Journal of Algebra 278 (2004) 134–147

www.elsevier.com/locate/jalgebra

Defining relations of the elliptic Lie algebra

(1,1)
of type A1
Tadayoshi Takebayashi
Department of Mathematical Science, School of Science and Engineering, Waseda University,
Ohkubo Shinjuku-ku, Tokyo, 169-8555, Japan
Received 21 July 2003
Available online 2 April 2004
Communicated by Peter Littelmann

Abstract
(1,1)
We give the defining relations of the elliptic Lie algebra of type A1 , which has finite generators
and infinite relations in terms of the elliptic Dynkin diagram, and isomorphic to the 2-toroidal Lie
algebra of type A1 .
 2004 Elsevier Inc. All rights reserved.

(1,1)
Keywords: Elliptic Lie algebra; Toroidal (Lie) algebra; Type A1

1. Introduction

A toroidal Lie algebra u is the universal central extension of a Lie algebra g ⊗

C[t1±1 , t2±1 , . . . , tm
±1 ], where g is one of the finite-dimensional simple Lie algebras over
±1 ±1 ±1 ] is the ring of Laurent polynomials in m variables t , . . . , t
C and C[t1 , t2 , . . . , tm 1 m
over C. Let A = (αi , αj )1
i,j
n be any simply-laced finite Cartan matrix of rank n
2,
and A[m] = (αi , αj )1
i,j
n+m be any m-fold affinization of A, then P. Slodowy [5]
introduced intersection matrix algebra im(A[m] ), and after that the following isomorphic
has been established: u  im(A[m] ) [4]. In vertex operator’s method, K. Saito and D. Yoshii
[7] constructed a Lie algebra g(Φ) attached to any m-extended homogeneous root system
Φ as a certain subalgebra of VQ(Φ) /DVQ(Φ) , here VQ is the lattice vertex algebra attached
to a lattice Q and D is the derivation (studied by Borcherds [1]). Especially, in the case
of 2-extended root system (also called elliptic root system), g(Φ) is isomorphic to the 2-

E-mail address: [email protected].

0021-8693/$ – see front matter  2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jalgebra.2004.02.009
 T. Takebayashi / Journal of Algebra 278 (2004) 134–147 135

toroidal algebra gtor . The 2-toroidal algebra gtor is presented in terms of infinite generators
and infinite relations by Moody, Rao, and Yokonuma [2,3]. In particular, for the simply-
laced elliptic root system Φ, Saito and Yoshii also defined a Lie algebra ẽ(Γ (Φ, G))
by Chevalley generators and generalized Serre relations attached to Γ (Φ, G), which are
finite generators and finite relations, where Γ (Φ, G) is the simply-laced elliptic Dynkin
diagram and (Φ, G) is a pair consisting of an elliptic root system Φ with a marking G.
Then ẽ(Γ (Φ, G)) is called the elliptic Lie algebra and isomorphic to g̃(Φ), where g̃(Φ)
is generated by g(Φ) and a nondegenerate h̃ extended from Cartan subalgebra h adding
two degree derivations. Further, the elliptic Lie algebra ẽ(Γ (Φ, G)) has been described in
terms of the extended elliptic Cartan matrix by the author [8]. In this paper, we describe
the 2-toroidal algebra gtor associated to the Lie algebra g = sl2 (which is not simply-laced)
(1,1)
in terms of the finite generators associated to the elliptic Dynkin diagram of type A1 ,
(1,1) (1,1)
and call it the elliptic Lie algebra g(A1 ). As a result, the elliptic Lie algebra g(A1 ) is
described by finite generators and infinite relations.

2. Definition of the 2-toroidal algebra gtor (g = sl2 )

Let g be a simple finite-dimensional Lie algebra over C of rank n with a nondegenerate

symmetric invariant bilinear form (· , ·). Let R := C[s ±1 , t ±1 ] be the ring of Laurent
polynomials in 2 variables s and t over C. Let ΩR1 := Rds ⊕ Rdt be an R-module with
generators df for ∀f ∈ R, and the relation d(fg) = f d(g) + gd(f ). Let ¯· : ΩR → ΩR1 /dR
be the canonical projection, in which there holds the relation d(fg) = f d(g) + gd(f ) = 0.
Then the Lie algebra gtor = g ⊗ R ⊕ (ΩR1 /dR), with Lie bracket [X ⊗ f, Y ⊗ g] = [X, Y ] ⊗
fg + (X, Y )(df )g, [c, gtor] = 0, for ∀c ∈ ΩR1 /dR, is the universal central extension of
g ⊗ R, and is called the 2-toroidal algebra. Let us recall the result by Moody et al. [2].
Let A = (aij )ni,j =0 be a Cartan matrix of affine type Xn(1) , ∆ be the root system associated
to A, and Q be the free Z-module on generators α0 , . . . , αn of ∆.

Definition 2.1 (Moody, Rao, and Yokonuma [2]). A Lie algebra τ (A) over C is defined by
the following presentation.
Generators: Hi,k , Ei,k , Fi,k (0  i  n, k ∈ Z), and c.
Relations: (0) c is central,

(I) [Hi,k , Hj,l ] = k(αi∨ | αj∨ )δk+l,0 c,

(II) (i) [Hi,k , Ej,l ] = (αi∨ | αj∨ )Ej,k+l ,
(ii) [Hi,k , Fj,l ] = −(αi∨ | αj∨ )Fj,k+l ,
(III) (i) [Ei,k , Ei,l ] = [Fi,k , Fi,l ] = 0,
(ii) [Ei,k , Fj,l ] = δij {Hi,k+l + 2((kδk+l,0)/(αi | αi ))c},
(iii) (ad Ei,0 )−aji +1 Ej,k = 0 (i
= j ),
(iv) (ad Fi,0 )−aji +1 Fj,k = 0 (i
= j ),

where 0  i, j  n, k, l ∈ Z, (· | ·) is a Z-valued symmetric bilinear form on Q normalized

by (α0 | α0 ) = 2, and αi∨ := (2αi )/(αi | αi ).
136 T. Takebayashi / Journal of Algebra 278 (2004) 134–147

Then the following has been established.

Proposition 2.2 [2,3]. τ (A)  gtor .

(1)
2 −2 
Definition 2.3. In the case of type A1 , its Cartan matrix is (aij ) = −2 2
and its 2-
toroidal algebra gtor (g = sl2 ) is presented as follows.
Generators: Hi,k , Ei,k , Fi,k , (i = 0, 1, k ∈ Z), and c.
Relations: (0) c is central,

(I) [Hi,k , Hj,l ] = kaij δk+l,0 c,

(II) (i) [Hi,k , Ej,l ] = aij Ej,k+l ,
(ii) [Hi,k , Fj,l ] = −aij Fj,k+l ,
(III) (i) [Ei,k , Ei,l ] = [Fi,k , Fi,l ] = 0,
(ii) [Ei,k , Fj,l ] = δij {Hi,k+l + kδk+l,0 c},
(iii) (ad Ei,0 )3 Ej,k = 0 (i
= j ),
(iv) (ad Fi,0 )3 Fj,k = 0 (i
= j ).

Then we have the following proposition.

Proposition 2.4. The 2-toroidal algebra gtor (g = sl2 ) can be described by finite generators
as follows.
Generators: Ei,0 , Fi,0 , Hi,p (i = 0, 1, p = 0, ±1), and c.
Relations: (0) c is central,

(A) [Hi,r , Hj,s ] = raij δr+s,0c,

(B) (i) [Hi,0 , Ej,0 ] = aij Ej,0 ,
(ii) [Hi,0 , Fj,0 ] = −aij Fj,0 ,
(iii) [Hi,p , Xi,0 ] = −[Hj,p , Xi,0 ] (i
= j ),
(iv) ad Hi,1 ad Hj,−1 Xj,0 = 2aij Xj,0 ,
(C) (i) ad Xi,0 (ad Hi,p )k Xi,0 = 0 (k
1),
(ii) [Ei,0 , Fj,0 ] = δij Hi,0 ,
(iii) ad Fi,0 ad Hi,p Ei,0 = −2Hi,p ,
(iv) ad Xi,0 (ad Hj,p )k Yj,0 = 0 (i
= j, k
1),
(v) (ad Xi,0 )3 (ad Hj,p )k Xj,0 = 0 (i
= j, k
0),
(D) (i) ad Hi,p ad Ej,0 (ad Hj,q )k Fj,0 = 0 (k
2),
(ii) ad Ei,0 (ad Hi,p )k Fi,0 = (−1)k+1 ad Fi,0 (ad Hi,p )k Ei,0 (k
2),
(iii) ad Xj,0 ad Xi,0 (ad Hi,p )k Yi,0 = 2(ad Hi,p )k Xj,0 (i
= j, k
1),

where i, j = 0, 1, r, s = 0, ±1, p, q = ±1, and Xi,0

= Yi,0 ∈ {Ei,0 , Fi,0 }.

Proof. The proof that the relations in Proposition 2.4 can be obtained from the relations in
Definition 2.3 is easy. Conversely, for a positive integer k and p = ±1, we set
 k   k 
1 1

Hi,pk = ad Ei,0 − ad Hi,p Fi,0 = − ad Fi,0 ad Hi,p Ei,0 by (D)(ii) ,
2 2
 T. Takebayashi / Journal of Algebra 278 (2004) 134–147 137

 k  k
1 1
Ei,pk = ad Hi,p Ei,0 , and Fi,pk = − ad Hi,p Fi,0 ,
2 2

then we show that the relations in Definition 2.3 can be obtained from the relations in
Proposition 2.4. At first, we note that the following relations hold:
  k 
1
for k
0, [Hi,1 , Hj,k ] = Hi,1 , ad Ej,0 − ad Hj,1 Fj,0 = 0
2


by (C)(iii) and (D)(i) ,
  k 
1
for k
2, [Hi,1 , Hj,−k ] = Hi,1 , ad Ej,0 − ad Hj,−1 Fj,0 = 0
2


by (D)(i) .

Therefore, considering the relation (A), we have

[Hi,1 , Hj,k ] = aij δk+1,0 c. (2.1)

Similarly, we have

[Hi,−1 , Hj,k ] = aij δk−1,0 c. (2.2)

We prove in turn of (III), (II), and (I), because we use each result.

Proof of (III)(i). [Ei,k , Ei,l ] = 0.

(i) The case of k
0, l
0.

We assume the case of k + l  n, and show the case of k + l = n + 1 by an induction
on k + l. If k = 0, it is obtained from (C)(i), so we assume k
1.
 k  l 
1 1
l.h.s. = ad Hi,1 Ei,0 , ad Hi,1 Ei,0
2 2
  k−1   l 
1 1 1
= Hi,1 , ad Hi,1 Ei,0 , ad Hi,1 Ei,0
2 2 2
 l   k−1 
1 1 1
=− ad Hi,1 Ei,0 , Hi,1 , ad Hi,1 Ei,0
2 2 2
 k−1  l  
1 1 1
− ad Hi,1 Ei,0 , ad Hi,1 Ei,0 , Hi,1
2 2 2
 
1
= [Ei,l+1 , Ei,k−1 ] − [Ei,k−1 , Ei,l ], Hi,1
2
= −[Ei,k−1 , Ei,l+1 ] (by the induction hypothesis)
138 T. Takebayashi / Journal of Algebra 278 (2004) 134–147

= (−1)k [Ei,0 , Ei,k+l ] (by iterating the same procedure)


= 0 by (C)(i) .

(ii) The case of k > 0, l < 0, i.e., [Ei,k , Ei,−l ] = 0, with k > 0, l > 0.
At first we show the following relation:

ad Hi,1 (ad Hj,−1 )l Xj,0 = 2aij (ad Hj,−1 )l−1 Xj,0 , (2.3)

where l > 0 and X = E, F.

The case of l = 1 is (B)(iv), and

ad Hi,1 (ad Hj,−1 )l+1 Xj,0


= ad Hj,−1 ad Hi,1 (ad Hj,−1 )l Xj,0 by (2.2) [Hi,1 , Hj,−1 ] = ai,j c

= ad Hj,−1 2aij (ad Hj,−1 )l−1 Xj,0 (by the induction hypothesis)

= 2aij (ad Hj,−1 )l Xj,0 . 2

Now we prove that [Ei,k , Ei,−l ] = 0, by the induction on k, l. The cases of k, l = 1, 2

are easily checked and,

 k+1  l 
1 1
ad Hi,1 Ei,0 , ad Hi,−1 Ei,0
2 2
  k   l 
1 1 1
= Hi,1 , ad Hi,1 Ei,0 , ad Hi,−1 Ei,0
2 2 2
 l   k 
1 1 1
=− ad Hi,−1 Ei,0 , Hi,1 , ad Hi,1 Ei,0
2 2 2
 k  l  
1 1 1
− ad Hi,1 Ei,0 , ad Hi,−1 Ei,0 , Hi,1
2 2 2
  l   k 
1 1 1
= Hi,1 , ad Hi,−1 Ei,0 , ad Hi,1 Ei,0 (by the induction hypothesis)
2 2 2
 l−1  k 
1 1

= ad Hi,−1 Ei,0 , ad Hi,1 Ei,0 by (2.3)
2 2
= 0 (by the induction hypothesis).

The other cases are similarly proved. 2

 T. Takebayashi / Journal of Algebra 278 (2004) 134–147 139

Proof of (III)(ii). [Ei,k , Fj,l ] = δij {Hi,k+l + kδk+l,0 c}.

(i) The case of i = j and k
0, l
0.

Assuming the case of k + l  n, we show the case of k + l = n + 1.
We note that the case of k = 0 is trivial, so we assume that k
1.
 k  l 
1 1
l.h.s. = ad Hi,1 Ei,0 , − ad Hi,1 Fi,0
2 2
  k−1   l 
1 1 1
= Hi,1 , ad Hi,1 Ei,0 , − ad Hi,1 Fi,0
2 2 2
 l   k−1 
1 1 1
=− − ad Hi,1 Fi,0 , Hi,1 , ad Hi,1 Ei,0
2 2 2
 k−1  l  
1 1 1
− ad Hi,1 Ei,0 , − ad Hi,1 Fi,0 , Hi,1
2 2 2
 
1
= [Ei,k−1 , Fi,l+1 ] − [Ei,k−1 , Fi,l ], Hi,1
2
 
1
= [Ei,k−1 , Fi,l+1 ] − Hi,k+l−1 + (k − 1)δk+l−1,0c, Hi,1
2
(by the induction hypothesis)


= [Ei,k−1 , Fi,l+1 ] by (2.1)
= [Ei,0 , Fi,k+l ] (by iterating the same procedure)
= Hi,k+l .

(ii) The case of i

= j and k
0, l
0.
If k = 0, it is easy, so we assume that k
1.

  k−1   l 
1 1 1
l.h.s. = Hi,1 , ad Hi,1 Ei,0 , − ad Hj,1 Fj,0
2 2 2
 l   k−1 
1 1 1
=− − ad Hj,1 Fj,0 , Hi,1 , ad Hi,1 Ei,0
2 2 2
 k−1  l  
1 1 1
− ad Hi,1 Ei,0 , − ad Hj,1 Fj,0 , Hi,1
2 2 2
  l  k−1 
1 1 1
= − − ad Hi,1 − ad Hj,1 Fj,0 , ad Hi,1 Ei,0
2 2 2
 
1
− [Ei,k−1 , Fj,l ], Hi,1
2
140 T. Takebayashi / Journal of Algebra 278 (2004) 134–147

 l+1  k−1 
1 1
= − ad Hj,1 Fj,0 , ad Hi,1 Ei,0 (by the induction hypothesis)
2 2
 k+l
1
= −[Ei,k−1 , Fj,l+1 ] = (−1)k [Ei,0 , Fj,k+l ] = (−1)k ad Ei,0 − ad Hj,1 Fj,0
2


= 0 by (C)(iv) .

(iii) The case of k > 0, l < 0, i.e. , [Ei,k , Fj,−l ] = δij {Hi,k−l + kδk−l,0 c}, with k > 0,
l > 0.

 k  l 
1 1
l.h.s. = ad Hi,1 Ei,0 , − ad Hj,−1 Fj,0
2 2
  k−1   l 
1 1 1
= Hi,1 , ad Hi,1 Ei,0 , − ad Hj,−1 Fj,0
2 2 2
 l   k−1 
1 1 1
=− − ad Hj,−1 Fj,0 , Hi,1 , ad Hi,1 Ei,0
2 2 2
 k−1  l  
1 1 1
− ad Hi,1 Ei,0 , − ad Hj,−1 Fj,0 , Hi,1
2 2 2
  l  k−1 
1 1 1
= − − ad Hi,1 − ad Hj,−1 Fj,0 , ad Hi,1 Ei,0
2 2 2
 
1
− [Ei,k−1 , Fj,−l ], Hi,1
2
  l−1  k−1 
aij 1 1
=− − ad Hj,−1 Fj,0 , ad Hi,1 Ei,0
2 2 2
 
1

− [Ei,k−1 , Fj,−l ], Hi,1 by (2.3)
2
 
aij   1
= − [Fj,−l+1 , Ei,k−1 ] − δij Hi,k−1−l + (k − 1)δk−1−l,0c , Hi,1
2 2
(by the induction hypothesis)
aij   1
= δij Hi,k−l + (k − 1)δk−l,0 c + δij aij δk−l,0 c
2 2


by the induction hypothesis and (2.1)
= δij {Hi,k−l + kδk−l,0 c}.

The other cases are similarly proved. 2

 T. Takebayashi / Journal of Algebra 278 (2004) 134–147 141

Proof of (II)(i). [Hi,k , Ej,l ] = aij Ej,k+l .

(i) The case of k
0, l
0.

If l = 0, then for i
= j ,
  k 
1
[Hi,k , Ej,0 ] = − ad Ej,0 ad Ei,0 − ad Hi,1 Fi,0
2


= −2(− ad Hi,1 ) Ej,0 by (D)(iii)
k


= −2(ad Hj,1 )k Ej,0 by (B)(iii)
= aij Ej,k ,

and for i = j ,
  k 
1
[Hi,k , Ei,0 ] = − ad Ei,0 − ad Fi,0 ad Hi,1 Ei,0
2
 k
1
= (ad Fi,0 ad Ei,0 + ad Hi,0 ) ad Hi,1 Ei,0
2
 k
1

=2 ad Hi,1 Ei,0 by (C)(i)
2
= 2Ei,k ,

so we have [Hi,k , Ej,0 ] = aij Ej,k .

If l
1, then
  l     l−1 
1 1 1
l.h.s. = Hi,k , ad Hj,1 Ej,0 = Hi,k , Hj,1 , ad Hj,1 Ej,0
2 2 2
 l−1     l−1 
1 1 1 1
=− ad Hj,1 Ej,0 , Hi,k , Hj,1 − Hj,1 , Hj,1 Ej,0 , Hi,k
2 2 2 2
  
1

= − Hj,1 , Ej,l−1 , Hi,k by (2.1)
2
 
1
= Hj,1 , aij Ej,k+l−1 (by the induction hypothesis)
2
= aij Ej,k+l .

(ii) The case of k > 0, l < 0, i.e ., [Hi,k , Ej,−l ] = aij Ej,k−l , with k > 0, l > 0.
At first we show the relation:

[Hi,−1 , Ej,k ] = aij Ej,k−1 . (2.4)

142 T. Takebayashi / Journal of Algebra 278 (2004) 134–147

If k > 0, then
  k    k
1 1 1
l.h.s. = Hi,−1 , ad Hj,1 Ej,0 = 2 ad Hi,−1 ad Hj,1 Ej,0
2 2 2
 k−1
1

= aij ad Hj,1 Ej,0 by (2.3)
2
= aij Ej,k−1 .

If k < 0, then we show [Hi,−1 , Ej,−k ] = aij Ej,−k−1 , with k > 0.

  k    k
1 1 1
l.h.s. = Hi,−1 , ad Hj,−1 Ej,0 = 2 ad Hi,−1 ad Hj,−1 Ej,0
2 2 2
 k+1
1

= aij ad Hj,−1 Ej,0 by (B)(iii)
2
= aij Ej,−k−1 . 2

Now we prove (ii) by the induction on l.

If l = 1, then
  
1
l.h.s. = Hi,k , Hj,−1 , Ej,0
2
    
1 1
= − Ej,0 , Hi,k , Hj,−1 − Hj,−1 , [Ej,0 , Hi,k ]
2 2
 
aij 1

= − Ej,0 , − δk−1,0 c + [Hj,−1 , aij Ej,k ] by (2.2) and the above case (i)
2 2


= aij Ej,k−1 by (2.4) .

Next,
  l+1 
1
Hi,k , ad Hj,−1 Ej,0
2
   l 
1 1
= Hi,k , Hj,−1 , ad Hj,−1 Ej,0
2 2
 l     l 
1 1 1 1
=− ad Hj,−1 Ej,0 , Hi,k , Hj,−1 − Hj,−1 , ad Hj,−1 Ej,0 , Hi,k
2 2 2 2
 l     l 
1 aij 1 1
=− ad Hj,−1 Ej,0 , − δk−1,0 c + Hj,−1 , Hi,k , ad Hj,−1 Ej,0
2 2 2 2


by (2.2)
 T. Takebayashi / Journal of Algebra 278 (2004) 134–147 143

1
= [Hj,−1 , aij Ej,k−l ] (by the induction hypothesis)
2


= aij Ej,k−l−1 by (2.4) .

The other cases are similarly proved.

Proof of (I). [Hi,k , Hj,l ] = kaij δk+l,0 c.

(i) The case of k
0, l
0.

  k 
1

l.h.s. = ad Ei,0 − ad Hi,1 Fi,0 , Hj,l = − [Hj,l , Ei,0 ], Fi,k − [Fi,k , Hj,l ], Ei,0
2
= −[aij Ei,l , Fi,k ] − [aij Fi,k+l , Ei,0 ] = −aij Hi,k+l + aij Hi,k+l = 0.

(ii) The case of k
0, l  0, i.e., [Hi,k , Hj,−l ] = kaij δk−l,0 c, with k
0, l
0.

From (2.1) and (2.2), we assume k
1, l
1.
  k 
1
l.h.s. = ad Ei,0 − ad Hi,1 Fi,0 , Hj,−l
2

= − [Hj,−l , Ei,0 ], Fi,k − [Fi,k , Hj,−l ], Ei,0


= [−aij Ei,−l , Fi,k ] − [aij Fi,k−l , Ei,0 ] = −aij Hi,k−l + (−l)δk−l,0 c + aij Hi,k−l
= kaij δk−l,0 c. 2

Therefore, we complete the proof. 2

We can describe gtor (g = sl2 ) by choosing another generators as follows.

Proposition 2.5. The algebra gtor (g = sl2 ) can be presented as follows.

Generators: Hi,0 , Ei,0 , Ei,1 , Fi,0 , Fi,−1 (i = 0, 1), and c.
Relations: (0) c is central,

(I) (i) [Hi,0 , Hj,0 ] = 0,

(ii) ad Hi,0 ad Ej,0 Fj,−1 = 0,
(iii) ad Hi,0 ad Ej,1 Fj,0 = 0,
(II) (i) [Hi,0 , Ej,r ] = aij Ej,r ,
(ii) [Hi,0 , Fj,s ] = −aij Fj,s ,
(iii) ad Ei,0 ad Fi,−1 Fj,0 = 2Fj,−1 (i
= j ),
(iv) ad Ei,1 ad Fi,0 Fj,−1 = 2Fj,0 (i
= j ),
(v) ad Ei,0 ad Ej,1 Fj,0 = 2Ei,1 (i
= j ),
(vi) ad Ei,1 ad Ej,0 Fj,−1 = 2Ei,0 (i
= j ),
(vii) ad Ei,0 ad Ej,0 Fj,−1 = − ad Ei,0 ad Ei,0 Fi,−1 (i
= j ),
(viii) ad Ei,1 ad Ej,1 Fj,0 = − ad Ei,1 ad Ei,1 Fi,0 (i
= j ),
144 T. Takebayashi / Journal of Algebra 278 (2004) 134–147

(ix) ad Fi,0 ad Ej,1 Fj,0 = − ad Fi,0 ad Ei,1 Fi,0 (i

= j ),
(x) ad Fi,−1 ad Ej,0 Fj,−1 = − ad Fi,−1 ad Ei,0 Fi,−1 (i
= j ),
(III) (i) [Ei,0 , Ei,1 ] = [Fi,0 , Fi,−1 ] = 0,
(ii) ad Xi,0 (ad Ai,p )k Xi,0 = 0 (k
1),
(iii) [Ei,r , Fj,s ] = 0 (i
= j ),
(iv) [Ei,r , Fi,−r ] = Hi,0 + δr,1 c,
(v) ad Xi,0 (ad Ai,p )k Yj,0 = 0 (i
= j, k
1),
(vi) (ad Xi,0 )3 (ad Ai,p )k Xj,0 = 0 (i
= j, k
0),
(IV) (i) ad Ai,p ad Ej,0 (ad Aj,q )k Fj,0 = 0 (k
2),
(ii) ad Ei,0 (ad Ai,p )k Fi,0 = (−1)k+1 ad Fi,0 (ad Ai,p )k Ei,0 (k
2),
(iii) ad Xj,0 ad Xi,0 (ad Ai,p )k Yi,0 = 2(ad Ai,p )k Xj,0 (i
= j, k
1),

where i, j, r = 0, 1, s = 0, −1, p, q = ±1, ad Ai,1 = ad Ei,1 ad Fi,0 − ad Fi,0 ad Ei,1 ,

ad Ai,−1 = ad Ei,0 ad Fi,−1 − ad Fi,−1 ad Ei,0 , and Xi,0
= Yi,0 ∈ {Ei,0 , Fi,0 }.

Proof. We set Hi,1 = ad Ei,1 Fi,0 , Hi,−1 = ad Ei,0 Fi,−1 , then we have Ei,1 = 12 ad Hi,1 Ei,0 ,
Fi,−1 = − 12 ad Hi,−1 Fi,0 , and

ad Hi,1 = ad Ei,1 ad Fi,0 − ad Fi,0 ad Ei,1 ,

ad Hi,−1 = ad Ei,0 ad Fi,−1 − ad Fi,−1 ad Ei,0 .

Proof.

1 1
1


ad Hi,1 Ei,0 = [Ei,1 , Fi,0 ], Ei,0 = − [Fi,0 , Ei,0 ], Ei,1 by (III)(i)
2 2 2
1

= [Hi,0 , Ei,1 ] = Ei,1 by (II)(i) .
2

For any X ∈ gtor ,

ad Hi,1 X = [ad Ei,1 Fi,0 , X] = − [X, Ei,1 ], Fi,0 − [Fi,0 , X], Ei,1
= (− ad Fi,0 ad Ei,1 + ad Ei,1 ad Fi,0 )X

therefore, we have

ad Hi,1 = ad Ei,1 ad Fi,0 − ad Fi,0 ad Ei,1 . 2

It is easy to see that the relations in Proposition 2.5 can be obtained from Proposition 2.4,
since Proposition 2.4 gives a presentation of gtor in Definition 2.3. We show that the
relations in Proposition 2.4 can be obtained from Proposition 2.5.

Proof of (A). [Hi,1 , Hj,−1 ] = aij c.

At first, we have the following relations:
 T. Takebayashi / Journal of Algebra 278 (2004) 134–147 145

[ ad Ei,1 Fi,0 , Ej,0 ] = aij Ej,1 , (2.5)

[ad Ei,1 Fi,0 , Fj,−1 ] = −aij Fj,0 . (2.6)

Proof of (2.5). When i

= j, it is trivial from (II)(v), and when i = j,


[ ad Ei,1 Fi,0 , Ei,0 ] = − ad Ei,0 ad Ei,1 Fi,0 = − ad Ei,1 Hi,0 by (III)(i) and (iv)


= 2Ei,1 by (II)(i) . 2

Using these, we prove (A).

l.h.s. = [ad Ei,1 Fi,0 , ad Ej,0 Fj,−1 ]

= − Fj,−1 , [ad Ei,1 Fi,0 , Ej,0 ] − Ej,0 , [Fj,−1 , ad Ei,1 Fi,0 ]


= −aij [Fj,−1 , Ej,1 ] − aij [Ej,0 , Fj,0 ] by (2.5) and (2.6)


= −aij (−Hj,0 − c) − aij Hj,0 by (III)(iv)
= aij c.

The other cases of (A) are similarly proved. 2

Proof of (B)(iii). [Hi,p , Xi,0 ] = −[Hj,p , Xi,0 ] (i

= j ).
(i) The case of [Hi,1 , Ei,0 ] = −[Hj,1 , Ei,0 ] (i
= j ).


l.h.s. = [ad Ei,1 Fi,0 , Ei,0 ] = 2Ei,1 by (2.5) ,


r.h.s. = −[ad Ej,1 Fj,0 , Ei,0 ] = 2Ei,1 by (2.6) .

(ii) The case of [Hi,−1 , Ei,0 ] = −[Hj,−1 , Ei,0 ] (i

= j ).
It is obtained from the relation (II)(vii), and the other cases are similarly obtained. 2

Proof of (B)(iv). ad Hi,1 ad Hj,−1 Ej,0 = 2aij Ej,0 .

Noting the relations [Hi,1 , Ej,0 ] = aij Ej,1 , [Hi,−1 , Ei,1 ] = 2Ei,0 ,


l.h.s. = ad Hj,−1 ad Hi,1 Ej,0 by (A)
= aij Hj,−1 Ej,1 = 2aij Ej,0 . 2

Proof of (C)(iii). ad Fi,0 ad Hi,1 Ei,0 = −2Hi,1 .

l.h.s. = 2 ad Fi,0 Ei,1 = −2Hi,1 = r.h.s. 2

The relations containing (ad Hi,p )k can be obtained from the relations containing
(ad Ai,p )k in Proposition 2.5. Therefore, we complete the proof. 2
146 T. Takebayashi / Journal of Algebra 278 (2004) 134–147

(1,1)
3. Defining relations of the elliptic Lie algebra of type A1

From the point of view of the elliptic Lie algebra, we describe the 2-toroidal algebra gtor
(1,1)
(g = sl2 ). The elliptic Dynkin diagram of the elliptic root system of type A1 is given
by [6]

α0∗
∗

α1





α0
∞

α1

where α0 and α1 are the simple roots of the affine root system of type A(1) 1 . Associated
to this diagram, we choose Chevalley generators of gtor (g = sl2 ) such as e0 = E0,0 ,
e1 = E1,0 , e0∗ = E0,1 , e1∗ = E1,1 , f0 = F0,0 , f1 = F1,0 , f0∗ = F0,−1 , f1∗ = F1,−1 , h0 =
H0,0 , h1 = H1,0, and c, then we call gtor the elliptic Lie algebra g(A1(1,1)) of type A1(1,1),
and from Proposition 2.5, we have the following theorem.

Theorem 3.1. The elliptic Lie algebra g(A1(1,1)) is described as follows.

Generators: ei , ei∗ , fi , fi∗ , hi (i = 0, 1), and c.
Relations: (0) c is central,

(I) (i) [hi , hj ] = 0,

(ii) [hi , [Xj , Yj∗ ]] = 0,
(II) (i) [hi , ej ] = aij ej , [hi , ej∗ ] = aij ej∗ ,
(ii) [hi , fj ] = −aij fj , [hi , fj∗ ] = −aij fj∗ ,
(iii) [ei , [fi∗ , fj ]] = 2fj∗ (i
= j ),
(iv) [ei∗ , [fi , fj∗ ]] = 2fj (i
= j ),
(v) [ei , [ej∗ , fj ]] = 2ei∗ (i
= j ),
(vi) [ei∗ , [ej , fj∗ ]] = 2ei (i
= j ),
(vii) [Xi , [Xj , Yj∗ ]] = −[Xi , [Xi , Yi∗ ]] (i
= j ),
(viii) [Xi∗ , [Xj∗ , Yj ]] = −[Xi∗ , [Xi∗ , Yi ]] (i
= j ),
(III) (i) [Xi , Xi∗ ] = 0,
(ii) ad Xi (ad Ai,p )k Xi = 0 (k
1),
(iii) [ei , fj ] = [ei , fj∗ ] = [ei∗ , fj ] = [ei∗ , fj∗ ] = 0 (i
= j ),
(iv) [ei , fi ] = hi , [ei∗ , fi∗ ] = hi + c,
(v) ad Xi (ad Ai,p )k Yj = 0 (i
= j, k
1),
(vi) (ad Xi )3 (ad Ai,p )k Xj = 0 (i
= j, k
0),
(IV) (i) ad Ai,p ad ej (ad Aj,q )k fj = 0 (k
2),
(ii) ad ei (ad Ai,p )k fi = (−1)k+1 ad fi (ad Ai,p )k ei (k
2),
(iii) ad Xj ad Xi (ad Ai,p )k Yi = 2(ad Ai,p )k Xj (i
= j, k
1),

where i, j = 0, 1, p, q = ±1, ad Ai,1 = ad ei∗ ad fi − ad fi ad ei∗ , ad Ai,−1 = ad ei ad fi∗ −

ad fi∗ ad ei , and Xi
= Yi ∈ {ei , fi }.
 T. Takebayashi / Journal of Algebra 278 (2004) 134–147 147

Further we can extend the algebra gtor to g̃tor by adding the degree operators d1 , d2 .
Then the commutation relations are given by [2]

[d1 , Ei,k ] = δi,0 Ei,k , [d1 , Fi,k ] = −δi,0 Fi,k , [d1, Hi,k ] = 0,
[d2 , Ei,k ] = kEi,k , [d2 , Fi,k ] = kFi,k , [d2 , Hi,k ] = kHi,k ,
[d1 , d2 ] = [c, d1] = [c, d2] = 0.
(1,1) (1,1)
From this, we can extend the Lie algebra g(A1 ) to g̃(A1 ) by adding d1 , d2 and we
call again it the elliptic Lie algebra in the sense of [7]. Then the commutation relations are
given as follows:

[d1 , hi ] = [d1 , c] = [d1 , d2 ] = [d2 , hi ] = [d2 , c] = 0,

[d1 , ei ] = δi,0 ei , [d1, ei∗ ] = δi,0 ei∗ , [d1, fi ] = −δi,0 fi , [d1 , fi∗ ] = −δi,0 fi∗ ,
[d2 , ei ] = 0, [d2 , ei∗ ] = ei∗ , [d2 , fi ] = 0, [d2 , fi∗ ] = −fi∗ (i = 0, 1).

References

[1] R. Borcherds, Vertex algebras, Kac–Moody algebras and the Monster, Proc. Nat. Acad. Sci. U.S.A. 83 (1986)
3068–3071.
[2] R.V. Moody, S. Eswara Rao, T. Yokonuma, Toroidal Lie algebra and vertex representation, Geom. Dedicata 35
(1990) 283–307.
[3] S. Eswara Rao, R.V. Moody, Vertex representations for n-toroidal Lie algebra and a generalization of the
Virasoro algebra, Comm. Math. Phys. 159 (1994) 239–264.
[4] S. Berman, R.V. Moody, Lie algebras graded by finite root systems and the intersection matrix algebras of
Sodowy, Invent. Math. 108 (1992) 323–347.
[5] P. Slodowy, Beyond Kac–Moody algebras and inside, Canad. Math. Soc. Proc. 5 (1986) 361–371.
[6] K. Saito, Extended affine root systems I. Coxeter transformations, Publ. Res. Inst. Math. Sci. 21 (1985) 75–
179.
[7] K. Saito, D. Yoshii, Extended affine root systems IV. Simply-laced elliptic Lie algebras, Publ. Res. Inst. Math.
Sci. 36 (2000) 385–421.
[8] T. Takebayashi, On the defining relations of the simply-laced elliptic Lie algebras, Proc. Japan Acad., Ser. A
Math. Sci. 77 (2001) 119–121.