Leibniz algebra




A Leibniz algebra is like a Lie algebra, but without the condition that the product, often still written as a bracket [,][-,-], is skew-symmetric. The Jacobi identity however is retained as a condition in its form as the derivation-property of the product over itself. In view of the analogous product law of differentiation (also a derivation-property) attributed to Gottfried Leibniz, this is then called the Leibniz identity which gives Leibniz algebras their modern name (Loday 93, Loday-Pirashvili 93) even though the concept itself is older (Blokh 65).

Leibniz algebras were motivated in Cuvier 91, Loday-Pirashvili 93 as generalizing the relation between Lie algebra cohomology and cyclic homology (Loday-Quillen 84) to one between Leibniz cohomology and Hochschild homology: Where the nilpotency of the differential in the Chevalley-Eilenberg algebras that compute Lie algebra cohomology is equivalent to the Jacobi identity in the corresponding Lie algebra, Leibniz cohomology is defined on non-skew symmetric dg-algebras where now it is the generalization of the Jacobi identity in form of the Leibniz rule (1) which still guarantees the nilpotency of the differential.

More recently, Leibniz algebras have been argued to clarify the nature of the embedding tensor and the resulting tensor hierarchies in gauged supergravity (Lavau 17).


Let kk be a commutative ring (typically a field).


(left Leibniz algebra)

A left Leibniz kk-algebra_ (or: left Loday algebra) is

equipped with

  • a linear map, the product

    AA A (v,w) vw \array{ A \otimes A &\longrightarrow& A \\ (v,w) &\mapsto& v \cdot w }

such that

  • the product satisfies the left Leibniz identity, saying that

    (1)v 1(v 2v 3)=(v 1v 2)v 3+v 2(v 1v 3) v_1 \cdot (v_ 2 \cdot v_3) \;=\; (v_1 \cdot v_2) \cdot v_3 \;+\; v_2 \cdot (v_1 \cdot v_3)

    for all v iAv_i \in A.

This says equivalently that the operations v():AAv \cdot (-) \colon A \to A of left-multiplication by elements vAv \in A via the given product is a derivation of the product itself, whence the name (paying tribute to Gottfried Leibniz‘s product rule of differentiation).

Analogously there is the concept of right Leibniz algebras in the evident way.


Relation to Lie algebras in Loday-Pirashvili category

There is a remarkable observation of Loday and Pirashvili that in the Loday–Pirashvili tensor category of linear maps with (exotic) “infinitesimal tensor product”, the category of internal Lie algebras has the category of, say left, Leibniz kk-algebras as a full subcategory.

Corepresentation, representation, crossed module

Both a representation and a corepresentation of a right Leibniz kk-algebra 𝔤\mathfrak{g} involve a kk-module MM and two kk-linear maps “actions” M𝔤MM\otimes\mathfrak{g}\to M and 𝔤MM\mathfrak{g}\otimes M\to M with 3 axioms.

For representations:

[m,[x,y]]=[[m,x],y][[m,y],x][m, [x, y]] = [[m, x], y] - [[m, y], x]
[x,[a,y]]=[[x,m],y][[x,y],m][x, [a, y]] = [[x, m], y] - [[x, y], m]
[x,[y,m]]=[[x,y],m][[x,m],y][x, [y, m]] = [[x, y], m] - [[x, m], y]

for x,y𝔤x,y\in\mathfrak{g} and for mMm\in M.

For corepresentatons:

[[x,y],m]=[x,[y,m]][y,[x,m]][[x, y], m] = [x, [y, m]] - [y, [x, m]]
[y,[a,x]]=[[y,m],x][m,[x,y]][y, [a, x]] = [[y, m], x] - [m, [x, y]]
[[m,x],y]=[m,[x,y]][[y,m],x].[[m, x], y] = [m, [x, y]] - [[y, m], x].

If the two “actions” are symmetric, i.e. [x,m]+[m,x]=0[x,m] + [m,x] = 0 for all mMm\in M, x𝔤x\in\mathfrak{g} then all the 6 axioms of representation or corepresentation are equivalent. If MM is underlying a Leibniz algebra then an action of 𝔤\mathfrak{g} on MM is by definition symmetric, hence all the 6 equivalent conditions hold.

A map t:𝔤𝔟t : \mathfrak{g}\to\mathfrak{b} together with an action of 𝔟\mathfrak{b} on 𝔤\mathfrak{g} is a Leibniz crossed module if

t([b,g])=[b,t(g)],t([g,b])=[t(g),b],forallb𝔟,g𝔤 t([b,g])= [b,t(g)],\,\,\,t([g,b])=[t(g),b],\,\,\,\, for all\,\,\, b\in\mathfrak{b}, g' \in\mathfrak{g}
[g,t(g)]=[g,g]=[t(g),g],forallg,g𝔤 [g, t(g')] = [g, g'] = [t(g), g'],\,\,\,\, for all\,\,\, g, g' \in\mathfrak{g}

Abelian extensions

Abelian extension of right Leibniz algebras is a split short exact sequence of kk-modules

0M𝔥𝔤0 0\to M \to \mathfrak{h}\to \mathfrak{g}\to 0

where the mapping 𝔥𝔤\mathfrak{h}\to\mathfrak{g} is a morphism of Leibniz algebras, and MM is equipped with induced action of 𝔤\mathfrak{g}. The isomorphisms of extensions of 𝔤\mathfrak{g} by MM with fixed action are defined as usual. This way we obtain a set of equivalence classes Ext(𝔤,M)Ext(\mathfrak{g},M). To classify the extensions one looks for compatible Leibniz brackets on M𝔤M\oplus \mathfrak{g}. The general form of a bracket is

[(m 1,x 1),(m 2,x 2)]=([m 1,x 2]+[x 1,m 2]+f(x 1,x 2),[x 1,x 2]), [(m_1,x_1),(m_2,x_2)] = ([m_1, x_2] + [x_1, m_2] + f(x_1, x_2), [x_1, x_2]),

where f(x 1,x 2)f(x_1,x_2) satisfy the following 2-cocycle identity:

[x,f(y,z)]+[f(x,z),y][f(x,y),z]=f([x,y],z)f([x,z],y)f(x,[y,z]) [x, f(y, z)] + [f(x, z), y] - [f(x, y), z] = f([x, y], z) - f([x, z], y) - f(x, [y, z])

The extension is split in the category of Leibniz algebras if ff is a boundary i.e. there exists a kk-module map g:𝔤Mg:\mathfrak{g}\to M such that

f(x,y)=[x,g(y)]+[g(x),y]g([x,y]),x,y,𝔤 f(x, y) = [x, g(y)] + [g(x), y] - g([x, y]), \,\,\,x,y,\in\mathfrak{g}

As for the Lie algebras, the group of abelian extensions agrees with the 2-cohomology HL 2(𝔤,M)HL^2(\mathfrak{g},M).

A kk-linear derivation of a right Leibniz algebra 𝔤\mathfrak{g} with values in its representation MM is a kk-linear map satisfying the Leibniz property with respect to the bracket:

δ([x,y])=[δ(x),y]+[x,δ(y)] \delta([x,y]) = [\delta(x),y]+[x,\delta(y)]

Such derivations form a kk-module Der(𝔤,M)Der(\mathfrak{g},M).

Homology and cohomology

The homology and cohomology of Leibniz algebra 𝔤\mathfrak{g} with abelian kk-module of coefficients, which is a corepresentation AA in the case of homology and a representation MM in the case of cohomology:

HL *(𝔤,A)=Tor * U𝔤(U(𝔤 Lie),A), HL_*(\mathfrak{g},A) = Tor^{U\mathfrak{g}}_*(U(\mathfrak{g}_{Lie}),A) ,
HL *(𝔤,M)=Ext U𝔤 *(U(𝔤 Lie),A) HL^*(\mathfrak{g},M) = Ext_{U\mathfrak{g}}^*(U(\mathfrak{g}_{Lie}),A)

where U(𝔤 Lie)U(\mathfrak{g}_{Lie}) is the universal enveloping of the maximal Lie algebra quotient 𝔤 Lie\mathfrak{g}_{Lie} of 𝔤\mathfrak{g} and U𝔤U\mathfrak{g} is the universal enveloping of a Leibniz algebra 𝔤\mathfrak{g}.

Fopr n0n\geq 0, the nn-cocycles are elements in C n(𝔤,M)=Hom k(𝔤 n,M)C^n(\mathfrak{g}, M) = Hom_k(\mathfrak{g}^{\otimes n}, M), satisfying the corresponding abelian cocycle condition determined by the differential

d n:C n(𝔤,M)C n+1(𝔤,M) d^n : C^n(\mathfrak{g}, M)\to C^{n+1}(\mathfrak{g}, M)
(d nf)(x 1,...,x n+1)=[x 1,f(x 2,,x n+1)]+ n+1 i=2(1) i[f(x 1,,x^ i,,x n+1),x i] (d^n f) (x_1, . . . , x_{n+1}) = [x_1,f(x_2,\ldots,x_{n+1})] +\sum_{n+1}^{i=2} (-1)^i [f(x_1,\ldots, \hat{x}_i, \ldots, x_{n+1}), x_i]

Notice a difference from the Lie algebra cocycles where instead of a tensor power we have an external power. Then HL *(𝔤,M)=H *(C *(𝔤,M),d *)HL^*(\mathfrak{g},M) = H^*(C^*(\mathfrak{g}, M),d^*).

There are standard interpretations of cocycles in low dimensions. For example for n=0n=0, HL 0(𝔤,M)HL^0(\mathfrak{g}, M) is the submodule of invariants. For n=1n=1 there is a natural projection Der(𝔤,M)HL 1(𝔤,M)Der(\mathfrak{g},M)\to HL^1(\mathfrak{g},M) whose kernel is generated by inner derivations.

Relation to Zinbiel algebras

The Leibniz operad is quadratic Koszul algebra whose Koszul dual operad is called the operad of dual Leibniz algebras or of Zinbiel algebras, see there.

Lie’s third theorem for Leibniz algebras

In complete analogy to the equivalence between the category of Lie algebras and the category of local Lie groups (Lie's third theorem), the category of Leibniz algebras is equivalent to the category of local pointed augmented Lie racks. See Covez 10.

This equivalence restricts to the equivalence between Lie algebras and local Lie groups.

Here a local pointed augmented Lie rack is a pointed augmented rack object in the category of germs of pointed smooth manifolds. An augmented rack is a triple (G,X,p:XG)(G,X,p\colon X\to G), where GG is a group, XX is a G-set, and pp is a map of sets such that p(gx)=gp(x)g 1p(g\cdot x)=g p(x)g^{-1}. An augmented rack is pointed if there is an element 1X1\in X such that p(1)=1p(1)=1 and g1=1g\cdot 1=1 for all gGg\in G.

If (G,X,p)(G,X,p) is an augmented rack, then XX can be made into a rack as follows: xy=p(x)yx\triangleright y = p(x)\cdot y. If (G,X,p)(G,X,p) is a pointed augmented rack, then XX is a pointed rack, meaning there is 1X1\in X such that 1x=x1\triangleright x=x and x1=1x\triangleright 1=1.


Basic examples

  • Every Lie algebra is a Leibniz algebra that happens to have skew-symmetric product. Conversely, a Leibniz algebra with skew-symmetric product is a Lie algebra.

From Lie modules and embedding tensors


(Leibniz algebra from Lie module with embedding tensor)


  • 𝔤\mathfrak{g} be a Lie algebra,

  • 𝔤VρV\mathfrak{g} \otimes V \overset{\rho}{\longrightarrow} V be a Lie algebra representation of 𝔤\mathfrak{g},

  • Θ:V𝔤\Theta \colon V \longrightarrow \mathfrak{g} be an embedding tensor, hence a linear map such that the “quadratic constraint” is satisfied: for all v 1,v 2Vv_1, v_2 \in V we have

    (2)[Θ(v 1),Θ(v 2)]=Θ(ρ Θ(v 1)(v 2)) [\Theta(v_1), \Theta(v_2)] \;=\; \Theta \big( \rho_{\Theta(v_1)}(v_2) \big)

Then VV becomes a Leibniz algebra with product defined by

(3)v 1v 2ρ Θ(v 1)(v 2) v_1 \cdot v_2 \;\coloneqq\; \rho_{\Theta(v_1)}(v_2)

and with respect to this the quadratic constraint (2) becomes the condition that Θ\Theta is a homomorphism of Leibniz algebras.

(Lavau 17, Example 3)


We directly check the Leibniz rule (1) as follows:

v 1(v 2v 3) =ρ Θ(v 1)(ρ Θ(v 2)(v 3)) =ρ [Θ(v 1),Θ(v 2)]=ρ Θ(v 1)(v 2)(v 3)+ρ Θ(v 2)(ρ Θ(v 1)(v 3)) =(v 1v 2)v 3+v 2(v 1v 3) \begin{aligned} v_1 \cdot (v_2 \cdot v_3) & = \rho_{\Theta(v_1)} \big( \rho_{\Theta(v_2)}(v_3) \big) \\ & = \rho_{ \underset{ = \rho_{\Theta(v_1)}(v_2) }{ \underbrace{ [\Theta(v_1), \Theta(v_2)] } } }(v_3) + \rho_{\Theta(v_2)} \big( \rho_{\Theta(v_1)}(v_3) \big) \\ & = (v_1 \cdot v_2) \cdot v_3 + v_2 \cdot (v_1 \cdot v_3) \end{aligned}

Here the first line is the definition (3), the second line is the Lie action property (here), under the brace we use the quadratic constraint (2) on the embedding tensor, and in the last line we observe again the definition (3).

From dg-Lie algebras


(Leibniz algebra from dg-Lie algebra)

Let ((V ,),[,])((V_\bullet, \partial), [-,-]) be a dg-Lie algebra with underlying chain complex (V ,)(V_\bullet, \partial) and with super Lie bracket [,][-,-].

On the graded vector space which is the direct sum

VnV nVect \mathbf{V} \;\coloneqq\; \underset{n}{\oplus} V_n \;\in\; Vect

of all the component vector spaces, consider the product given by the formula

(4)VV V (v,w) vw[v,w] \array{ \mathbf{V} \otimes \mathbf{V} &\longrightarrow& \mathbf{V} \\ (v,w) &\mapsto& v \cdot w \mathrlap{ \;\coloneqq\; [\partial v, w] } }

Then: Restricted to V 1VV_1 \subset \mathbf{V} this product gives a left Leibniz algebra (V 1,)(V_1, \cdot) (Def. ), i.e. satisfies the Leibniz condition (1).

This statement is highlighted in Lavau-Palmkvist 19, 2.1.


We directly compute as follows:

v 1(v 2v 3) =[v 1,[v 2,v 3]] =[[v 1,v 2],v 3]+(1) (deg(v 1)1)(deg(v 2)2)=0[v 2,[v 1,v 3]] =[[v 1,v 2],v 3]+[v 2,[v 1,v 3]] =(v 1v 2)v 3+v 2(v 1v 3). \begin{aligned} v_1 \cdot (v_2 \cdot v_3) & = \big[ \partial v_1 , [ \partial v_2, v_3 ] \big] \\ & = \big[ [ \partial v_1 , \partial v_2 ], v_3 \big] \;+\; (-1)^{ \overset{= 0}{ \overbrace{ (deg(v_1)-1) (deg(v_2)-2) } } } \big[ \partial v_2, [\partial v_1, v_3] \big] \\ & = \big[ \partial [ v_1 , \partial v_2 ], v_3 \big] \;+\; \big[ \partial v_2, [\partial v_1, v_3] \big] \\ & = (v_1 \cdot v_2) \cdot v_3 \;+\; v_2 \cdot (v_1 \cdot v_3) \,. \end{aligned}

Here the first line is the definition (4), the second line is the graded Jacobi identity, the third line uses the derivation-property and the nilpotency of the differential, and the last line invokes again the definition (4). Over the brace we used the assumption that v iV 1v_i \in V_1.


The construction in Prop. evidently extends to a functor from the category dgLieAlgdgLieAlg of dg-Lie algebras to the category LeibAlgLeibAlg of Leibniz algebras (both over the given ground ring/ground field):

(5)() 1:dgLieAlgLeibAlg. (-)_{1} \;\colon\; dgLieAlg \longrightarrow LeibAlg \,.

Notice the analogy to the evident functor that extract the Lie algebra in degree 0:

(6)() 0:dgLieAlgLieAlg. (-)_{0} \;\colon\; dgLieAlg \longrightarrow LieAlg \,.



The idea of Leibniz algebra, though not by this name, is given already in

  • A. Blokh, A generalization of the concept of Lie algebra, Dokl. Akad. Nauk SSSR, 165:471–473 (1965) (mathrunet:dan31825)

The concept was revived (and apparently the name Leibniz algebra was first chosen) in

Early review is in

See also

Relaization of Leibniz algebras as Lie algebra objects in a suitable tensor category:

Relation to Hochschild homology:

  • Christian Cuvier, Homologie de Leibniz et homologie de Hochschild, C.R. Acad. Sci. Paris, Ser. A-B313, 569-572 (1991)

  • Jerry M. Lodder, Leibniz homology, characteristic classes and K-theory, K-theory archive/0493;Leibniz cohomology and the calculus of variations_ (arXiv:math/9808036)

  • Jean-Louis Loday, Algebraic K-theory and the conjectural Leibniz K-theory, K-Theory 09/2003; 30(2):105-127, pdf doi

This is partly based on earlier insights of Kinyon and Weinstein:

  • Michael K. Kinyon, Leibniz algebras, Lie racks, and digroups, J. Lie Theory 17:1 (2007) 099–114, arxiv:math.GR/0403509

Relation to dg-Lie algebras and tensor hierarchies

Relation of Leibniz algebras to dg-Lie algebras such as the tensor hierarchies in gauged supergravity:

Lie integration

A generalization of Lie integration to conjectural Leibniz groups has been conjectured by J-L. Loday. A local version via local Lie racks has been proposed in

Last revised on May 24, 2020 at 11:56:30. See the history of this page for a list of all contributions to it.