n-Lie algebras

arXiv:0909.1419v1 [math.RA] 8 Sep 2009

Michel GOZE - Nicolas GOZE - Elisabeth REMM

[email protected]
Adress : LMIA. UHA
4 rue des Frères Lumière
F. 68093 Mulhouse Cedex
November 16, 2018

Abstract
The notion of n-ary algebras, that is vector spaces with a multiplica-
tion concerning n-arguments, n ≥ 3, became fundamental since the works
of Nambu. Here we first present general notions concerning n-ary alge-
bras and associative n-ary algebras. Then we will be interested in the
notion of n-Lie algebras, initiated by Filippov, and which is attached to
the Nambu algebras. We study the particular case of nilpotent or fili-
form n-Lie algebras to obtain a beginning of classification. This notion of
n-Lie algebra admits a natural generalization in Strong Homotopy n-Lie
algebras in which the Maurer Cartan calculus is well adapted.

This work has been presented during the 11eme Rencontre Nationale de Géométrie
Différentielle et Applications RNGDA11, Faculté des Sciences Ben-Msik, Casablanca.
This meeting was dedicated to Professor Younes BENSAID, Member of the Académie
Française de Chirurgie.

1 n-ary algebras

1.1 Basic definitions

Let K be a commutative field of characteristic zero and V a K-vector space. Let
n be in N, n ≥ 2. A n-ary algebra structure on V is given by a linear map
µ : V ⊗n → V.
We denote by (V, µ) such an algebra. Classical algebras (associative algebras,
Lie algebras, Leibniz algebras for example) are binary that is given by a 2-ary
product. In this paper, we are interested in the study of n-ary algebras for
n ≥ 3. A subalgebra of the n-ary algebra (V, µ) is a vector subspace W of V
such that the restriction of µ to W ⊗n satisfies µ(W ⊗n ) ⊂ W. In this case (W, µ)
is also a n-ary algebra.

1
Definition 1 Let (V, µ) be a n-ary algebra. An ideal of (V, µ) is a subalgebra
(I, µ) satisfying
µ(V ⊗p ⊗ I ⊗ V ⊗n−p−1 ) ⊂ I,
for all p = 0, · · · , n − 1 and where V ⊗0 ⊗ I = I ⊗ V ⊗0 = I.

Definition 2 Let (V1 , µ1 ) and (V2 , µ2 ) be n-ary algebras. A morphism of n-ary

algebras is a linear map ϕ : V1 → V2 satisfying

µ2 ◦ ϕ⊗n = ϕ ◦ µ1 .

In this case, the linear kernel Kerϕ of the morphism ϕ is an ideal of

(V1 , µ1 ). In fact, if v ∈ Kerϕ, then

ϕ(µ1 (v1 ⊗ · · · ⊗ v ⊗ · · · ⊗ vn−1 )) = µ2 (ϕ(v1 ) ⊗ · · · ⊗ ϕ(v) ⊗ · · · ⊗ ϕ(vn−1 )) = 0.

To simplify notations, we identify the linear map µ on V ⊗n with the cor-

responding n-linear map on V n . Then we write µ(v1 ⊗ · · · ⊗ vn ) as well as
µ(v1 · v2 · · · vn ).

1.2 Anticommutative n-ary algebras

Let (V, µ) be a n-ary algebra. It is called anticommutative if µ(v1 ⊗· · ·⊗vn ) = 0
whenever vi = vj for i 6= j. Since K is of characteristic 0, this is equivalent to

µ(vσ(1) ⊗ · · · ⊗ vσ(n) ) = (−1)ε(σ) µ(v1 ⊗ · · · ⊗ vn ),

for any σ in the symmetric group Σn where (−1)ε(σ) is the signum of the per-
mutation σ. If µ is an antisymmetric n-ary multiplication, we write

[v1 , · · · , vn ]

instead of µ(v1 ⊗ · · · ⊗ vn ).

1.3 Symmetric and commutative n-ary algebras

A n-ary algebra (V, µ) is called symmetric if it satisfies

µ(vσ(1) ⊗ · · · ⊗ vσ(n) ) = µ(v1 ⊗ · · · ⊗ vn ),

for all v1 , · · · , vn ∈ V and for all σ ∈ Σn . It is called commutative if

X
(−1)ε(σ) µ(vσ(1) ⊗ · · · ⊗ vσ(n) ) = 0,
σ∈Σn

for all v1 , · · · , vn ∈ V. Of course, any symmetric n-ary algebra is commutative.

2
1.4 Derivations
Let (V, µ) a n-algebra.
Definition 3 A derivation of the n-algebra (V, µ) is a linear map

D:V →V

satisfying
n
X
D(µ(v1 , · · · , vn )) = µ(v1 , · · · , D(vi ), · · · , vn ),
i=1

for any v1 , · · · , vn ∈ V .
All derivations of (V, µ) generate a subalgebra of Lie algebra gl(V ). It is called
the algebra of derivations of V and denoted by Der(V ).
Remark. For any v1 , · · · , vn−1 in V , let ad(v1 , · · · , vn−1 ) be the linear map
given by
ad(v1 , · · · , vn−1 )(v) = µ(v1 , · · · , vn−1 , v).
Then this linear map is a (inner) derivation if and only if the product µ satisfies
n
X
µ(v1 , · · · , vn−1 , µ(u1 , · · · , un )) = µ(u1 , · · · , µ(v1 , · · · , vn−1 , ui ), · · · , un ). (1)
i=1

We will study such a product for n-Lie algebras. If n = 2, this shows that the
maps ad(X) are derivations of (V, µ) if and only if the binary product satisfies

µ(v1 , µ(u1 , u2 )) = µ(µ(v1 , u1 ), u2 ) + µ(u1 , µ(v1 , v2 ))

and (V, µ) is a Leibniz algebra ([6]). Thus, for any n, a n-algebra (V, µ) satisfying
Equation (1) is called n-Leibniz algebra.

1.5 Simple, nilpotent n-ary algebras

Definition 4 A n-ary algebra (V, µ) is called simple if
• µ is not abelian (i.e µ(V · · · V ) 6= 0).
• Any ideal is isomorphic to V or is equal to 0.

We define the derived series by

 (1)
V = V,
V (k) = µ(V (k−1) , V (k−1) , V, · · · , V ),
and the lower central series by
 1
V = V,
V k = µ(V k−1 , V, V, · · · , V ).

3
Definition 5 The n-ary algebra (V, µ) is called
• Solvable if there is an integer k such that V (k) = 0.
• Nilpotent if there is an integer k such that V k = 0.
The definitions presented here are the definitions given in [15].

2 Particular classes of n-ary algebras

2.1 Associative n-ary algebras
The extension of the notion of associativity to n-ary product is not evident. The
most natural is perhaps the notion of totally associativity (see definition below).
Unfortunately, this notion is not auto-dual in the operadic point of view. This
means that it is necessary to introduce a most general notion of associativity
called the partially associativity.

Definition 6 The n-ary algebra (V, µ) is

• partially associative if µ satisfies
n−1
X
(−1)p(n−1) µ ◦ (Ip ⊗ µ ⊗ In−p−1 ) = 0, (2)
p=0

• totally associative if µ satisfies

µ ◦ (µ ⊗ In−1 ) = µ ◦ (Ip ⊗ µ ⊗ In−p−1 ), (3)

for any p = 0, · · · , n − 1.

2.2 σ−Associative n-ary algebras

We found for the first time this notion in the study of symmetric spaces with the
Jordan point of view in the work of Loos. Recently, we have studied a natural
generalization of the matrices product. This example gives directly the following
definition:
Let σ be an element of the symmetric group Σn of order n. It induces an
endomorphism of V ⊗n considering

ΦVσ (ei1 ⊗ · · · ⊗ ein ) = eiσ−1 (1) ⊗ · · · ⊗ eiσ−1 (n) .

Definition 7 The n-ary algebra (V, µ) is

• σ-partially associative if µ satisfies
n−1
X
(−1)p(n−1) (−1)npε(σ) µ ◦ (Ip ⊗ (µ ◦ ΦVσnp ) ⊗ In−p−1 ) = 0 (4)
p=0

4
 • σ-totally associative if µ satisfies

µ ◦ (µ ⊗ In−1 ) = µ ◦ (Ip ⊗ (µ ◦ ΦVσnp ) ⊗ In−p−1 ), (5)

for any p = 0, · · · , n − 1.

In particular, if σ = Id, we find the notion of partially and totally associa-

tivity again.
Examples.
• The Gerstenhaber products. Let A be a (binary) associative al-
gebra and H ∗ (A, A) its Hochschild cohomology. The space of k-cochains is
C k (A) = HomK (A⊗k , A). A graded pre-Lie algebra ⊕k C k (A) has been defined
by Gerstenhaber considering the product

•n,m : C n (A) × C m (A) → C n+m (A)

given by

(f •n,m g)(X1 ⊗ · · · ⊗ Xn+m−1 ) =

Pm (i−1)(m−1)
i=1 (−1) f (X1 ⊗ · · · ⊗ g(Xi ⊗ · · · Xi+m−1 ) ⊗ · · · ⊗ Xn+m−1 ).

A k-cochain µ satisfying µ•k,k µ = 0 provides A with a k-ary partially associative

structure.
• Some n-ary products of hypercubic matrices
These products are defined in [17]. We consider the space Tqp (E) of tensors
of q contravariant and p covariant type on a vector space E. If {e1 , · · · , en } is
a basis of E and ϕ ∈ Tqp (E), then
X j ,··· ,j
ϕ(ei1 ⊗ · · · ⊗ eip ) = Ci11,··· ,ipq ej1 ⊗ · · · ⊗ ejq .
j1 ,··· ,jq

e the tensor of Tpq (E) given by

We denote by ϕ
X j ,··· ,j
e j1 ⊗ · · · ⊗ ejq ) =
ϕ(e Ci11,··· ,ipq ei1 ⊗ · · · ⊗ eip .
i1 ,··· ,ip

The (2k + 1)-ary multiplication on Tqp (E) given by

µ(ϕ1 , · · · , ϕ2k+1 ) = ϕ1 ◦ ϕ
f2 ◦ ϕ3 ◦ · · · ◦ ϕ
g2k ◦ ϕ2k+1

is sk -totally associative where sk ∈ Σ2k+1 is the permutation

sk (1, · · · , 2k + 1) = (2k + 1, 2k, · · · , k, · · · , 2, 1).

j ,··· ,j
Identifying ϕ to its structural constants (Ci11,··· ,ipq ) and to the hypercubic matrix
(Ci1 ,··· ,ip ,j1 ,··· ,jq ) we obtain a ”natural” extension of the classical product of
matrices.

5
3 n-Lie algebras
Many notions of n-Lie algebras have been presented to generalize Lie algebras
for n-ary algebras. The first one is probably due to Filippov ([8]). These alge-
bras have been studied from an algebraic point of view (classification, simplicity,
nilpotency, representations) and because of their relations with the Nambu me-
chanic. The second one is the notion introduced with the strong homotopy
algebra point of view. In this paper we are concerned by the two approaches.
To distinguish these different definitions we will call n-Lie algebras the first one
and Lie n-algebras or sh-n-Lie algebras in the second one. In this section, we
study Filippov algebras.

Definition 8 An anticommutative n-ary algebra is a n-ary Lie algebra or sim-

pler n-Lie algebra if the following Jacobi identity holds:
n
X
[[u1 , · · · , un ], v1 , · · · , vn−1 ] = [u1 , · · · , ui−1 , [ui , v1 , · · · , vn−1 ], ui+1 , · · · , un ],
i=1

for any u1 , · · · , un , v1 , · · · , vn−1 ∈ V.

This last condition is called Jacobi identity for n-Lie algebras.

3.1 Fundamental examples

1. This example was given by Fillipov. Let A be a n-dimensional vector space
on K. Let {v1 , · · · , vn+1 } be a basis of V . The following product

[v1 , v2 , · · · , vˆi , · · · , vn+1 ] = (−1)n+1+i vi ,

for i = 1, · · · , n + 1 provides A with a n-Lie algebra structure. We denote this

algebra An+1 .
Theorem 9 If K = C, every simple n-Lie algebra is of dimension n + 1 and it
is isomorphic to An+1 .
2. Let A = K[X1 , · · · , Xn ] be the associative algebras of n indeterminates
polynomials. We consider the product

[P1 , · · · , Pn ] = Jac(P1 , · · · , Pn ),

where Jac denotes the Jacobian, that is the determinant of the Jacobian matrix
of partial derivatives of P1 , · · · , Pn . Provided with this product, A is an infinite
dimensional n-Lie algebra.
3. The Nambu brackets. It generalizes directly the previous example. Let A =
C ∞ (R3 ) be the algebra of differential functions on R3 . This algebra is considered
as classical observables on the three dimensional space R3 with coordinates
x, y, z. We consider on A the 3-product

{f1 , f2 , f3 } = Jac(f1 , f2 , f3 ).

6
This product is a 3-Lie algebra product which generalizes the usual Poisson
bracket from binary to ternary operations.
4. ([18]). Let A = K[X1 , · · · , Xn ] be the associative algebra of n indetermi-
nates polynomials. Let Ir be the linear subspace of A linearly generated by the
monomials of A of degree greater than or equal to r. Clearly Ir is a subspace of
I3 as soon as r ≥ 3. We define Jr = I3 /Ir for r > 3. For any Q1 , · · · , Qn ∈ Jr
we put
[Q1 , · · · , Qn ] = Jac(Q1 , · · · , Qn ).
This product is a n-Lie algebra bracket and Q is a finite dimensional nilpotent
n-Lie algebra.
5. Every n-Lie algebra of dimension n

• with n odd, is abelian, that is,

[e1 , · · · , en ] = 0,

• with n even, is abelian or isomorphic to

[e1 , · · · , en ] = e1 ,

where {e1 , · · · , en } is a basis of V.

In fact, we can write X

[e1 , · · · , en ] = ai e i ,
and if this product is not zero, we can consider that

[e1 , · · · , en ] = e1 .

Thus

[[e1 , · · · , en ], e2 , · · · , en ] = (−1)n−1 e1 .

But the Jacobi identity implies

[[e1 , · · · , en ], e2 , · · · , en ] = [[e1 , e2 , · · · , en ], e2 , · · · , en ] + · · ·
+[e1 , · · · , [en , e2 , · · · , en ]]
= e1 .

We obtain
(−1)n−1 e1 = e1 .

7
3.2 Nilpotent n-Lie algebras
In the first section we have defined nilpotency for general n-ary algebras. Since
any n-Lie algebra is a n-Leibniz algebra, any adjoint operator ad(v1 , · · · , vn−1 )
is a derivation.
Theorem 10 ([15]) For any finite dimensional nilpotent Lie algebras, the ad-
joint operators are nilpotent. Conversely, if the adjoint operators of the n-Lie
algebra V are nilpotent, then V is nilpotent.
Assume that V is a finite dimensional complex nilpotent n-Lie algebra. We will
generalize the notion of characteristic sequence of Lie algebras to n-Lie algebras.
We consider the set of generators of V which is isomorphic to V /V 2 .
Lemma 11
dim V /V 2 ≥ n.
Let us consider a free family {v1 , · · · , vn−1 } of n−1 vectors of V −V 2 . The opera-
tor ad(v1 , v2 , · · · , vn−1 ) is a linear nilpotent operator of V admitting v1 , · · · , vn−1
as eigenvectors associated to the eigenvalue 0. We consider now the ordered se-
quence of the similitude invariants (the dimensions of Jordan blocks ) of this
operator. It is of type (c1 , · · · , ck , 1, · · · , 1) with at least n − 1 invariant equal
to 1, corresponding to the dimension of the eigenspace generated by the eigen-
vectors vi . We assume that c1 ≥ · · · ≥ ck ≥ 0. We denote this sequence
c(v1 , · · · , vn−1 ).
Definition 12 The characteristic sequence of the nilpotent n-Lie algebra is the
sequence
c(V ) = max{c(v1 , · · · , vn−1 )},
where (v1 , · · · , vn−1 ) are n − 1 independent vectors of V − V 2 , the order relation
being the lexicographic order.
Assume that dim V = p. The possible extremal values of c(V ) are
• (1, · · · , 1) and V is an abelian n-Lie algebra,
• (p − n + 1, 1, · · · , 1). This sequence corresponds to a nilpotent operator
ad(v1 , v2 , · · · , vn−1 ) with a biggest nilindex.
Definition 13 A p-dimensional complex (or real) nilpotent n-Lie algebra is
called filiform is its characteristic sequence is equal to (p − n + 1, 1, · · · , 1).
| {z }
n−1

Examples
• We consider n = 3 and p = 4. The characteristic sequence is equal to
(2, 1, 1). Let {v1 , v2 , v3 , v4 } be a basis of V such that the characteristic
sequence of ad(v1 , v2 ) is (2, 1, 1). If {v3 v4 } is the Jordan basis of this
operator then we have
[v1 , v2 , v3 ] = v4 .
From the classification of [4], we deduce that we have obtained the only
filiform 3-Lie algebra of dimension 4.

8
 • We generalize easily this example. Let V be the p-dimensional 3-Lie alge-
bra given by 

 [X1 , X2 , X3 ] = X4 ,

[X1 , X2 , X4 ] = X5 ,

 ···

[X1 , X2 , Xp−1 ] = Xp .
It is also a filiform 3-Lie algebra. It is a model ([12]) of the filiform 3-Lie
algebras of dimension p, that is any p-dimensional filiform 3-Lie algebras
can be contracted on this algebra.
• Every filiform 5-dimensional 3-Lie algebra is isomorphic to


 [X1 , X2 , X3 ] = X4 ,

[X1 , X2 , X4 ] = X5 ,
 [X1 , X3 , X4 ] = aX5 ,


[X2 , X3 , X4 ] = bX5 .

3.3 Graded filiform n-Lie algebras

Let f be a derivation of a complex filiform n-Lie algebra V . We assume that
f is diagonalizable. The decomposition of eigenspaces of V gives a graduation
of this n-Lie algebra. We consider the maximal abelian subalgebra of Der(V )
given by the diagonalizable derivations of V which commute with f . We denote
this algebra T (f ).
Definition 14 The rank of V is the biggest dimension amongst the dimensions
of T (f ) for any diagonalizable derivation f .

Proposition 15 The rank of any filiform n-Lie algebra is equal to or smaller

than n.

Proof. We consider the model given by

[X1 , X2 , · · · , Xn−1 , Xi ] = Xi+1 ,

for i = n + 1, · · · , p − 1, with p = dim V . We can assume that X1 , X2 , · · · , Xn

are eigenvectors. If we put
f (Xt ) = λt Xt ,
for t = 1, · · · , n, then other eigenvalues are

λi = λ1 + · · · + λn−1 + λi−1

and this implies

λi = (n − i)(λ1 + · · · + λn−1 ) + λn .
Thus λ1 , · · · , λn are the independent roots of this algebra which is then of
rank n. Let V1 be any filiform n-Lie algebra of dimension p. There exists
(X1 , · · · , Xn−1 ) independent vectors in V1 − V12 such that the characteristic

9
sequence of V1 is given by the nilpotent operator ad(X1 , · · · , Xn−1 ). We consider
the corresponding Jordan basis of V1 . It satisfies

[X1 , X2 , · · · , Xn−1 , Xi ] = Xi+1

and other brackets are linear combinations of Xn+1 , · · · , Xp . Let ft be the

endomorphism given by ft (Xl ) = Xl if 1 ≤ l ≤ n and ft (Xl ) = tXl for n + 1 ≤
l ≤ p. This endomorphism generates a contraction of V1 in the model V . We
deduce that the rank of V1 is smaller than the rank of V .

• Let us consider the filiform 3-algebra


 [X1 , X2 , X3 ] = X4 ,


[X1 , X2 , X4 ] = X5 ,

 [X 1 , X3 , X4 ] = aX5 ,

[X2 , X3 , X4 ] = bX5 .

Its rank is equal to 2. In fact, in the basis {X1 , X2 , X3 − aX2 , X4 , X5 },

the algebra writes 

 [X1 , X2 , X3 ] = X4 ,

[X1 , X2 , X4 ] = X5 ,
 [X1 , X3 , X4 ] = 0,


[X2 , X3 , X4 ] = bX5 .
Any diagonalizable derivation which admits this basis as eigenvectors ba-
sis, satisfies
f (Xi ) = λi Xi
with
λ3 = λ1 , λ4 = 2λ1 + λ2 , λ5 = 3λ1 + 2λ2 .
Then the rank is 2.
• For n = 2, we have the following important result: any Lie algebra which
admits a nonsingular derivation is nilpotent. This is false as soon as n ≥ 3.
We have the interesting example ([18]): consider the n-Lie algebra given
by
[X1 , X2 , · · · , Xn ] = X2 .
This algebra admits a non singular derivation but it is not nilpotent.
• In a forthcoming paper we develop the classification of filiform 3-Lie alge-
bras whose rank is not 0.

4 sh-n-Lie algebras or Lie n-algebras

4.1 Definition
Definition 16 Let µ be a n-ary skewsymmetric product on a vector space A.
We say that (A, µ) is a sh-n-Lie algebra (or a Lie n-algebra) if µ satisfies the

10
(sh)-Jacobi’s identity:
X
(−1)ǫ(σ) µ(µ(xσ(1) , · · · , xσ(n) ), xσ(n+1) , · · · , xσ(2n−1) ) = 0,
σ∈Sh(n,n−1)

for any xi ∈ A, where Sh(n, n − 1) is the subset of Σ2n−1 defined by:

Sh(n, n − 1) = {σ ∈ Σ2n−1 , σ(1) < · · · < σ(n), σ(n + 1) < · · · < σ(2n − 1)}.

Moreover, we assume that µ is of degree n − 2.

For example, if n = 3, we have the following (sh)-Jacobi’s identity, writing
(123)45 in place of µ(µ(x1 , x2 , x3 ), x4 , x5 ):

(123)45 − (124)35 + (125)34 + (134)25 − (135)24 + (145)23 − (234)15 + (235)14

−(245)13 + (345)12 = 0.

4.2 n-Lie algebras and sh-n-Lie algebras

Proposition 17 Any n-Lie algebra is a sh-n-Lie algebra.

Proof. The Jacobi condition for n-Lie algebras writes

µ ◦ (µ ⊗ In−1 ) ◦ Φv = 0,

where v ∈ K[Σ2n−1 ], the algebra group of the symmetric group Σ2n−1 on 2n − 1

elements, given by
n
X
v = Id + (−1)i (i, n + 1, · · · , 2n − 1, 1, 2, · · · , i − 1, bi, i + 1, · · · , n),
i=1

where (i, n + 1, · · · , 2n − 1, 1, 2, · · · , i − 1, bi, i + 1, · · · , n) is the permutation

 
1 2 ··· n n+ 1 n+ 2 ··· ··· ··· · · · 2n − 1
.
i n + 1 · · · 2n − 1 1 2 ··· i − 1 i + 1 ··· n

Let X
w= (−1)ǫ(σ) σ.
σ∈Σ2n−1

We have in K[Σ2n−1 ], w ◦ v = α(n)w with α(n) = 1 − n if n is odd and

α(n) = 1 + n if n is even. This shows that the vector w is in the invariant
subspace of K[Σ2n−1 ] generated by the vector v. This means that the identity

µ ◦ (µ ⊗ In−1 ) ◦ Φv = 0

implies
µ ◦ (µ ⊗ In−1 ) ◦ Φw = 0
which is equivalent to the Jacobi identity for sh-n-Lie algebras.

11
Proposition 18 A sh-n-Lie algebra is a n-Lie algebra if and only if any adjoint
operator is a derivation.
Proof. We have seen that a n-Lie algebra is a n-Leibniz algebras and these last
are characterized by the fact that any adjoint operator is a derivation.
Remark. Colored Lie algebras, colored n-Lie algebras. Let us con-
sider a binary algebra with a skewsymmetric product satisfying a colored Jacobi
identity:

α[[Xi , Xj ], Xk ] + β[[Xj , Xk ], Xl + γ[[Xk , Xi ], Xj ] = 0,

for any i < j < k, the constants α, β, γ being in K. This identity is related to
the vector v = αId + βc + γc2 of K[Σ3 ]. Let w = Id − τ12 − τ13 − τ23 + c + c2 the
vector of K[Σ3 ]. Since K is of characteristic 0, the Jacobi identity, is equivalent
to
µ ◦ (µ ⊗ Id) ◦ Φw = 0.
But in K[Σ3 ] we have
w ◦ v = (α + β + γ)w.
Then, if α + β + γ 6= 0, the colored Lie algebra satisfies the (non colored) Jacobi
condition. This minimizes the interest of the notion of colored Lie algebras. It
is the same for colored n-Lie algebras.

4.3 3-Lie admissible algebras

To simplify notations, we consider the case n = 3. In this case the product is of
degree 1. A 3-ary algebra (A, ·) is called 3-Lie admissible if the antisymmetric
product X
[v1 , v2 , v3 ] = (−1)ε(σ) vσ(1) · vσ(2) · vσ(3)
σ∈Σ3

is a sh-3-Lie product.
Proposition 19 A n-ary algebra (A, ·) is 3-Lie admissible if and only if we
have
X
(−1)ε(σ) ((vσ(1) · vσ(2) · vσ(3) ) · vσ(4) · vσ(5) + vσ(1) · (vσ(2) · vσ(3) · vσ(4) ) · vσ(5)
σ∈Σ5
+(vσ(1) · vσ(2) · (vσ(3) · vσ(4) · vσ(5) )) = 0,

for any v1 , v2 , v3 , v4 , v5 ∈ A.
Exemples.
• Any 3-ary partially associative algebra is 3-Lie admissible.
• In [17], a notion of σ-associative algebra have been introduced in the space
of tensors (2, 1) based on a vector space. In case of symmetric tensor, this

12
 product can be simplified. A symmetric tensor is defined by its structure
constants Tijk which satisfy

Tijk = Tjki = Tkij .

The 3-product T · U · V whose structure constants are

X
(T · U · V )ijk = Tlij Ulki Vljk
l

is 3-Lie admissible. Moreover the associated sh-3-Lie algebra is a 3-Lie

algebra.

4.4 Maurer-Cartan equations

We assume in this section that any n-Lie algebras is of finite dimension. To
simplify the presentation, we assume also that n = 3. Let V be a finite dimen-
sional 3-Lie algebras. Let {v1 , · · · , vp } be a basis of V . The structure constants
of V related to this basis are given by
l
X
l
{vi , vj , vk } = Ci,j,k vl
l=1

and satisfy
l
Cijk = (−1)ε(σ) Cσ(i)σ(j)σ(k)
l
,
for any σ ∈ Σ3 . The Jacobi condition writes
t s t s t s t s t s t s
Cijk Ctlm − Cijl Ctkm + Cijm Ctjk + Cikl Ctjm − Cikm Ctjl + Cilm Ctjl
t s t s t s t s
−Cjkl Ctim + Cjkm Ctil − Cjlm Ctik + Cklm Ctij = 0,

for any i < j < k, l < m and s, t = 1, · · · , p. Let {ω1 , · · · , ωp } be the dual basis
of {v1 , · · · , vp }. We consider the graded exterior algebra Λ(V ) = ⊕Λk of V and
the linear operator
d : Λ1 (V ) = V ∗ → Λ3 (V )
given by X
l
dωl = Cijk ωi ∧ ωj ∧ ωk .
i<j<k

If we denote also by d the linear operator

d : Λ3 (V ) = V ∗ → Λ5 (V )

defined by

d(ωi ∧ ωj ∧ ωk ) = dωi ∧ ωj ∧ ωk + ωi ∧ dωj ∧ ωk + ωi ∧ ωj ∧ dωk ,

13
we obtain
X
l
d(dωl ) = Cijk (dωi ∧ ωj ∧ ωk + ωi ∧ dωj ∧ ωk + ωi ∧ ωj ∧ dωk
i<j<k
X
l i j
= Cijk (Clst ωl ∧ ωs ∧ ωt ∧ ωj ∧ ωk + Clst ωi ∧ ωl ∧ ωs ∧ ωt ∧ ωk
i<j<k
k
+Clst ωi ∧ ωj ∧ ωl ∧ ωs ∧ ωt ).

In this summand, all the products containing two equal factors are zero (this
justifies to use the exterior algebra). In the same way, the Jacobi condition
related to five vectors is trivial as soon as two vectors are equal. In fact, if we
compute the Jacobi condition for the vectors (X1 , X2 , X3 , X1 , X1 ) we find 0 = 0
and for the vector (X1 , X2 , X3 , X1 , X5 ) we find

[[X1 , X2 , X3 ], X1 , X5 ] + [[X1 , X2 , X5 ], X3 , X1 ] − [[X1 , X3 , X5 ], X2 , X1 ]

−[[X2 , X3 , X1 ], X1 , X5 ] − [[X2 , X1 , X5 ], X1 , X3 ] − [[X3 , X1 , X5 ], X1 , X2 ] = 0,

that is, 0 = 0. Thus the Jacobi condition concerns a family of 5 independent

vectors. Let us return to the computation of d(dω). The coefficient of d(dωl )
related for example to ω1 ∧ ω2 ∧ ω3 ∧ ω4 ∧ ω5 corresponds to the coefficient of
Xl in the Jacobi condition related to (X1 , X2 , X3 , X4 , X5 ). Thus

d(dωl ) = 0.

These relations can be called the Maurer-Cartan equations.

Remark. We cannot use the same calculus to obtain Maurer-Cartan equations
adapted to the structure of n-Lie algebras. This means that the Maurer-Cartan
equations of a n-Lie algebra are the Maurer-Cartan equations of this algebra
considered as a sh-n-Lie algebra. In the classical case of Lie algebras, we have
also such a situation. For example, when we consider the 2-step nilpotent Lie
algebras, defined by the 2-step Jacobi condition

[[Xi , Xj ], Xk ] = 0,

there is no exterior calculus adapted to this special Jacobi condition. The

Maurer-Cartan equations of a 2-step nilpotent algebra are the Maurer-Cartan
equations of this algebra considered as a Lie algebra.

References
[1] Awane A., k-symplectic structures. Journal of Mathematical physics 33
(1992) 4046-4052.
[2] Awane A., Goze M., Pfaffian systems, k-symplectic systems. Kluwer Aca-
demic Publishers. Dordrecht/boston/London 2000.
[3] Bai R., Zhang Z., Li H., Shi H., The inner derivation algebras of (n + 1)-
dimensional n-Lie algebras. Comm. Algebra 28 (2000), no. 6, 2927–2934.

14
 [4] Bai R., Guojie S., The classification of six dimensional 4-Lie algebras.
J.Phys. A. Math. Theor. (42), (2009), 035207.
[5] Bai R., Meng D., The simple n-ary Lie algebras. Hadronic J. 25 (2002),
no. 6, 713–723.
[6] Cuvier C., Leibnitz algebras. Ann. Sc. E.N.S. (1999).
[7] Dzhumadil’daev A.S., Wronskians as n-Lie multiplications. arXiv
math.RA/0202043.
[8] Filippov V.T., n-Lie algebras. Sib. Mat. Zh, 26(6), (1985), 126-140.
[9] Gnedbaye A. V. Opérades des algèbres k +1-aires. Operads: Proceedings of
Renaissance Conferences (Hartford, CT/Luminy, 1995), 83–113, Contemp.
Math., 202, Amer. Math. Soc., Providence, RI, 1997.
[10] Goze M., Remm E., Lie-admissible algebras and operads. J. Algebra 273
(2004), no. 1, 129–152.
[11] Goze M., Remm E., Lie admissible coalgebras. J. Gen. Lie Theory Appl. 1
(2007), no. 1, 19–28.
[12] Goze M., Hakimdjanov Y., Sur les algèbres de Lie nilpotentes admettant
un tore de dérivations, manuscripta mathematica, 84, Number 1, (1994).
[13] Hanlon P., Wachs M., On Lie k-algebras. ADvances in Mathematics, 113,
(1995), 206.
[14] Pei-Ming H., Ru-Chuen H., Yutaka M., Lie 3-Algebra and Multiple M2-
branes. arXiv : 0804.2110 (hep-th], (2008).
[15] Kasymaov Sh. M., On a theory of n-Lie algebras. Algebra Logka, 26(3),
(1987), 277-297.
[16] Lada T., Markl M., Strongly Homotopy Lie algebras. Comm. in . Algebra
23, (1995), 2147.
[17] Remm E., Goze N., The n-ary algebra of tensors and of cubic and hy-
percubic matrices. arXiv: math.RA 0902.2757. To appear in Multilinear
Algebra.
[18] Williams M. P., Nilpotent n-Lie algebras. Dissertation. North Carolina
State University. (2004).

15