1 s2.0 S0001870808002302 Main

Advances in Mathematics 220 (2009) 1–58
www.elsevier.com/locate/aim
R-matrix Poisson algebras and their deformations

Sebastian Zwicknagl
Department of Mathematics, University of California, Riverside, 92521 CA, USA
Received 22 December 2007; accepted 22 August 2008
Available online 1 October 2008
Communicated by Andrei Zelevinsky
Abstract
We classify in this paper Poisson structures on modules over semisimple Lie algebras arising from clas-
sical r-matrices. We then study their quantizations and the relation to classical invariant theory.
© 2008 Elsevier Inc. All rights reserved.
Keywords: Quantum groups; Invariant theory; r-Matrix; Poisson algebras; Quantization
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Quadratic Poisson brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1. Decorated spaces and bracketed algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. Symmetric algebras and bracketed symmetric algebras . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3. Bracketed Poisson algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3. Poisson modules over Lie algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1. Definition and basic properties of Poisson modules . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2. Classification of Poisson modules over semisimple Lie algebras . . . . . . . . . . . . . . . . . 19
4. Quantum symmetric algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1. The quantum group Uq (g) and its modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2. Braided symmetric and exterior powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3. The classical limit of braided algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4. Flat modules over reductive Lie algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5. Deformations of symmetric algebras of Poisson modules . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.1. Quantum radicals as symmetric algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
E-mail address: [email protected].
0001-8708/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.aim.2008.08.006
2 S. Zwicknagl / Advances in Mathematics 220 (2009) 1–58
5.2. Quantum Schubert cells: PBW-theorem and Levi action . . . . . . . . . . . . . . . . . . . . . . . 39

6.Proof of Theorem 3.12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.1. Necessary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.1.1. The case of g = sln . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.1.2. The case of g = so(2n + 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.1.3. Type g = so(2n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.1.4. Type g = sp(2n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.1.5. The exceptional Lie algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2. Sufficient conditions and rigidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.2.1. The case g = sln . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.2.2. The case g = so(k) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.2.3. The case g = sp(2n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.2.4. Proof of Theorems 6.20 and 3.12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7. Open questions and conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Appendix A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1. Introduction
We classify in the present paper Poisson brackets on modules over a semisimple complex Lie
algebra which are based on classical r-matrices. We then quantize these Poisson structures in the
spirit of the recent joint paper [2] with A. Berenstein and show that we recover many well-known
examples of quantized coordinate rings of classical varieties.
Let us briefly discuss the main results in the case of a simple Lie algebra g. Let g = n− ⊕
h ⊕ n+ be a simple complex Lie algebra and let ·.· be an invariant bilinear form on g such
that the square of the length of a long root is 2. Let {Eα | α ∈ R+ } (where R+ is the set of
positive roots of g) be the standard basis of n+ and {F−α | α ∈ R+ } a set of root vectors of n−
such that [Eα , Fα ] = α̌. Recall that r = α∈R+ (α,α) 2 Eα ⊗ Fα ∈ g ⊗ g is a classical r-matrix
−
(α,α)
and r = α∈R+ 2 (Eα ⊗ Fα − Fα ⊗ Eα ) ∈ g ∧ g the antisymmetrized r-matrix. For each
g-module V define a quadratic bracket {·,·} on the symmetric algebra S(V ) by the formula:
(α, α)
{a, b} = r − (a ∧ b) = Eα (a)Fα (b) − Eα (b)Fα (a) (1.1)
2
α∈R+
for a, b ∈ S(V ). In particular, if g = sl2 (C), then {a, b} = E(a)F (b) − E(b)F (a).
The bracket is, by construction, skew-commutative and satisfies the Leibniz rule. To determine
whether it is Poisson for a pair (g, V ) one has to verify whether it satisfies the Jacobi identity.
Our first main result is the following theorem.
Main Theorem 1.1 (See Theorem 3.12). Let V be a simple finite-dimensional g-module. Assume
that (g, V ) = (sp2n (C), Vω1 ). Then the following are equivalent:
(a) The bracket (1.1) on S(V ) is Poisson.

(b) c(Λ3 V ) = {0}, where c the canonical g-invariant in Λ3 g corresponding (under the identifi-
cation g∗ ∼
= g) to the Lie bracket [·,·] : g ∧ g → g.
S. Zwicknagl / Advances in Mathematics 220 (2009) 1–58 3
(c) V is geometrically decomposable as defined by R. Howe [12].

(d) Homg (Λ3 V , S 3 V ) = {0}.
(e) Λ2 V is simple.
If (g, V ) = (sp2n (C), Vω1 ), then parts (a), (b), and (d) of Theorem 1.1 hold, but parts (c)
and (e) fail. In Theorem 3.14 we classify all simple modules V over a complex semisimple Lie
algebra for which the bracket (1.1) on S(V ) is Poisson. We show that the natural modules of
g = sln × slm (C) for arbitrary m, n ∈ Z0 yield the only nontrivial examples of a simple module
with such a Poisson bracket over a semisimple Lie algebra with more than one simple factor.
We then continue to show that the deformation quantization of the r-matrix Poisson structure
q
on a g-module Vλ recovers the braided symmetric algebras of the Uq (g)-module Vλ . The braided
q
symmetric algebra Sq (Vλ ) is quadratic q-deformations of the symmetric algebra of the U (g)-
module Vλ (see [2] or Section 4.2) which are Uq (g)-module algebras. An important problem
in [2] is the question, for which Uq (g)-module V q the deformation is flat; i.e. one has

q dim(V ) + n − 1
dim Sq V n =
n
for all graded components (Sq (V q ))n . The following result completely classifies such flat mod-
ules.
Main Theorem 1.2 (See Theorem 4.23). A simple Uq (g)-modules V q is flat, if and only if the
bracket (1.1) defines a Poisson structure on the symmetric algebra of the classical limit V of V q .
If g is of type An , Bn or Dn , then the braided symmetric algebras of the flat modules are
the quantized coordinate rings of the classical varieties such as the quantum m × n-matrices,
the quantum Euclidean space (see e.g. [21]), quantum symmetric and quantum antisymmetric
matrices (see e.g. [20,24]). The braided exterior powers of the flat natural modules of quantized
enveloping algebras of types Bn , Cn and Dn agree with the q-wedge modules constructed by
Jing, Misra and Okado in [15]. Our approach, thus, provides a natural unifying construction for
these objects. Moreover, following the arguments of Goodearl and Yakimov [11, Chapter 5] one
obtains that the braided symmetric algebras are the quantizations of Poisson structures on open
Schubert cells of certain partial flag varieties. Additionally we obtain results similar to the ones
obtained by Donin, Gurevich and Majid (see e.g. [7]).
Theorem 1.1 shows that there is an apparent relation between r-matrix Poisson structures, flat
modules and maximal parabolic subalgebras whose radicals are Abelian Lie algebras or have
one-dimensional derived subalgebra (Heisenberg-type) and classical invariant theory as studied
by Howe in [12]. We use this connection to give the following explicit construction of braided
symmetric algebras of flat simple modules.
Theorem 1.3 (See Theorems 5.6, 5.12). In the notation of Theorem 1.1, if the bracket (1.1) is
Poisson (including the case (g, V ) = (sp2n (C), Vω1 )), then:
(a) There exists a unique simple Lie algebra g and a maximal parabolic subalgebra p ⊂ g such
that g is the semisimple part of the Levi factor of p and V is isomorphic (as a g-module) to
the nil-radical radp of p.
(b) The associated graded of the quantized enveloping algebra Uq (radp ) is the Kontsevich defor-
mation Sq (V q ) of the Poisson algebra S(V ) and carries a natural Uq (g)-module structure.
The paper is organized as follows. In Section 2 we study quadratic Poisson algebras. We

introduce the notions of decorated space and bracketed algebras and show that the categories
of decorated spaces and bracketed algebras are symmetric monoidal. We show how decorated
spaces define bracketed algebras and show under which conditions the bracket satisfies the Jacobi
identity (Theorem 2.21 ) extending well-known results of Gelfand and Fokas [10]. We also show
that if the bracketed algebra associated to a tensor product of two decorated spaces is Poisson,
then each of the factors must define a Poisson algebra (Theorem 2.25).
We use these results in Section 3 to classify all simple Poisson modules over simple Lie
algebras (Theorem 3.12) and finally all simple Poisson modules over semisimple Lie algebras
(Theorem 3.14). Since the proof of Theorem 3.12 is rather long we present it in Section 6.
Section 4 is devoted to the classification of flat simple modules (Theorem 4.23). For the con-
venience of the reader we provide brief introductions to the quantized enveloping algebras Uq (g),
the category of their finite-dimensional modules (Section 4.1), the definition and basic proper-
ties of braided symmetric and exterior algebras and powers (Section 4.2) and the classical limit
(Section 4.3).
In Section 5 we construct the braided symmetric algebras as the quantized enveloping algebra
of nilradicals (Theorem 5.6), respectively their associated graded algebras (Theorem 5.12). In
Section 5.2 we prove a PBW-type theorem for quantum Schubert cells and study Levi actions on
quantized nilradicals.
The results in this paper open up many questions and suggest connections between the theory
of braided symmetric algebras, cluster algebras, geometric crystals, equivariant Poisson struc-
tures and classical invariant theory. Appropriately, the paper concludes with a section on open
questions and conjectures (Section 7).
2. Quadratic Poisson brackets
2.1. Decorated spaces and bracketed algebras
In this section we will introduce the notion of a decorated space and relate it to bracketed al-
gebras, which are commutative algebras with a skew-commutative bracket satisfying the Leibniz
rule, but not necessarily the Jacobi identity. Consider a linear tensor category C, i.e. an Abelian
monoidal category, over a field k of characteristic char(k) = 0. We will view C as a symmet-
ric tensor category with the braiding given by the permutation of factors σ . Define a decorated
space to be a pair (V , Φ) of an object V of C and a morphism Φ : V ⊗ V → V ⊗ V such that
σ ◦ Φ = −Φ ◦ σ . Denote by D(C) the category of decorated spaces whose objects are decorated
spaces, and whose morphisms are structure preserving maps in C.
One has the following result.
Lemma 2.1. The category D(C) is a braided symmetric category. More precisely we have the
following:
(a) The unit object is (k, 0), where 0 : k ⊗ k → 0 ∈ k is the trivial map.
(b) The tensor product (V ⊗ V , Φ ) = (V , Φ) ⊗ (V , Φ ) of two decorated spaces (V , Φ) and

(V , Φ ) is defined for all u, v ∈ V and u , v ∈ V via
Φ = Φ13 + Φ24

, (2.1)
where Φ13 = σ23 ◦ (Φ ⊗ Id⊗2 ) ◦ σ23 and Φ24 = σ ◦ (Id ⊗2 ⊗ Φ ) ◦ σ and σ = Id ⊗ σ ⊗ Id.
23 23 23
(c) The symmetric braiding is given by the morphisms Dσ : (V , Φ) ⊗ (V , Φ ) → (V , Φ ) ⊗
(V , Φ) such that Dσ (V ⊗ V ) = V ⊗ V and Dσ (Φ13 + Φ24 ) = Φ + Φ .
13 24
(d) Moreover, the direct sum (V , Φ) ⊕ (V , Φ ) = (V ⊕ V , Φ + Φ ) of two decorated spaces

(V , Φ) and (V , Φ ) is a decorated space.
Proof. It is easy to see that the tensor product (V , Φ) ⊗ (V , Φ) of two decorated spaces is a
decorated space, and that the tensor product is indeed associative. One can now show that (k, 0)
satisfies the axioms of the unit. Parts (a) and (b) are proved. Parts (c) and (d) are obvious. The
lemma is proved. 2
Lemma 2.2. Let C be a symmetric linear category and let F : C → C be a covariant tensor
functor compatible with the braiding. Then, F defines a covariant functor DF : D(C) → D(C ),
which takes an object (V , Φ) to (F (V ), F (Φ)).
Proof. It suffices to show that for every object V of C and every Φ ∈ End(V ⊗ V ) such that
σ ◦ Φ = −Φ ◦ σ one has σ ◦ F (Φ) = −(F (Φ) ◦ σF (V ),F (V ) ). Since F is compatible with the
braiding we compute:
σF (V ),F (V ) ◦ F (Φ) = F (σV ,V ◦ Φ) = F (−Φ ◦ σV ,V ) = −(F (Φ) ◦ σF (V ),F (V ) ).
The lemma is proved. 2
We next introduce bracketed and Poisson algebras in the category C.
Definition 2.3. (a) A bracketed algebra is a pair (A, {·,·}), where A is a commutative algebra in
C and {·,·} is a structure preserving bilinear map {·,·} : A ⊗ A → A satisfying:
(i) anti-commutativity
{a, b} + {b, a} = 0, (2.2)
for any a, b ∈ A;
(ii) the Leibniz rule
{a, bc} = {a, b}c + b · {a, c} (2.3)
for any a, b, c ∈ A.
(b) A bracketed algebra is called Poisson, if {·,·} satisfies the Jacobi identity:

a, {b.c} + c, {a, b} + b, {c, a} = 0
for all a, b, c ∈ A.
We denote by BAlg(C) the category of bracketed algebras in C, whose morphisms are structure
preserving algebra homomorphisms in C.
Remark 2.4. Let (A, {·,·}) be a bracketed algebra in BAlg(C). The bracket {·,·} satisfies
{ca, b} = c · {a, b} + {c, b} · a. (2.4)
Note the following obvious fact.
Lemma 2.5. The Poisson algebras form a full subcategory of BAlg(C), the category of Poisson
algebras.
We need the following fact.
Lemma 2.6. The category BAlg(C) is a symmetric monoidal category. More precisely:
(a) The tensor product (A ⊗ B, { , }) := (A, {·,·}A ) ⊗ (B, {·,·}B ) of two objects (A, { , }A ) and
(B, { , }B ) is defined in the following way: for all a, a ∈ A, b, b ∈ B define,

{a ⊗ b, a ⊗ b } = (1 ⊗ μB ) ◦ {a, a } ⊗ b ⊗ b + (μA ⊗ 1) ◦ a ⊗ a ⊗ {b, b } ,
where μA denotes the multiplication in A and μB the multiplication in B.

(b) Its unit element is (k, 0).
(c) The symmetric braiding is defined as

σ A, { , }A ⊗ B, { , }B = B, { , }B ⊗ A, { , }A .
Proof. To prove part (a) we have to show that the tensor product is associative. We compute:

a ⊗ (b ⊗ c), a ⊗ (b ⊗ c ) = {·,·} ⊗ μB ⊗ μC + μA ⊗ {·,·} ⊗ μC

+ μA ⊗ μB ⊗ {·,·} τ a ⊗ (b ⊗ c) ⊗ a ⊗ (b ⊗ c ) ,
where τ = σ45 ◦ σ23,4 and σ45 = IdA⊗A⊗B ⊗ σC,B ⊗ IdC and σ23,4 = IdA ⊗ σB⊗C,A ⊗ IdB⊗C .
Similarly, we obtain that

(a ⊗ b) ⊗ c, (a ⊗ b ) ⊗ c

= {·,·} ⊗ μB ⊗ μC + μA ⊗ {·,·} ⊗ μC + μA ⊗ μB ⊗ {·,·} τ a ⊗ (b ⊗ c) ⊗ a ⊗ (b ⊗ c ) ,
where τ = σ23 ◦ σ3,45 and σ23 = IdA ⊗ σB,A and σ3,45 = IdA⊗B ⊗ σC,A⊗B ⊗ IdC . Since σ is a
braiding we have σ23,4 = σ23 ◦ σ34 and σ3,45 = σ45 ◦ σ34 with σ34 = IdA⊗B σC,A ⊗ IdB⊗C .
Since σ23 ◦ σ45 = σ45 ◦ σ23 it is now easy to verify that σ23 ◦ σ3,45 = σ45 ◦ σ23,4 and hence,
that the tensor product is indeed associative.
Part (a) is proved, and parts (b) and (c) are obvious. Lemma 2.6 is proved. 2
The following facts are obvious.

Lemma 2.7.
(a) The assignment A → (A, 0) defines a faithful tensor functor from the category of commuta-
tive algebras in C to the category of bracketed algebras.
(b) The assignment (A, {·,·}) → A defines a forgetful functor from BAlg(C) to the category of
algebras in C.
2.2. Symmetric algebras and bracketed symmetric algebras
Let V be an object of the symmetric linear category C. The category C is Abelian, hence
IdV ⊗V + σV ,V is a morphism in C and the symmetric square of S 2 V = Im(IdV ⊗V + σ ), respec-
tively the exterior square Λ2 V = Ker(IdV ⊗V + σ ) are objects in C. Define the nth symmetric
power as

S n V = S 2 V ⊗ V ⊗(n−2) ∩ V ⊗ S 2 V ⊗ V ⊗(n−2) ∩ · · · ∩ V ⊗(n−2) ⊗ S 2 V ⊂ V ⊗n .
Similarly, define the nth exterior power as

Λn V = Λ2 V ⊗ V ⊗(n−2) ∩ V ⊗ Λ2 V ⊗ V ⊗(n−2) ∩ · · · ∩ V ⊗(n−2) ⊗ Λ2 V ⊂ V ⊗n .
Define the symmetric algebra S(V ) = T (V )/Λ2 V as the quotient of the tensor algebra of
V by the two-sided ideal generated by the exterior square.
The following facts are immediate for the symmetric linear category C, where the braiding is
given by the permutation of factors.
Proposition 2.8.
(a) Let V be an object of C. The symmetric algebra S(V ) is a commutative algebra in C.

(b) The assignment V → S(V ) is functorial.
(c) The functor is exponential, i.e. S(U ⊕ V ) ∼
= S(U ) ⊗σ S(V ).
Define morphisms σi,i+1 ∈ EndC (V ⊗n ) for all 1 i (n − 1) by σi,i+1 = Id⊗(i−1) ⊗

σ ⊗ Id⊗(n−i−1) . Define for any permutation τ in the symmetric group Sn the morphism
στ ∈ End(V ⊗n ) as στ = σir ,ir +1 ◦ · · · ◦ σi1 ,i1 +1 , where τ = (ir , ir + 1) ◦ · · · ◦ (i1 , i1 + 1) is a
presentation of τ consisting of simple transpositions and σi,i+1 defined as above. Note that στ is
well defined, independent of the choice of presentation of τ .
Recall the definition of the nth braided factorial:

[n]!σ = [n]!V ,σ = [n]!σ : V ⊗n → V ⊗n , [n]!σ = στ .
τ ∈Sn
By definition [n]!σ ◦
σi = σi ◦ [n]!σ = [n]!σ . Similarly, define the nth braided skew-factorial
[n]!−σ = [n]!−σ,V = τ ∈Sn (−1) (τ ) στ , where (τ ) denotes the length of the permutation τ .
The following fact is well known.
Proposition 2.9. One has for any object V of C:
(a) S n V = Im([n]!V ,σ ) and Λn V = Im([n]!V ,−σ ).

(b) Λ2 V n = Ker[n]!σ , where Λ2 V n is the degree n component of the ideal Λ2 V . Equiv-
alently, the composition φn : S n V → T V S(V ) is an isomorphism between S n V and
S(V )n , where S(V )n is the nth graded component of S(V ).
Let V be an object in C and let Φ : V ⊗ V → V ⊗ V be a morphism in C. For 1 i < j n

define Φi,j : V ⊗n → V ⊗n as follows. First, Φi,i+1 : ∈ End(V ⊗n ) = Id⊗i−1 ⊗ Φ ⊗ Id⊗n−i−1 and,
recursively, for i < j − 1
Φi,j = σ(j −1,j ) ◦ Φi,j −1 ◦ σ(j −1,j ) . (2.5)
Proposition 2.10. For any i < j and any τ ∈ Sn such that τ (i) < τ (j ) one has Φτ (i),τ (j ) =
στ ◦ Φi,j ◦ στ −1 .
Proof. We first need the following fact.
Lemma 2.11. If 1 i < j n, m n − 1 and τ = (m, m + 1), then σm,m+1 ◦ Φi,j ◦ σm,m+1 =
Φτ (i),τ (j ) .
Proof. Clearly, we obtain
σm,m+1 ◦ Φi,j ◦ σm,m+1 = Φi,j , (2.6)
if {m, m + 1} ∩ {i, j } = ∅.
The assertion holds by definition (2.5) if m = j or m = j − 1. It remains the case, when m = i
or m + 1 = i. Since σm,m+1 is an involution, it suffices to prove the assertion for m + 1 = i; i.e.
we have to show that σi−1,i ◦ Φi,j ◦ σi−1,i = Φi−1,j . We need the following fact.
Lemma 2.12. Let τ = (j − 1, j ) ◦ (j − 2, j − 1) ◦ · · · ◦ (i − 1, i). Then we have for Φ ∈ End(V ⊗

V ) that
στ ◦ Φi,j ◦ στ −1 = Φi−1,j −1 = Φτ (i),τ (j ) .
Proof. The braid relation of the symmetric group Sn yields that σi,i+1 ◦ σi,i−1 ◦ Φi,i+1 =
Φi−1,i ◦ σi,i−1 ◦ σi,i+1 . We obtain through repeated application that στ ◦ Φi,j = Φi−1,j −1 ◦ στ .
Now, we compute
σi−1,i ◦ Φi,j ◦ σi−1,i

= σi−1,i ◦ (στ )−1 ◦ Φi−1,j −1 ◦ στ ◦ σi−1,i
= σi,i+1 ◦ · · · ◦ σj −2,j −1 ◦ σj −1,j ◦ Φi−1,j −1 ◦ σj −1,j ◦ σj −2,j −1 ◦ · · · ◦ σi,i+1
◦ σi,i+1 ◦ · · · ◦ σj −2,j −1 ◦ Φi−1,j ◦ σj −2,j −1 ◦ · · · ◦ σi,i+1 .
We obtain
σi,i+1 ◦ · · · ◦ σj −2,j −1 ◦ Φi−1,j ◦ σj −2,j −1 ◦ · · · ◦ σi,i+1 = Φi−1,j

by applying (2.6) multiple times, and hence, σi−1,i ◦ Φi−1,j ◦ σi−1,i = Φi,j as desired.
Lemma 2.11 is proved. 2
We can now prove Proposition 2.10 by induction on the length (τ ) of τ . The inductive base is
provided by Lemma 2.11. Next , let τ ∈ Sn such that (τ ) > 1. Write τ = τ ◦ σm,m+1 such that
(τ ) = (τ ) − 1. It is easy to see that τ (i) < τ (j ). Otherwise we would have τ (i) = m, τ (j ) =
m + 1, and τ (i) = m + 1 and τ (j ) = m. This implies that we would obtain a reduced expression
τ = τ ◦ σm ,m +1 ◦ τ with (τ ) = (τ ) + (τ ) + 1, such that τ (i) = m, τ (j ) = m + 1.
It is now easy to see that τ = τ ◦ τ , and hence (τ ) < (τ ). We can now apply the inductive
hypothesis to τ to obtain στ ◦ Φi,j ◦ σ(τ )−1 = Φτ (i),τ (j ) . Note that τ −1 = σm,m+1 ◦ (τ )−1 .
Lemma 2.11 implies that
σm,m+1 ◦ Φτ (i),τ (j ) ◦ σm,m+1 = Φτ (i),τ (j ) .
Proposition 2.10 is proved. 2
We now define a map Φ (m,n) : V ⊗(m+n) → V ⊗(m+n) by the formula:

m m+n
Φ (m,n) := Φi,j .
i=1 j =m+1
For any a ∈ S m V , b ∈ S n V , and â ∈ V ⊗m and b̂ ∈ V ⊗n such that m!

1
[m]!σ (â) = a and
n! [n]!σ (b̂) = b we define the morphism {·,·} :S V ⊗ S V → S
1 (m,n) m n m+n V by
(m,n) 1
{a, b}Φ := [m + n]!σ Φ (m,n) (â ⊗ b̂).
(m + n)!
The above morphism is well defined because of the following result. Recall that Sm ×Sn embeds
naturally in Sm+n .
Lemma 2.13. If τ = (τ1 , τ2 ) ∈ Sm × Sn ⊂ Sm+n , then στ ◦ Φ (m,n) = Φ (m,n) ◦ στ .
Proof. One has by Lemma 2.10 that for any τ ∈ Sn+m and any 1 i < j m + n that στ ◦
Φτ (i),τ (j ) = Φi,j ◦ στ . We compute

m m+n
m m+n
Φ (m,n) στ = Φi,j ◦ στ = στ ◦ Φτ −1 (i),τ −1 (j ) = στ ◦ Φ (m,n) .
1 2
i=1 j =m+1 i=1 j =m+1
We have the following result.
Proposition 2.14. Let (V , Φ) be a decorated space. The pair (S(V ), {·,·}Φ ), where {·,·}Φ =
(m,n)
m,n∈Z0 {·,·}Φ , is a bracketed algebra in C.
Proof. Prove anti-commutativity (2.2) first. We need the following fact.

Lemma 2.15.
(a) Let (V , Φ) be a decorated space. Then for all n 2 and i < j n one has
σ(i,j ) ◦ Φi,j = −Φi,j ◦ σ(i,j ) .
(b) If i < j n and τ ∈ Sn such that τ (i) > τ (j ), then
στ ◦ Φi,j ◦ στ −1 = −Φτ (j ),τ (i) .
Proof. Prove (a) first. Note that σ(i,i+1) ◦ Φi,i+1 ◦ σ(i,i+1) = −Φi,i+1 and that (i, j ) = (j −
1, j ) ◦ (j − 2, j − 1) ◦ · · · ◦ (i + 1, i + 2) ◦ (i, i + 1) ◦ (i + 1, i + 2) ◦ · · · ◦ (j − 1, j ). Denote
τ = (j − 1, j ) ◦ (j − 2, j − 1) ◦ (i + 1, i + 2). We compute using Proposition 2.10,
σ(i,j ) ◦ Φi,j ◦ σ(i,j )−1 = στ ◦ σ(i,i+1) ◦ στ −1 ◦ Φi,j ◦ στ ◦ σ(i,i+1) ◦ στ −1

= στ ◦ σ(i,i+1) ◦ Φi,i+1 ◦ σ(i,i+1) ◦ σ(τ )−1 = −στ ◦ Φi,i+1 ◦ στ = −Φi,j .
Part (a) is proved.

Prove (b) now. We clearly have τ (i) < τ (j ) for τ = (τ (j ), τ (i)) ◦ τ . Therefore, we have by
Lemma 2.10 and part (a),
−Φτ (j ),τ (i) = σ(τ (j ),τ (i)) ◦ Φτ (j ),τ (i) ◦ σ(τ (j ),τ (i)) = σ(τ (j ),τ (i)) ◦ στ ◦ Φi,j ◦ στ ◦ σ(τ (j ),τ (i))
= στ ◦ Φi,j ◦ στ −1 .
Part (b) is proved. The lemma is proved. 2
Let τ ∈ Sm+n be the permutation, which sends (1, 2, . . . , m, m + 1, . . . , n + m) → (n + 1, n +

2, . . . , m + n, 1, 2, . . . , n). Lemma 2.15 yields that for i n < j m + n one has Φi,j ◦ στ =
−στ ◦ Φτ (j ),τ (i) . Therefore,

m m+n
m m+n
Φ (n,m)
◦ στ = Φi,j ◦ στ = −στ Φτ (j ),τ (i) .
i=1 j =m+1 i=1 j =m+1
(n,m) (m,n)
This implies that {b, a}Φ = −{a, b}Φ for a ∈ S m V and b ∈ S n V , hence {a, b}Φ =
−{b, a}Φ . Anti-commutativity is proved.
It remains to verify the Leibniz identity (2.3). Let a ∈ S n V , b ∈ S m V and c ∈ S V and let
â ∈ V ⊗n , b̂ ∈ V ⊗m and ĉ ∈ V ⊗ be representatives of a, b and c, respectively. Denote by τ ∈
Sm+n+ the permutation τ (1, . . . , n + m + ) = (n + 1, . . . , n + m, 1, . . . , n, n + m + 1, . . . , n +
m + ). We compute
(n,m+ )
{a, b · c}Φ = {a, b · c}Φ
1
= [n + m + ]!σ Φ (n,m+ ) (â ⊗ b̂ ⊗ ĉ)
(n + m + )!

[n + m]!σ ⊗ [ ]!σ
n m+n
1
= [n + m + ]!σ Φi,j (â ⊗ b̂ ⊗ ĉ)
(n + m + )! (n + m)! !
i=1 j =n+1
n
[m]!σ ⊗ [n + ]!σ
n+m+
1
+ [n + m + ]!σ τ Φi,j
(n + m + )! m!(n + )!
i=1 j =n+m+1
× (â ⊗ b̂ ⊗ ĉ)
= {a, b} · c + b · {a, c}.
Therefore the Leibniz rule holds. Proposition 2.14 is proved. 2
We will denote by S(V , Φ) the bracketed algebra (S(V ), {·,·}Φ ) from Proposition 2.14 and
refer to it as the symmetric algebra of the decorated space (V , Φ). We have the following result.
Proposition 2.16. The correspondence (V , Φ) → S(V , Φ) defines a faithful exponential functor

from the category of decorated spaces to the category of bracketed algebras BAlg(C).
Proof. It is easy to verify that the correspondence is functorial and faithful. It remains to check
that it is exponential. By Proposition 2.8 one has S(V ⊕ V ) = S(V ) ⊗ S(V ), and we obtain the
bracket defined by

{u + u , v + v }Φ = (Φ + Φ ) (u + u ) ∧ (v + v )
for all u, v ∈ V and u , v ∈ V . The proposition is proved. 2
2.3. Bracketed Poisson algebras
Denote by J : A⊗3 → A the Jacobian map defined by
J = F + F ◦ σ12 ◦ σ23 + F ◦ σ23 ◦ σ12 , (2.7)
where F : A⊗3 → A, F (a, b, c) = {a, {b, c}} and σ12 = σ ⊗ Id and σ23 = Id ⊗ σ .
The following fact is obvious.
Lemma 2.17. (See also [2, Definition 2.23].) A bracketed algebra (A, {·,·}) is Poisson, if and
only if J (A⊗3 ) = 0.
Define the Jacobian ideal JΦ as the two-sided (bracketed) ideal in the bracketed algebra
S(V , Φ) generated by the image of the Jacobian map. We call the quotient of S(V )Φ by JΦ
the Poisson closure of S(V , Φ). The bracket {·,·}Φ induces a bracket {·,·}Φ on S(V )Φ , because
JΦ is by definition closed under the bracket.
Definition 2.18. The reduced symmetric algebra S(V , Φ) is the bracketed algebra S(V , Φ) =
(S(V )Φ , {·,·}Φ ).
We have the following result.

Proposition 2.19.
(a) The reduced symmetric algebra S(V , Φ) is Poisson.

(b) The reduced symmetric algebra S(V , Φ) has the following universal property:
Any homomorphism of bracketed algebras from S(V , Φ) to a Poisson algebra P factors

through S(V , Φ).
(c) The assignment (V , Φ) → S(V , Φ) defines a functor from the category of decorated spaces
to the category of Poisson algebras. Moreover, if (V , Φ ) = (V , Φ) ⊕ (V , Φ ), then there
exists a surjective homomorphism
S(V , Φ) ⊗ S(V , Φ ) S(V , Φ ).
Proof. Prove (a) first. Let a, b, c ∈ S(V , Φ) and let a, b, c ∈ S(V , Φ) be representatives of the
equivalence classes of a, b, c, respectively. Then J (a, b, c) is a representative of the class of
J (a, b, c), where J is the induced Jacobian map on S(V , Φ). By definition, J (a, b, c) ∈ JΦ ,
hence J (a, b, c) = 0 ∈ S(V , Φ). Part (a) is proved.
Prove (b) next. Let P be a Poisson algebra and ρ : S(V , Φ) → P a homomorphism of brack-
eted algebras. It is easy to see that J is contained in the kernel of ρ. Hence, ρ factors through
S(V , Φ). Part (b) is proved.
Prove (c) now. Let (V , Φ) and (V , Φ ) be decorated spaces, (V , Φ ) = (V , Φ) ⊕ (V , Φ ),
and J , respectively J , the Jacobian ideal in S(V , Φ), respectively S(V , Φ ). It is clear that
the ideals generated by J and J in S(V , Φ ) are contained in the Jacobian ideal J of
S(V , Φ ). Part (c) follows since S(V , Φ ) = S(V , Φ) ⊗ S(V , Φ ) by Proposition 2.16. The
proposition is proved. 2
Due to Proposition 2.19(a) we will sometimes refer to S(V , Φ) as the Poisson closure of
S(V , Φ) (see also [2, Section 3.1]).
We will now discuss conditions that are sufficient for S(V , Φ) to be Poisson; i.e., under which
S(V , Φ) = S(V , Φ).
For any Φ ∈ End(V ⊗ V ) define the Schouten square JΦ, ΦK ∈ End(V ⊗ V ⊗ V ) by
JΦ, ΦK = [Φ12 , Φ13 ] + [Φ12 , Φ23 ] + [Φ13 , Φ23 ],
where [a, b] = a ◦ b − b ◦ a denotes the usual commutator.

The Schouten square has the following very important property.
Lemma 2.20. Let V be an object of C and let Φ ∈ End(V ⊗ V ) such that Φ ◦ σ = −σ ◦ Φ. Then,
σi,i+1 ◦ JΦ, ΦK ◦ σi,i+1 = −JΦ, ΦK for i = 1, 2.
Proof. Straightforward computation yields:

σ12 ◦ JΦ, ΦK ◦ σ12 = σ12 ◦ [Φ12 , Φ13 ] + [Φ12 , Φ23 ] + [Φ13 , Φ23 ] ◦ σ12
= −[Φ12 , Φ23 ] − [Φ12 , Φ13 ] + [Φ23 , Φ13 ] = −JΦ, ΦK.
Similarly, we compute
σ23 ◦ JΦ, ΦK ◦ σ23 = [Φ13 , Φ12 ] − [Φ13 , Φ23 ] − [Φ12 , Φ23 ] = −JΦ, ΦK.
We call a decorated space (V , Φ) Poisson, if the symmetric algebra S(V , Φ) of the decorated
space (V , Φ) is Poisson.
Theorem 2.21. Let (V , Φ) be a decorated space. The following are equivalent:
(a) (V , Φ) is Poisson.
(b) Φ satisfies the equation
[3]!σ ◦ JΦ, ΦK = 0. (2.8)
(c) Φ satisfies the equation
JΦ, ΦK ◦ [3]!−σ = 0.
(d) Φ satisfies

JΦ, ΦK Λ3 V = 0,
where |Λ3 V denotes the restriction to Λ3 V .
Proof. The equivalence of (a) and (b) is well known and proved in [10, Theorem 3.1]. For the
convenience of the reader we nevertheless prove here that (a) equivalent (b).
Lemma 2.22. One has
Im(J ) ∩ S 3 (V ) = [3]!σ ◦ JΦ, ΦK.
Proof. Define the lifted Jacobian J : V ⊗3 → V ⊗3 by
J (x, y, z) := G + G ◦ σ12 ◦ σ23 + G ◦ σ23 ◦ σ12 ,
where G : V ⊗3 → V ⊗3 is the morphism given by
G = Φ12 ◦ (1 ⊗ 1 ⊗ 1 + σ23 ) ◦ Φ23 = (Φ23 ◦ Φ12 + σ12 ◦ Φ13 ◦ Φ12 ).
By definition, J (x, y, z) = [3]!

3! (J (x, y, z)) for all x, y, z ∈ S(V ) and all x, y, z ∈ T (V ) such
σ
that [3]!σ (x) = 3! · x, [3]!σ (y) = 3! · y and [3]!σ (z) = 3! · z.

One has [3]!σ ◦ σi,j = [3]!σ for all i, j = 1, 2, 3. Therefore, we obtain:
[3]!σ ◦ G = [3]!σ ◦ (Φ23 ◦ Φ12 + σ12 ◦ Φ13 ◦ Φ12 )

= [3]!σ ◦ (Φ23 ◦ Φ12 + Φ13 ◦ Φ12 ).
Similarly,
[3]!σ ◦ G ◦ σ12 ◦ σ23 = [3]!σ ◦ (−σ12 ◦ σ23 ◦ Φ12 ◦ Φ13 + σ23 ◦ Φ23 ◦ Φ13 )
= [3]!σ ◦ (−Φ12 ◦ Φ13 + Φ23 ◦ Φ13 ).
[3]!σ ◦ G ◦ σ23 ◦ σ12 = [3]!σ ◦ (−Φ13 ◦ Φ23 − Φ12 ◦ Φ23 ).
Combining these equations, we obtain
[3]!σ ◦ (G + G ◦ σ12 ◦ σ12 + G ◦ σ23 ◦ σ12 ) = −[3]!σ ◦ JΦ, ΦK
and thus

3! · J (x, y, z) = − [3]!σ ◦ JΦ, ΦK (x ⊗ y ⊗ z) (2.9)
for all x, y, z ∈ V . The lemma is proved. 2
We need the following fact which generalizes [2, Lemma 3.7].
Lemma 2.23. Let (A, {·,·}) be a bracketed Z0 -graded algebra in C generated by A1 and such
that A0 ∼
= k. Then A is Poisson if and only if the Jacobian (see (2.7)) vanishes on (A1 )3 .
Proof. We proceed by induction in homogeneity degrees = n + m + k of monomials u · v · w

with u ∈ An , v ∈ Am , w ∈ Ak . We start with the base of induction, which is the assumption:
suppose that for all u , v , w ∈ A1 one has J (u , v , w ) = 0. Now let u ∈ An , v ∈ Am , w ∈ Ak ,
z ∈ A1 and assume that the assertion holds for = n + m + k. We compute using the inductive
hypothesis and the Leibniz rules (2.4) and (2.3):

J (a, b, c · d) = J (a, b, c) · d + c J (a, b, d) = 0.
Since for all u, v, w ∈ A one has J (u, v, w) = J (w, u, v) = J (v, w, u), then the assertion holds
for n + m + k = + 1. This implies that A is indeed Poisson. The lemma is proved. 2
The above lemma implies that J (S(V )3 ) = 0 if and only if Φ satisfies (2.8). Therefore, (a) and
(b) are equivalent.
We will now prove the equivalence of (b) and (c). It follows from Lemma 2.20(a) that [3]!σ ◦
JΦ, ΦK = JΦ, ΦK ◦ [3]!−σ . Therefore, [3]!σ ◦ JΦ, ΦK = 0, if and only if JΦ, ΦK ◦ [3]!−σ = 0, and
(b) and (c) are equivalent.
Parts (c) and (d) are clearly equivalent. Theorem 2.21 is proved. 2
We will now employ Theorem 2.21 to study Poisson structures on subspaces and tensor prod-
ucts of decorated spaces.
First, note the following fact.
Proposition 2.24. Let V , V and V be objects of C such that V = V ⊕ V . Let additionally,

(V , Φ) and (V , Φ ) be decorated spaces such that for all v1 ⊗ v2 ∈ V ⊗ V one has Φ (v1 ⊗
v2 ) = πV ,V ◦ Φ(v1 ⊗ v2 ), where πV ,V : V ⊗ V → V ⊗ V denotes the canonical projection. If
(V , Φ) is Poisson, then (V , Φ ) is Poisson.

3 i 3−i V , and S 3 V =

3 i 3−i V .
Proof. One has Λ3 V = i=0 Λ V ⊗ Λ i=0 S V ⊗ S

Clearly, JΦ, ΦK defines a map JΦ, ΦK : Λ V → S V and it follows from our assertion
3 3
that JΦ, ΦK = JΦ , Φ K : Λ3 V → S 3 V ⊂ S 3 V . If (V , Φ) is Poisson, then JΦ, ΦK (Λ3 V ) =

JΦ , Φ K(Λ3 V ) = 0, and hence (V , Φ ) is Poisson. 2
The following result relates Poisson structures and tensor products.
Theorem 2.25. Let (U, Φ) and (V , Φ ) be decorated spaces. If their tensor product (U ⊗
V , Φ ) = (U, Φ) ⊗ (V , Φ ) is Poisson, then (U, Φ) and (V , Φ ) are Poisson.
∼
Proof. Let σ̃ be the “shuffle” U ⊗3 ⊗ V ⊗3 −
→ (U ⊗ V )3 . Abbreviating,

U 3,0 = S 3 U = S 2 U ⊗ U ∩ U ⊗ S 2 U , U 2,1 = S 2 U ⊗ U ∩ U ⊗ Λ2 U ,

U 0,3 = Λ3 U = Λ2 U ⊗ U ∩ U ⊗ Λ2 U , U 1,2 = Λ2 U ⊗ U ∩ U ⊗ S 2 U
and the same for V , we have the following containments for Λ3 (U ⊗ V ) and S 3 (U ⊗ V ):

Λ3 (U ⊗ V ) ⊇ σ̃ U i,j ⊗ V j,i , S 3 (U ⊗ V ) ⊇ σ̃ U i,j ⊗ V i,j .
i+j =3 i+j =3
] = 0 ∈ End((U ⊗ V )⊗2 ) we have that

Since Φ = Φ13 + Φ24 and [Φ13 , Φ24
q y q y
JΦ , Φ K = Φ13 + Φ24

, Φ13 + Φ24 = JΦ13 , Φ13 K + Φ24
, Φ24 .
Note that Φ(Λ2 U ) ⊆ S 2 U (respectively Φ(Λ2 V ) ⊆ S 2 V and Φ(S 2 U ) ⊆ Λ2 U (respectively

Φ(S 2 U ) ⊆ Λ2 U ). Hence

JΦ13 , Φ13 K σ̃ U i,j ⊗ V j,i ⊆ σ̃ U j,i ⊗ V j,i ,
q
y i,j
Φ24 , Φ24 σ̃ U ⊗ V j,i ⊆ σ̃ U i,j ⊗ V i,j
for all i + j = 3. This implies that

JΦ , Φ K σ̃ U i,j ⊗ V j,i ⊂ σ̃ U j,i ⊗ V j,i + U i,j ⊗ V i,j .
Theorem 2.25 now follows as the special cases i = 3, j = 0 and i = 0, j = 3 from the follow-
ing more general obvious result.
Lemma 2.26. If (U, Φ) ⊗ (V , Φ ) is Poisson, then

JΦ , Φ K σ̃ U i,j ⊗ V j,i = {0} ⊆ σ̃ U j,i ⊗ V j,i ⊕ U i,j ⊗ V i,j
for all i + j = 3.
Theorem 2.25 is proved. 2
The following example shows that the converse of Theorem 2.25 does not hold.
Example 2.27. Let V = V = C2 with standard basis {e1 , e2 }, and let Φ(ei ⊗ ej ) = sign(i −
j )(ej ⊗ ei ) and Φ (ei ⊗ ej ) = λ · sign(i − j )(ej ⊗ ei ). Clearly, both (V , Φ) and (V , Φ ) are
Poisson, because Λ3 C2 = {0}, but straightforward calculation shows that (V , Φ) ⊗ (V , Φ ) is
Poisson, if and only if λ = ±1.
We conclude this section with an apparently well-known and useful observation regarding a
general operator Φ : V ⊗ V → V ⊗ V that satisfies the identity JΦ, ΦK = 0, the classical Yang–
Baxter equation. For the reader’s convenience, however, we give a proof.
Proposition 2.28. Let Φ be an operator such that JΦ, ΦK = 0, and define Φ + = 12 (Φ + Φ op )

and Φ − = 12 (Φ − Φ op ), where Φ op = σ ◦ Φ ◦ σ . Then,
+ +
JΦ − , Φ − K = − Φ12 , Φ23 .
Proof. First note the following fact.
Lemma 2.29. Let Φ : V ⊗ V → V ⊗ V satisfy JΦ, ΦK = 0. Then Φ op satisfies JΦ op , Φ op K = 0.
Proof. If JΦ, ΦK = 0, then also σ13 ◦ JΦ, ΦK ◦ σ13 = 0. Hence,

q y
0 = σ13 ◦ JΦ, ΦK ◦ σ13 = [Φ32 , Φ31 ] + [Φ32 , Φ21 ] + [Φ31 , Φ21 ] = − Φ op , Φ op .
We need the following lemma.
Lemma 2.30. Let Φ be an operator such that JΦ, ΦK = 0. In the notation of Proposition 2.28
one has the following identity: JΦ − , Φ − K = −JΦ + , Φ + K.
Proof. Since JΦ, ΦK = [Φ op , Φ op K = 0 we obtain that JΦ − , Φ − K = 12 JΦ, −Φ op K = − 12 JΦ, Φ op K

and JΦ + , Φ + K = 12 JΦ, Φ op K. The lemma is proved. 2
By Lemma 2.20 we obtain that
JΦ − , Φ − K − σ13 ◦ JΦ − , Φ − K ◦ σ13 = 2JΦ − , Φ − K.
Using Lemma 2.30, we obtain

+ +
2JΦ − , Φ − K = −JΦ + , Φ + K + σ13 ◦ JΦ + , Φ + K ◦ σ13 = −2 Φ12 , Φ23 .
3. Poisson modules over Lie algebras
3.1. Definition and basic properties of Poisson modules
Let (g, (·,·)) be a quadratic complex Lie algebra; i.e. a complex Lie algebra g with a symmetric
invariant bilinear form (·,·) : g ⊗ g → C. Clearly (·,·) ∈ (g ⊗ g)∗ ∼
= g∗ ⊗ g∗ . The form (·,·) defines
an isomorphism between g∗ and g, and under this isomorphism we can identify the form with a
symmetric g-invariant element (·,·) = c ∈ S 2 (g) ⊂ g ⊗ g. In the case when g is semisimple and
(·,·) is the Killing form, c is known as the Casimir element. Similarly, note that the Lie bracket
[·,·] : g ∧ g → g defines an element [·,·] : g∗ ⊗ g∗ ⊗ g and we obtain under the isomorphism above
the canonical element c = [·,·] ∈ g3 . Observe the following facts.
Lemma 3.1.
(a) The canonical element c is g-invariant and totally skew symmetric; i.e., c ∈ (Λ3 g)g .
(b) The elements c ∈ S 2 g and c ∈ Λ3 g are related by
c = −[c12 , c23 ], (3.1)
where c12 = c ⊗ 1 and c23 = 1 ⊗ c.
Proof. Prove (a) first. By definition c = c(1) ⊗ c(1) ⊗ [c(2) , c(2) ] = [c13 , c23 ], where c = c(1) ⊗
c(2) . Note that [c(2) , c(2) ] = 0, despite Sweedler’s notation being very suggestive.
Since c is g-invariant, it is easy to see that c is g-invariant, as well. We have to prove that c
is anti-symmetric. We will show first that c indeed anti-commutes with the permutation σ13 ; i.e.,
[c13 , c12 ] = −[c13 , c23 ]. Since c is g-invariant we have [c, g ⊗ 1] = −[c, 1 ⊗ g] for all g ∈ g.
Now let c = c(1) ⊗ c(2) . We obtain,
[c13 , c12 ] = [c(1) , c(1) ] ⊗ c(2) ⊗ c(2) = −c(1) ⊗ c(1) ⊗ [c(2) , c(2) ] = −[c13 , c23 ].
We can show analogously that c anti-commutes with σ23 , as well. Part (a) is proved and (b)
follows immediately. The lemma is proved. 2
Note that c defines for each finite-dimensional g-module V a g-module homomorphism

c : Λ3 V → S 3 V . We make the following definition, and then explain, how it is connected to
the Poisson decorated spaces introduced in Section 2.3.
Definition 3.2. Let (g, (·,·)) be a Lie algebra with a symmetric invariant bilinear form. We say
that a finite-dimensional g-module V is Poisson, if

c Λ3 V = {0} ∈ S 3 V .
We immediately obtain the following sufficient condition guaranteeing that a g-module V is

Poisson.
Proposition 3.3. Let V be a finite-dimensional g-module. If Homg (Λ3 V , S 3 V ) = {0}, then V is

Poisson.
We make the following definition.
Definition 3.4. Let g be a complex semisimple Lie algebra. An element r ∈ g ⊗ g is called a

classical r-matrix if r satisfies:
(i) the classical Yang–Baxter equation, i.e.
Jr, rK = [r12 , r13 ] + [r12 , r23 ] + [r13 , r23 ] = 0, and
(ii) r + τ (r) = 2c, where c is the Casimir element of g.
Example 3.5. Let g = sl2 (C). The standard r-matrix is r = E ⊗ F , where E, F, H are the
elements of the standard basis of sl2 (C).
The classical r-matrices have been classified in the celebrated paper [1] in terms of Belavin–
Drinfeld triples.
Consider a classical r-matrix r and its antisymmetrized r-matrix r − = 12 (r − r op ), and a
finite-dimensional g-module V . The element r − ∈ g ⊗ g acts on V ⊗ V and the corresponding
decorated space (V , r − ) defines a bracket on the symmetric g-module algebra S(V ) defined as
{u, v}r − = r − (u∧v) on all u, v ∈ S(V ) as constructed in Proposition 2.14. We have the following
result.
Proposition 3.6. Let g be a complex semisimple Lie algebra, (·,·) the Killing form and r a
classical r-matrix, and let V be a finite-dimensional g-module. The decorated space (V , r − ) is
Poisson if and only if V is Poisson.
Proof. We have to prove that (V , r − ) is Poisson, if and only if c(Λ3 V ) = {0} ∈ S 3 V . We obtain
from Proposition 2.28 that
+ +
Jr − , r − K = r12 , r23 = [c12 , c23 ] = c.
The assertion now follows from Theorem 2.21. 2
We note the following facts.
Lemma 3.7. Let (g, (·,·)) be a quadratic algebra and let V = V g be a trivial g-module. Then V
is Poisson.
Proof. Obvious. 2
Lemma 3.8. Let (g1 , (·,·)1 ) and (g2 , (·,·)2 ) be quadratic Lie algebras and let c1 ∈ S 2 g1
and c2 ∈ S 2 g2 be the elements corresponding to (g1 , (·,·)1 ) and (g2 , (·,·)2 ). Then (g, (·,·)) =
(g1 ⊕ g2 , (·,·)) is a quadratic Lie algebra with form

g1 + g2 , g1 + g2 = g1 , g1 1 + g2 , g2 2
for g1 , g1 ∈ g1 and g2 , g2 ∈ g2 . Moreover, one has c = c1 + c2 .
Proof. The assertion follows from the fact that the subalgebras (g1 , 0) ∈ g and (0, g2 ) ∈ g com-
mute. 2
The following technical result will be of particular importance for the classification of Poisson
modules over a semisimple Lie algebra g, as it allows to restrict to certain good subalgebras, such
as Levi subalgebras (see Proposition 6.6).
Proposition 3.9. Let (g, (·,·)) be a quadratic Lie-algebra. Denote by c ∈ S 2 (g) the g-invariant
element corresponding to (·,·). Let gsub ⊂ g be a subalgebra such that gsub ∩ g⊥sub = {0}. Denote
by csub ∈ S (gsub ) the g-invariant element corresponding to (·,·)gsub .
2
(a) One has c = csub + c , where c ∈ g⊥

sub ∧ g ∧ g in the notation of Appendix A.
(b) Let V be a g-module and V1 ⊂ V a gsub -module such that g⊥ sub (V1 ) ∩ V1 = {0}. If the canon-
ical element c ∈ Λ3 g defined in (3.1) satisfies c(Λ3 V ) = {0}, then the element csub ∈ Λ3 gsub
satisfies csub (Λ3 V1 ) = {0}.
Proof. Prove (a) first. By definition we can express the element c ∈ g ⊗ g corresponding to (·,·)
∈ g⊥ ⊗ g ⊕ g ⊗ g⊥ . We obtain from (3.1)
as c = csub + crest , where csub ∈ gsub ⊗ gsub and crest sub sub
that

c = (csub + crest )12 , (csub + crest )23 = csub + c ,
where c ∈ g⊥
sub ∧ g ∧ g. Part (a) is proved.

3
Prove (b) now. Recall that S 3 V ∼ = i=0 S i V1 S 3−i V2 . One has csub (Λ3 V1 ) ⊂ S 3 V1 . Clearly,
c
c (V1 ⊗ V1 ⊗ V1 ) ⊂ S 3 V1 ,
in the notation of Appendix A. If V is Poisson, then c(Λ3 V1 ) = {0}, and hence csub (Λ3 V1 ) = {0}
and c (Λ3 V1 ) = {0}. This implies directly that V1 is Poisson as a gsub -module.
If g is a reductive Lie algebra we have the following fact.
Proposition 3.10. Let g be a reductive Lie algebra and κ(x, y) = tr(ad(x)ad(y)). The Lie
algebra g splits as g = g ⊕ z into a semisimple part g and a central subalgebra z. A finite-
dimensional (g, κ)-module V is Poisson, if and only if V is Poisson under the restriction to
(g , κg ), where κg denotes the restriction of κ to g , the Killing-form.
Proof. It is easy to see that cg = cg and cg = cg . The assertion now follows immediately. 2
3.2. Classification of Poisson modules over semisimple Lie algebras
In this section we will classify all simple Poisson modules over a semisimple Lie algebra g.
By Proposition 3.10 we immediately obtain a classification of all simple modules over reductive
Lie algebras. First we will introduce some notation. Choose a Borel subalgebra b ⊂ g and denote
by h and n+ the corresponding Cartan and upper nilpotent subalgebras, and, similarly, by b−
and n− the lower Borel and nilpotent subalgebras. By W (g) we shall denote the Weyl group of
g and by (·,·)h and (·,·)h∗ the standard inner product on h and h∗ , which we identify via the
inner product. Denote by R(g) ⊂ h∗ the set of roots, by R + (g) (respectively R − (g)) the set of
positive (respectively negative) roots and by Δ = {α1 , . . . , αn } the set of simple roots. Denote
by Eα for α ∈ R(g) and Hα ⊂ h, α ∈ R+ (g) the standard generators of g with the property that
[Eα , E−α ] = Hα = α̌ = 2 (α,α)
α
∈ h ⊂ g. We will sometimes write Fα = E−α for α ∈ R + (g) and
use the notation P (g) for the weight-lattice of g and ωi for the ith fundamental weight. The
numbering of fundamental weights and simple roots follows the tables in Bourbaki [3].
We now introduce the notion of geometrically decomposable modules following Howe
[12, Chapter 4]. Let g be a reductive Lie algebra and V a g-module and U ⊂ V be a b-module.
Denote by det(U ) the one-dimensional subspace det(U ) = Λtop U of Λ(U ). Clearly, det(U ) is a
b-submodule of Λ(V ), therefore, every u ∈ det(U ) is a highest weight vector in Λ(V ). In anal-
ogy to [12, Chapter 4.6], we call a highest weight vector v ∈ Λ(V ) geometric, if v ∈ det(U ) for
some b-module U ⊂ V .
Definition 3.11. (See [12, Chapter 4.6].) A g-module V is called geometrically decomposable,
if ΛV is generated as a g-module by geometric highest weight vectors.
The following result is the first main theorem of this section.
Main Theorem 3.12. Let g be a simple complex Lie algebra, and let V be a non-trivial simple
finite-dimensional g-module. Then the following are equivalent:
(a) The module V is Poisson.

(b) The decorated space (V , r − ) is Poisson for any classical r-matrix r ∈ g ⊗ g .
(c) Homg (Λ3 V , S 3 V ) = {0}.
(d) Λ2 V is simple or (g, V ) = (sp2n (C), C2n ) for some n.
(e) The module V is geometrically decomposable or (g, V ) = (sp(2n), Vω1 ).
(f) The pair (g, V ) is one of the following:
(i) (sln (C), Vλ ) where λ ∈ {ω1 , 2ω1 , ω2 , ωn−2 , ωn−1 , 2ωn−1 }.
(ii) (so(n), Vω1 ), (so(5), Vω2 ), (so(8), Vωi ), i = 3, 4 and (so(10), Vωj ), j = 4, 5.
(iii) (sp(2n), Vω1 ) and (sp(4), Vω2 ).
(iv) (E6 , Vω1 ) and (E6 , Vω6 ).
Remark 3.13. The Poisson structures listed here, with the exception of the case (sp(2n), Vω1 )
were also described by Goodearl and Yakimov in [11, Proposition 5.2]. Moreover, they compute
the corresponding symplectic foliations.
We prove the theorem using the following strategy. The equivalence of (a) and (b) is proved
in Proposition 3.6. The implication (c) implies (a) follows from Proposition 3.3. To prove that
(a) yields (c) and (f) and, we will give necessary conditions a dominant weight λ ∈ P + (g) has
to satisfy, if Vλ is Poisson. We will then show that Homg (Λ3 Vλ , S 3 Vλ ) = {0} for all λ ∈ P + (g)
satisfying these necessary conditions proving that (f) yields (c). In order to show that (a) and
(d) are equivalent, we prove that V is Poisson if Λ2 V is simple and that Λ2 V is simple for all
pair (g, V ) in (f). The equivalence of (a) and (e) then follows from the classification of simple
geometrically decomposable modules in [12]. Since the proof is rather lengthy we refer it to
Section 6.
We can generalize Theorem 3.12 to the case of semisimple Lie algebras.

Consider a semisim-
ple Lie algebra g and a finite-dimensional g-module V . If g = ni=1 gi , where gi are sim-
ple Lie algebras, denote by Vλ1 ,...,λn the simple g-module of highest weight (λ1 , . . . , λn ) ∈
P (g1 ) ⊕ · · · ⊕ P (gn ) ∼
= P (g). Denote by the support suppg (V ) of a g-module V the product
of all simple factors gi for which gi (V ) = {0}. We have the following classification result.
Theorem 3.14. Let g be a semisimple Lie algebra and V a simple g-module. The following are
equivalent:
(a) V is Poisson.
(b) The pair (supp(g), V ) is listed in Theorem 3.12(f) or (suppg (V ), V ) = (slm × sln , Vω1 ,ω1 )
where Vω1 ,ω1 is the natural slm × sln -module.
Proof. Recall that the m × n-matrices Matm×n (C) can be given a glm × glm -module struc-
ture such that Matm×n (C) ∼ = Vωm ,ωn ∼= Vω∗1 ,ω1 with glm acting on the left and gln acting on the
right. This action yields a glm × gln -module algebra structure on C[Matm×n ] = S(Vω1 ,ω1 ). It is
well known that the r-matrix bracket defines a Poisson structure on the algebra C[Matm×n ] =
S(Vω1 ,ω1 ) via

r − (xij ⊗ xk ) = sign(i − k) + sign(j − l) xkj xi .
It remains to show that if suppg (V ) is non-simple and (suppg (V ), V ) = (slm × sln , Vω1 ,ω1 ),
then V is not Poisson.
Let r1 , . . . , rn be classical r-matrices for g1 , . . . , gn . It is easy to see that r = r1 + · · · + rn is
a classical r-matrix for g. Recall that as a vector space V can be decomposed as a tensor product
V = Vλ1 ⊗ · · · ⊗ Vλn , where Vλi is a simple gi -module. The decorated space (V , r − ) decomposes
as a tensor product (V , r − ) = (Vλ1 , r1− ) ⊗ · · · ⊗ (Vλn , rn− ). It now follows from Theorem 2.25 that
if a simple g-module V is Poisson, then each Vλi is Poisson as a gi -module as are all the products
Vλi ⊗ Vλi+1 as gi ⊕ gi+1 -modules. It therefore suffices to show the following. First, let g1 and g2
be simple Lie algebras and let Vλ1 and Vλ2 be simple Poisson g1 - (respectively g2 )-modules and
(g2 , Vλ2 ) = (slk , Vω1 ) for some k 2. Then Vλ1 ,λ2 is not Poisson. Second, we have to prove that
the natural sl × slm × sln -module Vω1 ,ω1 ,ω1 is not Poisson for all , m, n 2.
We can further reduce the list of cases to investigate by considering the embedding of some
Levi subalgebra in g and making use of Proposition 6.6. We need the following result.
Proposition 3.15.
(a) If g = sl2 × sl2 , then V = V2,i is not Poisson, where Vi denotes the (n + 1)-dimensional
simple sl2 -module.
(b) If g = sl2 × sl2 × sl2 , then V1,1,1 is not Poisson.
(c) If g = sl2 × sp(4), then V = V1,ω1 is not Poisson.
Proof. Let g = g1 ⊕ g2 be a semisimple Lie algebra. We have c = c1 + c2 , where c, c1 and c2

are the Casimir elements of g, g1 and g2 , respectively, and c = c1 + c2 .
We will first prove case (a). Let g = sl2 × sl2 and let V = Vi,2 , i 1, be a simple g-module.
Denote by Vi and V2 the corresponding simple sl2 -modules. Since V ∼ = Vi ⊗ V2 as vector spaces
we can choose non-zero vectors (u ⊗ v), (u ⊗ v ), (u ⊗ v ) ∈ Vi ⊗ V2 ⊂ V such that u ∈ Vi (i),
u = F (u) and v ∈ V2 (2), v = F (v) and v ∈ V2 (−2), where Vk ( ) denotes the -weight space
of Vk . Abbreviate uv = (u ⊗ v) ∧ (u ⊗ v ) ∧ (u ⊗ v ) ∈ Λ3 V .
Note that if g = sl2 , then csl2 = E ∧ F ∧ H and csl2 ×sl2 = E1 ∧ F1 ∧ H1 + E2 ∧ F2 ∧ H2 . It

is easy to verify that

0 = E2 ∧ F2 ∧ H2 (uv) ∈ V (i, 2) · V (i , −2) · V (i, 0) ⊕ V (i, 0) · V (i , 2) · V (i, −2) ,
where i = i − 2. Similarly we obtain that

c
0 = E1 ∧ F1 ∧ H1 (uv) ∈ V (i, 2) · V (i , −2) · V (i, 0) ⊕ V (i, 0) · V (i , 2) · V (i, −2) .
Hence, c(uv) = 0 and V is not Poisson. Part (a) is proved.

Prove (b) next. Denote by V111 the 8-dimensional natural (sl2 )3 (C)-module. Choose a basis
of V111 with basis vectors xj1 ,j2 ,j3 , ji ∈ {0, 1}, such that xj1 ,j2 ,j3 is a weight vector of weight
(1 − 2ji ) of each subalgebra gi , the ith copy of sl2 (C) in g. Moreover, we can choose the basis
such that F1 (xij k ) = δi,0 x1,j,k and E1 (xij k ) = δi,1 x0,j k , and analogously for E2 , F2 , E3 and F3 .

We have c = 3i=1 −Ei ∧ Hi ∧ Fi . It is easy to compute that
0 = c(x111 ∧ x000 ∧ x100 )

= −x011 x100
2
+ x111 x100 x000 + x101 x010 x100 − x101 x000 x110 + x110 x001 x100 − x110 x000 x101 .
Therefore c(Λ3 V111 ) = {0}, hence V111 is not Poisson. Part (b) is proved.
It remains to prove part (c). Let Vω1 be the four-dimensional natural sp(4)-module and let
V1,ω1 be the natural sl2 × sp(4)-module. As in the proof of parts (a) and (b) choose u, u ∈ V1
such that u ∈ V1 (1) and u ∈ V1 (−1) and v ∈ Vω1 (ω1 ), v = Fα1 (v) and v ∈ Vω1 (−ω1 ). Denote
uv = (u ⊗ v) ∧ (u ⊗ v ) ∧ (u ⊗ v ) ∈ Λ3 V . We have by (6.2)
(α, α)(β, β)
c=E∧F ∧H + Eα ∧ Eβ ∧ [E−α , E−β ].
4
α,β∈R(sp(4))
We obtain for the simple root α1 ∈ R(sp(4))
Eα1 ∧ E−α1 ∧ Hα1 (uv) ∈ V (−1, ω1 ) · V (1, ω1 − α1 ) · V (1, −ω1 )
and observe that indeed
(α1 , α1 )2 c
c− Eα1 ∧ E−α1 ∧ Hα1 (uv) ∈ V (−1, ω1 ) · V (1, ω1 − α1 ) · V (1, −ω1 ) .
4
This implies that c(uv) = 0 and that V1,ω1 is not Poisson. Part (c) and the proposition are
proved. 2
We now return to the proof of Theorem 3.14. Now let g = g1 ⊕ g2 such that g1 and g2 are two
simple Lie algebras, and Vλ1 and Vλ2 simple Poisson g1 , respectively g2 -modules, and assume
that (g, V ) = (slm × sln , Vω1 ,ω1 ). We will list the semisimple part g ⊂ g of the Levi subalgebras
and the corresponding simple g -module V ⊂ Vλ1 ,λ2 verifying that Vλ1 ,λ2 is not Poisson. First,
note the following fact.
Proposition 3.16. Let g be a semisimple Lie algebra and V a simple g-module such that
suppg (V ) has at least three simple factors. Then V is not Poisson.
Proof. Note the following fact.
Lemma 3.17. Let g = (sl2 )3 and let V = Vi,j,k be a simple finite-dimensional g-module with
0∈
/ {i, j, k}. Then V is not Poisson.
Proof. Since V is not Poisson if 3 by Theorem 3.12(f), we obtain from Theorem 2.25 that
if Vi,j,k is Poisson, then i, j, k 2. If i = j = k = 1, then the assertion of the lemma agrees with
the assertion of Proposition 3.15(b). Now suppose, without loss of generality, that j = 2. Then
Vi,j is not Poisson by Proposition 3.15(b) and V = Vi,j,k is not Poisson by Theorem 2.25. The
lemma is proved. 2
Suppose suppg (V ) has at least three simple factors. We can find a Levi subalgebra g ∼
= sl32
∼
such that a highest weight vector v ∈ V generates a g -submodule V = Vi,j,k with i, j, k 1.
Hence V is not Poisson by the previous lemma and Proposition 6.6. The proposition is
proved. 2
Now we are able to complete the proof of Theorem 3.14. We assume that suppg (V ) has
two simple factors. In order to deal with most cases, it suffices to exhibit a Levi subalgebra
g ⊂ g with three simple factors or a Levi subalgebra g ⊂ g and a simple module V ⊂g V
such that (g , V ) ∈ {(sl2 × sl2 , Vi,2 ), (sl2 × sp(4), V1,ω1 )} to show that (V , g) is not Poisson by
Proposition 6.6. Since there are obvious choices in a large number of cases, and a complete list
would, therefore, be rather long, we will list only the non-obvious choices.
(a) If g = so(2n + 1) ⊕ g2 and V = Vω1 ,λ choose g = sl2 × sl2 generated by the second
node of the Dynkin diagram of so(2n + 1), respectively a node i of the diagram associated to g2
such that (λ, αi ) 1. Note that Fα1 (vω1 ) ∈ Vω1 generates a three-dimensional simple sl2 -module
for the subalgebra corresponding to the second node of the Dynkin diagram. Hence, we find a
(sl2 × sl2 )-submodule V ∼ = V2,i ⊂ Vω2 ,λ and V is not Poisson by Proposition 3.15(a).
(b) If g = sln × g2 , n 4 and V = Vω2 ,λ (respectively Vωn−2 ,λ ), choose g = (sl2 × sl2 ) × g2
generated by the first and third nodes of the Dynkin diagram An−1 (respectively the last and
third to last nodes) and g2 . Let vω2 be a highest weight vector in Vω2 . Note that Fα2 (vω2 ) ∈
Vω2 generates a four-dimensional simple sl2 × sl2 -module V1,1 . Hence, we find a sl2 × sl2 × g2
submodule V ∼ = V1,1,λ ⊂ Vω2 ,λ and V is not Poisson by Proposition 3.16. Similarly we obtain
that Vωn−2 ,λ is not Poisson.
(c) If g = so(2n) ⊕ g2 , and V = Vω1 ,λ , consider the Levi subalgebra of so(2n), isomorphic to
sl4 , generated by the (n − 2)nd, (n − 1)st and the nth nodes of the Dynkin diagram Dn . It can
be easily observed that if v ∈ Vω1 (ω1 ) is a highest weight vector, then v = Fαn−3 ◦ · · · ◦ Fα1 (v)
generates a simple sl4 -module Vω2 . We obtain that V is not Poisson by applying the argument in
case (b).
(d) If g = so(8) ⊕ g2 and V = Vωi ,λ , i = 3, 4, or g = so(10) ⊕ g2 and V = Vωi ,λ , i = 4, 5, we
argue analogous to case (c).
(e) If g = E6 ⊕ g2 and V = Vω1 ,λ consider the Levi subalgebra sl4 ⊂ E6 generated by the
second, third and fourth nodes of the Dynkin diagram E6 (in the notation of Bourbaki [3]). If
v ∈ Vω1 (ω1 ) is a highest weight vector, then v = Fα3 ◦Fα1 (v) generates a simple sl4 -module Vω2 .
We obtain that V is not Poisson by applying the argument in case (b).
The proof of Theorem 3.14 is now complete. 2
4. Quantum symmetric algebras
4.1. The quantum group Uq (g) and its modules
We start with the definition of the quantized enveloping algebra associated with a complex
reductive Lie algebra g (our standard reference here will be [4]). Let h ⊂ g be a Cartan subalge-
bra, P (g) the weight lattice, as introduced above, and let A = (aij ) be the Cartan matrix for g.
Additionally, let (·,·) be the standard non-degenerate symmetric bilinear form on h.
The quantized enveloping algebra U is a C(q)-algebra generated by the elements Ei and Fi
for i ∈ [1, r], and Kλ for λ ∈ P (g), subject to the following relations: Kλ Kμ = Kλ+μ , K0 = 1
for λ, μ ∈ P ; Kλ Ei = q (αi ,λ) Ei Kλ , Kλ Fi = q −(αi ,λ) Fi Kλ for i ∈ [1, r] and λ ∈ P ;
Kαi − K−αi
Ei , Fj − Fj Ei = δij (4.1)
q di − q −di
(αi ,αi )
for i, j ∈ [1, r], where di = 2 ; and the quantum Serre relations
1−aij
(1−aij −p) (p)
(−1)p Ei Ej Ei = 0,
p=0
1−aij
(1−aij −p) (p)
(−1)p Fi Fj Fi =0 (4.2)
p=0
(p)
for i = j , where the notation Xi stands for the divided power
(p) Xp q kdi − q −kdi

Xi = , (k)i = . (4.3)
(1)i · · · (p)i q di − q −di
The algebra U is a q-deformation of the universal enveloping algebra of the reductive Lie
algebra g, so it is commonly denoted by U = Uq (g). It has a natural structure of a bialgebra with
the co-multiplication Δ : U → U ⊗ U and the co-unit homomorphism ε : U → Q(q) given by
Δ(Ei ) = Ei ⊗ 1 + Kαi ⊗ Ei , Δ(Fi ) = Fi ⊗ K−αi + 1 ⊗ Fi ,

Δ(Kλ ) = Kλ ⊗ Kλ , (4.4)
ε(Ei ) = ε(Fi ) = 0, ε(Kλ ) = 1. (4.5)
In fact, U is a Hopf algebra with the antipode anti-homomorphism S : U → U given by
S(Ei ) = −K−αi Ei , S(Fi ) = −Fi Kαi , S(Kλ ) = K−λ . (4.6)
Let U − (respectively U 0 ; U + ) be the Q(q)-subalgebra of U generated by F1 , . . . , Fr (respec-

tively by Kλ (λ ∈ P ); by E1 , . . . , Er ). It is well known that U = U − · U 0 · U + (more precisely,
the multiplication map induces an isomorphism U − ⊗ U 0 ⊗ U + → U ).
We will consider the full sub-category Of of the category Uq (g)-Mod. The objects of Of are
finite-dimensional Uq (g)-modules V q having a weight decomposition

Vq = V q (μ),
μ∈P
where each Kλ acts on each weight space V q (μ) by the multiplication with q (λ|μ) (see e.g.,
q
[4, I.6.12]). The category Of is semisimple and the irreducible objects Vλ are generated by
highest weight spaces Vλ (λ) = C(q) · vλ , where λ is a dominant weight, i.e., λ belongs to P + =
q
{λ ∈ P : (λ|αi ) 0 ∀i ∈ [1, r]}, the monoid of dominant weights.

By definition, the universal R-matrix R ∈ Uq (g) ⊗ Uq (g)R can be decomposed as
R = R0 R1 = R1 R0 (4.7)
where R0 is “the diagonal part” of R, and R1 is unipotent, i.e., R1 is a formal power series
R1 = 1 ⊗ 1 + (q − 1)x1 + (q − 1)2 x2 + · · · , (4.8)
where all xk ∈ U − +
k ⊗C[q,q −1 ] U k , where U
− (respectively U + ) is the integral form of U + , i.e.,
U − is a C[q, q −1 ]-subalgebra of Uq (g) generated by all Fi (respectively by all Ei ) and U − k

(respectively U +
k ) is the kth graded component under the grading deg(F i ) = 1 (respectively
deg(Ei ) = 1).
By definition, for any U q , V q in Of and any highest weights elements uλ ∈ U q (λ), vμ ∈
V (μ) we have R0 (uλ ⊗ vμ ) = q (λ|μ) uλ ⊗ vμ .
q
Let R op be the opposite element of R, i.e., R op = τ (R), where τ : Uq (g) ⊗ Uq (g) → Uq (g) ⊗
op op
Uq (g) is the permutation of factors. Clearly, R op = R0 R1 = R1 R0 .
Following [8, Section 3], define D ∈ Uq (g) ⊗ Uq (g) by

op op
D := R0 R1 R1 = R1 R1 R0 . (4.9)
op
Clearly, D is well defined because R1 R1 is also unipotent as well as its square root. By
definition, D 2 = R op R, D op R = RD.
Furthermore, define

:= RD −1 = D op −1 R = R1 R op R1 −1 .
R (4.10)
1
It is easy to see that
R −1 .
op = R (4.11)
is a coboundary Hopf algebra.

According to [8, Proposition 3.3], the pair (Uq (g), R)
The braiding in the category Of is defined by RU q ,V q : U q ⊗ V q → V q ⊗ U q , where
RU q ,V q (u ⊗ v) = τ R(u ⊗ v)
for any u ∈ U q , v ∈ V q , where τ : U q ⊗ V q → V q ⊗ U q is the ordinary permutation of factors.

Denote by C ∈ Z(U q (g)) the quantum Casimir element which acts on any irreducible Uq (g)-
q
module Vλ in Of by the scalar multiple q (λ|λ+2ρ) , where 2ρ is the sum of positive roots.
The following fact is well known.
Lemma 4.1. One has R2 = Δ(C −1 )◦(C ⊗C). In particular, for each λ, μ, ν ∈ P+ the restriction
ν of the tensor product V q ⊗ V q is scalar multiplication
of R2 to the νth isotypic component Iλ,μ λ μ
by q (λ|λ)+(μ|μ)−(ν|ν)+(2ρ|λ+μ−ν) .
We now define the diagonalizable C(q)-linear map DU q ,V q : U q ⊗ V q → U q ⊗ V q by

DU q ,V q (u ⊗ v) = D(u ⊗ v) for any objects U q and V q of Of . It is easy to see that the op-
q q q q ν in V q ⊗ V q by
erator DV q ,Vμq : Vλ ⊗ Vμ → Vλ ⊗ Vμ acts on the νth isotypic component Iλ,μ λ μ
λ
1
the scalar multiplication with q 2 ((λ|λ)+(μ|μ)−(ν|ν))+(ρ|λ+μ−ν) .
For any U q and V q in Of define the normalized braiding σU q ,V q by
⊗ v).
σU q ,V q (u ⊗ v) = τ R(u (4.12)
Therefore, we have by (4.10):
σU q ,V q = DV−1q ,U q RU q ,V q = RU q ,V q DU−1q ,V q . (4.13)
We will sometimes write σU q ,V q in a more explicit way:

σU q ,V q = R−1 q
V ,U q R −1
q
U ,V q R U q ,V q = RU q ,V q R−1 −1
U q ,V q RV q ,U q . (4.14)
The following fact is an obvious corollary of (4.11).
Lemma 4.2. σV q ,U q ◦ σU q ,V q = idU q ⊗V q for any U q , V q in Of . That is, σ is a symmetric

commutativity constraint.
We also have the following coboundary relation (even though we will not use it).
Lemma 4.3. (See [8, Section 3].) Let Aq , B q , C q be objects of Of . Then, the following diagram
commutes:
σ12,3
Aq ⊗ B q ⊗ C q C q ⊗ Aq ⊗ B q
σ1,23 σ23 (4.15)
σ12
B q ⊗ C q ⊗ Aq C q ⊗ B q ⊗ Aq
where we abbreviated

σ12,3 := σAq ⊗B q ,C q : Aq ⊗ B q ⊗ C q → C q ⊗ Aq ⊗ B q ,

σ1,23 := σAq ,B q ⊗C q : Aq ⊗ B q ⊗ C q → B q ⊗ C q ⊗ Aq .
Remark 4.4. If one replaces the braiding R of Of by its inverse R−1 , the symmetric commuta-
tivity constraint σ will not change.
4.2. Braided symmetric and exterior powers
In this section we will use the notation and conventions of Section 4.1.
For any morphism f : V q ⊗ V q → V q ⊗ V q in Of and n > 1 we denote by f i,i+1 , i =
1, 2, . . . , n − 1, the morphism V q,⊗n → V q,⊗n which acts as f on the ith and the (i + 1)st
factors. Note that σVi,i+1
q ,V q is always an involution on V
q,⊗n .
Definition 4.5. For an object V q in Of and n 0 define the braided symmetric power Sσn V q ⊂
V q,⊗n and the braided exterior power Λnσ V q ⊂ V q,⊗n by

Sσn V q = (Ker σi,i+1 − id) = (Im σi,i+1 + id),
1in−1 1in−1

Λnσ V q = (Ker σi,i+1 + id) = (Im σi,i+1 − id),
1in−1 1in−1
where we abbreviate σi,i+1 = σVi,i+1

q ,V q .
Remark 4.6. Clearly, −R is also a braiding on Of and −σ is the corresponding normalized

braiding. Therefore, Λnσ V q = S−σ
n V q and S n V q = Λn V q . That is, informally speaking, the
σ −σ
symmetric and exterior powers are mutually “interchangeable.”
Remark 4.7. Another way to introduce the symmetric and exterior squares involves the well-
known fact that the braiding RV q ,V q is a semisimple operator V q ⊗ V q → V q ⊗ V q , and all the
eigenvalues of RV q ,V q are of the form ±q r , where r ∈ Z. Then positive eigenvectors of RV q ,V q
span Sσ2 V q and negative eigenvectors of RV q ,V q span Λ2σ V q .
Clearly, Sσ0 V q = C(q), Sσ1 V q = V q , Λ0σ V q = C(q), Λ1σ V q = V q , and

Sσ2 V q = v ∈ V q ⊗ V q σV q ,V q (v) = v ,

Λ2σ V q = v ∈ V q ⊗ V q σV q ,V q (v) = −v .
Proposition 4.8. For each n 0 the association V q → Sσn V q is a functor from Of to Of and the
association V q → Λnσ V q is a functor from Of to Of . In particular, an embedding U q → V q in
the category Of induces injective morphisms
Sσn U q → Sσn V q , Λnσ U q → Λnσ V q .
Definition 4.9. For any V q ∈ Ob(O) define the braided symmetric algebra Sσ (V q ) and the
braided exterior algebra Λσ (V q ) by

Sσ V q = T V q / Λ2σ V q , Λσ V q = T V q / Sσ2 V q , (4.16)
where T (V q ) is the tensor algebra of V q and I stands for the two-sided ideal in T (V v) gener-
ated by a subset I ⊂ T (V q ).
Note that the algebras Sσ (V q ) and Λσ (V q ) carry a natural Z0 -grading:

q q
Sσ V q = Sσ V n , Λσ V q = Λσ V n ,
n0 n0
since the respective ideals in T (V q ) are homogeneous.

Denote by Ogr,f the sub-category of Uq (g)-Mod whose objects are Z0 -graded:
q
Vq = Vn ,
n∈Z0
q
where each Vn is an object of Of ; and morphisms are those homomorphisms of Uq (g)-modules
which preserve the Z0 -grading.
Clearly, Ogr,f is a tensor category under the natural extension of the tensor structure of Of .
Therefore, we can speak of algebras and co-algebras in Ogr,f .
By the very definition, Sσ (V q ) and Λσ (V q ) are algebras in Ogr,f .
Proposition 4.10. The assignments V q → Sσ (V q ) and V q → Λσ (V q ) define functors from Of

to the category of algebras in Ogr,f .
We conclude the section with two important features of braided symmetric exterior powers
and algebras.
Proposition 4.11. (See [2, Proposition 2.11 and Eq. (2.3)].) Let V q be an object of Of and V ∗
its dual in Of . We have the following Uq (g)-module isomorphisms:
n q,∗ ∗ n q,∗ ∗
Sσ V ∼
= Sσ V q n , Λσ V ∼
= Λσ V q n . (4.17)
q
Proposition 4.12. (See [2, Proposition 2.13].) For any V q in Of each embedding Vλ → V q
q
defines embeddings Vnλ → Sσn V q for all n 2. In particular, the algebra Sσ (V q ) is infinite-
dimensional.
4.3. The classical limit of braided algebras
In this section we will discuss the specialization of the braided symmetric and exterior alge-
bras at q = 1, the classical limit. All of the results in this section are either well known or proved
in [2]. For a more detailed discussion of the classical limit we refer the reader to [2, Section 3.2].
We will first introduce the notion of an almost equivalence of categories:
Definition 4.13. We say that a functor F : C → D is an almost equivalence of C and D if:
(a) for any objects c, c of C an isomorphism F (c) ∼= F (c ) in D implies that c ∼

= c in C;
∼
(b) for any object d in D there exists an object c in C such that F (c) = d in D.
Denote by Of the full (tensor) sub-category category of U (g)-Mod, whose objects V are

finite-dimensional U (g)-modules having a weight decomposition V = μ∈P V (μ). The fol-

lowing fact will be the first result of this section.
Proposition 4.14. (See [2, Corollary 3.22].) The categories Of and Of are almost equivalent.
Under this almost equivalence a simple Uq (g)-module Vλ is mapped to the simple U (g)-module
V λ.

n
n
Let V ∼
= i=1 Vλi ∈ Of . We call V ∼
= i=1 V λi ∈ Of the classical limit of V under the
above almost equivalence.
Proof. First, we have to introduce the notion of (k, A)-algebras and investigate their properties.
Let k be a field and A be a local subring of k. Denote by m the only maximal ideal in A and by
k̃ the residue field of A, i.e., k̃ := A/m.
We say that an A-submodule L of a k-vector space V is an A-lattice of V if L is a free A-
module and k ⊗A L = V , i.e., L spans V as a k-vector space. Note that for any k-vector space
V and any k-linear basis B of V the A-span L = A · B is an A-lattice in V . Conversely, if L is
an A-lattice in V , then any A-linear basis B of L is also a k-linear basis of V .
Denote by (k, A)-Mod the category whose objects are pairs V = (V , L) of a k-vector space V
and an A-lattice L ⊂ V of V ; an arrow (V , L) → (V , L ) is any k-linear map f : V → V such
that f (L) ⊂ L .
Clearly, (k, A)-Mod is an Abelian category. Moreover, (k, A)-Mod is A-linear because each
Hom(U, V) in (k, A)-Mod is an A-module.
It can be easily verified that (k, A)-Mod is a symmetric tensor category [2, Lemma 3.14]. We
have the following fact.
Lemma 4.15. (See [2, Lemma 3.12].) The forgetful functor (k, A)-Mod → k-Mod given by
(V , L) → V is an almost equivalence of symmetric tensor categories.
Define a functor F : (k, A)-Mod → k̃-Mod by
F (V , L) = L/mL
for any object (V , L) of (k, A)-Mod and for any morphism f : (V , L) → (V , L ) we define
F (f ) : L/mL → L /mL to be a natural k̃-linear map.
Lemma 4.16. (See [2, Lemma 3.14].) F : (k, A)-Mod → k̃-Mod is a tensor functor and almost
equivalence.
Let U be a k-Hopf algebra and let UA be a Hopf A-subalgebra of U . This means that
Δ(UA ) ⊂ UA ⊗A UA (where UA ⊗A UA is naturally an A-sub-algebra of U ⊗k U ), ε(UA ) ⊂ A,
and S(UA ) ⊂ UA . We will refer to the above pair U = (U, UA ) as to (k, A)-Hopf algebra (please
note that UA is not necessarily a free A-module, that is, U is not necessarily a (k, A)-module).
Given (k, A)-Hopf algebra U = (U, UA ), we say that an object V = (V , L) of (k, A)-Mod is
a U -module if V is a U -module and L is an UA -module.
Denote by U -Mod the category which objects are U -modules and arrows are those morphisms
of (k, A)-modules which commute with the U -action.
Clearly, for (k, A)-Hopf algebra U = (U, UA ) the category U -Mod is a tensor (but not neces-
sarily symmetric) category.
For each (k, A)-Hopf algebra U = (U, UA ) we define U := UA /mUA . Clearly, U is a Hopf
algebra over k̃ = A/m.
Lemma 4.17. (See [2, Lemma 3.15].) In the notation of Lemma 4.16, for any (k, A)-Hopf algebra
U the functor F naturally extends to a tensor functor
U-Mod → U-Mod. (4.18)
Now let k = C(q) and A be the ring of all those rational functions in q which are defined at
q = 1. Clearly, A is a local PID with maximal ideal m = (q − 1)A (and, moreover, each ideal in
A is of the form mn = (q − 1)n A). Therefore, k̃ := A/m = C.
Recall from Section 4.1 the definition of the quantized universal enveloping algebra Uq (g).
λ −1
Denote hλ = Kq−1 and let UA (g) be the A-algebra generated by all hλ , λ ∈ P and all Ei , Fi .
Denote by Uq (g) the pair (Uq (g), UA (g)).
Lemma 4.18.
(a) The pair Uq (g) = (Uq (g), UA (g)) is a (k, A)-Hopf algebra [2, Lemma 3.16].
(b) Uq (g) = U (g) [2, Lemma 3.17].
Let Vλ ∈ Ob(Of ) be an irreducible Uq (g)-module with highest weight λ ∈ P + and let vλ ∈ Vλ

be a highest weight vector. Define Lvλ = UA (g) · vλ .
The following fact is obvious and, apparently, well known.
Lemma 4.19. (See [2, Lemma 3.19].)
(a) Each object (Vλ , Lvλ ) is irreducible in Of (Uq (g)); and each irreducible object of Of (Uq (g))
is isomorphic to one of (Vλ , Lvλ ).
(b) The category Of (Uq (g)) is semisimple.
(c) The forgetful functor (V , L) → V is an almost equivalence of tensor categories
Of (Uq (g)) → Of .
We also have the following fact.
Lemma 4.20. (See [2, Lemma 3.21].)
(a) The restriction of the functor Uq (g)-Mod → U (g)-Mod defined by (4.18) to the sub-category
Of (Uq (g)) is a tensor functor

Of Uq (g) → Of . (4.19)
(b) The functor (4.19) is an almost equivalence of categories.
Combining Lemmas 4.19 and 4.20 we obtain Proposition 4.14. 2
The following result relates the classical limit of braided symmetric algebras and Poisson
algebras.
Theorem 4.21. (See [2, Theorem 2.29].) Let V be an object of Of and let a V in Of be the
classical limit of V . Then:
(a) The classical limit Sσ (V ) of the braided symmetric algebra Sσ (V ) is a quotient of the sym-
metric algebra S(V ). In particular, dimC(q) Sσ (V )n = dimC (Sσ (V )) dimC S(V )n .
(b) Moreover, Sσ (V ) admits a Poisson structure defined by {u, v} = r − (u ∧ v), where r − is an
anti-symmetrized r-matrix.
4.4. Flat modules over reductive Lie algebras
In [2] we introduce the notion of flatness of a Uq (g)-module. In this section we will recall
the definition and basic properties of flat simple modules and then proceed to classify all flat
modules over Uq (g), where g is any semisimple Lie algebra.
We view Sσ (V q ) and Λσ (V q ) as deformations of the quadratic algebras S(V ) and Λ(V )
respectively, where V denotes the classical limit of V q . In [2] we show that

dim V q + n − 1
dim Sσn V q = dim Sσ V q n
n
for all n.
Therefore, it is natural to make the following definition.
Definition 4.22. A finite-dimensional Uq (g)-module is flat, if and only if

dim V q + n − 1
dim Sσn V q =
n
for all n 0; i.e., the braided symmetric power Sσn V q is isomorphic (as a vector space) to the
ordinary symmetric power S n V q .
The following theorem is our main result.
Main Theorem 4.23. Let g be a semisimple Lie algebra and Uq (g) its quantized enveloping
algebra. A simple Uq (g)-module V q is flat if and only if its classical limit V is Poisson as a
U (g)-module.
Proof. The “only if” assertion follows immediately from the following result.
Proposition 4.24. (See [2, Theorem 2.29].) If V q is an object of Of and V q is flat, then S(V ) is
Poisson.
Proof. Theorem 4.21 asserts that the classical limit of Sσ (V q ) is a Poisson algebra. If V q is flat
then (S(V ), r − ) is a Poisson algebra and V is Poisson by Proposition 3.6. 2
It therefore, remains to show that if a simple g-module V is Poisson, then V q is a flat Uq (g)-
module. Following the strategy of Section 3.2 we will first consider the case when g is a simple
Lie algebra and V is a simple g-module.
Employing a well-known result by Drinfeld [9, Theorem 1] about quadratic algebras it is

q
shown in [2, Proposition 2.33] that V q is flat if and only if dim(Sσ3 V q ) = dim(V3 )+2 .
In [2] we introduced a lower bound for the dimension of the symmetric and exterior cube,
and correspondingly we construct a minimal submodule contained in the symmetric and exterior
cubes. We will use the notation
Vμ (4.20)
for the space of highest weight vectors of weight μ in a finite-dimensional g-module V . For
dominant λ, μ, ν ∈ P + (g) denote

ν
cλ,μ = dim(Vλ ⊗ Vμ )ν = dimC Homg (Vν , Vλ ⊗ Vμ ) ;
ν is the tensor product multiplicity. And for any λ, μ ∈ P + (g) denote c+ = dim(S 2 V )μ
i.e., cλ,μ λ;μ λ
− + − μ
and cλ;μ = dim(Λ2 Vλ )μ , so that cλ;μ + cλ;μ = cλ,λ . Ultimately, define:

μ + − μ
dλ := cλ;ν − cλ;ν cν,λ .
ν∈P+
Since dequantization is an almost equivalence of the tensor categories Of and Of (g)

q q q + −
(Lemma 4.19(c)) the multiplicity of Vν in Sσ2 Vλ (respectively Λ2σ Vλ ) is cλ,ν (respectively cλ,ν ).
3,σ q q
3 ) the multiplicity of V in S 3 V (respectively of V in
Denote by cλ,μ (respectively cλ,μ μ σ λ μ
for all μ ∈ P + (g). Moreover, one has
μ3,σ
S 3 Vλ ). In [2, Proposition 2.35] we prove that dλ cλ,μ
for all μ ∈ P + and Vλ is flat. We prove in
3,σ μ 3,σ q
cλ,μ cλ,μ
3 Hence, if dλ = cλ,μ
3 , then cλ,μ= cλ,μ
3
μ
Section 6.2 that if a simple g-module Vλ is Poisson, then dλ = cλ,μ3 (Theorem 6.20 and Proposi-
q
tion 6.21) and hence Vλ is flat.
Now consider the case when g is semisimple and V a simple g-module. Theorem 3.14 asserts
that if supp(g) is not simple and V is Poisson, then supp(g) ∼ = sln × slm for some m, n 1
and V isomorphic to the natural module Vω1 ,ω1 . We show in [2, Proposition 2.38] that the natural
Uq (slm × sln )-module is flat, its braided symmetric algebra isomorphic to the algebra of quantum
m × n-matrices. Theorem 4.23 is proved. 2
Remark 4.25. A straightforward argument shows that if g is a reductive Lie algebra, then a
Uq (g )-module V q is flat if V q |Uq (g ) is a flat Uq (g )-module, where g ⊂ g is the maximal
semisimple subalgebra of g.
Most of the braided symmetric algebras obtained by the construction of Theorem 5.6 are well-
known examples of quantized coordinate rings. Our theory presents a unifying construction of
these important examples. We have the following list.
Corollary 4.26. Let g be a semisimple complex Lie algebra and V q a flat simple Uq (g)-module.
Then, the braided symmetric algebra Sσ (V q ) is isomorphic, as a Uq (g)-module algebra to one
of the following quantized coordinate rings:
q
• If g = slm × sln and V q = Vω1 , then Sσ (V q ) is isomorphic to the algebra of quantum m × n-
matrices.
q
• If g = sln and V q = V2ω1 , then Sσ (V q ) is isomorphic to the algebra of quantum symmetric
matrices introduced by Nouri in [20, Theorem 4.3 and Proposition 4.4] and by [16].
q
• If g = sln and V q = Vω2 , then Sσ (V q ) is isomorphic to the algebra of quantum antisymmet-
ric matrices introduced by Strickland in [24, Section 1].
q
• If g = so(n) and V q = Vω1 , then Sσ (V q ) is isomorphic to odd- and even-dimensional quan-
tum Euclidean space introduced by Reshetikhin et al. in [21] (see also [19]).
q
• If g = sp(2n) and V q = Vω1 , then Sσ (V q ) is isomorphic to quantum symplectic space intro-
duced by Reshetikhin et al. in [21].
• If g = so(10) and V = Vω5 , the spin representation or g = E6 and V = Vω1 , then we obtain
quantum algebras, which apparently have not been studied previously.
Proof. In order to prove the corollary one has to compare the Poisson structures defined on the
classical limits of the braided symmetric algebras with the Poisson structures obtained from the
classical limits of the quantized coordinate rings under consideration. This is done by Goodearl
q
and Yakimov in [11, Chapter 5] for all cases but the case g = sp(2n) and V q = Vω1 . In that case
one can easily compute that the r-matrix bracket and the bracket on quantum symplectic space
are as follows:
n
Let g = sp2n (C) and V = Vω1 ∼ = C2n = i=1 C · ei ⊕ C · e−i with the action Ek (ei ) =
δi ek + δi,−k e− for k, = ±1, . . . , ±n, k = . The bracket is given by

n
{ei , ej } = sign |j | − |i| · ei ej − 2δi,−j · sign(i) · e−k ek
k=|i|
for i, j = ±1, . . . , ±n. The corollary is proved. 2
Problem 4.27. Describe the braided symmetric algebras of the spin representation of so(10) and
the first fundamental module Vω1 of E6 by generators and relations as iterated twisted polynomial
algebras following [6].
5. Deformations of symmetric algebras of Poisson modules
In this section we will explicitly construct the braided symmetric algebras of flat modules, em-
ploying the relationship between geometrically decomposable modules and Abelian nil-radicals.
5.1. Quantum radicals as symmetric algebras
Our notation will be very closely related to the notation used by Jantzen in [14], where Chap-
ter 8 contains proofs to a number of known facts we shall employ. Let Uq (g) be the quantized
enveloping algebra corresponding to a Lie algebra g as introduced in Section 4.1. Denote, as
above, by W the Weyl group of g generated by the simple reflections si for i ∈ [1, r]. Corre-
sponding to each i ∈ [1, r] there exist maps Ti : Uq (g) → Uq (g) defined on the generators of
Uq (g) in the following way:
−aij
−aij −k
Ti (Ei ) = −Fi Kαi , Ti (Fi ) = −Kα−1
i
Ei , Ti (Ej ) = (−1)k−aij qi−k Ei Ej Eik ,
k=0
−aij
a −k
Ti (Fj ) = (−1)k−aij qik Fik Fj Fi ij , Ti (Kλ ) = Kσi (λ) . (5.1)
k=0
For every element w ∈ W with reduced expression w = si1 . . . sik we define Tw as Tw =

Tsi1 · · · Tsik . We will need the following well-known fact.
Lemma 5.1. (See e.g. [14, 8.18].) If w ∈ W , then Tw is independent of the choice of reduced
expression; i.e. if w = si1 . . . sik and w = sj1 . . . sjk are reduced expressions of w ∈ W , then
Tsi1 · · · Tsik = Tsj1 · · · Tsjk .
Recall from Section 4 that U + denotes the subalgebra of Uq (g) generated by the Ei for i ∈
[1, r], U − the subalgebra of Uq (g) generated by the Fi for i ∈ [1, r] and Uq (b− )—the subalgebra
of Uq (g) generated by all Kλ and all Fi .
Recall (see e.g. [14, Chapter 8]) that we can associate to each reduced expression of the
longest element w0 ∈ W a PBW-basis of Uq (g) in the following way. Let w0 = σi1 . . . σik be a
reduced expression of the longest element in W . It is well known that the set of positive roots
R + of the Lie algebra g can be ordered in the following way:
α(1) = αi1 < α(2) = si1 ◦ α2 < · · · < α(k) = si1 . . . sik−1 αik ,
where αj denotes the j th simple root.

We define for each presentation of w0 a set of positive roots spanning U + following
[14, Chapter 8]:
Eα(1) = Ei1 , Eα(2) = Ti1 (Ei2 ), ..., Eα(k) = Ti1 · · · Tik−1 (Eik ).
Similarly, we define a set of negative roots spanning U − by
Fα(1) = Fi1 , Fα(2) Ti1 (Fi2 ), ..., Fα(k) = Ti1 · · · Tik−1 (Fik ).
The following is the key definition for this section.
Definition 5.2. For every element w ∈ W in the Weyl group we define the quantum Schubert cell
U (w) ⊂ Uq (g) as

U (w) = Tw−1 Uq b− ∩ U + .
Note that this definition differs from other definitions of quantum Schubert cell such as in the
paper [5] by Concini and Procesi. We also have the following alternative description of quantum
Schubert cells.
Lemma 5.3. Let w ∈ W and w0 be the longest element in W . Denote w = w0 w. We have

U (w) = T(w )−1 Uq b+ ∩ U + .
Proof. We have
T(w )−1 = Tw−1 w0 = Tw−1 Tw0 ,
since w0−1 = w0 . Note that Tw0 (Uq (b+ )) = Uq (b− ). Therefore,

U (w) = Tw−1 Uq b− ∩ U + = Tw−1 Tw0 Uq b+ ∩ U + = T(w )−1 Uq b+ ∩ U + .
Definition 5.4. Let w, w ∈ W such that w0 = w w and (w) + (w ) = (w0 ), where (v)
denotes the length of an element v ∈ W . Then we denote by U (w) the twisted quantum Schubert
cell U (w) = Tw U (w).
Note the following fact which follows from the basic properties of the action of the Ti (see
e.g. [14, Chapter 8]).
Lemma 5.5.
(a) The twisted quantum Schubert cell U (w) is independent of the choice of reduced expression
for w .
(b) One has U (w) ⊂ U + for all w ∈ W .
Proof. Part (a) follows from the basic properties of the action of the Ti (see e.g. [14, Chapter 8]).
Part (b) follows from Proposition 5.23(b). 2
We will now consider quantum Schubert cells U (wΔ ) where wΔ ∈ W corresponds to subsets
Δ ⊂ [1, r] in the following way. Let WΔ be the subgroup of Delta generated by the simple re-
flections si for i ∈ Δ and denote by w0,Δ the longest element of WΔ . The element wΔ = w0,Δ w0
is commonly referred to as a parabolic element of W . Denote by pΔ the standard parabolic sub-
algebra of g associated with Δ. Recall that pΔ splits as a semi-direct product pΔ ∼ = lΔ radΔ
(see e.g. Humphreys [13]), where lΔ is a reductive Lie algebra, the Levi subalgebra, and radΔ
is nilpotent, the nil-radical. Note that WΔ ⊂ W can be identified with the Weyl group of lΔ .
Additionally recall that any Hopf algebra H algebra acts on itself via the adjoint action:
ad(a).b = a(1) bS(a(2) ). (5.2)
The following theorem is the first main result of this section.
Main Theorem 5.6.
(a) Let g be a reductive Lie algebra, pΔ a parabolic subalgebra with Levi lΔ and radical radΔ .
If radΔ is an Abelian Lie algebra, then U (wΔ ) is a flat quadratic q-deformation of the
symmetric algebra S(radΔ ).
(b) The twisted quantum Schubert cell U (wΔ ) is a Z0 -graded Uq (lΔ ) module algebra and
U (wΔ ) is the braided symmetric algebra of the Uq (lΔ )-module U (wΔ )1 .
(c) Moreover, let gΔ be the maximal semisimple Lie subalgebra of lΔ . Then, U (wΔ ) is a
Z0 -graded Uq (gΔ ) module algebra and U (wΔ ) is the braided symmetric algebra of the
Uq (gΔ )-module U (wΔ )1 .
Proof. In order to prove Theorem 5.6(a) we have to show that the classical limit q → 1 of
U (wΔ ) is S(radΔ ). We call a root α ∈ R(g) radical, if α ∈
/ R(g ) ∩ spanZ (Δ); i.e., Eα lies in the
radical (see discussion below). Recall that radΔ is spanned by Eα where α is radical. We obtain
the following well-known characterization of Abelian radicals.
Lemma 5.7. Let g be a reductive Lie algebra, pΔ a parabolic subalgebra with Levi lΔ and
radical radΔ . The radical radΔ is Abelian,if and only if Δ = [1, r] − {i} for some i ∈ [1, r] and
all radical roots are of the form α = αi + j =i cj αj .
Theorem 5.21 yields that U (wΔ ) is generated as an algebra by the Eα for which α ∈ R + (g)
is a radical root. Moreover we have the following fact.
Lemma 5.8. Let si1 . . . sik be a reduced expression of w0 such that w0,Δ = si1 . . . sid and wΔ =
sid+1 . . . sik for some 1 < d < k. The radical roots are the roots αj = si1 . . . sij , where d + 1
j k. The twisted quantum Schubert cell U (wΔ ) is generated as an algebra by the Eα for
α ∈ R + (g) radical.
Proof. Note that w0,Δ permutes the radical roots, since w0,Δ leaves all simple roots αi with
i ∈ Δ invariant. This proves the first assertion. The second assertion now follows from Theo-
rem 5.21(a) and the definition of twisted quantum Schubert cells. The lemma is proved. 2
We need the following well-known fact.
+ = w(R − ) ∩ R + . Then,
Lemma 5.9. (See e.g. [6, Lemma 2.2].) Let Rw
ad(Eα )(Eβ ) = [Eα , Eβ ]q = Eα Eβ − q (α,β) Eβ Eα ∈ span(Eγ1 . . . Eγk ), (5.3)
where α < γ1 · · · γk < β and γ1 + · · · + γk = α + β.
If radΔ is Abelian, then U (wΔ ) is a quadratic algebra, because of the following argument:
α < γ < β implies that γ is radical by Lemma 5.8; hence, α + β = γ1 + · · · + γk and α < γ1
· · · γk < β imply, by Lemma 5.7, that k = 2, so U (wΔ ) is indeed quadratic.
Finally, note that the classical limit of U (wΔ ) is S(radΔ ), as desired. Theorem 5.6(a) is
proved.
Let us now prove part (b). Note first that Uq (lΔ ) acts adjointly on U (wΔ ) by Theorem 5.21.
Recall that U (wΔ ) is naturally Z0 -graded as a quadratic algebra. If radΔ is Abelian, then
Lemmas 5.7 and 5.9 imply that

ad Uq (lΔ ) U (wΔ ) i ⊂ U (wΔ ) i ,
and that, therefore, U (wΔ ) is a graded Uq (lΔ )-module algebra. Denote by πΔ the canonical
Uq (lΔ )-module homomorphism πΔ : T ((U (wΔ ))1 ) → U (wΔ ). We obtain from Theorem 5.6
that the classical limit of the kernel ker(π) is equal to the U (lΔ )-ideal generated by Λ2 radΔ .
Howe proves in [12, Chapter 4.6] that if radΔ is Abelian then it is weight-multiplicity-free and
simple as a lΔ -module and gΔ -module, that means all weight-spaces are one-dimensional. Recall
the following well-known fact.
Lemma 5.10. Let g be a reductive Lie algebra. Then,

dim Homg (Vλ ⊗ Vλ , Vμ ) dim Vλ (μ − λ).
The lemma implies that radΔ ⊗ radΔ is multiplicity-free as a lΔ -module. Employing

Lemma 4.20, we obtain that (U (wΔ ))1 ⊗ (U (wΔ ))1 contains a unique Uq (lΔ )-submodule
Ext2q (radΔ ) such that its classical limit is isomorphic to Λ2 radΔ . This implies that
2
Λσ U (wΔ ) 1 = Ext2q (radΔ ) = ker(π) ⊂ T U (wΔ )1 .
Therefore, U (wΔ ) = Sσ (U (wΔ )1 ). Theorem 5.6(b) is proved.

Part (c) can be proved analogously to part (b). Theorem 5.6 is proved. 2
Call a Uq (g)-module geometrically decomposable if its classical limit is geometrically de-

composable as a U (g)-module. Theorem 5.6(c) has the following consequence.
Corollary 5.11. Let g be a semisimple Lie algebra and let V q be a simple geometrically decom-
posable Uq (g)-module. There exists a simple Lie algebra g and a parabolic element wΔ ∈ W (g )
such that Uq (g) ∼
= Uq (gΔ ) and the braided symmetric algebra Sσ (V q ) ∼ = U (wΔ ) as graded
Uq (gΔ )-module algebras.
Proof. The corollary follows immediately from Theorem 5.6 and the description of geometri-
cally decomposable modules as Abelian radicals in Section 3.2. 2
We will now extend the result of Theorem 5.6 to some subalgebras, when radΔ is of Heisen-
[radΔ , radΔ ] ⊂ radΔ is one-dimensional. Recall that an
berg type; i.e., the derived subalgebra
algebra U is called filtered, if U = ∞i=0 Ui with U
i ⊂ Ui+1 and Ui · Uj ⊂ Ui+j . The associated
graded algebra gr(U ) of U is defined as gr(U ) = ∞ i=0 Ui /Ui−1 , where we set U−1 = {0}. The
following result is the second main result of this section.
Theorem 5.12. Let Δ ⊂ [1, r] and let pΔ be the corresponding parabolic subalgebra, and radΔ
its nil-radical. If U (wΔ ) is a filtered Uq (lΔ )-module algebra, then:
(a) gr(U (wΔ )) is a Uq (lΔ )-module algebra and a flat q-deformation of S(radΔ ), and
(b) gr(U (wΔ )) is a Uq (gΔ )-module algebra, where gΔ is the maximal semisimple subalgebra
of lΔ .
Proof. The following facts are well known.
Lemma 5.13.

n
(a) If U is a filtered k-algebra, there are isomorphisms of vector spaces φ̃n : Un → i=0 Ũi
where Ũi = Ui /Ui−1 .
(b) The isomorphisms φ̃n induce an isomorphism of k-algebras φ : U → gr(U ).
Lemma 5.14. Let A be a Hopf algebra, and U be a filtered A-module algebra. Then φ : U →
gr(U ) is an isomorphism of A-modules.
Recall that an algebra A is called quadratic-linear, if A

is the quotient of a free al-
2
gebra Cx1 , . . . , xn by an ideal generated by elements of i=0 (Cx1 , . . . , xn )i , where
(Cx1 , . . . , xn )i denotes the ith graded component. The following proposition is a key step
to proving Theorem 5.12.
Proposition 5.15. Let Δ ∈ [1, r]. If U (wΔ ) is quadratic linear, i.e. [E α , E β ]q ∈ (U (wΔ ))2 for
all radical roots α, β and U (wΔ ) is a filtered Uq (lΔ )-module algebra, then gr(U (wΔ )) is a
Uq (lΔ )-module algebra.
Proof. If U (wΔ ) is quadratic linear, then U (wΔ ) is a filtered Hopf algebra. Hence, gr(U (wΔ ))
is a Uq (lΔ )-module algebra by Lemma 5.14. The proposition is proved. 2
Note first that if U (wΔ ) is filtered, then U (wΔ ) must be quadratic linear. Hence, gr(U (wΔ ))
is quadratic, and Theorem 5.21 and Lemma 5.9 yield that its classical limit is S(radΔ ). Theo-
rem 5.12(a) is proved.
Part (b) can be proved analogously. Theorem 5.12 is proved. 2
We now obtain the construction of the braided symmetric algebra of the natural module of
Uq (sp(2n)).
Corollary 5.16. Let g = sp(2n), g = sp(2n + 2) and Δ = {2, . . . , n + 1}. The braided sym-
metric algebra Sσ (Vω1 ) of the natural Uq (sp(2n))-module Vω1 is isomorphic to U (wΔ ) as a
Uq (sp(2n))-module.
Proof. Note that the maximal semisimple Lie subalgebra gΔ of lΔ is isomorphic to sp(2n).
Using Theorem 5.12, we have to show that U (wΔ ) is a filtered Uq (lΔ )-module algebra.
Lemma 5.17.
(a) The quantum Schubert cell U (wΔ ) is quadratic linear.

(b) The twisted quantum Schubert cell U (wΔ ) is quadratic linear.
j
Proof. Prove (a) first. The radical roots corresponding to Δ are of the form α1 + i=2 αi , α1 +
· · · + αj + 2αj +1 + · · · + 2αn + αn+1 for j n, and αmax = 2α1 + · · · + 2αn + αn+1 . For
convenience we will fix a reduced expression of w0 such that the roots of sp(2n) are ordered as

n+1
n−1
α1 < α1 + α2 < · · · < αi < αi + 2αn + αn+1 < · · · < αmax < · · · . (5.4)
i=1 i=1
It is easy to see from (5.4) that if α, β αmax one cannot find α < γ1 · · · γk < β with k 3
such that α + β = γ1 + · · · + γk . Hence U (wΔ ) is quadratic linear and part (a) is proved.
Part (b) follows directly, because the Ti are algebra automorphisms for all i ∈ [1, r]. The
lemma is proved. 2
The lemma implies that U (wΔ ) is filtered. It remains to investigate the Uq (lΔ )-action.
Lemma 5.18. Let g = sp(2n + 2) and Δ = {2, 3, . . . , n + 1} ⊂ [1, n + 1]. Then, U (wΔ ) is a
filtered Uq (lΔ )-module algebra.
Proof. One can easily verify using the definition of the adjoint action (5.2) and the defining
relations of Uq (g ) (see Section 4.1) that ad(Fi )(U (wΔ )m ) ⊂ U (wΔ )m for all m ∈ Z0 and
i ∈ Δ as well as ad(Kλ )(U (wΔ )m ) ⊂ U (wΔ )m for all λ ∈ P (g ). It remains to check that
ad(Ei )(Eαi1 . . . Eαim ) ∈ U (wΔ )m for i ∈ Δ and radical roots αi1 , . . . , αim . We prove this by
induction on m.
Let m = 1. Note that ad(Ei )(Eα ) ⊂ U (wΔ ) for all radical roots α and all i ∈ [2, n + 1] by
Theorem 5.21(b). If α = αmax , then and (5.3) and (5.4) imply that ad(Ei )(Eα ) = 0 ∈ U (wΔ ).
If α < αmax , then (α, ω1 ) = 1 and hence one cannot find γ1 · · · γk αmax with k 2 such
that α + αi = γ1 + · · · + γk , since (α + αi , ω1 ) = 1 and (γ1 + · · · + γk , ω1 ) k. We obtain that
ad(Ei )(U (wΔ )1 ) ⊂ U (wΔ )1 .
Let m > 1 and note that
ad(Ei )(Eαi1 . . . Eαim ) = ad(Ei )(Eαi1 )Eαi2 . . . Eαim + q r Eαi1 ad(Ei )(Eαi2 . . . Eαim ),
for some r ∈ Z. The assertion now follows immediately from the inductive hypothesis. The
lemma is proved. 2
Note that radΔ ∼ = Vω1 ⊕ V0 as a sp(2n)-module. Hence, (gr(U (wΔ )))1 ∼

q q
= Vω1 ⊕ V0 as
Uq (sp(2n))-modules. Therefore gr(U (wΔ )) is a flat deformation of S(Vω1 ⊕ V0 ) by Theo-
rem 5.12.
Let Ŝ(Vω1 ) be the quotient of gr(U (wΔ )) by the two-sided Uq (sp(2n)-module ideal gener-
q
ated by V0 ∈ gr(U (wΔ ))1 . Note that S 2 (Vω1 ) ∼

= V2ω1 and Λ2 Vω1 ∼
= Vω2 ⊕ V0 and V0 ⊗ Vω1 ∼=

Vω1 . This implies that the quotient map T (Vω1 ⊕ V0 ) → gr(U (wΔ )) factors through T (Vω1 )
q
and that the kernel of the map T (Vω1 ) → Ŝ(Vω1 ) is the ideal generated by Λ2σ V q , hence
q ∼ q q
Ŝ(Vω1 ) = Sσ (Vω1 ). The braided symmetric algebra Sσ (Vω1 ) is therefore a flat deformation of
S(Vω1 ). Corollary 5.16 is proved. 2
q
Remark 5.19. The braided symmetric algebra Sσ (Vω1 ) of the natural Uq (sp(2n))-module can be
obtained, by an argument similar to the proof of Corollary 5.16, as the quotient of gr(U (wΔ ))
by the two-sided ideal generated by the copy of the trivial module V0 in gr(U (wΔ ))1 .
Problem 5.20. Describe the q-deformed symmetric algebras associated to radicals of Heisenberg
type, where the quantum radical U (wΔ ) satisfies the assumptions of Theorem 5.12. These alge-
bras cannot be braided symmetric algebras, but should provide examples for some more general
concept of quantum symmetric algebra.
5.2. Quantum Schubert cells: PBW-theorem and Levi action
Let g be a complex reductive Lie algebra, and let W be the Weyl group of g. In this section
we prove a PBW-type theorem for quantum Schubert cells U (w) ⊂ Uq (g) associated to w ∈ W
and show that if wΔ is a parabolic element of the W , then the Hopf subalgebra Uq (lΔ ) ⊂ Uq (g)
acts adjointly on U (wΔ ) (for definitions see of Section 5.1).
The following theorem is the main result of this section.
Theorem 5.21.

(a) Let w ∈ W be an element of the Weyl group of g. The monomials Eα(1) (k)
(1) . . . Eα(k) , satisfying
(i) = 0 if w(α) ∈ R − (g), form a C(q)-linear basis of the quantum Schubert cell U (w).
(b) If w = wΔ ∈ W is parabolic and w0 = si1 . . . sik a reduced expression of w0 that can be
factored wo = w0,Δ wΔ , then Uq (lΔ ) acts adjointly on Tw0,Δ (U (wΔ )) = U (wΔ ), where
w0 = w0,Δ wΔ .
Proof. Prove (a) first. Recall the PBW-theorem for Uq (g).
Proposition 5.22. (See e.g. [14, 8.24].)
m m
(a) The monomials Eα(1) (k)
(1) . . . Eα(k) Fα1
(1) (k)
. . . Eα(k) K μ , with (i) , m(i) ∈ Z0 and μ ∈ P (g), form a
C(q)-linear basis of Uq (g).
+
(1) . . . Eα(k) with (i) ∈ Z0 form a C(q)-linear basis of U . Similarly, the
(b) The monomials Eα(1) (k)

(1)
monomials Fα(1) (k)
. . . Fα(k) with m(i) ∈ Z0 form a C(q)-linear basis of U − .
Denote by (w) the length of an element w ∈ W . The following fact relates quantum Schubert
cells.
Proposition 5.23.
(a) U (w0 ) = U + .
(b) Let w, w ∈ W such that w0 = ww and (w) + (w ) = (w0 ). Then

U + = U (w0 ) = U (w)Tw U (w ) , U (w) ∩ Tw U (w ) = C(q) · 1 ⊂ Uq (g).
Proof. We need the following fact.
Lemma 5.24. Let α ∈ R + (g) and let w ∈ W .
(1) If w(α) ∈ R + (g), then Tw (Eα ) ∈ U + .

(2) If w(α) ∈ R − (g), then Tw (Eα ) ∈ Uq (b− ).
Proof. Note first the following fact.
Lemma 5.25.
(a) Let β ∈ R + (g). Then, Ti (Eβ ) ∈ U + if si (β) ∈ R + (g).

(b) Let β ∈ R + (g). Then, Ti (Eβ ) ∈ U (b− ) if si (β) ∈ R − (g). Moreover, Ti (Eβ ) = −Fi Kαi .
(c) Let β ∈ R + (g). Then Ti (Fβ Kλ ) ∈ Uq (b− ) if si (β) ∈ R + (g).
Proof. Prove (a) first. Let β = si1 . . . sik (αj ) and let w = si si1 . . . sik (not necessarily reduced).
One has w(αj ) = si (β) ∈ R + (g). It is well known that w(αk ) ∈ R + (g) implies that (wsk ) =
(w) + 1. Hence there exist w, w ∈ W such that w0 = wsj w with (w0 ) = (w) + (w ) + 1.
That implies that for some choice of reduced expression, Ew(αj ) = Tw (Ej ) = Ti (Eα ) ∈ U + .
Part (a) is proved.
Prove (b) now. Note that if β ∈ R + (g) and si (β) ∈ R − (g), then β = αi . The assertion follows
from the definition of the Ti in (5.1). Part (b) is proved.
In order to prove part (c) note first that the Ti are algebra homomorphisms and that Ti (Kλ ) ∈
Uq (b− ). The assertion now follows from an argument analogous to the proof of (a). The lemma
is proved. 2
Now, let w = si1 . . . sik be a reduced expression of w. If w(α) ∈ R + (g), then sij . . . sik (α) ∈
R + (g) for all 1 j k. Indeed if sij . . . sik (α) ∈ R − (g) for some 1 j k, then there exists
j1 1 such that sij1 +1 . . . sik (α) = α and sij1 = s . Hence, si1 . . . sij1 −1 (−α ) = w(α) ∈ R + (g).
Recall the well-known exchange property of the Weyl group: let w = sim . . . sik−1 sik be a reduced
expression of w ∈ W . If w(αi ) ∈ R − (g), then there exists m r k such that
sir . . . sik+1 = sir+1 . . . sik si .
In our case we obtain that si1 . . . sij1 −1 has a reduced expression si1 . . . sij1 −1 = sm1 . . . smj1 −2 s ,
hence w has an expression
w = sm1 . . . smj1 −2 s s sij1 +1 . . . sik = sm1 . . . smj1 −2 sij1 +1 . . . sik ,
contradicting the assumption that w = si1 . . . sik was reduced. It follows now inductively from
Lemma 5.25 that Tw (Eα ) ∈ U + . Part (a) is proved.
Prove (b) now. If w(α) ∈ R − (g), and w = si1 . . . sik is a reduced expression, then we can find,
as in part (a) 1 j k such that w = si1 . . . sij −1 s sij +1 . . . sik , sij +1 . . . sik (α) = (α ). Employing
the exchange property as in part (a) we have sij1 . . . sij −1 (−α ) ∈ R − (g) and sij2 . . . sik (α) ∈
R + (g) for 1 j1 < j < j2 k. Arguing as in the proof of Lemma 5.25(a) we obtain that
Tij +1 . . . Tik (Eα ) = E , hence Lemma 5.25(b) and (c) yield that Tw (Eα ) ∈ Uq (b− ). Lemma 5.24
is proved. 2
Now we are ready to complete the proof of Proposition 5.23. Part (a) follows directly from
Lemma 5.24(b), since wo (α) ∈ R − (g) for all α ∈ R + (g).
Prove (b) now. Let wo = ww and (w0 ) = (w) + (w ). It follows from Lemma 5.24 that
Tw−1 (Eα ) ∈ U + or Tw−1 (Eα ) ∈ Uq (b− ). Since Tw0 (Eα ) ∈ Uq (b− ) we obtain that Tw−1 (Eα ) ∈
U (w ) and Eα ∈ Tw (U (w )), if Tw−1 (Eα ) ∈ U + . Similarly we obtain that Eα ∈ U (w) if
Tw−1 (Eα ) ∈ Uq (b− ). The fact that the Ti are algebra homomorphisms and the PBW-theorem
(Proposition 5.22(b)) now imply that U + = U (w0 ) = U (w)Tw (U (w )).
It is easy to see that U (w) ∩ Tw (U (w )) = C(q) · 1, because

C(q) · 1 ⊂ Tw−1 U (w) ∩ Tw U (w ) ⊂ Uq b− ∩ U + = C(q) · 1.
Part (b) and Proposition 5.23 are proved. 2
We can now complete the proof of Theorem 5.21(a). Let α ∈ R + (g). Note that Lemma 5.24
implies that Eα ∈ U (w) if w(α) ∈ R − (g). Since the Ti are algebra automorphisms we obtain that
the monomials Eα 11 . . . Eα kk with i = 0 if w −1 (αi ) ∈ R + (g) are elements of U (w). U (w) is an
algebra, hence the linear span of the above monomials is contained in U (w). The monomials are
linearly independent by Proposition 5.22, hence it remains to show that they span U (w). Choose
w ∈ W such that w w = w0 and (w ) + (w) = (w0 ). It follows by an analogous argument
from the proof of Proposition 5.23 that Eα 11 . . . Eαkk ∈ Tw (U (w )) if i = 0 whenever w −1 (αi ) ∈

R − (g). Proposition 5.22 allows us to write each u ∈ U + as u = ki=1 ui ui , where the ui ∈ U (w)
are linearly independent and ui ∈ Tw (U (w )). It follows immediately that Tw−1 (u) ∈ Uq (b− ),
if and only if Tw−1 ui ∈ Uq (b− ); this means that by Proposition 5.23(b) ui ∈ C(q) · 1 for all i.
Theorem 5.21(a) is proved.
We will now prove Theorem 5.21(b). It suffices to show that the Ei , Fi , i ∈ Δ and Kλ , λ ∈
P (g), which generate U (lΔ ) act adjointly on U (wΔ ).
Proposition 5.26.
(a) Let w, w ∈ W . Then Kλ , λ ∈ P (g) acts adjointly on Tw U (w).

(b) Let wΔ ∈ W be parabolic and let w0 = w0,Δ wΔ . If i ∈ Δ, then, Ei and Fi act adjointly on
Tw0,Δ U (wΔ ).
Proof. Prove (a) first. Let w0 = w w w. In order to prove the assertion it suffices by Lemma 5.3
to show that Tw −1 (ad(Kλ )u) ∈ U + for all x ∈ U i and u ∈ Tw U (w). Since Tv (Kλ )Tv (K−λ ) = 1
for all v ∈ W and λ ∈ P (g) we obtain that

Tw −1 ad(Kλ )u = Tw −1 (K−λ )Tw −1 uTw −1 (Kλ ) ∈ U +
for all u ∈ Tw U (w). Part (a) is proved.

Prove (b) now. We need the following fact.
Lemma 5.27. Let w ∈ W be an element of the Weyl group W , and αi , αj simple roots such that
w(αi ) = −αj . Then, Tw (Ei ) = −Fj Kαj and Tw (Fi ) = −K−αj Ei .
Proof. Recall that by the exchange property (see proof of Lemma 5.24) we can choose a re-
duced expression w = w si = si1 . . . sik si of w such that si+m . . . sik (αi ) ∈ R + (g). Note that if
w (αi ) = αj , then Tw (Ei ) = Ej and Tw (Fi ) = Fj . We compute using (5.1)

Tw (Ei ) = Tw Ti (Ei ) = Tw (−Fi Kαi ) = −Fj Kαj ,

Tw (Fi ) = −Tw Ti (Fi ) = Tw (K−αi Ei ) = −K−αj Ej .
−1
Suppose that i ∈ Δ. Note that w0,Δ (αi ) = −αj with j ∈ Δ, and hence Tw−1 (Fi ) = Kαj Ej by
0,Δ
Lemma 5.27.
Let u ∈ U (wΔ ). We show that if u ∈ Tw0,Δ U (w) and u = Tw0,Δ u for u ∈ U (wΔ ), then

Tw−1 ad(Fi )u = Tw−1 (Fi )u Tw−1 (K−αi ) − u Tw−1 (Fi )Tw−1 (Kαi )
0,Δ 0,Δ 0,Δ 0,Δ 0,Δ
+
= −K−αj Ej u Kαj − u K−αj Ej Kαj ∈ U ,
because K−αi mKαi = q r m for every monomial m in U + . Hence, ad(Fi )u ∈ Tw0,Δ (U wΔ ). Note
that the proof does not require wΔ to be a parabolic element. However, the assumption will be
needed to prove the assertion for the action of Ei .
Choose i ∈ Δ and choose a reduced expression of w0 compatible with the factorization w0 =
w0,Δ w such that Eih = Ei , where h = (w0,Δ ). To complete the proof of the proposition it
suffices by Theorem 5.21(a) to show that ad(Ei )(Tw0,Δ (Eαi1 . . . Eαi )) ∈ Tw0,Δ (U (wΔ )), if i ∈ Δ,
and Eαij ∈ U (w) for j ∈ [1, ]. We use induction on .
Consider the case = 1; i.e., we have to show that ad(Ei )(Tw0,Δ (Eα )) ∈ Tw0,Δ (U (wΔ )) if
i ∈ Δ and Eα ∈ U (wΔ ). If Eα ∈ U (wΔ ), then wo,Δ (α) ∈ R + (g). Moreover, we have that αi <
w0,Δ (α) and Tw0,Δ (Eα ) ∈ U (wΔ ). Employing the choice of reduced expression and Lemma 5.9
we obtain that

ad(Ei ) Tw0,Δ (Eα ) ∈ span(Eγ1 . . . Eγk ),
where αi < γ1 · · · γk < Tw0,Δ (α), with Eγi ∈ Tw0,Δ (Eα ), and therefore ad(Ei )(Tw0,Δ (Eα )) ∈
Tw0,Δ (U (w)) as desired.
Now consider the case when > 1. Note that by a straightforward calculation

ad(Ei ) Tw0,Δ (Eαi1 . . . Eαi )
= Ei Tw0,Δ (Eαi1 . . . Eαi ) − Kαi Tw0,Δ (Eαi1 . . . Eαi )K−αi Ei

= ad(Ei ) Tw0,Δ (Eαi1 ) Tw0,Δ (Eαi2 . . . Eαi )

+ ad(Kαi ) Tw0,Δ (Eαi1 ) ad(Ei ) Tw0,Δ (Eαi2 . . . Eαi ) .
We have ad(Kαi )(Tw0,Δ (Eαi1 ) ∈ Tw0,Δ U (wΔ ) by Proposition 5.26(a) and hence
ad(Ei )(Eαi1 . . . Eαi ) ∈ Tw0,Δ U (wΔ ) by the inductive hypothesis. Part (b) is proved. Proposi-
tion 5.26 is proved. 2
Theorem 5.21(b) is proved. 2
6. Proof of Theorem 3.12
6.1. Necessary conditions
In this section we establish necessary conditions a weight λ ∈ P (g) has to satisfy if the simple
module Vλ is Poisson; i.e., we prove the “only if” assertion of the equivalence of (a) and (f) in
Theorem 3.12. Before we proceed with the proof of Theorem 3.12 we will have to introduce
some convenient notation. Since any finite-dimensional

module V over a semisimple Lie algebra
g splits as a direct sum of weight spaces V = μ∈P (g) V (μ), we will use the abbreviation

V (μ)c = ν=μ V (ν), as the “standard” complement of V (μ). Additionally we will use the
notation and results from Appendix A.
First we have to calculate c explicitly.
Lemma 6.1. Let g be a semisimple Lie algebra of rank r and c its Casimir element. Then (up to
a constant multiple)
(α, α)(β, β)
c = [c12 , c23 ] = Eα ∧ [E−α , Eβ ] ∧ E−β . (6.1)
+
4
α,β∈R
Proof. Choose a basis H1 , . . . , Hr for h which is orthonormal with respect

to the Killing form.
It
r is well known that the Casimir element c is up to a constant c = α∈R (α, α)Eα ⊗ E−α +
i=1 Hi ⊗ Hi . We calculate
c = [c12 , c23 ]

= (α, α)(β, β)Eα ⊗ [E−α , Eβ ] ⊗ E−β
α,β∈R

r

+ (α, α) Eα ⊗ [E−α , Hi ] ⊗ Hi + (α, α)Hi ⊗ [Hi , Eα ] ⊗ E−α . (6.2)
α∈R, i=1
It is easy to see that for all the summands X ⊗ Y ⊗ Z we have {X, Y, Z} ∩ {Eα : α ∈ R + } = ∅
and {X, Y, Z} ∩ {E−α : α∈ R + } = ∅. The element c ∈ g⊗3 is skew-symmetric by Lemma 3.1(a),
hence we can write c = α,β∈R + Eα ∧ Xα,β ∧ E−β .
It follows from (6.2) that for all α, β ∈ R + :
Eα ∧ Xα,β ∧ E−β = (α, α)(β, β)Eα ⊗ [E−α , Eβ ] ⊗ E−β + other terms.

This yields that Eα ∧ Xα,β ∧ E−β = 6 α,β∈R + (α, α)(β, β)Eα ∧ [E−α , Eβ ] ∧ E−β . Rescaling
shows that the assertion of the lemma follows. 2
The following result will now allow us to observe that large classes of simple modules are
“too big” to be Poisson.
Lemma 6.2. Let g be a complex simple Lie-algebra, P (g) its weight-lattice and R(g) ⊂ P (g) the
corresponding root-system with basis S = {α1 , . . . , αn }. Denote by w0 ∈ W the longest element
of the Weyl group W . Let λ ∈ P + (g) be a dominant weight, such that w0 (λ) = −λ. If Vλ is
Poisson, then (2λ − αi ) ∈ R(g) ∪ {0} for all αi such that (λ, αi ) = 0.
Proof. Let vλ ∈ Vλ (λ) be a highest weight vector, and suppose that 2λ − αi ∈ / R(g) ∪ {0}. Let αi
be a simple root such that (λ, αi ) = 0. Since λ is dominant, (λ, αi ) > 0. Set v = F−αi (v) = 0.
We have, by assumption, Vλ (−λ) = 0, since w0 (λ) = −λ, and clearly (αi |λ) < 0. Therefore, we
obtain for all v ∈ Vλ (−λ)
Eαi ∧ Fαi ∧ Hαi (v ∧ v ∧ v ) = cv · v · v + u,
where u ∈ (V (λ) · V (λ − α) · V (−λ))c . If 2λ − αi ∈ / R(g) ∪ {0}, then Eα (v ) ∈/ V (λ − αi ) for

all α ∈ R(g). Additionally, if 2λ − αj ∈ / R(g) ∪ {0} for all j ∈ [1, n], then 2λ is not a root,
either, and hence Eα (v ) ∈
/ V (λ) for all α ∈ R(g). Therefore, c(v ∧ v ∧ v ) = cv · v · v + u ,
where u ∈ (V (λ) · V (λ − α) · V (−λ))c , as defined in Lemma A.1 and (A.1). We obtain that
c(v ∧ v ∧ v ) = 0 and, hence, Vλ is not Poisson. The lemma is proved. 2
Lemma 6.2 has the following consequence.
Proposition 6.3. Let g be a complex simple Lie algebra, not isomorphic to E6 or sln (C), and let
λ ∈ P + (g). If Vλ is Poisson, then 2λ − αi ∈ R(g) for all simple roots αi such that (λ, αi ) = 0.
Proof. Recall the following well-known fact.
Lemma 6.4. Let g be a simple Lie algebra of rank r and let P (g) be its weight lattice, spanned
by the fundamental weights ωi , i = 1, . . . , r, labeled according to [3, Tables]. Denote by W (g)
the Weyl group and by w0 the longest element of W (g). We have w02 = 1 ∈ W (g) and:
(a) if g = slr+1 (C), then w0 (ωi ) = −ωr−i ;

(b) if g = E6 , then w0 (ω1 ) = −ω6 , w0 (ω3 ) = −ω5 and w0 (ω2 ) = −ω2 and w0 (ω3 ) = −ω3 ;
slr+1 , E6 , then w0 (ωi ) = −ωi for all i.
(c) if g =
Lemma 6.4 yields that the assumptions of Lemma 6.2 are satisfied for all λ ∈ P + (g), if g is
not isomorphic to either sln or E6 . Therefore, Proposition 6.3 now follows from Lemma 6.2. 2
We will also use the following very useful result.
Lemma 6.5. Let g = sl2 and V be a ( + 1)-dimensional simple sl2 -module. It is Poisson if and
only if 2.
Proof. If = 0, 1, then V is Poisson, because Λ3 V = 0. If = 2, then Λ3 V2 ∼ = V0 is a trivial

module and since S 3 V2 ∼ = 6 V ⊕ V2 we obtain that Hom g (Λ 3 V , S 3 V ) = {0} and hence V is
2 2 2
Poisson. Now, suppose that 3. Let E, F, H be a standard basis of sl2 . Since Λ3 (sl2 ) =
span(E ∧ F ∧ H ) we have c = E ∧ F ∧ H . Choose a weight basis v0 , . . . , v of V such that
H (vi ) = ( − 2i)vi , E(vi ) = ivi−1 and F (vi ) = ( − i)vi+1 . If, however, 3, then it is easy to
see that

c(v1 ∧ v0 ∧ v ) = E ∧ F ∧ H (v1 ∧ v0 ∧ v ) = − 2 v1 · v0 · v = 0.
Hence, V is not Poisson, if 3. The lemma is proved. 2
Another, very powerful tool will be a special case of Proposition 3.9. Recall that a parabolic
subalgebra p of a semisimple Lie algebra g is a Lie subalgebra containing a Borel subalgebra
of g, and that p splits as a semidirect product g = l n of a reductive Lie-algebra l, the Levi
subalgebra and a nilpotent Lie algebra n, the nil-radical. Let l ∼
= g ⊕ z, where g is a semisimple
Lie algebra and z is the center of l. Recall that if W is a simple g-module, then the W -isotypic
component of a g-module V is the submodule of V isomorphic to (Homg (W, V )) ⊗C W .
Proposition 6.6. Let g be a semisimple Lie algebra, and let p be a parabolic subalgebra with
Levi subalgebra l ∼
= g ⊕ z, where g is semisimple and z is the center of l. Let V be

a g-module,
and let V split as an l-module into a direct sum of isotypic components V ∼ =l Vi . If V is
a Poisson g-module, then each Vi must be a Poisson l-module. Moreover, Vi must be Poisson as
a g -module.
Proof. The Lie algebra g splits, in the notation of Proposition 3.9, as a vector space into g =
l ⊕ l⊥ where l⊥ is spanned by the Eα , α ∈ R(g)/R(g ). Choose an l-isotypic component Vi .
Denote by B be the set of all weights β ∈ h∗ such

that the weight spaces
Vi (β) = 0. Consider
the decomposition V ∼= VB ⊕ VC where VB ∼ = β∈B V (β) and VC ∼ = γ ∈B / V (γ ). We need
the following fact.
Lemma 6.7. Let g and g be as assumed in Proposition 6.6. Let V be a finite-dimensional g-

module, and U ⊂ V an isotypic g -module component of V . Let β ∈ B and u ∈ U (β) be a
weight vector. Then Eα (u) ∈ VB /{0} implies that α ∈ R(g ).
Proof. Let α ∈ R(g), and u ∈ U (β) for some β ∈ B. Then, Eα (u) ∈ V (α + β). Clearly
(α + β) ∈ B implies that α = (α + β) − β lies in the Z-linear span of R(g ), or, equivalently, in
the Z-linear span of a basis of R(g ). Since a root-system is determined uniquely by its basis, this
implies that α ∈ R(g ). Therefore, Eα (U ) ⊂ VC , if α ∈ (R(g)/R(g )). Lemma 6.7 is proved. 2
We can now apply Proposition 3.9(c) by choosing gsub = l and V1 = Vi . Therefore V is Pois-
son if and only if V1 is Poisson as a l-module. The l-module V1 is Poisson if and only if V1 is
Poisson as a g -module (Lemma 3.10). Proposition 6.6 is proved. 2
We can now derive the following criterion, which allows us to reduce the classification prob-
lem to a few cases.
Lemma 6.8. Let g be a simple Lie algebra, and let λ ∈ P + (g) be a dominant weight. If Vλ is
Poisson, then (λ|α) 2 for all roots α ∈ R + (g).
Proof. Consider the subalgebra gα = Eαi ⊕ E−αi ⊕ h ⊂ g. Clearly, gα is isomorphic to sl2 ⊕

Crank(g)−1 . Denote by g⊥α the vector space complement of gα spanned by the Eβ for α = ±β ∈
R(g). Let vλ ∈ Vλ (λ) be a highest weight vector in Vλ . Then, v generates a simple ((λ, αi ) +
1)-dimensional g -module V and Vλ = V ⊕ V ⊥ as gα -modules. We have g⊥ ⊥
α (V ) ⊂ V by
Lemma 6.7. We can now apply Proposition 3.9 and obtain that if V is Poisson, then V is Poisson
by Proposition 3.9 as a g -module. Hence, V is Poisson as a sl2 -module by Proposition 3.10. This
implies that (λ|α) 2 by Lemma 6.5 as desired. The lemma is proved. 2
We will now address the necessary conditions on λ ∈ P + (g) by type of Lie algebra.
6.1.1. The case of g = sln

Claim 6.9. Let g = sln . If Vλ is Poisson then λ ∈ {ω1 , 2ω1 , ω2 , ωn−2 , ωn−1 , 2ωn−1 }.
Proof. Lemma 6.8 has the following consequence.

n−1
Lemma 6.10. Let g = sln and let λ = i ωi be a dominant weight. If Vλ is Poisson, then
n−1 i=1
i=1 i 2.

Proof. Recall that if g = sln , i.e. of type An−1 , then the highest root αmax = n−1 i=1 αi . Since
n−1
(αi , ωj ) = δij for all i, j ∈ [1, n − 1], we obtain that (αmax |λ) = i=1 i . The assertion now
follows from Lemma 6.8(b). 2
It remains to prove that Vλ is not Poisson if λ = ωi + ωj , i = j , λ = 2ωk , 2 k n − 2, or

λ = ω with 3 n − 3. We will consider them case by case.
Lemma 6.11.
(a) Let g = sln , n 3 and let λ = ω1 + ωn−1 . Then, Vλ , the adjoint module, is not Poisson.
(b) Let g = sln , n 3, and let λ = ωi + ωj , 1 i < j n − 1. Then Vλ is not Poisson.
Proof. Prove (a) first. Denote by w0 the longest element of the Weyl group W . It is well known
that w0 (ωi ) = −ωn−i , and hence w0 (λ) = −λ. We know that λ = ω1 + ωn−1 = αmax , the highest
root, since Vλ is the adjoint module. It is clear that 2αmax − αi is not a root for all i ∈ [1, n − 1].
Therefore, Vλ is not Poisson by Lemma 6.2. Part (a) is proved.
Prove (b) next. Consider the Levi subalgebra li,j of g obtained by removing the first i − 1
nodes and the last n − j − 1 nodes from the Dynkin diagram An−1 . Clearly, li,j ∼ = slj −i+1 ⊕
Cn−j +i−1 , where C denotes the trivial sln -module. Any vector 0 = vλ ∈ Vλ (λ) generates an
adjoint slj −i+1 -module, which is not Poisson by part (a). Therefore, Vλ is not Poisson by Propo-
sition 6.6. The lemma is proved. 2
Now, consider simple modules of highest weight ωi and 2ωi for 3 i n − 3.
Lemma 6.12.
(a) Let g = sl6 and λ = ω3 . The simple module Vkω3 is not Poisson for all k 1.
(b) Let g = sln , n 6, and let 3 i n − 3. If Vλ is Poisson, then λ = kωi for 3 i n − 3.
Proof. Prove (a) first. We have ω3 = 12 α1 + α2 + 32 α3 + α4 + 12 α5 . It is easy to see that 2kω3 − αi

is not a root for all i ∈ [1, 5], since the highest root is αmax = 5i=1 αi . Part (a) is proved.
Prove (b) next. Consider the Levi subalgebra li−2,i+2 of g obtained by removing the first i − 3
nodes and the last n − i − 3 nodes from the Dynkin diagram An−1 . Clearly, li−2,i+2 ∼ = sl6 ⊕ Cn−6 .
Any vector 0 = vkωi ∈ Vkωi (kωi ) generates an sl6 -module isomorphic to Vkω3 , which is not
Poisson by part (a). Therefore, Vkωi is not Poisson by Proposition 6.6. The lemma is proved. 2
The following lemma addresses the last remaining case.
Lemma 6.13.
(a) If g = sl4 , then V2ω2 is not Poisson.

(b) If g = sln , n 4, then V2ω2 and V2ωn−2 are not Poisson.
Proof. Prove (a) first. We have 2ω2 = α1 + 2α2 + α3 , and w0 (2ω2 ) = −2ω2 by Lemma 6.4. It is
easy to see that 4ω3 − αi is not a root for all i ∈ [1, 3], since the highest root is αmax = 3i=1 αi .
Part (a) is proved.
Prove (b) now. Consider the Levi subalgebras l1,3 and ln−3,n−1 of g obtained by removing
the last n − 3 nodes (respectively the first n − 3) from the Dynkin diagram An−1 . Clearly,
l1,3 ∼
= ln−3,n−1 ∼
= sl3 ⊕ Cn−3 . Any vector 0 = v2ω2 ∈ V2ω2 (2ω2 ) (respectively 0 = v2ωn−2 ∈
V2ωn−2 (2ωn−2 )) generates an sl3 -module V2ω2 which is not Poisson by part (a). Therefore, V2ω2
is not Poisson by Proposition 6.6. The lemma is proved. 2
Claim 6.9 is proved. 2
6.1.2. The case of g = so(2n + 1)

Claim 6.14. Let g = so(2n + 1). If Vλ is Poisson, then λ = ω1 or (g, Vλ ) = (so(5), Vω2 ).
Proof. Consider first the case of g = so(5). Let {α1 , α2 } be a basis of the root system R(g).
The fundamental weights are ω1 = α1 + α2 and ω2 = α21 + α2 and the highest root is αmax =
α1 + 2α2 = 2ω2 . It is easy to verify that if λ ∈ P + (g) and the weight 2λ − αi ∈ R(g) for some
i = 1, 2 imply that λ ∈ {ω1 , ω2 }. Employing Proposition 6.3 we obtain immediately that if Vλ is
Poisson, then λ ∈ {ω1 , ω2 }.
We now consider g = so(2n + 1), n 3. Let R(g) be the corresponding root system and let
{α1 , α2 , . . . , αn } be a basis. The fundamental weights are, for 1 i n − 1,
ωi = α1 + 2α2 + · · · + (i − 1)αi−1 + i(αi + · · · + αn )

1
ωn = (α1 + 2α2 + · · · + nαn ).
2
The highest root is αmax = α1 + 2α2 + · · · + 2αn . It is easy to verify that for λ ∈ P + (g) and
2λ − αi ∈ R(g) for some i ∈ [1, n] imply that λ = ω1 or, in the case n = 3, λ = ω3 . We now
obtain immediately from Proposition 6.3 that if Vλ is Poisson and n 4 then λ = ω1 .
Consider the case n = 3. We have to show that Vω3 is not Poisson. Note that λ = ω3 = 12 α1 +
α2 + 32 α3 and αmax = α1 + 2α2 + 2α3 . Let v ∈ Vω3 (ω3 ), v = Eα3 (v) and v = Eαmax (v ). Clearly
v = 0, v = 0, v ∈ Vω3 ( 12 α1 + α2 + 12 α3 ) and v ∈ Vω3 (−ω3 ). Consider roots, α, β ∈ R + (g).
We have for all α, β ∈ R(g)

k
Eα ∧ [E−α , Eβ ] ∧ E−β (v ∧ v ∧ v ) = w · w · w ,
i=1
where w ∈ V (γi ), w ∈ V (γi ), and w ∈ V (γi ) such that γi + γi + γi = 12 α1 + α2 +

2 α3 . We obtain that (γi , γi , γi ) = (ω3 , 2 α1 + α2 + 2 α3 , −ω3 ), if and only if (α, β) ∈
1 1 1
{(α3 , α3 ), (αmax , αmax )}. Denote by
(α3 , α3 )2 (αmax , αmax )2

c = Eα3 ∧ α̌3 ∧ E−α3 + Eαmax ∧ α̌max ∧ E−αmax .
4 4
We obtain that
c
1 1
(c − c )(v ∧ v ∧ v ) ∈ Vω3 (ω3 ) · Vω3 α1 + α2 + α3 · Vω3 (−ω3 ) ⊂ S 3 Vω3 .
2 2
A straightforward calculation shows that c (v · v · v ) = 0.

This implies that c(v ∧ v ∧ v ) = 0 and proves that Vω3 is not Poisson. Claim 6.14 is
proved. 2
6.1.3. Type g = so(2n)

Claim 6.15. Let g = so(2n). If Vλ is Poisson, then λ = ω1 or n ∈ {4, 5} and λ ∈ {ωn , ωn−1 }.
Proof. Let {α1 , α2 , . . . , αn } be a basis of R(g) as in [3, Tables]. The fundamental weights have
the form
1
ωi = α1 + 2α2 + (i − 1)αi−1 + i(αi + · · · + αn−2 ) + i(αn−1 + αn ),
2

1 1 1
ωn−1 = α1 + 2α2 + · · · + (n − 2)αn−2 + nαn−1 + (n − 2)αn ,
2 2 2

1 1 1
ωn = α1 + 2α2 + · · · + (n − 2)αn−2 + (n − 2)αn−1 + nαn .
2 2 2
The highest root is αmax = α1 + 2α2 + · · · + 2αn−2 + αn−1 + αn . It is easy to verify that if
the weight 2λ − αi ∈ R(g) for some λ ∈ P + (g), then λ = ω1 or, in the case of n = 4, 5, λ = ωn
and λ = ωn−1 . Therefore, we obtain immediately from Proposition 6.3 that if Vλ is Poisson, then
λ = ω1 , or n ∈ {4, 5} and λ ∈ {ωn , ωn−1 }.
6.1.4. Type g = sp(2n)

Claim 6.16. Let g = sp(2n). If Vλ is Poisson, then λ = ω1 or (g, V ) = (sp(4), Vω2 ).
Proof. If n = 2, we have sp(4) ∼

= so(5) and we obtain from Section 6.1.2 that Vλ is not Poisson,
unless λ = ωi for i = 1, 2.
Now, let us continue with the case of g = sp(2n), n 3. Let {α1 , . . . , αn } be a basis of R(g).
The fundamental weights have the form

1
ωi = α1 + 2α2 + · · · + (i − 1)αi−1 + i αi + · · · + αn−1 + αn
2
for i n, and the highest root is αmax = 2α1 +2α2 +· · ·+αn . It is easy to verify that if λ ∈ P + (g)
and 2λ − αi ∈ R(g), then λ = ω1 . Therefore, we obtain immediately from Proposition 6.3 that if
Vλ is Poisson, then λ = ω1 . Claim 6.16 is proved. 2
6.1.5. The exceptional Lie algebras

Claim 6.17. Let g be an exceptional complex simple Lie algebra. If Vλ is a nontrivial simple
Poisson module, then g = E6 and λ ∈ {ω1 , ω6 }.
Proof. The Dynkin diagram E6 contains two subdiagrams of type D5 and one of type A5 and
Levi subalgebras isomorphic to so(10) ⊕ C and sl6 ⊕ C. Let λ = 6i=1 i ωi ∈ P + (g). We obtain
that vλ ∈ Vλ (λ) generates a so(10)-module
5 Vλ , where λ = 1 ω1 + 3 ω2 + 4 ω4 + 2 ω5 and

an sl6 -module Vλ , where λ = 1 ω1 + i=2 i+1 ωi . Note that we are only adjusting notations
from [3] for the various root systems A5 , D5 and E6 . Applying Proposition 6.6 to the correspond-
ing Levi subalgebras and the results of Sections 6.1.1 and 6.1.3, we obtain that if Vλ is Poisson,
then λ = ωi and with i = 1, 2, 6. When considering the case E6 , we cannot apply Proposition 6.3
to all weights λ, because the longest word in the Weyl-group associated to E6 does not send all
dominant weights λ ∈ P + (g) to −λ ∈ P (g). Since ω2 = αmax we immediately see that Vω2 is the
adjoint module. Clearly, 2αmax − αj is not a root for all j ∈ [1, 6], and hence Vω2 is not Poisson
by Proposition 6.3.
Next, let g = E7 . Denote by R(g) the corresponding root system and by P (g) the weight
lattice. It is easy to derive from the tables in [3, Tables] that there exists no λ ∈ P + (g) such that
2λ − αi ∈ R(g) for any i ∈ [1, 7]. Therefore, Proposition 6.3 yields that there is no λ ∈ P + (g)
such that Vλ is Poisson.
Now, let g = E8 . Denote by R(g) the corresponding root system and by P (g) the weight
lattice. It is easy to derive from the tables in [3, Tables] that there exists no λ ∈ P + (E7 ) such that
2λ − αi ∈ R(g) for any i ∈ [1, 8]. Therefore, Proposition 6.3 yields that there is no λ ∈ P + (g)
such that Vλ is Poisson.
As the second to last case, we will consider the case of g = F4 . Denote by R(g) the corre-
sponding root system and by P (g) the weight lattice. It is easy to derive from the tables in [3,
Tables] that there exists no λ ∈ P + (g) such that 2λ − αi ∈ R(g) for any i ∈ [1, 4]. Therefore,
Proposition 6.3 yields that there is no λ ∈ P + (g) such that Vλ is Poisson.
Finally, let g = G2 . Denote by R(g) the corresponding root system and by P (g) the weight
lattice, {α1 , α2 } a basis of R(g). The fundamental weights are ω1 = 2α1 +α2 and ω2 = 3α1 +2α2 ,
and the highest root is αmax = ω2 = 3α1 + 2α2 . It is easy to verify that if λ ∈ P + (g) and 2λ − αi ∈
R(g) for some i ∈ {1, 2} then λ = ω1 . Therefore, if Vλ is Poisson then λ = ω1 by Proposition 6.3.
Now, let λ = ω1 . Let v ∈ Vω1 (ω1 ) and v = Eα1 (v) ∈ Vω1 (α1 + α2 ) and v = E3α1 +2α2 (v ) ∈
Vω1 (−ω1 ). It is easy to see that v = 0 and v = 0. As in the discussion of the so(7)-case we have
for all α, β ∈ R(g),

k
Eα ∧ [E−α , Eβ ] ∧ E−β (v ∧ v ∧ v ) = w · w · w ,
i=1
where w ∈ V (γi ), w ∈ V (γi ), w ∈ V (γi ) such that γi + γi + γi = α1 + α2 . We obtain that

(γi , γi , γi ) = (ω1 , α1 + α2 , −α1 − α2 ) if and only if (α, β) ∈ {(α1 , α1 ), (αmax , αmax ). Denote by
(α1 , α1 )2 (αmax , αmax )2

c = Eα1 ∧ α̌1 ∧ E−α1 + Eαmax ∧ α̌max ∧ E−αmax .
4 4
This implies that
c
c − c ∈ Vω1 (ω1 ) · Vω1 (α1 + α2 ) · Vω1 (−ω1 ) ⊂ S 3 Vω3 .
A straightforward computation yields that c(v ∧ v ∧ v ) = 0. Hence, Vω1 is not Poisson. That
concludes our discussion of the case g = G2 .
6.2. Sufficient conditions and rigidity
In order to prove the remaining assertions of Theorem 3.12 it now suffices to show that for
all the modules Vλ listed in Theorem 3.12(f), we have that Homg (Λ3 Vλ , S 3 Vλ ) = {0}, since
this implies that Vλ is Poisson by Proposition 3.3. The decomposition of the symmetric and
exterior powers of the geometrically decomposable modules and the sp(2n)-module Vω1 are well
known. As a reference see Howe [12]. In particular, in the case of g = sln the decompositions of
symmetric powers of the geometrically decomposable modules Vλ are well-known results from
classical invariant theory, in the remaining cases the symmetric powers have been computed by
multiple authors (e.g. Schmid [22]).
The decomposition of the exterior powers of the simple geometrically decomposable mod-
ules and the sp(2n)-module Vω1 are computed in a beautiful way by Stembridge in [23];
the decomposition however was already well known through the calculations of Lie algebra
cohomology of nilradicals by Kostant [18]. One immediately obtains from these results that
Homg (Λ3 Vλ , S 3 Vλ ) = {0}.
We are, however, interested in proving a stronger result which will be useful in the classifica-
tion of flat modules in Section 4.4.
In [2] we introduced a lower bound for the dimension of the symmetric and exterior cube,
and correspondingly we construct a minimal submodule contained in the symmetric and exterior
cubes. We will use the notation
Vμ (6.3)
for the set of highest weight vectors of weight μ in a finite-dimensional g-module V . For domi-
nant λ, μ, ν ∈ P + (g) denote

ν
cλ,μ = dim(Vλ ⊗ Vμ )ν = dimC Homg (Vν , Vλ ⊗ Vμ ) ;
ν is the tensor product multiplicity. And for any λ, μ ∈ P + (g) denote c+ = dim(S 2 V )μ
i.e., cλ,μ λ;μ λ
− + − μ
and cλ;μ = dim(Λ Vλ ) , so that cλ,μ + cλ,μ = cλ,λ . Ultimately, define:
2 μ

μ + − μ
dλ := cλ;ν − cλ;ν cν,λ .
ν∈P+
We need the following definition.
3 V =
μ
μ∈P + C
Definition 6.18. We will call the g-module Slow max(dλ ,0) ⊗ V the “lower sym-
λ μ

μ
metric cube” of Vλ , and similarly Λ3low Vλ = μ∈P + Cmax(−dλ ,0) ⊗ Vμ the “lower exterior cube.”
The definition is motivated by the following fact.
3 V → S 3 V and
Lemma 6.19. There exist injective homomorphisms of g-modules from Slow λ λ
Λ3low Vλ → Λ3 Vλ .
Proof. By definition of S 3 Vλ , one has (S 3 Vλ )μ = (S 2 Vλ ⊗ Vλ )μ ∩ (Vλ ⊗ S 2 Vλ )μ . Therefore, we

obtain the inequality:
μ μ μ
dim S 3 Vλ dim S 2 Vλ ⊗ Vλ + dim Vλ ⊗ S 2 Vλ − dim(Vλ ⊗ Vλ ⊗ Vλ )μ
μ μ μ
= dim S 2 Vλ ⊗ Vλ − dim Vλ ⊗ Λ2 Vλ = dλ .
3 V → S 3 V follows and the

The existence of injective homomorphisms of g-modules from Slow λ λ
assertion for Λ3 Vλ can be proved analogously. 2
3 V ∼ S 3 V and Λ3 V ∼ Λ3 V . The following

We say that a g-module Vλ is rigid, if Slow λ= λ low λ = λ
theorem is the main result of this section.
Theorem 6.20. Let g be a simple Lie algebra. A simple g-module V is rigid, if and only if (g, V )
is one of the pairs listed in Theorem 3.12(f).
Proof. We will first prove the “only if” assertion. Note the following fact connecting rigid and
Poisson modules.
Proposition 6.21. If a simple g-module Vλ is rigid, then Homg (Λ3 Vλ , S 3 Vλ ) = {0} and Vλ is
Poisson.
Proof. It is easy to see that

μ μ
dim Homg Λ3 Vλ , S 3 Vλ = max dλ , 0 · max −dλ , 0 = 0.
μ∈P +
We immediately obtain Homg (Λ3 Vλ , S 3 Vλ ) = {0}, and hence, that Vλ is Poisson. 2
We obtain from Proposition 6.21 and the arguments in Section 6.1 that if V is rigid, then
(g, V ) must be one of the pairs of Theorem 3.12(f).
The proof of the converse will consist of the following steps for each of the listed simple
modules Vλ .
(1) Determine the decomposition S 2 Vλ and Λ2 Vλ , as well as S 3 Vλ and Λ3 Vλ . Recall from above
that the symmetric and exterior powers of the modules in question are well known and indeed
multiplicity-free (see e.g. [12, Chapter 4]).
μ
(2) Suppose Vμ appears with multiplicity one in S 3 Vμ . Show that dλ = 1. Similarly, if Vμ
μ
appears with multiplicity one in Λ Vμ . Show that dλ = −1.
3
In order to further simplify our computations note the following fact. Recall that if g is a
semisimple Lie algebra and τ a graph automorphism of the Dynkin diagram, then τ induces
automorphisms τg : g → g, τP : P (g) → P (g).
Lemma 6.22. Let g be a semisimple Lie algebra and let τ be a graph automorphism of the
Dynkin diagram of g.
λ =c τ (λ)
(a) We have for the tensor product multiplicities cμ τ (μ) . Similarly, if S n Vλ = Vνi then

i
S Vτ (λ) = i Vτ (νi ) and if Λ Vλ = i Vνi , then Λ Vτ (λ) = i Vτ (νi ) .
n n n
(b) If the simple g-module Vλ is rigid, then Vτ (λ) is rigid.

Proof. Part (a) is well known, and (b) follows immediately from (a). 2
In order to accomplish step (2) we will utilize a generalization of the tensor product stabi-
lization of the Littlewood–Richardson rule to simple Lie algebras of classical type by Kleber and
Vishwanath ([17] and [26]). Let g be a complex simple Lie algebra of type Xn , X ∈ {A, B, C, D}.
We denote for a triple of dominant weights λ, μ, ν ∈ P + (g) = P + (Xn ) the tensor multiplicity
ν (X ) = dim(Hom (V , V ⊗ V )).
cλ,μ n g ν λ μ
rule for g = sl
We first recall the Littlewood–Richardson n+1 (C). I.e. the Dynkin
diagram of
g is of type An : for dominant weights λ = ni=1 i ωi , μ = ni=1 mi ωi and ν = ni=1 ni ωi one
ν (A ) = cν (A ) for all m n if
has cλ,μ n λ,μ m

n
n
i( i + mi ) = ini .
i=1 i=1
This phenomenon is commonly referred to as tensor product stabilization (see e.g. [17]).
Next, let g be a complex simple Lie algebra of type Xn , X ∈ {B, C, D}. Following ideas of
Vishwanath
ωn−2 [26, Chapter 6, Corollary 3] we call a weight γ ∈ P (g) = P (Xn ) A-supported if γ =
i=1 i i This means that
c ω . γn is supported entirely in the Ak -part of the Dynkin diagram Xn .
Note that if a weight γ = ωi=1 ci ωi is A-supported in P (Xn ), then it is A-supported in P (Xm )
for all m n. One obtains the following fact.
Proposition 6.23. (See [26, Chapter 6, Corollary 3, Remark 10.1].) Let g be a complex simple
Lie algebras of type Xn where X ∈ {B, C, D}. If λ, μ, ν ∈ P + (Xn ) are A-supported, then
ν
cλ,μ (Xn ) = cλ,μ
ν
(Xm ), m n.
We will now show, case by case, that the modules in question are indeed rigid. Since the
computations are rather long, but straightforward, we include the complete proof in only one
nontrivial case ((g, V ) = (sln , V2ω1 )).
6.2.1. The case g = sln

Proposition 6.24. If g = sln , then the simple module V = Vλ is rigid, if λ ∈ {ω1 , 2ω1 ω2 , ωn−2 ,
ωn−1 , 2ωn−1 }.
Proof. By Lemma 6.22 it suffices to consider the case of λ ∈ {ω1 , 2ω1 ω2 }.
Lemma 6.25. Let g = sln , n 7. We have the following:
(a) Let V = Vω1 . We have S 2 V = V2ω1 , S 3 V = V3ω1 , Λ2 V = Vω2 and Λ3 V = Vω3 .

(b) (See [12, Theorems 3.1, 4.4.2].) Let V = V2ω1 . We have S 2 V ∼ = V4ω1 ⊕ V2ω2 , Λ2 V ∼ =
∼ ∼
V2ω1 +ω2 , S V = V6ω1 ⊕ V2ω1 +2ω2 ⊕ V2ω−3 and Λ V = V3ω1 +ω3 ⊕ V3ω2 .
3 3
(c) (See [12, Theorems 3.8.1, 4.4.4].) Let V = Vω2 . We have S 2 V ∼

= V2ω2 ⊕ Vω4 , Λ2 V ∼
= Vω1 +ω3 ,
∼
S V = V3ω2 ⊕ Vω2 +ω4 ⊕ V2ω3 .
3
It remains to show the following.
Lemma 6.26. Let g = sln , n 7.

(a) If λ = ω1 , then dωλ1 = 1 if λ = 3ω1 , and dωλ1 = −1 if λ = ω3 .

(b) If λ = 2ω1 , then d2ω λ = 1 if λ ∈ {6ω , 2ω + 2ω , 2ω } and d λ = −1 if λ ∈ {3ω +
1
1 1 2 3 2ω1 1
ω3 , 3ω2 }.
λ = 1, if λ ∈ {3ω , ω + ω , ω } and d λ = −1, if λ ∈ {2ω + ω , 2ω }.
(c) If λ = ω2 , then d2ω 1
2 2 4 6 2ω1 1 4 3
Proof. Computing the decompositions manually (or using the computer algebra system
LIE [25]) for n = 7, and applying the Littlewood–Richardson rule for all n 7 we obtain:
(a) For S 2 Vω1 ⊗ Vω1 and Λ2 Vω1 ⊗ Vω1 :

1 if λ = 3ω1 , 1 if λ = ω3 ,
c2ω1 ,ω1 =
λ
cωλ 2 ,ω1 =
0 else, 0 else.
Part (a) follows.

(b) For S 2 V2ω1 ⊗ V2ω1 and (λ, ωi ) = 0 for some i 7:

1 if λ ∈ {6ω1 , 2ω1 + 2ω2 , 4ω1 + ω2 },
λ
c4ω 1 ,2ω1
=
0 else,

1 if λ ∈ {2ω1 + 2ω2 , ω1 + ω2 + ω3 , 2ω3 },
λ
c2ω 2 ,2ω1
=
0 else.
For Λ2 V2ω1 :

1 if λ ∈ {4ω1 + ω2 , 2ω1 + 2ω2 , 3ω1 + ω3 , 3ω2 , ω1 + ω2 + ω3 },
λ
c2ω 1 +ω2 ,2ω1
=
0 else.
μ
We can now read off the d2ω1 and observe that (b) holds.
(c) Proceed analogous to part (b). The lemma is proved. 2
We obtain that the simple sln -modules Vω1 , V2ω1 and Vω2 are rigid for n 7. In the case
of n < 7 one can prove the assertion of the proposition by direct computation (e.g. using LIE).
6.2.2. The case g = so(k)

Proposition 6.27. Let g = so(k) and V = Vω1 . Then V is rigid. Moreover, V = Vλ is rigid, if
g = so(8) and λ = {ω3 , ω4 } or if g = so(10) and λ = {ω4 , ω5 }.
Proof.
Lemma 6.28. (See [22,23].) Let g = so(k), k 9. Then S 2 Vω1 ∼

= V2ω1 ⊕ V0 , Λ2 Vω1 ∼
= Vω2 ,
∼ ∼
S Vω1 = V3ω1 ⊕ Vω1 and Λ Vω1 = Vω3 .
3 3
μ μ
Lemma 6.29. Let g = so(k), k 9. If μ ∈ {3ω1 , ω1 }, then dω1 = 1. If μ = ω3 , then dω1 = −1.
Proof. Analogous to the proof of Lemma 6.26. 2
We obtain the assertion of Proposition 6.27 for k 9 from Lemmas 6.28 and 6.29. The asser-
tion for the remaining cases of k 8 and the case k = 10, λ ∈ {ω4 , ω5 } can be verified directly
(e.g. using LIE). Proposition 6.27 is proved. 2
6.2.3. The case g = sp(2n)

Proposition 6.30. Let g = so(2n) and V = Vω1 . Then V is rigid.
Proof.
Lemma 6.31. (See [12,23].) Let g = sp(2n) with n 4 and let V = Vω1 . We have S 2 V ∼
= V2ω1
and Λ2 V ∼
= Vω2 ⊕ V0 . Moreover, S 3 V ∼
= V3ω1 and Λ3 V ∼
= Vω1 ⊕ Vω3 .
We have the following additional fact.

μ μ
Lemma 6.32. Let g = sp(2n), n 4. We have that dω1 = 1, if μ = 3ω1 and dω1 = −1, if μ ∈
{ω3 , ω1 }.
Proof. Analogous to the proof of Lemma 6.26. 2
This proves the proposition in the case n 4. The assertion for the remaining cases n 3 can
be verified directly using LIE. Proposition 6.30 is proved. 2
6.2.4. Proof of Theorems 6.20 and 3.12

The assertion of Theorem 6.20 in the cases of g = E6 and λ = {ω1 , ω6 } can also be verified
through direct computation (e.g. using LIE). Theorem 6.20 then follows from Propositions 6.24,
6.27 and 6.30. 2
Theorem 6.20 and Proposition 6.21 imply that (f) yields (c) and (a). Thus, we have so far
proved the equivalence of parts (a)–(c), (e) and (f) of Theorem 3.12. Additionally, we obtain that
(f) implies (d) from the explicit computation of the exterior squares in the proof of Theorem 6.20.
We complete the proof of Theorem 3.12 by showing that (d) implies (a).
Proposition 6.33. Let g be a semisimple Lie algebra and let Vλ be a simple g-module. If Λ2 Vλ
is simple, then Vλ is Poisson.
Proof. Suppose that Vλ is simple. Then c acts as multiplication by a constant μ on Λ2 Vλ . There-

fore, c12 (v) = μv for all v ∈ Λ2 Vλ ⊗ Vλ and c23 (v ) = μv for all v ∈ Vλ ⊗ Λ2 Vλ . This implies
that

[c12 , c23 ](v) = μ2 − μ2 (v) = 0
for all v ∈ Λ2 Vλ ⊗ Vλ ∩ Vλ ⊗ Λ2 Vλ = Λ3 Vλ and hence, Vλ is Poisson. Proposition 6.33 is

proved. 2
Proposition 6.33 completes the proof of Theorem 3.12.
Theorems 3.12 and 6.20 have the following corollary, which can be applied in the classifica-
tion of flat modules in Section 4.
Corollary 6.34. Let g be a simple Lie algebra and V a simple g-module. V is Poisson, if and
only if V is rigid.
7. Open questions and conjectures
In this final chapter we will present a number of conjectures and questions which will be
interesting for future research in the area of braided symmetric algebras. The classification of flat
modules lets us expect that the following conjecture holds.
Conjecture 7.1. Let g be a reductive Lie algebra and V q a finite-dimensional Uq (g)-module.

The braided symmetric algebra Sσ (V q ) is a flat deformation of the reduced symmetric algebra
S(V , r − ) of the classical limit V of V q .
This conjecture is of particular interest because it opens the possibility to address the following
not yet investigated question.
Problem 7.2. Quantize a commutative Poisson algebra A; e.g. a reduced symmetric algebra.
While this conjecture gives rise to the question of quantization of manifolds with a brack-
eted structure, the classification of flat modules in Theorem 4.23 and our computation of the
braided symmetric and exterior cubes of simple Uq (sl2 )-modules in [2, Theorem 2.40] suggest
that braided symmetric and exterior cubes are rigid in the following sense.
Conjecture 7.3. Let g be a simple Lie algebra and V q a simple Uq (g)-module and V its classical
3 V (respectively
limit. The classical limit of Sσ3 V q (respectively Λ3σ V q ) is isomorphic to Slow
3
Λlow V ) as a U (g)-module.
It is easy to see that the assertion does not hold for nonsimple V q or g semisimple, as the
example of the natural Uq (slm × sln )-module shows.
Acknowledgments
I would like to acknowledge the advice and support of A. Berenstein. Additionally, I am

thankful for stimulating discussions with P. Etingof, A. Polishchuk and B. Shelton.
Appendix A
We develop in this appendix notation for products and complements of weight-spaces in sym-
metric and exterior powers of vector spaces. Let V be a finite-dimensional vector space, and let
V1 , V2 , . . . , Vn be subspaces of V . We define V1 ·V2 ·· · ··Vn ⊂ S n V and V1 ∧V2 ∧. . .∧Vn ⊂ Λn V
the subspaces generated by elements v1 · v2 · · · · · vn , respectively v1 ∧ v2 ∧ . . . ∧ vn , where vi ∈ Vi
for i ∈ [1, n].
Lemma A.1. Let

nV be a finite-dimensional vector space and let V1 , . . . , Vn be subspaces
such that V ∼ = i=1 Vi . Denote by P(m) the set of all increasing m-element sequences p =
(p1 , . . . , pm ), 1 p1 · · · pm n in [1, n]. The mth symmetric and exterior powers now
admit the following decomposition:

SmV ∼= Vp1 · · · · · Vpm , Λm V ∼
= Vp1 ∧ . . . ∧ Vpm .
p∈P p∈P

m
Proof.
m Recall that if V ∼
= V1 ⊕ V2 , then S m V ∼
= i=0 S i V1 S m−i V2 , respectively m ∼

n that Λ V =
i V Λn−i V . We can now use induction in n to prove that for V ∼
i=0 Λ 1 2 = i=1 i V we have

n
n
SmV ∼
= S i Vi , Λm V ∼
= Λ i Vi ,
∈Lm i=1 ∈Lm i=1
where Lm denotes the set of n-element sequences ( 1 , . . . , n ) such that 1 + · · · + n = m.

Using the notation V · · · · · V = S m V and V ∧ V ∧ . . . ∧ V = Λm V and the fact that S i V1
S V2 ∼
j
= S i V1 · S j V2 , respectively Λi V1 Λj V2 ∼
= Λi V1 ∧ Λj V2 , we obtain the desired result.
We will use the notation (Vp1 · · · · · Vpm )c ⊂ S m V to denote

(Vp1 · · · · · Vpm )c = V (λp ) · · · · · V (λpm ), (A.1)
1
p =p∈P
and similarly, (Vp1 ∧ . . . ∧ Vpm )c ⊂ Λm V for

(Vp1 ∧ . . . ∧ Vpm )c = V (λp ) ∧ . . . ∧ V (λpm ). (A.2)
1
p =p∈P
References
[1] A. Belavin, V. Drinfeld, Triangle equations and simple Lie algebras, Soviet Sci. Rev. Sect. C, Math. Phys. Rev. 4
(1984) 93–165.
[2] A. Berenstein, S. Zwicknagl, Braided symmetric and exterior algebras, Trans. Amer. Math. Soc. 360 (2008) 3429–
3472.
[3] N. Bourbaki, Groupes et algèbres de Lie, Masson, Paris, 1981.
[4] K. Brown, K. Goodearl, Lectures on Algebraic Quantum Groups, Birkhäuser, Basel, 2002.
[5] C. De Concini, C. Procesi, Quantum Schubert cells and representations at roots of 1, in: Algebraic Groups and Lie
Groups, in: Austral. Math. Soc. Lect. Ser., vol. 9, Cambridge Univ. Press, Cambridge, 1997, pp. 127–160.
[6] C. De Concini, V.G. Kac, C. Procesi, Some quantum analogues of solvable Lie groups, in: Geometry and Analysis,
Bombay, 1992, Tata Inst. Fund. Res. Bombay, 1995, pp. 41–65.
[7] J. Donin, D. Gurevich, S. Majid, R-matrix brackets and their quantization, Ann. Inst. H. Poincaré Phys. Théor. 58 (2)
(1993) 235–246.
[8] V. Drinfel’d, Quasi-Hopf algebras, Leningrad Math. J. 1 (6) (1990) 1419–1457.
[9] V. Drinfel’d, Commutation relations in the quasi-classical case, Selecta Math. Soviet. 11 (4) (1992) 317–326.
[10] A. Fokas, I. Gel’fand, Quadratic Poisson algebras and their infinite-dimensional extensions, J. Math. Phys. 35 (6)
(June 1994).
[11] K.R. Goodearl, M. Yakimov, Poisson structures on affine spaces and flag varieties. II. General case, Trans. Amer.
Math. Soc., in press.
[12] R. Howe, Perspectives on invariant theory: Schur duality, multiplicity-free actions and beyond, Israel Math. Conf.
Proc. 8 (1995) 1–182.
[13] J. Humphreys, Introduction to Lie Algebras and Representation Theory, Grad. Texts in Math., vol. 9, Springer,
Berlin, 1972.
[14] J.C. Jantzen, An Introduction to Quantum Groups, Grad. Stud. Math., Amer. Math. Soc., Providence, RI, 1996.
[15] N. Jing, K. Misra, M. Okado, q-Wedge modules for quantized enveloping algebras of classical type, J. Alge-
bra 230 (2) (2000) 518–539.
[16] A. Kamita, Quantum deformations of certain prehomogeneous vector spaces III, Hiroshima Math. J. 30 (2000)
79–105.
[17] M. Kleber, S. Viswanath, Tensor product stabilization in Kac–Moody Lie algebras, Adv. Math. 201 (1) (2006) 1–35.
[18] B. Kostant, Lie algebra cohomology and the generalized Borel–Weil theorem, Ann. of Math. 74 (1961) 329–387.
[19] I. Musson, Ring theoretic properties of the coordinate rings of quantum symplectic and Euclidean space, in:
S.K. Jain, S.T. Rizvi (Eds.), Ring Theory, Proc. Biennial Ohio State–Denison Conf. 1992, World Scientific, Sin-
gapore, 1993, pp. 248–258.
[20] M. Noumi, Macdonald’s symmetric polynomials as zonal spherical functions on some quantum homogeneous
spaces, Adv. Math. 123 (1996) 16–77.
[21] N. Reshitikhin, L.A. Takhtadzhyan, L.D. Fadeev, Quantization of Lie groups and Lie algebras, Leningrad Math. J. 1
(1990) 193–225.
[22] W. Schmid, Die Randwerte holomorpher Funktionen auf hermitesch symmetrischen Räumen, Invent. Math. 91
(1969/1970) 61–80.
[23] J. Stembridge, On the classification of multiplicity-free exterior algebras, Int. Math. Res. Not. (40) (2003) 2181–
2191.
[24] E. Strickland, Classical invariant theory for the quantum symplectic group, Adv. Math. 123 (1996) 78–90.
[25] M. VanLeeuwen, LiE, A software package for Lie group computations, http://young.sp2mi.univ-poitiers.fr/marc/
LiE/.
[26] S. Vishwanath, Dynkin diagram sequences and stabilization phenomena, Comm. Algebra 34 (11) (2006) 3903–
3933.

1 s2.0 S0001870808002302 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0001870808002302 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0001870808002302 Main

Uploaded by

Copyright:

Available Formats

Advances in Mathematics 220 (2009) 1–58

R-matrix Poisson algebras and their deformations

Keywords: Quantum groups; Invariant theory; r-Matrix; Poisson algebras; Quantization

E-mail address: [email protected].

5.2. Quantum Schubert cells: PBW-theorem and Levi action . . . . . . . . . . . . . . . . . . . . . . . 39

(a) The bracket (1.1) on S(V ) is Poisson.

(c) V is geometrically decomposable as defined by R. Howe [12].

The paper is organized as follows. In Section 2 we study quadratic Poisson algebras. We

2. Quadratic Poisson brackets

2.1. Decorated spaces and bracketed algebras

(b) The tensor product (V ⊗ V , Φ ) = (V , Φ) ⊗ (V , Φ ) of two decorated spaces (V , Φ) and

(V , Φ) and (V , Φ ) is a decorated space.

σF (V ),F (V ) ◦ F (Φ) = F (σV ,V ◦ Φ) = F (−Φ ◦ σV ,V ) = −(F (Φ) ◦ σF (V ),F (V ) ).

The lemma is proved. 2

We next introduce bracketed and Poisson algebras in the category C.

{a, b} + {b, a} = 0, (2.2)

{a, bc} = {a, b}c + b · {a, c} (2.3)

{ca, b} = c · {a, b} + {c, b} · a. (2.4)

Note the following obvious fact.

We need the following fact.

where μA denotes the multiplication in A and μB the multiplication in B.

The following facts are obvious.

2.2. Symmetric algebras and bracketed symmetric algebras

Similarly, define the nth exterior power as

(a) Let V be an object of C. The symmetric algebra S(V ) is a commutative algebra in C.

Define morphisms σi,i+1 ∈ EndC (V ⊗n ) for all 1 i (n − 1) by σi,i+1 = Id⊗(i−1) ⊗

Proposition 2.9. One has for any object V of C:

(a) S n V = Im([n]!V ,σ ) and Λn V = Im([n]!V ,−σ ).

Let V be an object in C and let Φ : V ⊗ V → V ⊗ V be a morphism in C. For 1 i < j n

Φi,j = σ(j −1,j ) ◦ Φi,j −1 ◦ σ(j −1,j ) . (2.5)

Proof. We first need the following fact.

Proof. Clearly, we obtain

σm,m+1 ◦ Φi,j ◦ σm,m+1 = Φi,j , (2.6)

Lemma 2.12. Let τ = (j − 1, j ) ◦ (j − 2, j − 1) ◦ · · · ◦ (i − 1, i). Then we have for Φ ∈ End(V ⊗

στ ◦ Φi,j ◦ στ −1 = Φi−1,j −1 = Φτ (i),τ (j ) .

σi−1,i ◦ Φi,j ◦ σi−1,i

σi,i+1 ◦ · · · ◦ σj −2,j −1 ◦ Φi−1,j ◦ σj −2,j −1 ◦ · · · ◦ σi,i+1 = Φi−1,j

σm,m+1 ◦ Φτ (i),τ (j ) ◦ σm,m+1 = Φτ (i),τ (j ) .

Proposition 2.10 is proved. 2

We now define a map Φ (m,n) : V ⊗(m+n) → V ⊗(m+n) by the formula:

For any a ∈ S m V , b ∈ S n V , and â ∈ V ⊗m and b̂ ∈ V ⊗n such that m!

Lemma 2.13. If τ = (τ1 , τ2 ) ∈ Sm × Sn ⊂ Sm+n , then στ ◦ Φ (m,n) = Φ (m,n) ◦ στ .

The lemma is proved. 2

We have the following result.

Proof. Prove anti-commutativity (2.2) first. We need the following fact.

σ(i,j ) ◦ Φi,j = −Φi,j ◦ σ(i,j ) .

(b) If i < j n and τ ∈ Sn such that τ (i) > τ (j ), then

στ ◦ Φi,j ◦ στ −1 = −Φτ (j ),τ (i) .

σ(i,j ) ◦ Φi,j ◦ σ(i,j )−1 = στ ◦ σ(i,i+1) ◦ στ −1 ◦ Φi,j ◦ στ ◦ σ(i,i+1) ◦ στ −1

Part (a) is proved.

Part (b) is proved. The lemma is proved. 2

Let τ ∈ Sm+n be the permutation, which sends (1, 2, . . . , m, m + 1, . . . , n + m) → (n + 1, n +

Therefore the Leibniz rule holds. Proposition 2.14 is proved. 2

Proposition 2.16. The correspondence (V , Φ) → S(V , Φ) defines a faithful exponential functor

for all u, v ∈ V and u , v ∈ V . The proposition is proved. 2

2.3. Bracketed Poisson algebras

Denote by J : A⊗3 → A the Jacobian map defined by

J = F + F ◦ σ12 ◦ σ23 + F ◦ σ23 ◦ σ12 , (2.7)

We have the following result.

Let τ ∈ Sm+n be the permutation, which sends (1, 2, . . . , m, m + 1, . . . , n + m) → (n + 1, n +

Proposition 2.16. The correspondence (V , Φ) → S(V , Φ) defines a faithful exponential functor

Hence, c(uv) = 0 and V is not Poisson. Part (a) is proved.

0 = c(x111 ∧ x000 ∧ x100 )