nLab
L-infinity-algebra

Context

\infty-Lie theory

∞-Lie theory

Background

Smooth structure

Higher groupoids

Lie theory

∞-Lie groupoids

∞-Lie algebroids

Formal Lie groupoids

Cohomology

Homotopy

Examples

\infty-Lie groupoids

\infty-Lie groups

\infty-Lie algebroids

\infty-Lie algebras

Higher algebra

Rational homotopy theory

Contents

Idea

L L_\infty-algebras (or strong homotopy Lie algebras) are a higher generalization (a “vertical categorification”) of Lie algebras: in an L L_\infty-algebra the Jacobi identity is allowed to hold (only) up to higher coherent homotopy.

An L L_\infty-algebra that is concentrated in lowest degree is an ordinary Lie algebra. If it is concentrated in the lowest two degrees is is a Lie 2-algebra, etc.

From another perspective: an L L_\infty-algebra is a Lie ∞-algebroid with a single object.

L L_\infty-algebras are infinitesimal approximations of smooth ∞-groups in analogy to how an ordinary Lie algebra is an infinitesimal approximation of a Lie group. Under Lie integration every L L_\infty-algebra 𝔤\mathfrak{g} “exponentiates” to a smooth ∞-group Ωexp(𝔤)\Omega \exp(\mathfrak{g}).

Definition

Abstract definition in terms of algebras over an operad

An L L_\infty-algebra is an algebra over an operad in the category of chain complexes over the L-∞ operad.

In the following we spell out in detail what this means in components.

Definition in terms of higher brackets

For VV a graded vector space, for v iV |v i|v_i \in V_{\vert v_i\vert} homogenously graded elements, and for σ\sigma a permutation of nn elements, write χ(σ,v 1,,v n){1,+1}\chi(\sigma, v_1, \cdots, v_n)\in \{-1,+1\} for the product of the signature of the permutation with a factor of (1) |v i||v j|(-1)^{\vert v_i \vert \vert v_j \vert} for each interchange of neighbours (v i,v j,)(\cdots v_i,v_j, \cdots ) to (v j,v i,)(\cdots v_j,v_i, \cdots ) involved in the permutation.

Definition

An L L_\infty-algebra is

  1. a graded vector space VV;

  2. for each nn \in \mathbb{N} a multilinear map called the nn-ary bracket

    l n()[,,,]:V nVl_n(\cdots) \coloneqq [-,-, \cdots, -] \colon V^{\wedge n} \to V

    of degree n2n-2

such that

  1. each l nl_n is graded antisymmetric, in that for every permutation σ\sigma and homogeneously graded elements v iV |v i|v_i \in V_{\vert v_i \vert} then

    l n(v σ(1),v σ(2),,v σ(n))=χ(σ,v 1,,v n)l n(v 1,v 2,v n) l_n(v_{\sigma(1)}, v_{\sigma(2)},\cdots ,v_{\sigma(n)}) = \chi(\sigma,v_1,\cdots, v_n) \cdot l_n(v_1, v_2, \cdots v_n)
  2. the generalized Jacobi identity holds:

    (1) i+j=n+1 σUnShuff(i,ni)χ(σ,v 1,,v m)(1) i(j1)l j(l i(v σ(1),,v σ(i)),v σ(i+1),,v σ(n))=0, \sum_{i+j = n+1} \sum_{\sigma \in UnShuff(i,n-i)} \chi(\sigma,v_1, \cdots, v_m) (-1)^{i(j-1)} l_{j} \left( l_i \left( v_{\sigma(1)}, \cdots, v_{\sigma(i)} \right), v_{\sigma(i+1)} , \cdots , v_{\sigma(n)} \right) = 0 \,,

    for all nn, all and homogeneously graded elements v iV iv_i \in V_i (here the inner sum runs over all (i,j)(i,j)-unshuffles σ\sigma).

There are various different conventions on the gradings possible, which lead to similar formulas with different signs.

Example

In lowest degrees the generalized Jacobi identity says that

  1. for n=1n = 1: the unary map l 1\partial \coloneqq l_1 squares to 0:

    ((v 1))=0 \partial (\partial(v_1)) = 0

1: for n=2n = 2: the unary map \partial is a graded derivation of the binary map

[v 1,v 2](1) |v 1||v 2|[v 2,v 1]+[v 1,v 2]=0 - [\partial v_1, v_2] - (-1)^{\vert v_1 \vert \vert v_2 \vert} [\partial v_2, v_1] + \partial [v_1, v_2] = 0

hence

[v 1,v 2]=[v 1,v 2]+(1) |v 1|[v 1,v 2]. \partial [v_1, v_2] = [\partial v_1, v_2] + (-1)^{\vert v_1 \vert}[v_1, \partial v_2] \,.
Example

When all higher brackets vanish, l k>2=0l_{k \gt 2}= 0 then for n=3n = 3:

[[v 1,v 2],v 3]+(1) |v 1|(|v 2|+|v 3|)[[v 2,v 3],v 1]+(1) |v 2|(|v 1|+|v 3|)[[v 1,v 3],v 2]=0 [[v_1,v_2],v_3] + (-1)^{\vert v_1 \vert (\vert v_2 \vert + \vert v_3 \vert)} [[v_2,v_3],v_1] + (-1)^{\vert v_2 \vert (\vert v_1 \vert + \vert v_3 \vert)} [[v_1,v_3],v_2] = 0

this is the graded Jacobi identity. So in this case the L L_\infty-algebra is equivalently a dg-Lie algebra.

Example

When l 3l_3 is possibly non-vanishing, then on elements x ix_i on which =l 1\partial = l_1 vanishes then the generalized Jacobi identity for n=3n = 3 gives

[[v 1,v 2],v 3]+(1) |v 1|(|v 2|+|v 3|)[[v 2,v 3],v 1]+(1) |v 2|(|v 1|+|v 3|)[[v 1,v 3],v 2]=[v 1,v 2,v 3]. [[v_1,v_2],v_3] + (-1)^{\vert v_1 \vert (\vert v_2 \vert + \vert v_3 \vert)} [[v_2,v_3],v_1] + (-1)^{\vert v_2 \vert (\vert v_1 \vert + \vert v_3 \vert)} [[v_1,v_3],v_2] = - \partial [v_1, v_2, v_3] \,.

This shows that the Jacobi identity holds up to an “exact” term, hence up to homotopy.

Reformulation in terms of semifree differential coalgebra

A little later it was realized that the above huge sum expressions above just expresses the fact that the differential DD in a semifree dg-coalgebra squares to 0, D 2=0D^2 = 0:

An L L_\infty-algebra is

  • an +\mathbb{N}_+-graded vector space 𝔤\mathfrak{g};

  • equipped with a differential D: 𝔤 𝔤D : \vee^\bullet \mathfrak{g} \to \vee^\bullet \mathfrak{g} of degree 1-1 on the free graded co-commutative coalgebra over 𝔤\mathfrak{g} that squares to 0

D 2=0. D^2 = 0 \,.

Here the free graded co-commutative co-algebra 𝔤\vee^\bullet \mathfrak{g} is, as a vector space, the same as the graded Grassmann algebra 𝔤\wedge^\bullet \mathfrak{g} whose elements we write as

3t 1t 2+t 3+t 3t 4t 5 3 t_1 \vee t_2 + t_3 + t_3 \vee t_4 \vee t_5

etc (where the \vee is just a funny way to write the wedge \wedge, in order to remind us that:…)

but throught of as equipped with the standard coproduct

Δ(v 1v 2v n) i±(v 1v i)(v i+1v n) \Delta (v_1 \vee v_2 \cdots \vee v_n) \propto \sum_i \pm (v_1 \vee \cdots \vee v_i) \otimes (v_{i+1} \vee \cdots \vee v_n)

(work out or see the references for the signs and prefacors).

Since this is a free graded co-commutative coalgebra, one can see that any differential

D: 𝔤 𝔤 D : \vee^\bullet \mathfrak{g} \to \vee^\bullet \mathfrak{g}

on it is fixed by its value “on cogenerators” (a statement that is maybe unfamiliar, but simply the straightforward dual of the more familar statement to which we come below, that differentials on free graded algebras are fixed by their action on generators) which means that we can decompose DD as

D=D 1+D 2+D 3+, D = D_1 + D_2 + D_3 + \cdots \,,

where each D iD_i acts as l il_i when evaluated on a homogeneous element of the form t 1t nt_1 \vee \cdots \vee t_n and is then uniquely extended to all of 𝔤\vee^\bullet \mathfrak{g} by extending it as a coderivation on a coalgebra.

For instance D 2D_2 acts on homogeneous elements of word lenght 3 as

D 2(t 1t 2t 3)=D 2(t 1,t 2)t 3±permutations. D_2(t_1 \vee t_2 \vee t_3) = D_2(t_1, t_2)\vee t_3 \pm permutations \,.

exercise for the reader: spell this all out more in detail with all the signs and everyrthing. Possibly by looking it up in the references given below.

Using this, one checks that the simple condition that DD squares to 0 is precisely equivalent to the infinite tower of generalized Jacobi identities:

(D 2=0)(n: i+j=n shufflesσ±l i(l j(v σ(1),,v σ(j),v σ(j+1),,v σ(n)))=0). (D^2 = 0) \Leftrightarrow \left( \forall n : \sum_{i+j = n} \sum_{shuffles \sigma} \pm l_i (l_j (v_{\sigma(1)}, \cdots, v_{\sigma(j)} , v_{\sigma(j+1) , \cdots , v_{\sigma(n)}} ) ) = 0 \right) \,.

So in conclusion we have:

An L L_\infty-algebra is a dg-coalgebra whose underlying coalgebra is cofree and concentrated in negative degree.

Reformulation in terms of semifree differential algebra

The reformulation of an L L_\infty-algebra as simply a semi-co-free graded-co-commutative coalgebra ( 𝔤,D)(\vee^\bullet \mathfrak{g}, D) is a useful repackaging of the original definition, but the coalgebraic aspect tends to be not only unfamiliar, but also a bit inconvenient. At least when the graded vector space 𝔤\mathfrak{g} is degreewise finite dimensional, we can simply pass to its degreewise dual graded vector space 𝔤 *\mathfrak{g}^*. Its Grassmann algebra 𝔤 *\wedge^\bullet \mathfrak{g}^* is then naturally equipped with an ordinary differential d=D *d = D^* which acts on ω 𝔤 *\omega \in \wedge^\bullet \mathfrak{g}^* as

(dω)(t 1t n)=±ω(D(t 1t n)). (d \omega) (t_1 \vee \cdots \vee t_n) = \pm \omega(D(t_1 \vee \cdots \vee t_n)) \,.

When the grading-dust has settled one finds that with

𝔤 *=k𝔤 1 *(𝔤 1 *𝔤 1 *𝔤 2 *) \wedge^\bullet \mathfrak{g}^* = k \oplus \mathfrak{g}^*_1 \oplus (\mathfrak{g}^*_1 \wedge \mathfrak{g}^*_1 \oplus \mathfrak{g}^*_2) \oplus \cdots

with the ground field in degree 0, the degree 1-elements of 𝔤 *\mathfrak{g}^* in degree 1, etc, that dd is of degree +1 and of course squares to 0

d 2=0. d^2 = 0 \,.

This means that we have a semifree dga

CE(𝔤):=( 𝔤 *,d). CE(\mathfrak{g}) := (\wedge^\bullet \mathfrak{g}^*, d) \,.

In the case that 𝔤\mathfrak{g} happens to be an ordinary Lie algebra, this is the ordinary Chevalley-Eilenberg algebra of this Lie algebra. Hence we should generally call CE(𝔤)CE(\mathfrak{g}) the Chevalley-Eilenberg algebra of the L L_\infty-algebra 𝔤\mathfrak{g}.

One observes that this construction is bijective: every (degreewise finite dimensional) cochain semifree dga generated in positive degree comes from a (degreewise finite dimensional) L L_\infty-algebra this way.

This means that we may just as well define a (degreewise finite dimensional) L L_\infty-algebra as an object in the opposite category of (degreewise finite dimensional) commutative dg-algebras that are semifree dgas and generated in positive degree.

And this turns out to be one of the most useful perspectives on L L_\infty-algebras.

In particular, if we simply drop the condition that the dg-algebra be generated in positive degree and allow it to be generated in non-negative degree over the algebra in degree 0, then we have the notion of the (Chevalley-Eilenberg algebra of) an L-infinity-algebroid.

Details

We discuss in explit detail the computation that shows that an L L_\infty-algebra structure on 𝔤\mathfrak{g} is equivalently a dg-algebra-structure on 𝔤 *\wedge^\bullet \mathfrak{g}^*.

Let 𝔤\mathfrak{g} be a degreewise finite-dimensional +\mathbb{N}_+graded vector space equipped with multilinear graded-symmetric maps

[,,] k:Sym k𝔤𝔤 [-,\cdots,-]_k : Sym^k \mathfrak{g} \to \mathfrak{g}

of degree -1, for each k +k \in \mathbb{N}_+.

Let {t a}\{t_a\} be a basis of 𝔤\mathfrak{g} and {t a}\{t^a\} a dual basis of the degreewise dual 𝔤 *\mathfrak{g}^*. Equip the Grassmann algebra Sym 𝔤 *Sym^\bullet \mathfrak{g}^* with a derivation

d:Sym 𝔤 *Sym 𝔤 * d : Sym^\bullet \mathfrak{g}^* \to Sym^\bullet \mathfrak{g}^*

defined on generators by

d:t a k=1 1k![t a 1,,t a k] k at a 1t a k. d : t^a \mapsto - \sum_{k = 1}^\infty \frac{1}{k!} [t_{a_1}, \cdots, t_{a_k}]^a_k \, t^{a_1} \wedge \cdots \wedge t^{a_k} \,.

Here we take t at^a to be of the same degree as t at_a. Therefore this derivation has degree +1.

We compute the square d 2=ddd^2 = d \circ d:

ddt a =d(1) k=1 1k![t a 1,,t a k] k at a 1t a k = k,l=1 1(k1)!l![[t b 1,,t b l],t a 2,,t a k] at b 1t b lt a 2t a k. \begin{aligned} d d t^a &= d (-1)\sum_{k = 1}^\infty \frac{1}{k!} [t_{a_1}, \cdots, t_{a_k}]^a_k \, t^{a_1} \wedge \cdots \wedge t^{a_k} \\ & = \sum_{k,l = 1}^\infty \frac{1}{(k-1)! l!} [[t_{b_1}, \cdots, t_{b_l}], t_{a_2}, \cdots, t_{a_k}]^a \, t^{b_1} \wedge \cdots \wedge t^{b_l} \wedge t^{a_2} \wedge \cdots \wedge t^{a_{k}} \end{aligned} \,.

Here the wedge product on the right projects the nested bracket onto its graded-symmetric components. This is produced by summing over all permutations σΣ k+l1\sigma \in \Sigma_{k+l-1} weighted by the Koszul-signature of the permutation:

= k,l=1 1(k+l1)! σΣ k+l1(1) sgn(σ)1(k1)!l![[t b 1,,t b l],t a 2,,t a k] at b 1t b lt a 2t a k. \cdots = \sum_{k,l = 1}^\infty \frac{1}{(k+l-1)!} \sum_{\sigma \in \Sigma_{k+l-1}} (-1)^{sgn(\sigma)} \frac{1}{(k-1)! l!} [[t_{b_1}, \cdots, t_{b_l}], t_{a_2}, \cdots, t_{a_k}]^a \, t^{b_1} \wedge \cdots \wedge t^{b_l} \wedge t^{a_2} \wedge \cdots \wedge t^{a_{k}} \,.

The sum over all permutations decomposes into a sum over the (l,k1)(l,k-1)-unshuffles and a sum over permutations that act inside the first ll and the last (k1)(k-1) indices. By the graded-symmetry of the bracket, the latter do not change the value of the nested bracket. Since there are (k1)!l!(k-1)! l! many of them, we get

= k,l=1 1(k+l1)! σUnsh(l,k1)(1) sgn(σ)[[t a 1,,t a l],t a l+1,,t a k+l1]t a 1t a k+l1. \cdots = \sum_{k,l = 1}^\infty \frac{1}{(k+l-1)!} \sum_{\sigma \in Unsh(l,k-1)} (-1)^{sgn(\sigma)} [[t_{a_1}, \cdots, t_{a_l}], t_{a_{l+1}}, \cdots, t_{a_{k+l-1}}] \, t^{a_1} \wedge \cdots \wedge t^{a_{k+l-1}} \,.

Therefore the condition d 2=0d^2 = 0 is equivalent to the condition

k+l=n1 σUnsh(l,k1)(1) sgn(σ)[[t a 1,,t a l],t a l+1,,t a k+l1]=0 \sum_{k+l = n-1} \sum_{\sigma \in Unsh(l,k-1)} (-1)^{sgn(\sigma)} [[t_{a_1}, \cdots, t_{a_l}], t_{a_{l+1}}, \cdots, t_{a_{k+l-1}}] = 0

for all nn \in \mathbb{N} and all {t a i𝔤}\{t_{a_i} \in \mathfrak{g}\}. This is equation (1) which says that {𝔤,{[,,] k}}\{\mathfrak{g}, \{[-,\dots,-]_k\}\} is an L L_\infty-algebra.

In terms of algebras over an operad

L L_\infty-algebras are precisely the algebras over an operad of the cofibrant resolution of the Lie operad.

Examples

Special cases

  • An L L_\infty-algebra for which VV is concentrated in the first nn degree is a Lie nn-algebra (sometimes also: “L nL_n-algebra”).

  • An L L_\infty-algebra for which only the unary operation and the binary bracket are non-trivial is a dg-Lie algebra: a Lie algebra internal to the category of dg-algebras. From the point of view of higher Lie theory this is a strict L L_\infty-algebra: one for which the Jacobi identity does happen to hold “on the nose”, not just up to nontrivial coherent isomorphisms.

  • So in particular

  • if 𝔤\mathfrak{g} is a Lie algebra over K\mathbf{K}, and b k1𝕂b^{k-1}\mathbb{K} is the complex consisting of the field 𝕂\mathbb{K} in degree 1k1-k, then an L L_\infty-algebra morphism from 𝔤\mathfrak{g} to b k1𝕂b^{k-1}\mathbb{K} is precisely a degree kk Lie algebra cocycle.

  • The skew-symmetry of the Lie bracket is retained strictly in L L_\infty-algebras. It is expected that weakening this, too, yields a more general vertical categorification of Lie algebras. For n=2n=2 this has been worked out by Dmitry Roytenberg: On weak Lie 2-algebras.

  • The horizontal categorification of L L_\infty-algebras are L L_\infty-algebroids.

  • An L L_\infty-algebra with only D nD_n non-vanishing is called an n-Lie algebra – to be distinguished from a Lie nn-algebra ! However, in large parts of the literature nn-Lie algebras are considered for which D nD_n is not of the required homogeneous degree in the grading, or in which no grading is considered in the first place. Such nn-Lie algebras are not special examples of L L_\infty-algebras, then. For more see n-Lie algebra.

  • An L L_\infty-alghebra internal to super vector spaces is a super L-∞ algebra.

Classes of examples

Specific examples

Properties

Model category structure

See model structure for L-∞ algebras.

Relation to dg-Lie algebras

Every dg-Lie algebra is in an evident way an L L_\infty-algebra. Dg-Lie algebras are precisely those L L_\infty-algebras for which all nn-ary brackets for n>2n \gt 2 are trivial. These may be thought of as the strict L L_\infty-algebras: those for which the Jacobi identity holds on the nose and all its possible higher coherences are trivial.

Theorem

Let kk be a field of characteristic 0 and write L Alg kL_\infty Alg_k for the category of LL\infty-algebras over kk.

Then every object of L Alg kL_\infty Alg_k is quasi-isomorphic to a dg-Lie algebra.

Moreover, one can find a functorial replacement: there is a functor

W:L Alg kL Alg k W : L_\infty Alg_k \to L_\infty Alg_k

such that for each 𝔤L Alg k\mathfrak{g} \in L_\infty Alg_k

  1. W(𝔨)W(\mathfrak{k}) is a dg-Lie algebra;

  2. there is a quasi-isomorphism

    𝔤W(𝔤). \mathfrak{g} \stackrel{\simeq}{\to} W(\mathfrak{g}) \,.

This appears for instance as (KrizMay, cor. 1.6).

Relation to \infty-Lie groupoids

In generalization to how a Lie algebra integrates to a Lie group, L L_\infty-algebras integrate to ∞-Lie groups.

See

Lie integration

and

Lie integrated ∞-Lie groupoids.

References

General

The original references are:

A quick web entry is:

See also

A discussion in terms of resolutions of the Lie operad is for instance in

See also for instance section 3.1 of:

A detailed reference for Lie 2-algebras is:

  • John Baez and Alissa Crans, Higher-dimensional algebra VI: Lie 2-algebras, TAC 12, (2004), 492–528. (arXiv)

For more general ‘weak Lie 2-algebras’, see:

Equipped with bilinear invariant polynomial

  • Yunfeng Jiang, Motivic Milnor fiber of cyclic L L_\infty-algebras (arXiv:0909.2858)

Revised on August 17, 2015 18:31:09 by Urs Schreiber (82.69.70.127)