nLab
(infinity,1)-operad

Context

Higher algebra

(,1)(\infty,1)-Category theory

Contents

Idea

The notion of (,1)(\infty,1)-operad is to that of (∞,1)-category as operad is to category.

So, roughly, an (,1)(\infty,1)-operad is an algebraic structure that has for each given type of input and one type of output an ∞-groupoid of operations that take these inputs to that output.

There is a fairly evident notion of ∞-algebra over an (∞,1)-operads. Examples include

(,1)(\infty,1)-Operads form an (∞,2)-category (∞,1)Operad.

Definitions

Two models for (,1)(\infty,1)-operads exist to date, one by CisinskiMoerdijkWeiss, the other by Lurie. It is expected though not yet entirely proved that the two are equivalent (Higher Algebra, draft, Remark 2.0.0.8).

The first one models (,1)(\infty,1)-operads as dendroidal sets in close analogy to (in fact as a generalization of how) simplicial sets model (∞,1)-categories.

The second models the (∞,1)-category version of a category of operators of an operad.

In terms of dendroidal sets

Here simplicial sets are generalized to dendroidal sets. The theory of (,1)(\infty,1)-operads is then formulated in terms of dendroidal sets in close analogy to how the theory of (∞,1)-categories is formulated in terms of simplicial sets.

There is a model structure on dendroidal sets whose fibrant objest are the quasi-operads in direct analogy to the notion of quasi-category.

So the model structure on dendroidal sets is a presention of the (∞,1)-category of (,1)(\infty,1)-operads. It is Quillen equivalent to the standard model structure on operads enriched over Top or sSet. Therefore, conversely, the traditional homotopy-theoretic constructions on topological and chain operads (such as cofibrant resolutions in order to present homtopy algebras such as A-∞ algebras, L-∞ algebras, homotopy BV-algebras and the like) are also indeed presentations of (,1)(\infty,1)-operads.

In terms of (,1)(\infty,1)-categories of operators

Every operad AA encodes and is encoded by its category of operators C AC_A. In the approach to (,1)(\infty,1)-operators described below, the notion of category of operators is generalized to an (∞,1)-category of operators.

In this approach an (,1)(\infty,1)-operad C C^\otimes is regarded as an (∞,1)-category CC – the unary part of the (,1)(\infty,1)-operad to be described– with extra structure that determines (∞,1)-functors C ×nCC^{\times n} \to C.

This and the conditions on these are encoded in requiring that C C^\otimes is an (,1)(\infty,1)-functor C ΓC^\otimes \to \Gamma over Segal's category Γ\Gamma of pointed finite sets, satisfying some conditions.

In particular, any symmetric monoidal (∞,1)-category yields an example of an (,1)(\infty,1)-operad in this sense. In fact, symmetric monoidal (,1)(\infty,1)-categories can be defined as (,1)(\infty,1)-operads such that the functor C ΓC^\otimes \to \Gamma is a coCartesian fibration. (For the moment, see monoidal (infinity,1)-category for more comments and references on higher operads in this context.)

This is the approach described in (LurieCommutative)

Basic definitions

We are to generalize the following construction from categories to (∞,1)-categories.

Definition

For 𝒪\mathcal{O} a symmetric multicategory, write 𝒪 FinSet */\mathcal{O}^\otimes \to FinSet^{*/} for its category of operators.

Here 𝒪 \mathcal{O}^\otimes is the category whose

  • objects are finite sequences (tuples) of objects of 𝒪\mathcal{O};

  • morphisms (X 1,,X n 1)(Y 1,,Y n 2)(X_1, \cdots, X_{n_1}) \to (Y_1, \cdots, Y_{n_2}) are given by a morphism α:n 1n 2\alpha \colon \langle n_1\rangle \to \langle n_2\rangle in FinSet *FinSet_* together with a collection of multimorphisms

    {ϕ j𝒪({X i} iα 1{j},Y j)} 1jn 2. \left\{ \phi_j \in \mathcal{O}\left( \left\{ X_i\right\}_{i \in \alpha^{-1}\left\{j\right\}} , Y_j \right) \right\}_{1 \leq j \leq n_2} \,.

The functor p:𝒪 FinSet */p \colon \mathcal{O}^\otimes \to FinSet^{*/} is the evident forgetful functor.

In (Lurie) this is construction 2.1.1.7.

This motivates the following definition of the generalization of this situation to (∞,1)-category theory.

Definition

Write FinSet */FinSet^{*/} for the category of pointed finite set (Segal's Gamma-category).

For nn \in \mathbb{N} we write

n*[n]FinSet */ \langle n\rangle \coloneqq {*} \coprod [n] \in FinSet^{*/}

for the pointed set with n+1n+1 elements.

A morphism in FinSet */FinSet^{*/}

For nn \in \mathbb{N} and 1in1 \leq i \leq n write

ρ i:n1 \rho^i \colon \langle n\rangle \to \langle 1\rangle

for the inert morphism that sends all but the iith element to the basepoint.

Notice that for each nn \in \mathbb{N} there is a unique active morphism n1\langle n\rangle \to \langle 1\rangle.

(Lurie, def. 2.1.1.8)

Definition

The (,1)(\infty,1)-category of operators of an (,1)(\infty,1)-operad is a morphism

p:𝒪 FinSet */ p \colon \mathcal{O}^\otimes \to FinSet^{*/}

of quasi-categories such that the following conditions hold:

  1. For every inert morphism in FinSet */FinSet^{*/} and every object over it, there is a lift to a pp-coCartesian morphism in 𝒪 \mathcal{O}^\otimes. In particular, for f:n 1n 2f \colon \langle n_1\rangle \to \langle n_2\rangle inter, there is an induced (∞,1)-functor

    f !:𝒪 n 1 𝒪 n 2 . f_! \colon \mathcal{O}^\otimes_{\langle n_1\rangle} \to \mathcal{O}^\otimes_{\langle n_2\rangle} \,.
  2. The coCartesian lifts of the inert projection morphisms induce an equivalence of derived hom-spaces in 𝒪 \mathcal{O}^{\otimes} between maps into multiple objects and the products of the maps into the separete objects:

    For f:n 1n 2f \colon \langle n_1 \rangle \to \langle n_2 \rangle write 𝒪 f (,)𝒪 (,)\mathcal{O}^\otimes_f(-,-) \hookrightarrow \mathcal{O}^\otimes(-,-) for the components of the derived hom-space covering f,thenthef, then the (\infty,1)$-functor

    𝒪 f (C 1,C 2)1kn 2𝒪 ρ if (C 1,(C 2) i) \mathcal{O}^\otimes_f(C_1,C_2) \to \underset{1 \leq k \leq n_2}{\prod} \mathcal{O}^\otimes_{\rho^i\circ f}(C_1,(C_2)_i)

    induced as above is an equivalence.

  3. For every finite collection of objects C 1,c n𝒪 1 C_1, \cdots c_n \in \mathcal{O}^\otimes_{\langle 1\rangle} there exists a multiobject C𝒪 nC \in \mathcal{O}^\otimes_\langle n\rangle and a collection of pp-coCartesian morphisms {CC i}\{C \to C_i\} covering ρ i\rho^i.

    Equivalently (given the first two conditions): for all nn \in \mathbb{N} the (,1)(\infty,1)-functors {(ρ i) !} 1in\{(\rho^i)_!\}_{1 \leq i \leq n} induce an equivalence of (∞,1)-categories

    𝒪 n (𝒪 1 ) × n \mathcal{O}^\otimes_{\langle n\rangle} \to (\mathcal{O}^\otimes_{\langle 1\rangle})^{\times^n}

(Lurie, def. 2.1.1.10, remark 2.1.1.14)

Remark

The conditions in def. 3 mean that p:𝒪 FinSet */p \colon \mathcal{O}^\otimes \to FinSet^{*/} encodes

  1. an (∞,1)-category 𝒪𝒪 1 \mathcal{O} \coloneqq \mathcal{O}^\otimes_{\langle 1\rangle};

  2. for each nn \in \mathbb{N} an nn-ary operation given by the (,1)(\infty,1)-functor

    𝒪 n=(𝒪 1 ) ×n𝒪 n 𝒪 1 =𝒪 \mathcal{O}^{n} = (\mathcal{O}^\otimes_{\langle 1\rangle})^{\times n} \simeq \mathcal{O}^{\otimes}_{\langle n\rangle} \to \mathcal{O}^\otimes_{\langle 1\rangle} = \mathcal{O}

    induced by the unique active morphism n1\langle n\rangle \to \langle 1\rangle

  3. a coherently associative multicomposition of these operations.

Remark

The notion of def. 3 may also be called a symmetric (,1)(\infty,1)-multicategory or colored (,1)(\infty,1)-operad. The colors are the objects of 𝒪\mathcal{O}.

We now turn to the definition of homomorphisms of (,1)(\infty,1)-operads.

Definition

Given an (,1)(\infty,1)-operad p:𝒪 FinSet */p \colon \mathcal{O}^\otimes \to FinSet^{*/} as in def. 3, a morphism ff in 𝒪 \mathcal{O}^\otimes is called an inert morphism if

  1. p(f)p(f) is an inert morphism in FinSet */FinSet^{*/} by def. 2;

  2. ff is a pp-coCartesian morphism.

Definition

A morphism of (∞,1)-operads is a map between their (∞,1)-categories of operators over FinSet */FinSet^{*/} that preserves the inert morphisms of def. 4.

Morphisms of operads 𝒪 1𝒪 2\mathcal{O}_1 \to \to \mathcal{O}_2 can be understood equivalently as exhibiting an 𝒪 1\mathcal{O}_1-algebra in 𝒪 2\mathcal{O}_2. Therefore:

Definition

For 𝒪 1,𝒪 2\mathcal{O}_1, \mathcal{O}_2 to (,1)(\infty,1)-operads, write

Alg 𝒪 1(𝒪 2)qCat /FinSet */(𝒪 1,𝒪 2) Alg_{\mathcal{O}_1}(\mathcal{O}_2) \hookrightarrow qCat_{/FinSet^{*/}}(\mathcal{O}_1, \mathcal{O}_2)

for the full sub-(∞,1)-category of the (∞,1)-functor (∞,1)-category on those that are morphisms of (∞,1)-operads by def. 5.

(Lurie, def. 2.1.2.7).

We also have the notion of

See there for more details.

Model for (,1)(\infty,1)-categories of operators

There is a model category that presents the (∞,1)-category (,1)Cat Oper(\infty,1)Cat_{Oper} of (,1)(\infty,1)-categories of operations.

Proposition

There exists a

model category 𝒫Op (,1)\mathcal{P} Op_{(\infty,1)}

  • whose underlying category has

    • objects are marked simplicial set SS equipped with a morphism SN(FinSet *)S \to N(FinSet_*) such that marked edges map to inert morphisms in FinSet *FinSet_* (those for which the preimage of the marked point contains just the marked point)

    • morphisms are morphisms of marked simplicial sets STS \to T such that the triangle

      S T N(FinSet *) \array{ S &&\to&& T \\ & \searrow && \swarrow \\ && N(FinSet_*) }

      commutes;

  • which is canonically an SSet-enriched category;

  • and whose model structure is given by

    • cofibrations are those morphisms whose underlying morphisms of simplicial sets ate cofibrations, hence monomorphisms

    • weak equivalences are those morphisms STS \to T such that for all AN(FinSet *)A \to N(FinSet_*) that are (,1)(\infty,1)-categories of operations by the above definition, the morphism of SSet-hom objects

      𝒫Op (T,A)𝒫Op (S,A) \mathcal{P}Op_\infty(T,A) \to \mathcal{P}Op_\infty(S,A)

      is a homotopy equivalence of simplicial sets.

    • an object is fibrant if and only if it is an (,1)(\infty,1)-category of operations, by the above definition.

This is prop 1.8 4 in

Examples

We list some examples of (,1)(\infty,1)-operads incarnated as their (∞,1)-categories of operators by def. 3.

The first basic examples to follow are in fact all given by 1-categories of operators.

Example

The identity functor on FinSet */FinSet^{*/} exhibits an (.1)(\infty.1)-operad. This is the commutative operad

Comm =FinSet */idFinSet */. Comm^\otimes = FinSet^{*/} \stackrel{id}{\to} FinSet^{*/} \,.

The (∞,1)-algebras over an (∞,1)-operad over this (,1)(\infty,1)-operad are E-∞ algebras.

Example

The associative operad has Assoc Assoc^\otimes the category whose objects are the natural numbers, whose nn-ary operations are labeled by the total orders on nn elements, equivalently the elements of the symmetric group Σ n\Sigma_n, and whose composition is given by forming consecutive total orders in the obvious way.

The (∞,1)-algebras over an (∞,1)-operad over this (,1)(\infty,1)-operad are A-∞ algebras

In (Lurie) this is remark 4.1.1.4.

Example

The operad for modules over an algebra LMLM is the colored symmetric operad whose

  • objects are two elements, to be denoted 𝔞\mathfrak{a} and 𝔫\mathfrak{n};

  • multimorphisms (X i) i=1 nY(X_i)_{i = 1}^n \to Y form

    • if Y=𝔞Y = \mathfrak{a} and X i=𝔞X_i = \mathfrak{a} for all ii then: the set of linear orders on nn elements, equivalently the elements of the symmetric group Σ n\Sigma_n;

    • if Y=𝔫Y = \mathfrak{n} and exactly one of the X i=𝔫X_i = \mathfrak{n} then: the set of linear order {i 1<<i n}\{i_1 \lt \cdots \lt i_n\} such that X i n=𝔫X_{i_n} = \mathfrak{n}

    • otherwise: the empty set;

  • composition is given by composition of linear orders as for the associative operad.

The (∞,1)-algebras over an (∞,1)-operad over this (,1)(\infty,1)-operad are pairs consisting of A-∞ algebras with (∞,1)-modules over them.

In (Lurie) this appears as def. 4.2.1.1.

Definition

The operad for bimodules over algebras BModBMod is the colored symmetric operad whose

  • objects are three elements, to be denoted 𝔞 ,𝔞 +\mathfrak{a}_-, \mathfrak{a}_+ and 𝔫\mathfrak{n};

  • multimorphisms (X i) i=1 nY(X_i)_{i = 1}^n \to Y form

    • if Y=𝔞 Y = \mathfrak{a}_- and all X i=𝔞 X_i = \mathfrak{a}_- then: the set of linear orders of nn elements;

    • if Y=𝔞 *Y = \mathfrak{a}_* and all X i=𝔞 *X_i = \mathfrak{a}_* then again: the set of linear orders of nn elements;

    • if Y=𝔫Y = \mathfrak{n}: the set of linear orders {i 1<<i n}\{i_1 \lt \cdots \lt i_n\} such that there is exactly one index i ki_k with X i k=𝔫X_{i_k} = \mathfrak{n} and X i j=𝔞 X_{i_j} = \mathfrak{a}_- for all j<kj \lt k and X i j=𝔞 +X_{i_j} = \mathfrak{a}_+ for all k>kk \gt k.

  • composition is given by the composition of linear orders as for the associative operad.

The (∞,1)-algebras over an (∞,1)-operad over this (,1)(\infty,1)-operad are pairs consisting of two A-∞ algebras with an (∞,1)-bimodule over them.

Properties

Relation between the two definitions

At the time of this writing there is no discussion in “the literature” of the relation between the definition of (,1)(\infty,1)-operads in terms of dendroidal sets (Cisinski, Moerdijk, Weiss) and (,1)(\infty,1)-categories of operators (Lurie). The following are some tentative observations. - Urs

update: meanwhile this has been worked out by some people. Results should appear in preprint form soon.

There is an obvious way to regard a tree as an (,1)(\infty,1)-category of operators:

Definition

(dendroidal (,1)(\infty,1)-category of operators)

Let

ω:ΩOpC ()Cat/FinSet *N𝒫Op (,1) \omega : \Omega \hookrightarrow Op \stackrel{C_{(-)}}{\to} Cat/FinSet_* \stackrel{N}{\to} \mathcal{P}Op_{(\infty,1)}

be the dendroidal object given by the following composition:

  • ΩOp\Omega \hookrightarrow Op is the functor from the tree category Ω\Omega to the category of symmetric colored operads (over Set) that sends a tree to the operad freely generated from it;

  • OpC ()Cat/FinSet *Op \stackrel{C_{(-)}}{\to} Cat/FinSet_* sends an operad to its category of operators;

  • Cat/FinSet *N𝒫Op (,1)Cat/FinSet_* \stackrel{N}{\to} \mathcal{P}Op_{(\infty,1)} takes the nerve of this category, regarded as a marked simplicial set over N(FinSet *)N(FinSet_*), whose marked edges are the inert morphisms in the category of operations.

Following the general pattern of nerve and realization, we get:

Definition

(dendroidal nerve of Lurie-\infty-operad)

The functor

N d:=Hom 𝒫Op (,1)(ω(),):𝒫Op (,1)dSet N_d := Hom_{\mathcal{P}Op_{(\infty,1)}}(\omega(-), -): \mathcal{P}Op_{(\infty,1)} \to dSet

that sends a marked simplicial set AN(FinSet *)A \to N(FinSet_*) to the dendroidal set which sends a tree TT to the set of morphisms of ω(T)\omega(T) into AA

N d(A):THom 𝒫Op (,1)(ω(T),A) N_d(A) : T \mapsto Hom_{\mathcal{P}Op_{(\infty,1)}}(\omega(T), A)

is the dendroidal nerve of AA.

One expects that N dN_d induces a Quillen adjunction and indeed a Quillen equivalence between the above model category structure on 𝒫Op (,1)\mathcal{P}Op_{(\infty,1)} and the model structure on dendroidal sets. The following is as far as I think I can prove aspects of this. -Urs.

Proposition

The dendroidal nerve functor has the following properties:

  • it is the right adjoint of a SSet-enriched adjunction

    C ():dSet𝒫Op (,1):N d C_{(-)} : dSet \stackrel{\leftarrow}{\to} \mathcal{P}Op_{(\infty,1)} : N_d
  • it sends fibrant objects to fibrant objects

    i.e. it sends (,1)(\infty,1)-categories of operations to (,1)(\infty,1)-operads in their incarnation as “quasi-operads”;

  • it sends objects π:AN(FinSet *)\pi : A \to N(FinSet_*) that come from grouplike symmetric monoidal ∞-groupoids to fully Kan dendroidal sets (that have the extension property with respect to all horns)

  • it sends objects π:AN(FinSet *)\pi : A \to N(FinSet_*) that come from symmetric monoidal (∞,1)-categories to dendroidal sets that have the extension property with respect to at least one outer horn Λ vT\Lambda_{v} T for vTv \in T an nn-corolla, for all nn \in \mathbb{N}.

  • its left adjoint sends cofibrations to cofibrations and acyclic cofibrations with cofibrant domain to acyclic cofibrations.

Proof

respect for fibrant objects. If AN(FinSet *)A \to N(FinSet_*) is fibrant, then in particual AA is a weak Kan complex hence has the extension property with respect to all inner horn inclusions of simplices. We need to show that this implies that N d(A)N_d(A) has the extension property with respect to all inner horn inclusions of trees.

By an (at the moment unpublished) result by Moerdijk, right lifting property with respect to inner horn inclusions of trees is equivalend to right lifting property with respect to inclusions of spines of trees: the union over all the corollas in a tree.

For this the extension property means that if we find a collection {C k iN d(A)}=Sp(T)\{C_{k_i} \to N_d(A)\} = Sp(T) of corollas in N d(A)N_d(A) that match at some inputs and output, then these can be composed to an image TN d(A)T \to N_d(A) of the corresponding tree TT in N d(A)N_d(A).

An image of TT in N d(A)N_d(A) is an image of ω(T)\omega(T) in AA. In the category of operators ω(A)\omega(A) every tree may be represented as the composite of a sequence of morphisms each of which consists of precisely one of the corollas C k iC_{k_i} in parallel to identity morphisms. This way gluing the tree from the corollas is a matter of composing a sequence of edges in AA. But this is guaranteed to be possible if AA is a weak Kan complex.

symmetric monoidal product and outer horn lifting

As described at cartesian morphism, an edge f:Δ 1Af : \Delta^1 \to A in AA is coCartesian if for all diagrams

Δ 0,1 f Λ 0 n A Δ n N(FinSet *) \array{ \Delta^{0,1} \\ \downarrow & \searrow^f \\ \Lambda^n_0 &\to & A \\ \downarrow && \downarrow \\ \Delta^n &\to& N(FinSet_*) }

of 0-horn lifting problems where the first edge of the horn is ff itself, there exists a lift

Δ 0,1 f Λ 0 n A Δ n N(FinSet *). \array{ \Delta^{0,1} \\ \downarrow & \searrow^f \\ \Lambda^n_0 &\to & A \\ \downarrow &\nearrow & \downarrow \\ \Delta^n &\to& N(FinSet_*) } \,.

For ff the parallel application of an nn-corolla with a collection of identity morphisms this implies that any outer horn Λ vTN d(A)\Lambda_v T \to N_d(A) for which the vertex v:C nN d(A)v : C_n \to N_d(A) maps to ff, the dendroidal set N d(A)N_d(A) has the extension property with respect to the inclusion Λ dTT\Lambda_d T \hookrightarrow T.

the left adjoint and its respect for cofibrations

By general nonsense the left adjoint to N dN_d is given by the coend

C ():dSet𝒫Op (,1) C_{(-)} : dSet \to \mathcal{P}Op_{(\infty,1)}
C P= TΩω(T)P(T), C_P = \int^{T \in \Omega} \omega(T) \cdot P(T) \,,

where in the integrand we have the tautological tensoring of 𝒫Op (,1)\mathcal{P}Op_{(\infty,1)} over Set.

Notice that ω:Ω𝒫Op (,1)\omega : \Omega \to \mathcal{P}Op_{(\infty,1)} is an SSet-enriched functor for the ordinary category Ω\Omega regarded as a simplicially enriched category by the canonical embedding SetSSetSet \hookrightarrow SSet. Therefore this adjunction FN dF \dashv N_d is defined entirely in SSet-enriched category theory and hence is a simplicial adjunction.

The model structure on dendroidal sets has a set of generating cofibrations given by the boundary inclusions of trees. Ω[T]Ω[T]\partial \Omega[T] \hookrightarrow \Omega[T]. Tese evidenly map to monomorphisms of underlying simplicial sets under FF, hence to cofibrations.

For f:PQf : P \hookrightarrow Q an acyclic cofibration with cofibrant domain, we need to check that C f:C XC YC_f : C_X \to C_Y is a weak equivalence in 𝒫Op (,1)\mathcal{P}Op_{(\infty,1)}. This is by definition the case if for every fibrant object AA the morphism

𝒫Op (,1)(C Y,A)𝒫Op (,1)(C X,A) \mathcal{P}Op_{(\infty,1)}(C_Y,A) \to \mathcal{P}Op_{(\infty,1)}(C_X,A)

is a weak equivalence in the standard model structure on simplicial sets. By the simplicial adjunction FN dF \dashv N_d this is equivalent to

dSet(f,N d(A)):dSet(Y,N d(A))dSet(X,N d(A)) dSet(f,N_d(A)) : dSet(Y,N_d(A)) \to dSet(X,N_d(A))

being a weak equivalence. By the above N d(A)N_d(A) is fibrant. By section 8.4 of the lecture notes on dendroidal sets cited at model structure on dendroidal sets a morphism between cofibrant dendroidal sets is a weak equivalence precisely if homming it into any fibrant dendroidal set produces an equivalence of homotopy categories.

Since ff is a weak equivalence between cofibrant objects by assumption, it follows that indeed dSet(f,N d(A))dSet(f,N_d(A)) is a weak equivalence for all fibrant AA.

(AHM, or does it? there is a prob here, but I need to run now…)

Hence C fC_f is a weak equivalence.

References

The formulation in terms of dendroidal sets is due to

Here are two blog entries on talks on this stuff:

The formulation in terms of an (,1)(\infty,1)-version of the category of operators is introduced in

and further discussed in

Now in section 2 of the textbook

The equivalence between the dendroidal set-formulation and the one in terms of (,1)(\infty,1)-categories of operators is shown in

Revised on May 18, 2013 04:52:33 by Urs Schreiber (82.113.99.177)