Quantum Affine Algebras and Affine Hecke Algebras

Vyjayanthi Chari1 and Andrew Pressley

arXiv:q-alg/9501003v1 3 Jan 1995

August 30, 1993

1 Introduction
One of the most beautiful results from the classical period of the representation theory of
Lie groups is the correspondence, due to Frobenius and Schur, between the representations
of symmetric groups and those of general or special linear groups. If V0 is the natural
irreducible (n + 1)–dimensional representation of SLn+1 (CI), the symmetric group Sℓ acts
on V0⊗ℓ by permuting the factors. This action obviously commutes with the action of
SLn+1 (CI). It follows that one may associate to any right Sℓ –module M a representation
of SLn+1 (CI), namely
FS(M ) = M ⊗S V0⊗ℓ ,
ℓ

the action of SLn+1 (C I) on FS(M ) being induced by its natural action on V0⊗ℓ . The
main result of the Frobenius–Schur theory is that, if ℓ ≤ n, the assignment M → FS(M )
defines an equivalence from the category of finite–dimensional representations of Sℓ to
the category of finite–dimensional representations of SLn+1 (C I), all of whose irreducible
components occur in V0⊗ℓ .
Around 1985, Drinfeld and Jimbo independently introduced a family of Hopf algebras
Uq (g), depending on a parameter q ∈ C I × , associated to any symmetrizable Kac–Moody
algebra g. Assuming that q is not a root of unity, Jimbo [7] proved an analogue of the
Frobenius–Schur correspondence in which SLn+1 (C I) is replaced by Uq (sln+1 ), V0 by the
natural (n + 1)–dimensional irreducible representation V of Uq (sln+1 ), and Sℓ by its Hecke
algebra Hℓ (q 2 ).
In [5], Drinfeld announced an analogue of the Frobenius–Schur theory for the Yangian
Y (sln+1 ), which is a “deformation” of the universal enveloping algebra of the Lie algebra
of polynomial maps C I → sln+1 . The role of Sℓ in this theory is played by the degenerate
affine Hecke algebra Λℓ , an algebra whose defining relations are obtained from those of
the affine Hecke algebra Ĥℓ (q 2 ) by letting q → 1 in a certain non–trivial fashion.
1
Both authors were partially supported by the NSF, DMS-9207701.

1
 In the same paper, Drinfeld conjectured that there should be an analogue of the
Frobenius–Schur theory relating the quantum affine algebra Uq (sl ˆ n+1 ) and Ĥℓ (q 2 ). In this
paper, we construct a functor from the category of finite–dimensional Ĥℓ (q 2 )–modules to
the category of finite–dimensional Uq (sl ˆ n+1 )–modules W of ‘type 1’ (a mild spectral con-
dition) with the property that every irreducible Uq (sln+1 )–type which occurs in W also
occurs in V ⊗ℓ (we assume that q is not a root of unity). We prove that this functor is
an equivalence if ℓ ≤ n. Drinfeld’s theory can be obtained from ours by taking a suitable
limit q → 1. Related results were obtained by Cherednik in [4].
We give a precise description of our functor at the level of irreducible representations,
using the known parametrizations of such representations of Uq (sl ˆ n+1 ) and of Ĥℓ (q 2 ).
Namely, in [2], [3] we showed that the finite–dimensional irreducible Uq (sl ˆ n+1 )–modules
of type 1 are in one to one correspondence with n–tuples of monic polynomials in one
variable. On the other hand, Zelevinsky [13] and Rogawski [12] have given a one to one
correspondence between the finite–dimensional irreducible Ĥℓ (q 2 )–modules and the set of
(unordered) collections of ‘segments’ of complex numbers, the sum of whose lengths is ℓ.
(A segment of length k is a k–tuple of the form (a, q 2 a, . . . , q 2k−2 a), for some a ∈ C I × .) We
compute explicitly the n–tuple of polynomials associated under our functor to any such
collection of segments.
The affine Lie algebra sl ˆ n+1 is a central extension, with one-dimensional centre, of the
Lie algebra of Laurent polynomial maps f : C I × → sln+1 . An obvious way to construct
ˆ
representations of sln+1 is to pull back a representation of sln+1 by the one-parameter
family of homomorphisms eva0 : sl ˆ n+1 → sln+1 which annihilate the centre and evaluate the
×
maps f at a ∈ C I . In [7], Jimbo defined a one-parameter family of algebra homomorphisms
ˆ n+1 ) → Uq (sln+1 ) which are quantum analogues of the ev 0 (actually, eva takes
eva : Uq (sl a
values in an ‘enlargement’ of Uq (sln+1 )). On the other hand, in [4] Cherednik defined a
one-parameter family of homomorphisms ev ˜ a : Ĥℓ (q 2 ) → Hℓ (q 2 ) which are the identity
on Hℓ (q 2 ) ⊂ Ĥℓ (q 2 ). Pulling back representations of Uq (sln+1 ) (resp. Hℓ (q 2 )) under eva
(resp. ev˜ a ) gives a one-parameter family of representations of Uq (sl ˆ n+1 ) (resp. Ĥℓ (q 2 )).
We show that these ‘evaluation’ representations correspond to
each other under our functor.

Acknowledgement We would like to thank I. V. Cherednik for several discussions related

to this work.

2 Quantum Kac–Moody algebras

Let A = (aij ) be a symmetric generalized Cartan matrix, where the indices i, j lie in
some finite set I. Thus, aij ∈ Z, aii = 2, and aij ≤ 0 if i 6= j. To A one can associate a

2
Kac–Moody Lie algebra g(A) (see [8]).
Let q be a non–zero complex number, assumed throughout this paper not to be a root
of unity. For n, r ∈ N, n ≥ r, define
q n − q −n
[n]q = ,
q − q −1
n [n]q [n − 1]q . . . [n − r + 1]q
 
= .
r q [r]q [r − 1]q . . . [1]q

2.1
Definition The quantum Kac–Moody algebra Uq (g(A)) associated to a symmetric gener-
alized Cartan matrix A = (aij )i,j∈I is the unital associative algebra over C
I with generators
x±
i , ki
±1
(i ∈ I) and the following defining relations:

ki ki−1 = 1 = ki−1 ki ,
ki kj = kj ki ,
± −1
ki xj ki = q ±aij x±
j ,
ki − ki−1
[x+ −
i , xj ] = δij ,
q − q −1
1−aij
1 − aij
 
r
(x± r ± ± 1−aij −r
X
(−1) i ) xj (xi ) = 0 , i 6= j.
r=0
r q

It is well–known that Uq (g(A)) is a Hopf algebra with comultiplication ∆ given on

generators by

∆(ki±1 ) = ki±1 ⊗ki±1 ,

∆(x+ + +
i ) = xi ⊗ki + 1⊗xi ,
∆(x− − −1 −
i ) = xi ⊗1 + ki ⊗xi

(we shall not need the formulas for the counit and antipode of Uq (g(A))).

2.2
By a representation of a quantum Kac–Moody algebra Uq (g(A)) we shall mean a left
Uq (g(A))–module. A representation W is said to be of type 1 if
M
W = Wµ ,
µ∈ZI

3
where Wµ = {w ∈ W |ki .w = q µ(i) w}. If Wµ is non–zero, then Wµ is called the weight
space of W with weight µ. Restricting consideration to type 1 representations results in
no essential loss of generality, for any finite–dimensional irreducible representation can be
obtained by twisting a type 1 representation with a suitable automorphism of Uq (g(A))
(cf. [10]).

2.3
Assume that dim(g(A)) < ∞. A representation W of Uq (g(A)) is said to be highest weight
with highest weight λ ∈ ZI if W is generated as a Uq (g(A))–module by an element wλ
satisfying
x+
i .wλ = 0, ki .wλ = q
λ(i)
wλ ,
for all i ∈ I.
A weight λ ∈ ZI is said to be dominant if λ(i) is non–negative for all i ∈ I.

Proposition ([10]) Assume that dim (g(A)) < ∞.

(i) Every finite–dimensional Uq (g(A))–module is completely reducible.
(ii) Every irreducible finite–dimensional Uq (g(A))–module of type 1 is highest weight. As-
signing to such a representation its highest weight defines a one to one correspondence
between the set of isomorphism classes of finite–dimensional irreducible representations of
type 1 and the set of dominant weights.
(iii) The finite–dimensional irreducible Uq (g(A))–module V (λ) of type 1 and highest weight
λ has the same character (in particular, the same dimension) as the irreducible g(A)–
module of the same highest weight.
(iv) The multiplicities of the irreducible components in a tensor product V (λ)⊗V (µ) of
irreducible finite–dimensional Uq (g(A))–modules is the same as in the tensor product of
the irreducible g(A)–modules of the same highest weights. ✷

2.4
The case of most interest to us is when A is the matrix
2 −1 0 0 ··· 0 −1
 

 −1 2 −1 0 ··· 0 0 


 0 −1 2 −1 ··· 0 0 

.. .. .. .. .. ..,

..

 . . . . . . .



0 ··· −1 2 −1 

 0 0
−1 0 0 ··· 0 −1 2

4
 ˆ n+1 . Fix a square root q 1/2
where i, j ∈ {0, 1, . . . , n}. Then g(A) is the affine Lie algebra sl
of q. For any elements a, b of an associative algebra over C I, set

[a, b]q1/2 = q 1/2 ab − q −1/2 ba.

ˆ n+1 ) can be written

Since aij = 0 or −1 if i 6= j, the quantized Serre relations in Uq (sl

[x± ±
i , xj ] = 0 if i − j 6= 0, ± 1 (mod n),
[x± ± ±
i , [xj , xi ]q 1/2 ]q 1/2 = 0 if i − j = ± 1 (mod n).

Deleting the 0th row and column of A gives the Cartan matrix of sln+1 . Thus, there is
a natural Hopf algebra homomorphism from Uq (sln+1 ) to Uq (sl ˆ n+1 ); this homomorphism
is injective (this follows from Proposition 5.4 below).
If g(A) = sln+1 , then I = {1, . . . , n} and so weights are identified with n–tuples of
integers. It is useful to introduce the weights ǫi , for 1 ≤ i ≤ n, defined by

 −1 if j = i − 1,

ǫi (j) = 1 if j = i,
0 otherwise.



Note that n+1

P
i=1 ǫi = 0.
Set αi = ǫi − ǫi+1 . If λ, µ ∈ ZI , we write λ ≥ µ if λ − µ = ni=1 ri αi for some
P

non-negative integers ri .
The elements λi = ij=1 ǫj , 1 ≤ i ≤ n, are called fundamental weights and the corre-
P

sponding irreducible representations V (λi ) the fundamental representations of Uq (sln+1 ).

The representation V (λ1 ) is called the natural representation of Uq (sln+1 ); it will be
denoted by V from now on. It has a basis {v1 , . . . , vn+1 } on which the action is given by:

x+
i .vr = δr,i+1 vr−1 ,
x−
i .vr = δr,i vr+1 ,
ki .vr = q ǫr (i) vr

(we set v−1 = vn+2 = 0).

Let x±θ be the operators on V defined by

x+ −
θ .vr = δr,n+1 v1 , xθ .vr = δr,1 vn+1 ,

ˆ n+1 )–module
and let kθ = k1 k2 . . . kn . It is easy to see that V can be made into a Uq (sl
× −1 ± ±1 ∓
V (a), for all a ∈ C
I , by letting k0 act as kθ and x0 as a xθ .

5
2.5
Definition If ℓ ≤ n, a finite–dimensional Uq (sln+1 )–module W is said to be of level ℓ if
every irreducible component of W is isomorphic to an irreducible component of V ⊗ℓ .

Note that every level ℓ representation of Uq (sln+1 ) is of type 1.

The next result follows immediately from Proposition 2.3 and the corresponding clas-
sical result (which is well–known and easy to prove).

Proposition Assume that ℓ ≤ n. Then, the finite–dimensional Uq (sln+1 )–module V (λ)

is of level ℓ ≤ n iff ni=1 iλ(i) = ℓ. ✷
P

Remark This proposition shows that the concept of level is well–defined. The assumption
that ℓ ≤ n is necessary, for if ℓ1 or ℓ2 is greater than n, it is possible for V ⊗ℓ1 and V ⊗ℓ2
to have an irreducible component in common even if ℓ1 6= ℓ2 .

2.6
ˆ n+1 ).
It is easy to check that c = k0 k1 . . . kn is central in Uq (sl
ˆ n+1 ) acts as 1 on every finite–dimensional
Proposition The central element c of Uq (sl
ˆ
Uq (sln+1 )–module W of type 1.

Proof. This was proved in [2] when n = 1 and W is irreducible. Essentially the same
proof works for all n and the extension to arbitrary finite–dimensional W follows by an
easy argument using Jordan–Hölder series. ✷

3 Hecke algebras and affine Hecke algebras

In this section, we collect some well–known definitions and results concerning (affine)
I × is not a root of unity.
Hecke algebras (cf. [9], [12]). We continue to assume that q ∈ C

3.1
Definition Fix ℓ ≥ 1. The affine Hecke algebra Ĥℓ (q 2 ) is the unital associative algebra
over CI with generators σi±1 , i ∈ {1, . . . , ℓ − 1}, yj±1 , j ∈ {1, . . . , ℓ}, and the following
defining relations:

σi σi−1 = σi−1 σi = 1,
σi σi+1 σi = σi+1 σi σi+1 ,

6
 σi σj = σj σi if |i − j| > 1,
2
(σi + 1)(σi − q ) = 0,
yj yj−1 = yj−1 yj = 1,
yj yk = yk yj ,
y j σi = σi y j if j 6= i or i + 1,
2
σi yi σi = q yi+1 .
The unital associative algebra with generators σi±1 , i ∈ {1, . . . , ℓ − 1}, defined by the
first four sets of relations above is called the Hecke algebra Hℓ (q 2 ).

There is an obvious homomorphism of Hℓ (q 2 ) onto the subalgebra of Ĥℓ (q 2 ) generated

by the σi .
Lemma The multiplication map C[y1±1 , . . . , yℓ±1 ]⊗Hℓ (q 2 ) → Ĥℓ (q 2 ) is an isomorphism
of vector spaces. ✷

3.2
The following well-known result provides an analogue for affine Hecke and Hecke algebras
of the canonical homomorphism Sℓ1 × Sℓ2 → Sℓ1 +ℓ2 .
Proposition There exists a unique homomorphism of algebras
ι̂ℓ1 ,ℓ2 : Ĥℓ1 (q 2 ) ⊗ Ĥℓ2 (q 2 ) → Ĥℓ1 +ℓ2 (q 2 )
such that
ι̂ℓ1 ,ℓ2 (σi ⊗1) = σi , ι̂ℓ1 ,ℓ2 (yj ⊗1) = yj , i = 1, . . . , ℓ1 − 1, j = 1, . . . , ℓ1 ,
ι̂ℓ1 ,ℓ2 (1⊗σi ) = σi+ℓ1 , ι̂ℓ1 ,ℓ2 (1⊗yj ) = yj+ℓ1 , i = 1, . . . , ℓ2 − 1, j = 1, . . . , ℓ2 . ✷
Clearly the restriction of ι̂ℓ1 ,ℓ2 to Hℓ1 (q 2 ) ⊗ Hℓ2 (q 2 ) induces a homomorphism ιℓ1 ,ℓ2 :
Hℓ1 (q 2 ) ⊗ Hℓ2 (q 2 ) → Hℓ1 +ℓ2 (q 2 ).
Let Mi be a right Hℓi (q 2 )–module for i = 1, 2, and let M1 ⊗M2 be their outer tensor
product (an Hℓ1 (q 2 )⊗Hℓ2 (q 2 )–module). Then, the Hℓ1 +ℓ2 (q 2 )–module M1 ⊙M2 , sometimes
called the Zelevinsky tensor product of M1 and M2 , is defined by
Hℓ (q 2 )
Hℓ1 +ℓ2 (q 2 ).
+ℓ
O
M1 ⊙ M2 = indHℓ1 (q22)⊗Hℓ (q 2 ) (M1 ⊗M2 ) = (M1 ⊗M2 )
1 2
Hℓ1 (q 2 )⊗Hℓ2 (q 2 )

The Zelevinsky tensor product ⊙ˆ for affine Hecke algebra modules is defined similarly.
Standard properties of induced modules show that the Zelevinsky tensor products are
associative up to isomorphism.

7
3.3
Proposition Let Mi be a finite–dimensional Ĥℓi (q 2 )–module, i = 1, 2. Then, there is a
canonical isomorphism of Hℓ1 +ℓ2 (q 2 )–modules

ˆ 2 )|H
(M1 ⊙M (q 2 )
∼
= M1 |Hℓ1 (q2 ) ⊙ M2 |Hℓ2 (q2 ) ,
ℓ 1 +ℓ2

where Mi |Hℓ (q 2 ) means Mi regarded as an Hℓi (q 2 )–module by restriction, etc.

i

Proof. It is easy to see that the canonical map

M1 |Hℓ (q 2 ) ⊙ M2 |Hℓ (q 2 )
ˆ 2 )|H
→ (M1 ⊙M (q 2 )
1 2 ℓ 1 +ℓ2

given by
(m1 ⊗m2 )⊗h 7→ (m1 ⊗m2 )⊗h (mi ∈ Mi , h ∈ Hℓ1 +ℓ2 (q 2 ))
is a well–defined surjective homomorphism of Hℓ1 +ℓ2 (q 2 )–modules. But, by Lemma 3.1,
the rank of Ĥℓ1 +ℓ2 (q 2 ) as an Ĥℓ1 (q 2 )⊗Ĥℓ2 (q 2 )–module is the same as that of Hℓ1 +ℓ2 (q 2 )
as an Hℓ1 (q 2 )⊗Hℓ2 (q 2 )–module. It follows that

dimC ˆ 2 ) = dimC
I (M1 ⊙M I (M1 ⊙ M2 ). ✷

3.4
Affine Hecke algebras have a family of universal modules, defined as follows. Let a =
I × )ℓ and set
(a1 , a2 , . . . , aℓ ) ∈ (C
Ma = Ĥℓ (q 2 )/Ha ,
the quotient of Ĥℓ (q 2 ) by the right ideal Ha generated by yj − aj , j = 1, . . . , ℓ.

Proposition ([12])
(a) Every finite–dimensional irreducible Ĥℓ (q 2 )–module is isomorphic to a quotient of some
Ma .
(b) For all a ∈ (CI × )ℓ , Ma is isomorphic as an Hℓ (q 2 )–module to the right regular repre-
sentation.
(c) Ma is reducible as an Ĥℓ (q 2 )–module iff aj = q 2 ak for some j, k. ✷

4 ˆ n+1) and Ĥℓ (q 2)

Duality between Uq (sl
We begin by recalling the duality, established by Jimbo [7], between representations of
Uq (sln+1 ) and Hℓ (q 2 ).

8
4.1
Let V be the natural (n + 1)–dimensional representation of Uq (sln+1 ) defined in 2.4, and
let Ř : V ⊗V → V ⊗V be the linear map given by

2
 q vr ⊗vs
 if r = s,
Ř(vr ⊗vs ) = qvs ⊗vr if s > r, (1)
qvs ⊗vr + (q 2 − 1)vr ⊗vs if r > s.



Fix ℓ > 1 and let Ři ∈ EndC ⊗ℓ th th

I (V ) be the map which acts as Ř on the i and (i + 1)
factors of the tensor product, and as the identity on the other factors.
Proposition ([7]) Fix ℓ, n ≥ 1. There is a unique left Hℓ (q 2 )–module structure on V ⊗ℓ
such that σi acts as Ři for i = 1, . . . , ℓ − 1. Moreover, the action of Hℓ (q 2 ) commutes with
the natural action of Uq (sln+1 ) on V ⊗ℓ .
If M is a right Hℓ (q 2 )–module, define
J (M ) = M ⊗Hℓ (q2 ) V ⊗ℓ ,

equipped with the natural left Uq (sln+1 )–module structure induced by that on V ⊗ℓ . Then,
if ℓ ≤ n, the functor M → J (M ) is an equivalence from the category of finite–dimensional
Hℓ (q 2 )–modules to the category of finite-dimensional Uq (sln+1 )–modules of level ℓ. ✷

4.2
We now state the main result of this section, which is an analogue of Proposition 4.1 for
quantum affine algebras. Recall the operators kθ , x±
θ ∈ EndCI (V ) defined in Section 2.4.
Theorem Fix ℓ, n ≥ 1. There is a functor F from the category of finite–dimensional
ˆ n+1 )–modules of type
right Ĥℓ (q 2 )–modules to the category of finite–dimensional left Uq (sl
1 which are of level ℓ as Uq (sln+1 )–modules, defined as follows. If M is an Ĥℓ (q 2 )–module,
then F(M ) = J (M ) as a Uq (sln+1 )–module and the action of the remaining generators of
ˆ n+1 ) is given by
Uq (sl
ℓ
x± [Link]±1 ⊗Yj± .v,
X
0 .(m⊗v) = (2)
j=1

k0 .(m⊗v) = m⊗(kθ−1 )⊗ℓ .v, (3)

where m ∈ M , v ∈ V ⊗ℓ and the operators Yj± ∈ EndC ⊗ℓ
I (V ), j = 1, . . . , ℓ, are defined by
Yj+ = 1⊗j−1 ⊗x− −1 ⊗ℓ−j
θ ⊗(kθ ) ,
Y− = k ⊗ j−1
⊗x+ ⊗1⊗ℓ−j .
j θ θ

9
 The functor F is an equivalence of categories if ℓ ≤ n.
Proof. We first show that the formulas (2) and (3) are well–defined. We do this for the
action of x+ −
0 , leaving the verification for x0 and k0 to the reader. Thus, we must prove
that
x+ +
0 .(m.σi ⊗v) = x0 .(m⊗σi .v)
for i = 1, . . . , ℓ, v ∈ V ⊗ℓ . This is equivalent to proving that, as operators on J (M ) =
M⊗ 2 V
⊗ℓ ,
Hℓ (q )
ℓ ℓ
σi yj ⊗Yj+ = yj ⊗ Yj+ σi .
X X
(4)
j=1 j=1

If j =6 i, i + 1, the j th terms on the left and right-hand sides of (4) are equal, since
σi yj = yj σi and σi Yj+ = Yj+ σi . Hence we must show that

σi yi ⊗Yi+ + σi yi+1 ⊗Yi+1

+
= yi ⊗Yi+ σi + yi+1 ⊗Yi+1
+
σi .

Using the relation σi − (q 2 − 1) = q 2 σi−1 , this reduces to

q 2 yi+1 ⊗(σi−1 Yi+ − Yi+1

+ −1 +
σi ) + yi ⊗(σi Yi+1 − Yi+ σi ) = 0.

Thus, it suffices to prove that

+
σi Yi+1 = Yi+ σi ,
i.e. that
Ř(1⊗x− − −1
θ ) = (xθ ⊗kθ )Ř (5)
as operators on V ⊗V . But this is easily checked by using the formula for Ř in (1) and
that for x−
θ in 2.4.
ˆ n+1 ), we shall
In proving that the formulas (2) and (3) define a representation of Uq (sl
assume that n > 1. The proof for the sl2 case is similar (the difference arises because the
Dynkin diagram of slˆ 2 has a double bond).
The only relations to be checked are those involving x+ −
0 , x0 and k0 . This is straight-
forward except for the quantized Serre relations:

[x± ± ±
i , [x0 , xi ]q 1/2 ]q 1/2 = 0, (6)

[x± ± ±
0 , [xi , x0 ]q 1/2 ]q 1/2 = 0, (7)
for i = 1, n. We verify (7) for x+
1 , leaving the other cases to the reader.
Applying the left–hand side of (7) to J (M ) and considering the terms involving yj yk ,
one sees that it suffices to prove that

[Yj+ , [∆(ℓ) (x+ +

1 ), Yk ]q 1/2 ]q 1/2 + (j ↔ k) = 0, (8)

10
where (j ↔ k) means the result of interchanging j and k in the first term and ∆(ℓ) is the
ℓth iterated comultiplication (so that ∆(2) = ∆). Equation (8) will be proved by induction
+(ℓ)
on ℓ, and we accordingly denote Yk+ by Yk . If ℓ = 1, then (8) becomes

[x− − −
θ , [x1 , xθ ]q 1/2 ]q 1/2 = 0,

which holds by the remarks at the end of 2.4.

For the inductive step we distinguish three cases:
(i) j, k < ℓ,
(ii) j < ℓ, k = ℓ or j = ℓ, k < ℓ,
(iii) j = k = ℓ.
For the first case, notice that the left–hand side of (8) is
+(ℓ−1) +(ℓ−1)
[Yj ⊗kθ−1 , [∆(ℓ−1) (x+ +
1 )⊗k1 + 1⊗x1 , Yk ⊗kθ−1 ]q1/2 ]q1/2 + (j ↔ k)
+(ℓ−1) +(ℓ−1)
= [Yj , [∆(ℓ−1) (x+ +
1 ) + 1⊗x1 , Yk ]q1/2 ]q1/2 ⊗k1 kθ−2 + (j ↔ k)
+(ℓ−1) +(ℓ−1)
+[Yj ⊗kθ−1 , Yk ⊗[x+ −1
1 , kθ ]q 1/2 ]q 1/2 + (j ↔ k).

The sum of the first two terms on the right-hand side vanishes by the induction hypothesis,
and the sum of the last two terms is a multiple of
+(ℓ−1) +(ℓ−1) +(ℓ−1) +(ℓ−1)
[Yj ⊗kθ−1 , Yk ⊗kθ−1 x+
1 ]q 1/2 + (j ↔ k) = q
1/2
[Yj , Yk ]⊗kθ−2 x+
1 + (j ↔ k).

+(l−1) +(l−1)
But the expression on the right-hand side is zero since [Yj , Yk ] = 0, so the
induction step is established in this case. The other two cases are similar; we omit the
details.
ˆ n+1 ). If
We have thus proved that formulas (2) and (3) define a representation of Uq (sl
f : M → M is a homomorphism of Ĥℓ (q )–modules, we define F(f ) : F(M ) → F(M ′ ) by
′ 2

F(f )(m⊗v) = f (m)⊗v.

ˆ n+1 )–modules is completely
The proof that F(f ) is a well–defined homomorphism of Uq (sl
straightforward. It is now obvious that F is a functor between the appropriate categories
of representations.

4.3
Assume for the remainder of the proof that ℓ ≤ n. To prove that F is an equivalence, we
must prove that
ˆ n+1 )–module W of type 1 which is of level ℓ as a
(a) every finite–dimensional Uq (sl
Uq (sln+1 )–module is isomorphic to F(M ) for some Ĥℓ (q 2 )–module M ;

11
(b) F is bijective on sets of morphisms.
(See [11], p.91.)
To prove (a), note that by Proposition 4.1, we may assume that W = J (M ) for some
Hℓ (q 2 )–module M . We shall reconstruct the action of the yj±1 on M from the known
action of x±0 and k0 on W .
We need the following lemma.

Lemma (a) Let M be a finite–dimensional Hℓ (q 2 )–module, and let v ∈ V ⊗ℓ . The linear

map M → J (M ) given by m → m⊗v is injective if v has non–zero component in each
isotypical component of J (M ).
(b) If {v1 , . . . , vn+1 } is the standard basis of V , i1 , . . . , iℓ ∈ {1, . . . , n + 1} are distinct, and
v = vi1 ⊗ · · · ⊗viℓ , then V ⊗ℓ = Uq (sln+1 ).v. In particular, v satisfies the condition in part
(a).

Proof. Part (a) follows easily from Proposition 4.1, and part (b) is elementary. ✷

4.4
For 1 ≤ j ≤ n, let

v(j) = v2 ⊗ · · · ⊗vj ⊗vn+1 ⊗vj+1 ⊗ · · · ⊗vℓ ,

w(j) = v2 ⊗ · · · ⊗vj ⊗v1 ⊗vj+1 ⊗ · · · ⊗vℓ .
(j) (j)
Let wτ be the result of permuting the factors of w(j) by τ ∈ Sℓ . Since {wτ }τ ∈Sℓ clearly
spans the subspace of V ⊗ℓ of weight λ , we get, for any m ∈ M ,
ℓ

x− (j)
mτ ⊗wτ(j)
X
0 .(m⊗v ) =
τ ∈Sℓ

(j)
for some mτ ∈ M . By (1), wτ is a (non–zero) scalar multiple of σ.w(j) for some σ ∈
Hℓ (q 2 ) (depending on τ ). It follows that

x− (j) ′
0 .(m⊗v ) = m ⊗w
(j)

for some m′ ∈ M . By Lemma 4.3, there exists α− j ∈ EndC

′ −
I (M ) such that m = αj (m) for
all m ∈ M . By a similar argument, there exists α+
j ∈ EndCI (M ) such that
x+
0 .(m⊗vn−ℓ+2 ⊗ · · · ⊗vn−ℓ+j ⊗v1 ⊗vn−ℓ+j+1 ⊗ · · · ⊗vn )
= α+
j (m)⊗vn−ℓ+2 ⊗ · · · ⊗vn−ℓ+j ⊗vn+1 ⊗vn−ℓ+j+1 ⊗ · · · ⊗vn

for all m ∈ M .

12
4.5
We need to prove the following lemma. The proof of the theorem itself continues in Section
4.6.

Lemma For all m ∈ M , v ∈ V ⊗ℓ , we have

ℓ
x± α± ±
X
0 .(m⊗v) = j (m)⊗Yj .v.
j=1

Proof. Let v = vi1 ⊗ · · · ⊗viℓ . If {i1 , . . . , iℓ } ⊂ {1, . . . , n}, it is clear that x− 0 .(m⊗v) = 0,
since ǫi1 + . . . + ǫiℓ + ǫ1 + . . . + ǫn cannot be a weight of V . ⊗ ℓ

Let r ≥ 0, s ≥ 1, 1 ≤ j1 < j2 < . . . < jr ≤ ℓ, 1 ≤ j ′ 1 < j ′ 2 < . . . < j ′ s ≤ ℓ, and

assume that {j1 , . . . , jℓ } ∩ {j ′ 1 , . . . , j ′ s } = ∅. Write j = (j1 , . . . , jr ), j′ = (j ′ 1 , . . . , j ′ s ), and
let V (j,j ) be the subspace of V ⊗ℓ spanned by vectors which have v1 in positions j1 , . . . , jr ,
′

vn+1 in positions j ′ 1 , . . . , j ′ s , and vectors from {v2 , . . . , vn } in the remaining positions. We

′
shall prove the lemma when v ∈ V (j,j ) for all such j, j′ in two steps:
(i) for s = 1, by induction on r;
(ii) for all r, by induction on s.
Observe that, by Lemma 4.3 (b) applied to the subalgebra of Uq (sln+1 ) generated by
the x± ±1 ′
i , ki for i ∈ {2, . . . , n}, to prove Lemma 4.5 for all v ∈ V (j,j ) , it suffices to prove it
′
for one v ∈ V (j,j ) with the property that no vector from the set {v2 , . . . , vn } is repeated.
(Note that such vectors v exist since ℓ + 1 − r − s ≤ ℓ ≤ n.)
Proof of Step (i). If r = 0 (and s = 1), there is nothing to prove, for we can take
v = v2 ⊗ · · · ⊗vj ′ 1 ⊗vn+1 ⊗vj ′ 1 +1 ⊗ · · · ⊗vℓ and use the definition of α−
j ′ 1 . Assume that the
′
result holds for r − 1, and let j̃ = (j1 , . . . , jr−1 ). Let v′ ∈ V (j̃,j ) have v2 in the jrth position,
and distinct vectors from {v3 , . . . , vn } in the remaining positions. Then,

v = x+ ′
1 .v .

Let v′′ (resp. v′′′ ) be the element obtained from v′ by replacing vn+1 by v1 (resp. v2 by
v1 ). We then get, for all m ∈ M ,

x− + − ′
0 .(m⊗v) = x1 x0 .(m⊗v )
′
= q |{t<r|jt <j 1 }| α− ⊗jr −1 ⊗x+ ⊗kℓ−jr ).v′′
j ′ 1 (m)⊗(1 1 1
′ δjr <j ′
= q |{t<r|jt <j 1 }| q 1 α−
j ′ 1 (m)⊗v
′′′

′
= q |{t≤r|jt <j 1 }| α−
j ′ 1 (m)⊗v
′′′

= α− −
j ′ 1 (m)⊗Yj ′ 1 .v.

13
 ′
Proof of Step (ii). Assume that the result holds for all v ∈ V (j,j ) with fewer than s
′
vn+1 s. It suffices, as in step 1, to prove the result for one element v ∈ V (j,j ) which has
distinct entries from {v3 , . . . , vn } in the remaining positions. Fix such a v and let v′ be
the element obtained from v by replacing vn+1 in positions j1′ and j2′ by vn . Then,
(x−
n)
2
v= .v′ .
q + q −1
Using a quantized Serre relation we get
(x− 2 −
n ) x0
x− − − − ′
0 .(m⊗v) = xn x0 xn .(m⊗v ) − .(m⊗v′ ).
q + q −1
Since x− ′th ′th ′
n operates in the j 1 and j 2 positions in v , we obtain, using the induction
hypothesis,
2 − s
(x−n ) x0 ′ 2
α− −
X
.(m⊗v ) = q jk′ (m)⊗Yj ′ k .v.
q + q −1 k=3
On the other hand,
x− ′ ′′ −1 ′′′
n .(m⊗v ) = m⊗v + q m⊗v ,
where v′′ (resp. v′′′ ) is obtained from v′ by replacing the vn in its j ′th ′th
1 position (resp. j 2
position) by vn+1 . Using the induction hypothesis, we get
x− − ′
α− − ′′ −1
α− − ′′′
X X
0 xn .(m⊗v ) = j ′ k (m)⊗Yj ′ k .v + q j ′ k (m)⊗Yj ′ k .v .
k6=2 k6=1

Noting that v′′′ has vn only in the j ′th

2 position, we find that
x− α− − ′′ − − ′ 2
α− − ′
X X
n. j ′ k (m)⊗Yj ′ k .v = αj ′ 1 (m)⊗Yj ′ 1 .v + q j ′ k (m)⊗Yj ′ k .v .
k6=2 k>2

Similarly,
x− α− − ′′′
α− − ′
X X
n. j ′ k (m)⊗Yj ′ k .v = q j ′ k (m)⊗Yj ′ k .v .
k6=1 k6=1
Combining these computations we obtain finally,
x− 2
α− −
X
0 .(m⊗v) = −q j ′ k (m)⊗Yj ′ k .v
k>2
2
α− − − −
X
+ q j ′ k (m)⊗Yj ′ k .v + αj ′ 1 (m)⊗Yj ′ 1 .v
k>2

α− −
X
+ j ′ k (m)⊗Yj ′ k .v
k6=1
s
α− −
X
= j ′ k (m)⊗Yj ′ k .v,
k=1

14
as required.
This proves Lemma 4.5 for x− +
0 . The proof for x0 is similar. ✷

4.6
We can now complete the proof of the theorem. We show that setting

[Link]±1 = α±
j (m)

defines a right Ĥℓ (q 2 )–module structure on M , extending its Hℓ (q 2 )–module structure.

We have to check the following relations:
(i) yj yj−1 = yj−1 yj = 1,
(ii) yj yk = yk yj ,
(iii) q 2 yj+1 = σj yj σj .
Relations (i) and (ii) are proved by computing both sides of the equation

k0 − k0−1
!
[x+ −
0 , x0 ].(m⊗v) = .(m⊗v),
q − q −1

where in the first case we take v to be a vector with vn+1 in the j th place and vn−ℓ+2 , . . . , vn
in the remaining places (in any order), and in the second case we take v to be a vector
with v1 in the j th place, vn+1 in the kth place and distinct vectors from {v2 , . . . , vn } in the
other places. Notice that since the central element c ∈ Uq (sl ˆ n+1 ) acts as 1 on W we have
−1 ⊗ℓ
k0 .(m⊗v) = m⊗(kθ ) .v.
To prove (iii), let v = vi1 ⊗ · · · ⊗viℓ ∈ V ⊗ℓ , where ij = 2, ij+1 = 1, and the remaining
ik are distinct elements from {3, . . . , n} (this is possible since ℓ ≤ n). Let v′ be the result
of replacing v1 in the ithj+1 position in v by vn+1 . Since

Ř(v2 ⊗vn+1 ) = qvn+1 ⊗v2 , Ř(v1 ⊗v2 ) = qv2 ⊗v1 ,

we have, for all m ∈ M ,

m.σj yj σj ⊗v′ = qm.σj yj ⊗v′′ ,
where v′′ is obtained from v′ by interchanging its j th and (j + 1)th factors, which

= qx+ ′′′
0 .(m.σj ⊗v ),

where v′′′ is obtained from v by interchanging its j th and (j + 1)th factors, which

= q 2 x+ 2 ′
0 .(m⊗v) = q [Link]+1 ⊗v .

15
Since v′ has distinct components, Lemma 4.3 implies that

q 2 [Link]+1 = m.σj yj σj ,

for all m ∈ M .
The proof that W ∼ ˆ n+1 )–modules is now complete. To show that F
= F(M ) as Uq (sl
is an equivalence, we must prove that it is bijective on sets of morphisms. Injectivity of
F follows from that of J . For surjectivity, let F : F(M ) → F(M ′ ) be a homomorphism
ˆ n+1 )–modules. By Proposition 4.1 again, F = J (f ) for some homomorphism
of Uq (sl
f : M → M ′ of Hℓ (q 2 )–modules. The fact that F commutes with the action of x+ 0 gives

ℓ ℓ
f ([Link] )⊗Yj+ .v = f (m).yj ⊗Yj+ .v
X X

j=1 j=1

for all m ∈ M , v ∈ V ⊗ℓ . By choosing v suitably, as in the preceding part of the proof, it

is easy to see that this implies
f ([Link] ) = f (m).yj
for all j = 1, . . . , ℓ. ✷

4.7
The functor F is clearly one of C
I–linear categories. The following result shows that it also
ˆ n+1 )–modules.
captures part of the tensor structure of the category of Uq (sl

Proposition Let Mi be a finite–dimensional Ĥℓi (q 2 )–module, i = 1, 2. Then, there is a

ˆ n+1 )–modules
canonical isomorphsm of Uq (sl

ˆ 2) ∼
F(M1 ⊙M = F(M1 )⊗F(M2 ).

Proof. We recall the following elementary fact: if ι : B → A is a homomorphism of unital

associative algebras over a field, M is a right B–module, W a left A–module, and W |B is
W regarded as a left B–module via ι, there is a canonical isomorphism of vector spaces

indA ∼
O
B (M )⊗W = M W |B .
B

In fact, the isomorphism is given by

(m⊗a)⊗w → m⊗aw (m ∈ M, a ∈ A, w ∈ W ).

16
 Taking A = Hℓ1 +ℓ2 (q 2 ), B = Hℓ1 (q 2 )⊗Hℓ2 (q 2 ), ι = ιℓ1 ,ℓ2 , M = M1 ⊗M2 and W =
V ⊗ℓ1 +ℓ2 , and noting that W ∼
= (V ⊗ℓ1 )⊗(V ⊗ℓ2 ) as an Hℓ1 (q 2 )⊗Hℓ2 (q 2 )–module, we get a
canonical isomorphism of vector spaces

(V ⊗ℓ1 ⊗V ⊗ℓ2 ).
O
ˆ 2 ) → (M1 ⊗M2 )
F(M1 ⊙M
Hℓ1 (q 2 )⊗Hℓ2 (q 2 )

The right–hand side is obviously isomorphic to F(M1 )⊗F(M2 ) as a vector space. To

complete the proof, one must check that the resulting isomorphism of vector spaces
ˆ 2 ) → F(M1 )⊗F(M2 )
F(M1 ⊙M
ˆ n+1 ). This is completely straightforward. ✷
commutes with the action of Uq (sl

4.8
We analyze the functor F in more detail in Section 7, when the parametrizations of the
ˆ n+1 ) have been described.
finite–dimensional irreducible representations of Ĥℓ (q 2 ) and Uq (sl
The following result is, however, easy to prove now.. Recall the universal Ĥℓ (q 2 )–modules
ˆ n+1 )–modules V (a) defined in Sections 2.4 and 3.4 respectively.
Ma and the Uq (sl
I × )ℓ , ℓ, n ≥ 1. There is a canonical isomorphism
Proposition Let a = (a1 , . . . , aℓ ) ∈ (C
ˆ n+1 )–modules
of Uq (sl
F(Ma ) ∼ = V (a1 )⊗ · · · ⊗V (aℓ ).
Proof. As an Hℓ (q 2 )–module, Ma is the right regular representation. It follows that the
map
V ⊗ℓ → J (Ma ) (9)
given by v → 1⊗v is an isomorphism of Uq (sln+1 )–modules. Now,
ℓ ℓ
x+ [Link] ⊗Yj+ .v = ( aj Yj+ ).v.
X X
0 .(1⊗v) =
j=1 j=1

On the other hand,

ℓ
∆(ℓ) (x+
X
1⊗j−1 ⊗x+ ⊗ℓ−j
0)= 0 ⊗k0
j=1

acts on V (a1 )⊗ · · · ⊗V (aℓ ) as

ℓ ℓ
1⊗j−1 ⊗aj x− −1 ⊗ℓ−j
aj Yj+ .
X X
θ ⊗(kθ ) =
j=1 j=1

17
One checks in the same way that the map in (9) commutes with the action of x−
0 and k0 .
✷
Corollary Let 1 ≤ ℓ ≤ n.
(a) Every finite–dimensional Uq (sl ˆ n+1 )–module of type 1 and level ℓ as a Uq (sln+1 )–module
I ×.
is isomorphic to a quotient of V (a1 )⊗ · · · ⊗V (aℓ ), for some a1 , . . . , aℓ ∈ C
ˆ n+1 )–module iff
I × , then V (a1 )⊗ · · · ⊗V (aℓ ) is reducible as a Uq (sl
(b) If a1 , . . . , aℓ ∈ C
2
aj = q ak for some j, k.
Proof. This follows immediately from Proposition 3.4 and the fact that F is an equivalence
of categories. ✷

4.9
Theorem 4.2 has a classical analogue, in which Uq (sl ˆ n+1 ) is replaced by (the universal
enveloping algebra of) the affine Lie algebra sl ˆ n+1 , and Ĥℓ (q 2 ) by (the group algebra
of) the affine Weyl group of GLℓ (CI), i.e. the semi–direct product Sℓ ×Z ˜ ℓ , where Sℓ acts
on the additive group Zℓ by permuting the coordinates. We recall that sl ˆ n+1 is the
universal central extension (with one–dimensional centre) of the Lie algebra L(sln+1 ) of
Laurent polynomial maps C I × → sln+1 . We identify sln+1 with the subalgebra of L(sln+1 )
consisting of the constant maps.
˜ ℓ –modules
Theorem There is a functor F0 from the category of finite–dimensional Sℓ ×Z
to the category of finite–dimensional L(sln+1 )–modules which are of level ℓ as sln+1 –
modules, defined as follows. One takes

V0⊗ℓ
O
F0 (M ) = M
Sℓ
with the action of f ∈ L(sln+1 ) given by
ℓ
[Link] ⊗(1⊗j−1 ⊗f (1)⊗1⊗ℓ−j ).v,
X
f.(m⊗v) =
j=1

where zj = (0, . . . , 0, 1, 0, . . . , 0) ∈ Zℓ ⊂ Sℓ ×Z
˜ ℓ (with 1 in the j th position). If ℓ ≤ n, F0 is
an equivalence. ✷
The proof of this theorem is analogous to (but simpler than) that of Theorem 4.2.
Remark The finite–dimensional irreducible representations of L(sln+1 ) were classified
I × , there is a homomorphism of Lie algebras
in [1]. For any a ∈ C

eva0 : L(sln+1 ) → sln+1

18
given by eva0 (f ) = f (a). If W is an irreducible sln+1 –module, pulling back by eva0 gives an
irreducible L(sln+1 )–module W (a). It is not difficult to prove that every finite–dimensional
irreducible representation of L(sln+1 ) is isomorphic to a tensor product of W (a)s.
It is easy to identify the corresponding representations of Sℓ ×Z ˜ ℓ . There is a homo-
morphism
˜ 0a : Sℓ ×Z
ev ˜ ℓ → Sℓ
which is the identity on Sℓ and for which ev˜ 0a (zj ) = a for all j. If M is an irreducible
0 ˜ ℓ –module M (a). It is clear
˜ a gives an irreducible Sℓ ×Z
Sℓ –module, pulling M back by ev
that
F0 (M (a)) ∼
= FS(M )(a).
˜ ℓ –module is isomorphic to a
By Theorem 4.9, every finite–dimensional irreducible Sℓ ×Z
Zelevinsky tensor product of M (a)s.

5 Evaluation Representations
ˆ n+1 ) and Ĥℓ (q 2 ) of the representations of
In this section, we construct analogues for Uq (sl
˜ ℓ described in Remark 4.9, and show how these representations are related
sln+1 and Sℓ ×Z
by the functor F.

5.1
The following result was observed by Cherednik [4]. The proof is straightforward.

I × , there exists a homomorphism ev

Proposition For every a ∈ C ˜ a : Ĥℓ (q 2 ) → Hℓ (q 2 )
such that

ev
˜ a (σi ) = σi ,
˜ a (yj ) = aq −2(j−1) σj−1 σj−2 . . . σ2 σ12 σ2 . . . σj−1 ,
ev

for i = 1, . . . ℓ − 1, j = 1, . . . , ℓ. ✷

Note that ev˜ a can be characterized as the unique homomorphism Ĥℓ (q 2 ) → Hℓ (q 2 )

which is the identity on Hℓ (q 2 ) ⊂ Ĥℓ (q 2 ) and which maps y1 to a.
If M is any Hℓ (q 2 )–module, pulling back M by ev ˜ a gives an Ĥℓ (q 2 )–module M (a)
which is isomorphic to M as an Hℓ (q 2 )–module.

19
5.2
ˆ n+1 → sln+1 . To
In [7], Jimbo defined a quantum analogue of the homomorphism eva0 : sl
describe it, we need the following

Definition Uq (gln+1 ) is the associative algebra over C I with generators x±

i , i = 1, . . . , n,
±1
tr , r = 1, . . . , n + 1, and the following defining relations:

tr t−1
r = 1 = t−1
r tr ,
tr ts = ts tr ,
± −1
t r xi t r = q ±(δr,i −δr,i+1 ) x±
i ,
[x± ± ±
i , [xj , xi ]q 1/2 ]q 1/2 = 0 if |i − j| = 1,
[x± ±
i , xj ] = 0 if |i − j| > 1,
ki − ki−1
[x+ −
i , xj ] = δij ,
q − q −1

where ki = ti t−1
i+1 .

The algebra Uq (gln+1 ) has a Hopf algebra structure, but we shall not make any use of
it.
Note that there is an obvious homomorphism Uq (sln+1 ) → Uq (gln+1 ).

5.3
Fix an (n+1)th root q 1/(n+1) of q. We shall say that a finite–dimensional Uq (gln+1 )–module
W is of type 1 if
(a) W is of type 1 regarded as a Uq (sln+1 )–module,
(b) the tr act semisimply on W with eigenvalues which are integer powers of q 1/(n+1) ,
(c) t1 t2 . . . tn+1 acts as 1 on W .
It is easy to see that restriction to Uq (sln+1 ) is an equivalence from the category
of finite–dimensional Uq (gln+1 )–modules of type 1 to the category of finite–dimensional
Uq (sln+1 )–modules of type 1. In particular the functor J of Proposition 4.1 may be viewed
as taking values in the category of finite–dimensional Uq (gln+1 )–modules of type 1.

5.4
We can now state

20
 ˆ n+1 ) →
I × , there exists a homomorphism eva : Uq (sl
Proposition ([7]) For any a ∈ C
Uq (gln+1 ) such that
eva (x± ±
i ) = xi , eva (ki ) = ki , i = 1, . . . , n,
eva (k0 ) = (k1 k2 . . . kn )−1 ,
eva (x±
0 ) = ( ± 1)
(n−1) ∓(n+1)/2 ±1
q a (t1 tn+1 )±1 [x∓ ∓ ∓ ∓
n , [xn−1 , . . . , [x2 , x1 ]q 1/2 . . .]q 1/2 ]q 1/2 . ✷

If W is a Uq (sln+1 )–module of type 1, we may regard W as a Uq (gln+1 )–module by

(5.3). The pull-back of W by the homomorphism eva is a Uq (sl ˆ n+1 )–module which we
denote by W (a).

5.5
The main result of this section is
Theorem Let 1 ≤ ℓ ≤ n, and let M be a finite–dimensional right Hℓ (q 2 )–module. Then
ˆ n+1 )–modules,
there is a canonical isomorphism of Uq (sl

F(M (q −2ℓ/(n+1) a)) ∼

= J (M )(a),
I ×.
for all a ∈ C
Proof. By Theorem 4.2 we know that J (M )(a) ∼= F(N ), for some Ĥℓ (q 2 )–module N which
2
is isomorphic to M as an Hℓ (q )–module. It suffices to prove that y1 acts as the scalar a
on N . To prove this, we compute the action of x+0 on m⊗v1 ⊗vn−ℓ+2 ⊗vn−ℓ+3 ⊗ · · · ⊗vn ∈
F(N ) in two different ways, for all m ∈ M .
First, by the definition of F, we have
x+
0 .(m⊗v1 ⊗vn−ℓ+2 ⊗vn−ℓ+3 ⊗ · · · ⊗vn ) = m.y1 ⊗vn+1 ⊗vn−ℓ+2 ⊗vn−ℓ+3 ⊗ · · · ⊗vn . (10)
− − −
On the other hand, let fn = [x−
n , [xn−1 , . . . , [x2 , x1 ]q 1/2 . . .]q 1/2 ]q 1/2 . Then,

x+ +
0 .(m⊗v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn ) = m⊗eva (x0 ).(v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn )
= aq −(n−1)/2−2ℓ/(n+1) m⊗fn .(v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn ).(11)
We prove by induction on n that
fn .(v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn ) = q (n−1)/2 vn+1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn .
The result is obvious if n = 1. Assuming it for n − 1, note that fn = [x−
n , fn−1 ]q 1/2 , so by
the induction hypothesis,
fn .(v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn ) = q (n−1)/2 x−
n .(vn ⊗vn−ℓ+2 ⊗ · · · ⊗vn )
−q −1/2 fn−1 .(v1 ⊗vn−ℓ+2 ⊗ . . . ⊗vn−1 ⊗vn+1 ).

21
Since x−
i .vn+1 = 0 for 1 ≤ i ≤ n − 1, we see that

fn−1 .((v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn−1 )⊗vn+1 ) = (fn−1 .(v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn−1 ))⊗vn+1
= q (n−2)/2 vn ⊗vn−ℓ+2 ⊗ · · · ⊗vn−1 ⊗vn+1 ,

by the induction hypothesis again. Hence,

fn .(v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn ) = q (n−1)/2 (vn+1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn + q −1 vn ⊗vn−ℓ+2 ⊗ · · · ⊗vn−1 ⊗vn+1 )
−q (n−3)/2 vn ⊗vn−ℓ+2 ⊗ · · · ⊗vn−1 ⊗vn+1
= q (n−1)/2 vn+1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn ,

as required.
Hence, from (11), we obtain

x+
0 .(m⊗v1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn ) = aq
−2ℓ/(n+1)
m⊗vn+1 ⊗vn−ℓ+2 ⊗ · · · ⊗vn .

Comparing with (10), and using Lemma 4.3, we obtain

m.y1 = aq −2ℓ/(n+1) m

for all m ∈ M . ✷

6 ˆ n+1)–modules
Classification of finite–dimensional Uq (sl
6.1
The finite–dimensional irreducible Uq (slˆ n+1 )–modules of type 1 were classified in [2], [3].
ˆ n+1 ) given in [6]. By
To describe this result, we need an alternative presentation of Uq (sl
Proposition 2.6, we need only consider the quantum loop algebra Lq (sln+1 ), the quotient
ˆ n+1 ) by the two sided ideal generated by c − 1.
of Uq (sl
±
Proposition Lq (sln+1 ) is isomorphic as an algebra to the algebra A with generators Xi,r
(i ∈ {1, . . . , n}, r ∈ Z), Hi,r (i ∈ {1, . . . , n}, r ∈ Z\{0}), and Ki±1 , (i ∈ {1, . . . , n}), and
the following defining relations:

Ki Ki−1 = 1 = Ki−1 Ki ,
Ki Hj,r = Hj,r Ki ,
[Hi,r , Hj,s ] = 0 ,
Ki Xj± Ki−1 = q ±aij Xj± ,

22
 ± 1 ±
[Hi,r , Xj,s ] = ± [raij ]q Xj,r+s ,
r
± ± ± ± ± ± ± ±
Xi,r+1 Xj,s − q ±aij Xj,s Xi,r+1 = q ±aij Xi,r Xj,s+1 − Xj,s+1 Xi,r ,

+ −
Φ+ −
i,r+s − Φi,r+s
[Xi,r , Xj,s ] = δij ,
q − q −1
p
p
 
k ± ± ± ± ±
X X
(−1) Xi,r . . . Xi,r Xj,s Xi,r . . . Xi,r = 0 , i 6= j,
π∈Sp k=0
k q
π(1) π(k) π(k+1) π(p)

for all sequences (r1 , . . . , rp ) ∈ Zp , where p = 1 − aij and the elements Φ±

i,r are determined
by equating coefficients of powers of u in the formal power series
∞ ∞
Φ± ±r
= Ki±1 exp( ± (q − q −1 ) Hi,±s u±s ).
X X
i,±r u
r=0 s=1

The isomorphism f : Lq (sln+1 ) → A is given by

f (x± ± ±1 ±1
i ) = Xi,0 , f (ki ) = Ki ,

for i ∈ {1, . . . , n}, and

f (k0±1 ) = (K1 K2 . . . Kn )∓1 ,

f (x+
0 ) = (−1)
m−1 −(n−3)/2
q −
[Xn,0 −
, [Xn−1,0 −
, . . . , [Xm+1,0 −
, [X1,0 −
, . . . , [Xm−1,0 −
, Xm,1 ]q1/2 . . .]q1/2 f (k0 ),
f (x− −1 + + + + + +
0 ) = µf (k0 )[Xn,0 , [Xn−1,0 , . . . , [Xm+1,0 , [X1,0 , . . . , [Xm−1,0 , Xm,−1 ]q 1/2 . . .]q 1/2 ,

I × is determined by
where µ ∈ C

f (k0 ) − f (k0−1 )
[f (x+ −
0 ), f (x0 )] = . ✷
q − q −1

Remark Using the relations in A, it is not difficult to see that the isomorphism f is
independent of the choice of m ∈ {1, 2, . . . , n}.

6.2
The following result is proved in [2], [3].

Proposition Let W be a finite–dimensional irreducible Lq (sln+1 )–module of type 1.

Then,
(a) W is generated by a vector w0 satisfying

23
 +
Xi,r .w0 = 0, Φ± ±
i,r .w0 = φi,r w0

for all i ∈ {1, . . . , n}, r ∈ Z, and some φ± i,r ∈ CI.

(b) There exist unique monic polynomials P1 (u), . . . , Pn (u) (depending on W ) such that
the φ±
i,r satisfy
∞ −2 ∞
deg Pi Pi (q u)
φ+ r
φ− −r
X X
i,r u = q = i,r u ,
r=0
P i (u) r=0
in the sense that the left and right-hand sides are the Laurent expansions of the middle
term about 0 and ∞ respectively. Assigning to W the corresponding n–tuple of polynomials
defines a one to one correspondence between the isomorphism classes of finite–dimensional
irreducible Lq (sln+1 )–modules of type 1 and the set of n–tuples of monic polynomials in
one variable u. ✷
A consequence of this proposition is:
Corollary Let W be a finite-dimensional irreducible representation of Uq (sl ˆ n+1 ) with
associated polynomials Pi . Set λ = (deg P1 , . . . , deg Pn ). Then W contains the irreducible
Uq (sln+1 )–module V (λ) with multiplicity one. Further, if V (µ) is any other Uq (sln+1 )–
module occurring in W , then λ ≥ µ. ✷

6.3
The next proposition can be proved by studying the action of the comultiplication ∆ of
ˆ n+1 ) on the generators X + etc., as in [2].
Uq (sl i,r

Proposition Let W and W ′ be two finite–dimensional irreducible Uq (sl ˆ n+1 )–modules

′ ′
with associated monic polynomials Pi and Pi , i = 1, . . . , n. Let w0 and w0 be the generating
vectors of W and W ′ as in Proposition 6.2. Then, in W ⊗W ′ we have
+
Xi,r .(w0 ⊗w0′ ) = 0

for all i ∈ {1, . . . , n}, r ∈ Z. Further, w0 ⊗w0′ is a common eigenvector of the Φ±

i,r with
eigenvalues given as in Proposition 6.2 (b) by the polynomials Pi Pi′ . ✷
This result suggests the following
I × , the irreducible finite–dimensional representation
Definition If i ∈ {1, . . . , n}, a ∈ C
ˆ
of Uq (sln+1 ) with associated polynomials
(
u−a if j = i,
Pj (u) =
1 otherwise,

24
 ˆ n+1 ) with parameter a, and is denoted
is called the ith fundamental representation of Uq (sl
by V (λi , a).

Remark Note that it follows from Corollary 6.2 that V (λi , a) ∼

= V (λi ) as Uq (sln+1 )–
modules.

6.4
We shall need the following result in Section 7.

Lemma Let vλm be the Uq (sln+1 )–highest weight vector in V (λm , a), where m ∈ {1, . . . , n},
a∈CI × . Then,

x+
0 .vλm = (−1)
m−1 −1 − −
a xn xn−1 . . . x− − −
m+1 x1 . . . xm .vλm .

Proof. By Proposition 2.3 and the preceding remark, we know that the weight spaces of
V (λm , a) as a Uq (sln+1 )–module are all one–dimensional and that the weights are precisely
ǫi1 + ǫi2 + . . . + ǫim , 1 ≤ i1 < i2 < . . . < im ≤ n + 1. It follows that
−
Xm,1 .vλm = bx−
m .vλm

for some b ∈ C
I. Using Proposition 6.1 we get

Φ+ −1
m,1 .vλm = b(q − q )vλm .

Hence, from Proposition 6.2 (b), we get

q(q −2 u − a) = (u − a)(q + b(q − q −1 )u + O(u2 )),

so that b = a−1 . Finally, from Proposition 6.1 again, we find that

x+
0 .vλm = (−1)
m−1 −1 − −
a xn xn−1 . . . x− − −
m+1 x1 . . . xm .vλm . ✷

7 Comparison with results of Zelevinsky and Rogawski

In this section, we describe a parametrization, due to Zelevinsky [13] and Rogawski [12],
of the finite–dimensional irreducible Ĥℓ (q 2 )–modules. We then relate this, via the functor
F defined in Theorem 4.2, to the parametrization of the finite–dimensional irreducible
ˆ n+1 )–modules given in Section 6.
Uq (sl

25
7.1
∼C
Since q is not a root of unity, Hℓ (q 2 ) = I[Sℓ ] as an algebra. It follows that the finite–
2
dimensional Hℓ (q )–modules are completely reducible and that the irreducibles are in one
to one correspondence with the partitions of ℓ. We now describe this correspondence.
The defining relations of Hℓ (q 2 ) imply that, if w ∈ Sℓ and if
w = τ i1 τ i2 . . . τ ik
is any reduced expression for w in terms of the simple transpositions τi = (i, i + 1), the
element
σw = σi1 σi2 . . . σik ∈ Hℓ (q 2 )
depends only on w.
Let ≤ be the Bruhat order on Sℓ , and for w′ ≤ w, let Pw′ ,w (q) be the Kazhdan–Lusztig
polynomial (see [9]). Define elements Cw ∈ Hℓ (q 2 ) by
′ ′
Cw = q ℓ(w) (−1)ℓ(w)−ℓ(w ) q −2ℓ(w ) Pw,w′ (q −2 )σw .
X

w ′ ≤w

We write Ci for Cτi . Note that Ci = q −1 σi − q. It is known (see [9]) that {Cw }w∈W is a
basis of Hℓ (q 2 ), and that
Cw σi = −Cw if wτi < w. (12)
Let ℓ = ℓ1 + ℓ2 + · · · + ℓp be a partition π of ℓ, with each ℓr > 0, and let Sℓπ be
the subgroup Sℓ1 × Sℓ2 × · · · × Sℓp of Sℓ which fixes π. Let wr be the longest element of
the subgroup Sℓr , i.e. the permutation which reverses the order of (ℓ1 + ℓ2 + · · · + ℓr−1 +
1, . . . , ℓ1 +· · ·+ℓr ), and set wπ = w1 w2 . . . wp . Let Iπ be the right ideal in Hℓ (q 2 ) generated
by Cwπ .
Proposition ([12]) For every partition π of ℓ, Iπ has a unique irreducible quotient Jπ
in which Cwπ has non–zero image. Conversely, every finite–dimensional irreducible right
Hℓ (q 2 )-module is isomorphic to some Jπ . ✷

7.2
Using Jimbo’s functor J , we can compare this parametrization of the finite–dimensional
irreducible representations of Hℓ (q 2 ) with that of the representations of Uq (sln+1 ) given
by their highest weights.
Proposition Let 1 ≤ ℓ ≤ n and let ℓ1 + ℓ2 + · · · + ℓp be a partition π of ℓ. Then,
J (Jπ ) ∼
= V (λℓ1 + λℓ2 + · · · + λℓp )
as Uq (sln+1 )–modules.

26
Proof. We need the following lemma, which follows from (1).

Lemma Let π be as in the prceding proposition, and let 1 ≤ i ≤ ℓ be such that i 6= rj=1 ℓj
P

for any 1 ≤ r < p. Let v ∈ V ⊗ℓ have vr ⊗vs in the ith and (i + 1)th positions, and let v′
be the result of interchanging the vectors in these positions. Then, in J (Jπ ), we have

−1
 −q Cwπ ⊗v
 if r < s,
′
Cwπ ⊗v = −qCwπ ⊗v if r > s, ✷
 0

if r = s.

Returning to the proof of the proposition, note that the weight space of V ⊗ℓ of weight
λℓ1 + λℓ2 + · · · + λℓp is spanned by the permutations of the vector

vπ = v1 ⊗v2 · · · ⊗vℓ1 ⊗v1 ⊗v2 ⊗ · · · ⊗vℓ2 ⊗v1 · · · ⊗vℓp .

By Proposition 4.1, there exists a partition π ′ of ℓ, say ℓ = ℓ′1 + ℓ′2 + · · · + ℓ′r , such that

J (Jπ′ ) ∼
= V (λℓ1 + · · · + λℓr ). (13)

By the lemma, if vi1 ⊗ · · · ⊗viℓ is any permutation of vπ ,

Cwπ ⊗vi1 ⊗ · · · ⊗viℓ = 0

unless the first ℓ′1 vectors in the sequence vi1 , . . . , viℓ are distinct, together with the next
ℓ′2 , . . ., and the last ℓ′r . It follows that, if ≤ is the usual lexicographic ordering on the set
of partitions of ℓ, we have π ′ ≤ π. But the map π → π ′ defined by (13) is a bijection since
J is an equivalence. Since ≤ is a total ordering it follows that this bijection is the identity
map, i.e. π ′ = π. ✷

7.3
We now turn to the representations of affine Hecke algebras. Recall the universal modules
Ma defined in Section 3.4. We begin with the following elementary result.

I × )ℓ , w ∈ Sℓ , j ∈ {1, . . . , ℓ}. Then, in Ma , we

Lemma ([12]) Let a = (a1 , . . . , aℓ ) ∈ (C
have X
Cw .yj = aw−1 (j) Cw + αw′ Cw′
w ′ <w

for some αw′ ∈ C

I. ✷

27
7.4
Following Rogawski [12] and Zelevinsky [13], we make the following definition.

Definition The segment s with centre a ∈ C I × and length |s| = k is the ordered sequence
−k+1 −k+3 k−1 × k
s = (aq , aq , . . . , aq ) ∈ (C
I ) .

If s = {s1 , s2 , . . . , sp } is any (unordered) collection of segments, and if |sr | = ℓr , then

ℓ = ℓ1 + ℓ2 + · · · + ℓp is a partition π(s) of ℓ.

Proposition ([12]) Let ℓ ≥ 1 and let s = {s1 , . . . , sp } be any collection of segments, the
I × )ℓ be the result of juxtaposing the
sum of whose lengths is ℓ. Let a = (s1 , . . . , sp ) ∈ (C
segments in s. Then,
(a) Iπ(s) is an Ĥℓ (q 2 )–submodule of Ma (this statement makes sense in view of Proposition
3.4 (b));
(b) with the Ĥℓ (q 2 )–module structure from Ma , Iπ(s) has a unique irreducible subquotient
Va in which Cwπ(s) has non–zero image.
Moreover, every finite–dimensional irreducible right Ĥℓ (q 2 )–module is isomorphic to
some Va . ✷

7.5
To prove the main result of this section, we shall need another description of Iπ(s) (we
continue to use the notation of Section 7.4). Let Σπ(s) ⊂ Sℓ be the set of transpositions
τi = (i, i + 1) for i ∈ {1, . . . , ℓ}\{ℓ1 , ℓ1 + ℓ2 , . . . , ℓ1 + · · · + ℓp−1 }. For τi ∈ Σπ(s) , let aτi be
the result of interchanging the ith and (i + 1)th components of a, and let

Aa,i : Maτi → Ma

be the map given by left multiplication by Ci (we identify Ma and Maτi with Ĥℓ (q 2 ) in
the usual way).

Proposition ([12]) With the above notation:

(a) Aa,i is a homomorphism of Ĥℓ (q 2 )–modules;
(b) regarded as an Ĥℓ (q 2 )–submodule of Ma ,
\
Iπ(s) = (image of Aa,i ). ✷
τi ∈Σπ(s)

28
7.6
We can now state the main result of this section.

Theorem Let s = {s1 , . . . , sp } be a collection of segments, the sum of whose lengths is

I × )ℓ be the result
ℓ, let ar be the centre of sr and ℓr its length, and let a = (s1 , . . . , sp ) ∈ (C
of juxtaposing s1 , . . . , sp , as in Proposition 7.4. Then, if ℓ ≤ n, F(Va ) is the irreducible
ˆ n+1 )–module defined by the polynomials
Uq (sl

(u − a−1
Y
Pi (u) = j ), i = 1, . . . , n.
{j|ℓj =i}

Proof. We first prove the result in the special case p = 1, so that a = (aq −ℓ+1 , aq −ℓ+3 , . . . , aq ℓ−1 )
(we drop the subscripts for simplicity). Note that wπ(s) = w0 , the longest element of Sℓ ,
and that Iπ(s) (= Jπ(s) = Va ) is one-dimensional and spanned by Cw0 . By Proposition 7.2,

J (Iπ(s) ) ∼
= V (λℓ ),

the highest weight vector being

vλℓ = Cw0 ⊗v1 ⊗v2 ⊗ · · · ⊗vℓ .

ˆ n+1 )–module, F(Va ) is therefore defined by the polynomials

As a Uq (sl
(
u − a′ if i = ℓ,
Pi (u) =
1 otherwise,

I × . To compute a′ , note first that, by the definition of F,

for some a′ ∈ C

x+
0 .vλℓ = Cw0 .y1 ⊗vn+1 ⊗v2 ⊗ · · · ⊗vℓ .

Since Iπ(s) is one–dimensional, Lemma 7.3 implies that

x+
0 .vλℓ = q
ℓ−1
aCw0 ⊗vn+1 ⊗v2 ⊗ · · · ⊗vℓ . (14)

On the other hand Lemma 6.4 gives

x+
0 .vλℓ = (−1)ℓ−1 (a′ )−1 x− − − − − −
n xn−1 · · · xℓ+1 x1 x2 · · · xℓ .vλℓ
= (−1)ℓ−1 (a′ )−1 (Cw0 ⊗v2 ⊗ · · · ⊗vℓ ⊗vn+1 ).

Now by (12),
Cw0 σi−1 = −Cw0 ,

29
and by (1),
vr ⊗vn+1 = q Ř−1 (vn+2 ⊗vr ), if r ≤ n.
Hence,
Cw0 ⊗v2 ⊗ · · · ⊗vℓ ⊗vn+1 = (−1)ℓ−1 q ℓ−1 Cw0 ⊗vn+1 ⊗v2 ⊗ · · · ⊗vℓ ,
and so
x+
0 .vλℓ = q
ℓ−1 ′ −1
(a ) Cw0 ⊗vn+1 ⊗v2 ⊗ · · · ⊗vℓ .
Comparing with (14) gives a′ = a−1 . (It follows from the proof of Proposition 7.2 that
Cw0 ⊗vn+1 ⊗v2 ⊗ · · · ⊗vℓ 6= 0.)
Suppose now that r is arbitrary. From Proposition 7.5 (b),
\
F(Iπ(s) ) = (image of F(Aa,i )). (15)
τi ∈Σπ(s)

To compute F(Aa,i ), note that v 7→ 1⊗v defines an isomorphism of Uq (sln+1 )–modules

V ⊗ℓ → F(Ma ), and that

F(Aa,i )(1⊗v) = Ci ⊗v = 1⊗Ci .v.

It follows that
F(Aa,i ) = q −1 Ři − q ∈ EndC ⊗ℓ
I (V ).
From (15) and the r = 1 case, it follows that

F(Iπ(s) ) = V (λℓ1 , a−1 −1

1 )⊗ · · · ⊗V (λℓp , ap ).

By Propositions 7.2 and 7.4 (b), F(Va ) is the unique irreducible subquotient of F(Iπ(s) )
in which the tensor product of the highest weight vectors in the V (λℓr , a−1
r ) has non–
zero image. The theorem now follows from the multiplicativity of the polynomials in
Proposition 6.3. ✷

30
References

1. Chari, V. and Pressley, A.N., New unitary representations of loop groups, Math. Ann.
275 (1986), 87–104.
2. Chari V. and Pressley A.N., Quantum affine algebras, Comm. Math. Phys. 142 (1991),
261–283.
3. Chari V. and Pressley A.N., A Guide to Quantum Groups, Cambridge University Press,
Cambridge (in press).
4. Cherednik, I. V., A new interpretation of Gelfand-Tzetlin bases, Duke Math. J. 54
(1987), 563–577.
5. Drinfeld, V. G., Degenerate affine Hecke algebras and Yangians, Func. Anal. Appl. 20
(1986), 62–64.
6. Drinfeld, V. G., A new realization of Yangians and quantized affine algebras, Sov.
Math. Dokl. 36 (1988), 212–216.
7. Jimbo, M., A q–analogue of U (gl(N +1)), Hecke algebra and the Yang–Baxter equation,
Lett. Math. Phys. 11 (1986), 247–252.
8. Kac, V. G., Infinite Dimensional Lie Algebras, Birkhäuser, Boston, 1983.
9. Kazhdan, D. and Lusztig, G., Representations of Coxeter groups and Hecke algebras,
Invent. Math. 53 (1979), 165–184.
10. Lusztig, G., Quantum deformations of certain simple modules over enveloping algebras,
Adv. Math. 70 (1988), 237-249.
11. MacLane, S., Categories for the Working Mathematician, Springer, 1971.
12. Rogawski, J.D., On modules over the Hecke algebra of a p–adic group, Invent. Math.
79 (1985), 443–465.
13. Zelevinsky, A.V., Induced representations of reductive p–adic groups II. On irreducible
representations of GLn , Ann. Sci. Ec. Norm. Sup. 4e Sér. 13 (1980), 165–210.

Vyjayanthi Chari,
Department of Mathematics,
University of California,
Riverside, CA 92521, U.S.A.

Andrew Pressley,
Department of Mathematics,
King’s College, Strand,
London WC2R 2LS, U.K.

31