Todd Trimble
Notes on operads and the Lie operad

Contents

The plug-in product

Let FB be the monoidal groupoid of finite sets and bijections, where the monoidal product is obtained by restricting the coproduct on the category of finite sets. Let F,G:FBV be species valued in a bicomplete symmetric monoidal closed category V. The plug-in product FG is defined by the formula

(FG)[S]= TSF[S/T]G[T](F \star G)[S] = \sum_{T \subseteq S} F[S/T] \otimes G[T]

where S/T denotes the pushout of T1 along the inclusion TS. This includes the degenerate case where T=0; in this case S/0 is the result of freely adjoining a point to S.

The set S/T has T/T as distinguished basepoint, and if U is complementary to T, we have a natural bijection S/TU+{T/T}. It follows that

(FG)[S] U+T=SF[U+*]G[T] U+T=SF[U]G[T] (FG)[S]\array{ (F \star G)[S] & \cong & \sum_{U + T = S} F[U + *] \otimes G[T] \\ & \cong & \sum_{U + T = S} F'[U] \otimes G[T] \\ & \cong & (F' \otimes G)[S] }

where FG refers to the usual Day convolution product, induced from the tensor product on FB. Since differentiation of species,

V FB+*V FB,V^{FB} \stackrel{- + *}{\to} V^{FB},

is cocontinuous, and since Day convolution is cocontinuous in each of its arguments, we see that

FGFGF \star G \cong F' \otimes G

is separately cocontinuous in each of its arguments F, G. We remark below on the fundamental implications of this observation.

For V=Set, a structure of species FG is given by three data:

Lax monoidal structure of plug-in product

We note that the plug-in poduct is not associative up to isomorphism, but there is a lax associativity. Specifically, we calculate

(FG)H (FG)H FGH+FGH FGH+F(GH)\array{ (F \star G) \star H & \cong & (F' \otimes G)' \otimes H \\ & \cong & F'' \otimes G \otimes H + F' \otimes G' \otimes H \\ & \cong & F'' \otimes G \otimes H + F \star (G \star H) }

so that there is a noninvertible associativity (an inclusion)

α FGH:F(GH)(FG)H\alpha_{F G H}: F \star (G \star H) \to (F \star G) \star H

which is natural in each of its arguments F, G, and H, and which satisfies an evident pentagon coherence condition. Additionally, there is a lax unit, defined as the species X for which X[S] is the monoidal unit of V if card(S)=1, else X[S] is initial, for which we have evident natural maps

λ F:XFF,ρ F:FXF\lambda_F: X \star F \to F, \qquad \rho_F: F \star X \to F

The first map λ F is invertible, but the second is not: its component at a finite set S is the codiagonal

:SF[S]F[S]\nabla: S \cdot F[S] \to F[S]

However, the standard unit coherence conditions for monoidal categories hold. We may call such a structure, relaxing the condition of invertibility of α and ρ but retaining the naturality and coherence conditions, a lax monoidal category.

In the sequel, F n will denote the iterated plug-in product defined recursively by

F 0=X,F (n+1)=F nFF^{\star 0} = X, \qquad F^{\star (n+1)} = F^{\star n} \star F

so that all parentheses in F n are to the left. We have a partial coherence theorem:

Proposition

Any two maps

F mF nF (m+n)F^{\star m} \star F^{\star n} \to F^{\star (m+n)}

definable in the language of lax monoidal categories are equal. This map by α mn.

For F a Set-valued species, an element of F n is described by data as follows:

  1. A structure of tree constructed from n sprouts in a prescribed plugging order;

  2. An element of F[σ] for each sprout σ used in the construction of the tree in 1.

Thus we expect some connection between the “geometric series” nF n and the free operad on F. The only difference between the two is that the standard construction of free operads via trees is in no way dependent on plugging order. Thus, to obtain the free operad, we have to mod out by isomorphisms on F n induced by “inessential” differences in plugging order. This will be described in detail in the section on trees and hierarchies.

Interlude: the importance of separate cocontinuity of plug-in

The are various well-traveled roads to operads and free operads. For example, there is a well-known concise definition of operad, as monoid in the category of species with respect to a monoidal product known as the substitution product (denoted ). It may therefore be wondered what is the point of introducing another approach, based on the somewhat outré “lax monoidal product” we have named the plug-in product.

A key point for us is that the substitution product is only partially adapted to developing the analogy between bar/cobar constructions for algebras and bar/cobar constructions for operads, and that the plug-in product is actually better adapted. On the substitution product side, we have the analogy

Now the bar construction of an algebra is based on a simplicial resolution of A whose components A n are the graded components of the free (tensor) algebra nA n. (There are in fact several versions of bar resolutions; another is based on the tensor algebra of an augmentation ideal.) At this point however the analogy on the side of operads begins to break down, because the free operad on (the underlying species) M is not nM n.

Analyzing the reasons for failure of the analogy, we recall the following classical result:

Proposition

Let (C,,I) be a monoidal category. Then the free monoid on an object c is provided by the geometric series nc n provided that the tensor product ab distributes over countable coproducts, i.e., preserves countable coproducts in each of the separate arguments a and b.

The substitution product FG does not distribute over countable coproducts. It is true that the functor G preserves countable products, and this partial result can be exploited to yield simplicial resolutions (as we recount below in our summary of Joyal’s approach to the Lie species), but the bar constructions on operads (as in the work on Koszul operads initiated by Ginzburg and Kapranov) are not of this character.

At this point we recall our earlier observation that does distribute over countable coproducts (and over colimits generally). We wish to pursue the very rough analogy:

Operads are like certain monoids in the lax monoidal category of species under the plug-in product. The free operad on a species is like a geometric series nF n. The bar resolution of an operad has components based on the graded components F n.

As is obvious from the phrasing (and as already noted), the analogy is still somewhat imperfect, but the imperfections are easier to smooth over than is the case with the earlier analogy based on the substitution product. The analogy is actually worth pursuing for its own sake, as it lays bare other connections between algebras and operads which as far as I know have attracted little notice in the literature.

Trees and hierarchies

It is interesting to resolve F n into a sum of monomials, each of which is a tensor product of derivatives of F. For example, we have

FF FF (FF)F FF 2+FFF F 4 FF 3+2FFF+FFFF+FFF 2+(F) 3F\array{ F \star F & \cong & F' \otimes F \\ (F \star F) \star F & \cong & F'' \otimes F^{\otimes 2} + F' \otimes F' \otimes F \\ F^{\star 4} & \cong & F''' \otimes F^{\otimes 3} + 2 F'' \otimes F' \otimes F + F'' \otimes F \otimes F' \otimes F + F' \otimes F'' \otimes F^{\otimes 2} + (F')^{\otimes 3} \otimes F }

There are in fact (n1)! monomial terms, each a tensor product of n derivative terms, and each corresponding to a function f:[2,n][1,n] such that f(k)<k for all k[2,n]. We call such an f a hierarchy. In detail, if the S th derivative of a species F is defined by F (S)[T]=F[S+T], we have

F n= fHier(n)F (f 1(1))F (f 1(2))F (f 1(n1))FF^{\star n} = \sum_{f \in Hier(n)} F^{(f^{-1}(1))} \otimes F^{(f^{-1}(2))} \otimes \ldots \otimes F^{(f^{-1}(n-1))} \otimes F

as may be proved by an easy induction.

Each fHier(n) endows the set [1,n] with a rooted tree structure. A rooted tree structure on a set S is precisely a function f:SS with a fixed point r (the root) such that for every sS, some finite iterate of f takes s to r, i.e., f (k)(s)=r for some k. (The function f simply moves a vertex one step closer to the root along the unique path to the root.) Given a hierarchy f:[2,n][1,n], the corresponding rooted tree structure on [1,n] is the extension of f which fixes the point 1. The well-ordering of the vertices 1<<n serves to prescribe a plugging order on sprouts, starting from the sprout at the root 1, plugging the sprout rooted at 2 to a leaf of the initial sprout, and so on.

Somewhat more generally, given any finite well-ordered set S={s 1<<s n}, a hierarchy on S is a function f:S{s 1}S such that f(s i)<s i for 1<in, and this also gives rise to a rooted tree structure. A rooted tree whose vertices are equipped with a well-ordering and whose tree structure arises from a hierarchy function will be called a hierarchical tree. The root r is of course the initial element s 1.

Now by removing the root of a finite rooted tree (S,f,r), we obtain a forest of finite rooted trees S/s whose roots s range over f 1(r). An isomorphism Φ of rooted trees (S,f,r)(S,f,r) may be described recursively as consisting of

  1. A bijection ϕ:f 1(r)(f) 1(r);

  2. Isomorphisms of rooted trees Φ s:S/sS/ϕ(s) where sf 1(r).

If (S,f) is hierarchical, then the subtrees S/s inherit hierarchical tree structures from S by restricting both the well-ordering and f:SS to a function f:S/sS/s. Thus isomorphisms of hierarchical trees S (defined to be isomorphisms of the underlying rooted trees) can also be described recursively.

If (S={s 1<<s n},f) is a hierarchical tree and F is a species, we define a species

F S,f= k=1 nF (f 1(s k))F_{S, f} = \bigotimes_{k=1}^{n} F^{(f^{-1}(s_k))}

Observe that there is a well-defined isomorphism

F S,fF (f 1(r))( sf 1(r)F S/s,f)F_{S, f} \cong F^{(f^{-1}(r))} \otimes (\bigotimes_{s \in f^{-1}(r)} F_{S/s, f|})

by permuting tensor factors. (The order of tensor factors on either side is fixed by the well-ordering of vertices.)

Now if Φ:(S,f)(S,f) is an isomorphism of hierarchical trees, we define by recursion a species isomorphism to be the composite of

F (f 1(s 1)) sf 1(s 1)F S/s,fF (ϕ)( sf 1(s 1)F Φ s)F ((f) 1(r))( s(f) 1(r)F S/s,f)F^{(f^{-1}(s_1))} \otimes \bigotimes_{s \in f^{-1}(s_1)} F_{S/s, f|} \stackrel{F^{(\phi)} \otimes (\bigotimes_{s \in f^{-1}(s_1)} F_{\Phi_s})}{\to} F^{((f')^{-1}(r'))} \otimes (\bigotimes_{s' \in (f')^{-1}(r')} F_{S'/s', f'|})

with factor-permuting isomorphisms

F S,fF (f 1(r))( sf 1(r)F S/s,f)F ((f) 1(r))( s(f) 1(r)F S/s,f)F S,fF_{S, f} \cong F^{(f^{-1}(r))} \otimes (\bigotimes_{s \in f^{-1}(r)} F_{S/s, f|}) \qquad F^{((f')^{-1}(r'))} \otimes (\bigotimes_{s' \in (f')^{-1}(r')} F_{S'/s', f'|}) \cong F_{S', f'}

on either side. The result is an isomorphism F Φ:F S,fF S,f.

There is a groupoid consisting of hierarchical trees and isomorphisms between them. Let Hier n be the full subgroupoid whose objects are hierarchical tree structures on [1,n]. (The isomorphisms in Hier n can be thought of as permutations on [1,n] that relabel vertices to give a new plugging order.) Given a V-species F, there is a functor

Hier nV FB:(Φ:([1,n],f)([1,n],g))F [1,n],fF [1,n],gHier_n \to V^{FB}: (\Phi: ([1, n], f) \to ([1, n], g)) \mapsto F_{[1, n], f} \to F_{[1, n], g}

which defines a diagram in the category of species. The colimit of this diagram is a certain quotient of

F n= fHier(n)F [1,n],fF^{\star n} = \sum_{f \in Hier(n)} F_{[1, n], f}

which we will denote as F n/Hier n. By taking this quotient, we eliminate dependence on a specified hierarchy or plugging order.

The constructions are made in view of the following result.

Theorem

The species

𝒪(F)= n0F n/Hier n\mathcal{O}(F) = \sum_{n \geq 0} F^{\star n}/Hier_n

is the underlying species of the free operad on F.

This will be explored further in the next section.

Operads

The notion of (permutative) V-valued operad may be formulated in terms of the plug-in product. Specifically, if M is an operad, there is a natural map

ψ:MMMM\psi: M \star M \to M \circ M$ so that an operad, defined as a monoid $(M, \mu: M \circ M \to M, u: X \to M)$ in the monoidal category, gives rise to a monoid-like structure with respect to the plug-in product, whose multiplication $m: M \star M \to M$ is the composite $$M \star M \stackrel{\psi}{\to} M \circ M \stackrel{\mu}{\to} M

and whose unit is again u:XM. The structure of an operad M may be described entirely in terms of m:MMM, u:XM, and the notion of operad may be recast as appropriate -monoids (with a slight extra twist).

The map ψ is not hard to describe: the plug-in product FG may be alternatively described by the formula

(FG)[S]= k 1kk(F[k]G^ j)[S](F \star G)[S] = \sum_k \sum_{1 \leq k \leq k} (F[k] \otimes \widehat{G}_j)[S]

where G^ j is a tensor product of k factors

XGX,X \otimes \ldots \otimes G \otimes \ldots \otimes X,

all factors being X except for a single instance of G as the j th tensor factor. If there is in addition a map i:XG (for example, a unit map u:XM of an operad), there is an induced map

i j=i1 Gi:G^ jG ki_j = i \otimes \ldots \otimes 1_G \otimes \ldots \otimes i: \widehat{G}_j \to G^{\otimes k}

and under these circumstances there is an induced map

FG kF[k]G k kF[k] S kG k=FGF \star G \to \sum_k F[k] \otimes G^{\otimes k} \to \sum_k F[k] \otimes_{S_k} G^{\otimes k} = F \circ G

These must satisfy the following axioms:

  1. F(FF) 1m FF α m (FF)F m1FF m F\array{ F \star (F \star F) & \stackrel{1 \star m}{\to} & F \star F & \\ \alpha \downarrow & & & \searrow m \\ (F \star F) \star F & \underset{m \star 1}{\to} F \star F & \underset{m}{\to} & F }

    commutes;

  2. The two composites FF 2F named in

    F(FF)1σF(FF)i(FF)Fm(m1)FF'' \otimes (F \otimes F) \stackrel{\overset{\sigma}{\to}}{\underset{1}{\to}} F'' \otimes (F \otimes F) \stackrel{i}{\to} (F \star F) \star F \stackrel{m(m \star 1)}{\to} F

    commute. Here i denotes the inclusion complementary to α:F(FF)(FF)F, and σ is the involution

    σ 1σ 2:FF 2FF 2\sigma_1 \otimes \sigma_2: F'' \otimes F^{\otimes 2} \to F'' \otimes F^{\otimes 2}

    where σ:FF is the natural involution on the second derivative

    ()=(+2)*:V FBV FB(-)'' = (- + 2)*: V^{FB} \to V^{FB}

    induced by the nonidentity involution 22, and σ 2 is the symmetry isomorphism F 2F 2.

Revised on July 20, 2010 01:43:03 by Todd Trimble