν̃, )|G1∩G2 =1δSym(f̂)(µ̃, )|G1∩G2 =1 = |G1 ∩G2||Rad0 | d2|ΛR/Λ′|2 SymG ( f̂ ) (µ, ν)δRadµ 6=∅δRadν 6=∅ and this is invertible if an only if Rad0 ∼= Rad ( SymG ( f̂ )) = 0. � 5.2 Factorizability for symmetric R0(f) For R0 = ∑ µ,ν f(µ, ν)Kµ ⊗ Kν being the Cartan part of an R-matrix, assume that f̂ = |G|f on G is symmetric. We have shown in Example 5.11 that factorizability is equivalent to |G| being odd. We now want to give a necessary and sufficient condition for this: Lemma 5.12. Let ΛR ⊆ Λ ⊆ ΛW be an arbitrary intermediate lattice for a certain irreducible root system. Then the order of the group G = Λ/CentΛR(Λ) is odd if and only if both of the following conditions are satisfied: 1) |Λ/ΛR| is odd, 2) ` is either odd or (` ≡ 2 mod 4, g = Bn, Λ = ΛR) including A1. Proof. We saw that in all our cases, there exists an isomorphism Λ/ΛR ∼= CentΛ(ΛR)/CentΛR(Λ). Moreover, from Lemma 4.2 we know that |Λ/CentΛ(ΛR)| = det(D`), where D` was the diagonal matrix Diag ( ` gcd(`,dΛ i ) ) ) with dΛ i being the invariant factors of the lattice Λ (i.e., the diagonal entries of the Smith normal form of a basis matrix of Λ). Thus, |G| = |Λ/CentΛR(Λ)| = |Λ/CentΛ(ΛR)||CentΛ(ΛR)/CentΛR(Λ)| = |Λ/CentΛ(ΛR)||Λ/ΛR| = det(D`)|Λ/ΛR| = n∏ i=1 ` gcd(`, dΛ i ) |Λ/ΛR|. Clearly, this term is odd if ` and |Λ/ΛR| are odd. In the case (` ≡ 2 mod 4, g = Bn, Λ = ΛR), the Smith normal form SR of the basis matrix AR is given by 2 · id. Thus, |G| is odd in this case. On the other hand, let |G| be odd: We first consider the case ` even. A necessary condition for |Λ/Λ′| odd is that the multipli- city m` of the prime 2 in n∏ i=1 ` gcd(`,dΛ i ) is at most the multiplicity mπ1 of the prime 2 in |π1|. We check this condition for rank n > 1: Factorizable R-Matrices for Small Quantum Groups 21 • For g simply-laced (or triply-laced g = G2) we have all di = 1, hence n |m` (equality for ` = 2 mod 4). The cases Dn with mπ1 = 2 have rank n ≥ 4, all others except An have mπ1 = 0, 1, so the necessary condition m` ≤ mπ1 is never fulfilled. The cases An have 2mπ1 |(n+ 1) ≤ (m` + 1) ! ≤ (mπ1 + 1) which can only be true in rank n = 1 treated above. • For g doubly-laced of rank n > 1, we always have always mπ1 = 0, 1 but m` can be considerably smaller than above, namely for ` = 2 mod 4 equal to the number of short simple roots dαi = 1 (otherwise m` again increases by n for every factor 2 in `), hence the necessary condition m` ≤ mπ1 can be fulfilled only for Bn (which would also include A1 above for n = 1). More precisely, since m` = mπ1 and the decomposition for Λ/Λ′ has an additional factor |Λ/ΛR|, it can only be odd for Λ = ΛR. On the other hand, if ` is odd, then the whole product term is odd. But since |G| was assumed to be odd, also |Λ/Λ′| must be odd. � Corollary 5.13. Let Λ = ΛR. In the previous section we have seen that f̂ = q−(·,·) gives always an R-matrix in this case. By the proof of the previous lemma, we have Rad0 ∼= n∏ i=1 Z gcd ( 2, ` gcd(`,dR i ) ), where the dRi denote the invariant factors of Λ∨W /ΛR. 5.3 Factorizability for D2n, R0 antisymmetric The split case g = D2n, G = G1 × G2 is clearly factorizable, so the only remaining case for which we have to check factorizabilty is g = D2n, Λ = ΛW for f̂ being not symmetric. We know that in this case, the corresponding form g on Λ/ΛR is uniquely defined by a 2× 2-matrix K ∈ gl(2,F2), s.t. g(λ2(n−1)+i, λ2(n−1)+j) = exp (2πiKij 2 ) for i, j ∈ {1, 2}. From this we see that if g is not symmetric, it must be antisymmetric, i.e., g(µ, ν) = g(ν, µ)−1. Thus, the following lemma applies in this case, and hence there are no factorizable R-matrices for D2n, Λ = ΛW . Lemma 5.14. For g simply-laced and Λ = ΛW , let f̂ = q−(·,·)g : G×G→ C× be a non-degenerate form as in Theorem 3.3 and Lemma 3.5, s.t. the form g : π1 × π1 → C× is asymmetric. Then, Rad0 ∼= n⊕ i=1 Zgcd(2,`dRi ), where the dRi denote the invariant factors of π1. In particular, Rad0 = 0 holds if and only if gcd(2, `|π1|) = 1. Proof. We recall the definition of Rad0 ( SymG ( f̂ )) in this case: Rad0(SymG(f̂)) = { µ ∈ G | f(ν, µ)−1 = f(µ, ν) ∀ ν ∈ G } = { µ ∈ G | q(ν,µ)g(ν, µ)−1 = q−(µ,ν)g(µ, ν) ∀ ν ∈ G } = { µ ∈ G | q(ν,µ) = q−(µ,ν) ∀ ν ∈ G } = { µ ∈ G | q(2µ,ν) = 1 ∀ ν ∈ G } = { µ ∈ G | 2µ ∈ Cent2ΛW (ΛW )/2 CentΛR(ΛW ) } = (∗). For g is simply-laced, we have ΛW = Λ∨W , thus (∗) ∼= Cent2ΛW (ΛW )/2 CentΛR(ΛW ) = (2ΛW ∩ `ARΛW )/2`ARΛW = PR Diag ( lcm ( 2, `dRi )) ΛW /PR2`SRΛW = ΛW /Diag ( gcd ( 2, `dRi )) ΛW . This proves the claim. � 22 S. Lentner and T. Ohrmann 5.4 Transparent objects in non-factorizable cases In this section, we determine the transparent objects in the representation category of uq(g,Λ) with our R-matrix given by R0Θ̄ and R0 = 1 |Λ/Λ′| ∑ µ,ν∈Λ/Λ′ f̂ with f̂ a group pairing Λ1/Λ ′ × Λ2/Λ ′ → C×. Definition 5.15. Let C be a braided monoidal category with braiding c. An object V ∈ C is called transparent if the double braiding cW,V ◦ cV,W is the identity on V ⊗W for all W ∈ C. The following theorem by Shimizu gives a very important characterization of factorizable categories: Theorem 5.16 ([18, Theorem 1.1]). A braided finite tensor category is factorizable if and only if the transparent objects are direct sums of finitely many copies of the unit object. Corollary 5.17. In particular, for a Hopf algebra H the representation category H −modfd is factorizable if and only if the transparent objects are multiples of the trivial representation and vice versa. Since in our cases Λ1 6= Λ2 can only appear in D2n, and we know those are factorizable, we shall in the following restrict ourselves to the case Λ1 = Λ2 = Λ. The proof below works also in the more general case, but requires more notation. As usual we first reduce the Hopf algebra question to the group ring and then solve the group theoretical problem. Lemma 5.18. If a uq(g)-module V , with a highest-weight vector v and Kµv = χ(Kµ)v, is a transparent object, then necessarily the 1-dimensional Λ/Λ′-module Cχ is a transparent object over the Hopf algebra C[Λ/Λ′] with R-matrix R0. If V is 1-dimensional, then V is transparent if and only if Cχ is. Proof. Let V be transparent. For every ψ : Λ/Λ′ → C× we have another finite-dimensional module W := uq(g) ⊗uq(g)+ Cψ with highest weight vector w = 1 ⊗ 1ψ which we can test this assumption against c2 : V ⊗W →W ⊗ V → V ⊗W. We calculate the effect of c2 on the highest-weight vectors v ⊗ w: c2(v ⊗ w) = τW⊗VR0Θ̄τV⊗WR0Θ̄(v ⊗ w). Because v, w were assumed highest-weight vectors, the Θ̄ act trivially. Hence follows that Cχ, Cψ have a trivial double braiding over the Hopf algebra C[Λ/Λ′] with R-matrix R0. Because we could achieve this result for any ψ this means that Cχ is transparent as asserted. Now, let V = Cχ be 1-dimensional over uq(g) and transparent over C[Λ/Λ′], and let w be any element in any module W , then again the two Θ act trivially, one time because v = 1χ is a highest weight vector, and one time because it is also a lowest weight vector. But if the double-braiding of v = 1χ with any element w is trivial, then V = Cχ is already transparent over uq(g). � Lemma 5.19. Cχ is a transparent object over the Hopf algebra C[Λ/Λ′] with R-matrix R0 iff it is an f -transformed of the radical of SymG ( f̂ ) , i.e., χ(µ) = f(µ, ξ), ξ ∈ Rad0 . Factorizable R-Matrices for Small Quantum Groups 23 Proof. Since f is nondegenerate, we can assume χ(µ) = f(µ, ξ) and wish to prove Cχ is transparent iff ξ ∈ Rad0. We test transparency against any module Cψ and also write ψ(µ) = f(λ, µ) (note the order of the argument). We evaluate the double-braiding on 1χ ⊗ 1ψ and get the following scalar factor, which needs to be = 1 for all ψ in order to make Cχ transparent: 1 |G|2 ∑ µ,ν χ(µ)ψ(ν) ∑ µ1+µ2=µ ν1+ν2=µ Sym ( f̂ ) ((µ1, µ2), (ν1, ν2)) = 1 |G|2 ∑ µ,ν f(µ, ξ)f(λ, ν) ∑ µ1+µ2=µ ν1+ν2=µ f(µ1, ν1)f(ν2, µ2) = 1 |G|2 ∑ µ,ν f(µ, ξ)f(λ, ν) ∑ ν1,µ1 f(µ1, ν1)f(ν, µ)f−1(ν1, µ)f−1(ν, µ1)f(ν1, µ1) = 1 |G| ∑ ν f(λ, ν) ∑ ν1,µ1 f(µ1, ν1)δξ=−ν+ν1f −1(ν, µ1)f(ν1, µ1) = 1 |G| ∑ ν f(λ, ν) ∑ µ1 f(µ1, ξ + ν)f(ξ, µ1) = f−1(λ, ξ)f−1(ξ, λ) = SymG ( f̂ ) (λ, ξ). This scalar factor of the double braiding is equal +1 for all λ (and hence all Cψ) iff ξ ∈ Rad0 as asserted. � The previous two lemmas combined imply that any irreducible transparent uq(g)-module has necessarily the characters χ(µ) = f(µ, ξ), ξ ∈ Rad0 as highest-weights, and conversely if such a character χ gives rise to 1-dimensional uq(g)-modules (i.e., χ|2ΛR = 1), then these are guaranteed transparent objects. Hence the final step is to give more closed expressions for the f -transformed characters χ of the radical depending on the case and check the 1-dimensionality condition. In all cases where f is symmetric we have seen in Example 5.11 that Rad0 ( SymG ( f̂ )) is the 2-torsion subgroup of Λ/Λ′, so in these cases χ gives rise to a 1-dimensional object. Corollary 5.20. If f is symmetric (true for all cases except D2n) then the transparent objects are all 1-dimensional Cχ where the characters χ are the f -transformed of the elements in the radical of the bimultiplicative form Sym ( f̂ ) |G on G = Λ/Λ′. In particular the group of transparent objects is isomorphic to this radical as an abelian group. Corollary 5.21. In the case of symmetric f (all cases except D2n) the fact that Rad0 is the 2-torsion of Λ/Λ′ and f -transformation is a group isomorphism shows: The group T of transparent objects consists of Cχ where χ|2Λ = 1, i.e., the two-torsion of the character group. The remaining case in D2n with f nonsymmetric and has been done by hand in Lemma 5.14. 6 Quantum groups with a ribbon structure In [16, Theorem 8.23], the existence of ribbon structures for uq(g,Λ) is proven. In this section we construct a ribbon structure for all cases. In the proof, we use several auxiliary results from [16]. Theorem 6.1. Let uq(g,Λ) be quasitriangular Hopf algebra, with an R-matrix satisfying the conditions in Theorem 2.2 and let u := S(R(2))R(1). Then v := K−1 ν0 u is a ribbon element in uq(g,Λ). 24 S. Lentner and T. Ohrmann Proof. We consider the natural N0[αi | i ∈ I]-grading on the Borel parts u± := uq(g,Λ)± [13]. Since u± is finite-dimensional, there exists a maximal ν0 ∈ N0[αi | i ∈ I], s.t. the homogeneous component u±ν0 is non-trivial. More explicitly ν0 is of the form ν0 = ∑ α∈Φ+ (`α − 1)α, where `α := ` gcd(`,2dα) . Using the formulas u = (∑ f(µ, ν)Kµ+ν )−1 ϑ and S(u) = (∑ f(µ, ν)Kµ+ν )−1 S(ϑ), where ϑ = ∑ Θ̄(2)S−1 ( Θ̄(2) ) , Müller proves the formula K2 −ν0 = u−1S(u). Using the fact that u commutes with all grouplike elements, this implies v2 = uS(u). In order to show that v is central, we first show that K−1 ν0+2ρ is a central element. By the K,E-relations, this is equivalent to ν0 + 2ρ ∈ CentΛ(ΛR), where ρ = 1 2 ∑ α∈Φ+ α is the Weyl vector. We calculate directly that this is always the case: (ν0 + 2ρ, β) = q ∑ α∈Φ+ (`α−1+1)(α,β) = q ` ∑ α∈Φ+ 1 gcd(`,2dα) ·2dα(α∨,β) = 1. Since K2ρux = xK2ρu holds for all x ∈ uq(g,Λ) (see [16, Lemmas 8.22 and 8.19]), we have vx = K−1 ν0 ux = K−1 ν0+2ρK2ρux = K−1 ν0+2ρxK2ρu = xK−1 ν0+2ρK2ρu = xv, hence v is central. � 7 Open questions Question 7.1. It was surprising to us that the case D2n = so4n(C) has so many more solutions that the other cases, in particular with non-symmetric R0, due to the non-cyclic fundamental group. Do these additional modular tensor categories appear elsewhere? Does the non-symmetry have interesting implications on the category? Question 7.2. Our procedure would be similarly possible for any diagonal Nichols algebra. The Lusztig ansatz can in these cases be found in [1]. Question 7.3. In each case where uq(g,Λ), R is not factorizable, we can modularize (see [2]) the corresponding representation category and get a modular tensor category, which should be representations over some “quasi-quantum group” uq(g, Λ̃, ω), R which is a quasi-Hopf algebra where the group ring C[Λ̃] is deformed by a 3-group-cocycle ω. Can we describe this quasi-Hopf algebra in a closed form? Moreover, is every factorizable quasi-quantum group the modularization of a quasi-triangular quantum group from our list? More technically: Question 7.4. The centralizer transfer map A` in Definition 3.8 (and correspondingly the form a`) had a very general characterization, but we could only prove existence by a construc- tion using the classification of simple Lie algebras (and distinguishing three cases). We strongly suspect that these maps exist under rather general assumptions. Also the result Theorem 3.3 from our previous article [11] has only been proven there for cyclic groups (and by hand for Z2 × Z2) although we strongly suspect it holds for every abelian group. Factorizable R-Matrices for Small Quantum Groups 25 Acknowledgements Both authors thank Christoph Schweigert for helpful discussions and support. They also thank the referees, who gave a relevant contribution to improve the article with their comments. The first author was supported by the DAAD P.R.I.M.E program funded by the German BMBF and the EU Marie Curie Actions as well as the Graduiertenkolleg RTG 1670 at the University of Hamburg. The second author was supported by the Collaborative Research Center SFB 676 at the University of Hamburg. References [1] Angiono I., Yamane H., The R-matrix of quantum doubles of Nichols algebras of diagonal type, J. Math. Phys. 56 (2015), 021702, 19 pages, arXiv:1304.5752. [2] Bruguières A., Catégories prémodulaires, modularisations et invariants des variétés de dimension 3, Math. Ann. 316 (2000), 215–236. [3] Etingof P., Gelaki S., Nikshych D., Ostrik V., Tensor categories, Mathematical Surveys and Monographs, Vol. 205, Amer. Math. Soc., Providence, RI, 2015. [4] Feigin B.L., Gainutdinov A.M., Semikhatov A.M., Tipunin I.Yu., Modular group representations and fusion in logarithmic conformal field theories and in the quantum group center, Comm. Math. Phys. 265 (2006), 47–93, hep-th/0504093. [5] Feigin B.L., Tipunin I.Yu., Logarithmic CFTs connected with simple Lie algebras, arXiv:1002.5047. [6] Gainutdinov A.M., Runkel I., Symplectic fermions and a quasi-Hopf algebra structure on Uqs`(2), J. Algebra 476 (2017), 415–458, arXiv:1503.07695. [7] Kerler T., Lyubashenko V.V., Non-semisimple topological quantum field theories for 3-manifolds with cor- ners, Lecture Notes in Mathematics, Vol. 1765, Springer-Verlag, Berlin, 2001. [8] Kondo H., Saito Y., Indecomposable decomposition of tensor products of modules over the restricted quan- tum universal enveloping algebra associated to sl2, J. Algebra 330 (2011), 103–129, arXiv:0901.4221. [9] Lentner S., A Frobenius homomorphism for Lusztig’s quantum groups for arbitrary roots of unity, Commun. Contemp. Math. 18 (2016), 1550040, 42 pages, arXiv:1406.0865. [10] Lentner S., Nett D., A theorem of roots of unity and a combinatorial principle, arXiv:1409.5822. [11] Lentner S., Nett D., New R-matrices for small quantum groups, Algebr. Represent. Theory 18 (2015), 1649–1673, arXiv:1409.5824. [12] Lusztig G., Quantum groups at roots of 1, Geom. Dedicata 35 (1990), 89–113. [13] Lusztig G., Introduction to quantum groups, Modern Birkhäuser Classics, Birkhäuser/Springer, New York, 2010. [14] Müller E., Quantengruppen im Einheitswurzelfall, Ph.D. Thesis, Ludwig-Maximilians-Universität München, 1998. [15] Müller E., Some topics on Frobenius–Lusztig kernels. I, J. Algebra 206 (1998), 624–658. [16] Müller E., Some topics on Frobenius–Lusztig kernels. II, J. Algebra 206 (1998), 659–681. [17] Schneider H.-J., Some properties of factorizable Hopf algebras, Proc. Amer. Math. Soc. 129 (2001), 1891– 1898. [18] Shimizu K., Non-degeneracy conditions for braided finite tensor categories, arXiv:1602.06534. [19] Turaev V.G., Quantum invariants of knots and 3-manifolds, De Gruyter Studies in Mathematics, Vol. 18, Walter de Gruyter & Co., Berlin, 1994. https://doi.org/10.1063/1.4907379 https://doi.org/10.1063/1.4907379 https://arxiv.org/abs/1304.5752 https://doi.org/10.1007/s002080050011 https://doi.org/10.1007/s002080050011 https://doi.org/10.1090/surv/205 https://doi.org/10.1007/s00220-006-1551-6 https://arxiv.org/abs/hep-th/0504093 https://arxiv.org/abs/1002.5047 https://doi.org/10.1016/j.jalgebra.2016.11.026 https://arxiv.org/abs/1503.07695 https://doi.org/10.1007/b82618 https://doi.org/10.1016/j.jalgebra.2011.01.010 https://arxiv.org/abs/0901.4221 https://doi.org/10.1142/S0219199715500406 https://doi.org/10.1142/S0219199715500406 https://arxiv.org/abs/1406.0865 https://arxiv.org/abs/1409.5822 https://doi.org/10.1007/s10468-015-9555-6 https://arxiv.org/abs/1409.5824 https://doi.org/10.1007/BF00147341 https://doi.org/10.1007/978-0-8176-4717-9 https://doi.org/10.1006/jabr.1997.7364 https://doi.org/10.1006/jabr.1998.7423 https://doi.org/10.1090/S0002-9939-01-05787-2 https://arxiv.org/abs/1602.06534 1 Introduction 2 Preliminaries 2.1 Lie theory 2.2 Lusztig's ansatz for R-matrices 3 Conditions for the existence of R-matrices 3.1 A first set of conditions on /' 3.2 A natural form on the fundamental group 4 Explicit calculation for every g 4.1 Technical tools 4.2 Case =W 4.3 Case An 4.4 Case Dn 4.5 Table of all quasitriangular quantum groups 5 Factorizability of quantum group R-matrices 5.1 Monodromy matrix in terms of R0 5.2 Factorizability for symmetric R0(f) 5.3 Factorizability for D2n, R0 antisymmetric 5.4 Transparent objects in non-factorizable cases 6 Quantum groups with a ribbon structure 7 Open questions References
id nasplib_isofts_kiev_ua-123456789-148764
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1815-0659
language English
last_indexed 2025-12-07T16:50:27Z
publishDate 2017
publisher Інститут математики НАН України
record_format dspace
spelling Lentner, S.
Ohrmann, T.
2019-02-18T18:47:57Z
2019-02-18T18:47:57Z
2017
Factorizable R-Matrices for Small Quantum Groups / S. Lentner, T. Ohrmann // Symmetry, Integrability and Geometry: Methods and Applications. — 2017. — Т. 13. — Бібліогр.: 19 назв. — англ.
1815-0659
2010 Mathematics Subject Classification: 17B37; 20G42; 81R50; 18D10
DOI:10.3842/SIGMA.2017.076
https://nasplib.isofts.kiev.ua/handle/123456789/148764
Representations of small quantum groups uq(g) at a root of unity and their extensions provide interesting tensor categories, that appear in different areas of algebra and mathematical physics. There is an ansatz by Lusztig to endow these categories with the structure of a braided tensor category. In this article we determine all solutions to this ansatz that lead to a non-degenerate braiding. Particularly interesting are cases where the order of q has common divisors with root lengths. In this way we produce familiar and unfamiliar series of (non-semisimple) modular tensor categories. In the degenerate cases we determine the group of so-called transparent objects for further use.
Both authors thank Christoph Schweigert for helpful discussions and support. They also thank&#xd; the referees, who gave a relevant contribution to improve the article with their comments. The&#xd; first author was supported by the DAAD P.R.I.M.E program funded by the German BMBF&#xd; and the EU Marie Curie Actions as well as the Graduiertenkolleg RTG 1670 at the University&#xd; of Hamburg. The second author was supported by the Collaborative Research Center SFB 676&#xd; at the University of Hamburg.
en
Інститут математики НАН України
Symmetry, Integrability and Geometry: Methods and Applications
Factorizable R-Matrices for Small Quantum Groups
Article
published earlier
spellingShingle Factorizable R-Matrices for Small Quantum Groups
Lentner, S.
Ohrmann, T.
title Factorizable R-Matrices for Small Quantum Groups
title_full Factorizable R-Matrices for Small Quantum Groups
title_fullStr Factorizable R-Matrices for Small Quantum Groups
title_full_unstemmed Factorizable R-Matrices for Small Quantum Groups
title_short Factorizable R-Matrices for Small Quantum Groups
title_sort factorizable r-matrices for small quantum groups
url https://nasplib.isofts.kiev.ua/handle/123456789/148764
work_keys_str_mv AT lentners factorizablermatricesforsmallquantumgroups
AT ohrmannt factorizablermatricesforsmallquantumgroups