nLab dendroidal homotopy coherent nerve



Homotopy theory

homotopy theory, (∞,1)-category theory, homotopy type theory

flavors: stable, equivariant, rational, p-adic, proper, geometric, cohesive, directed

models: topological, simplicial, localic, …

see also algebraic topology



Paths and cylinders

Homotopy groups

Basic facts


Higher algebra



The dendroidal homotopy coherent nerve is an operadic generalization of the standard homotopy coherent nerve. It is a functor

hcN d:TopOperaddSet hcN_d : Top Operad \to dSet

from the category of Top-operads to that of dendroidal sets, given by

hcsN d(P):TdSet(W H(T),P), hcsN_d(P) : T \mapsto dSet(W_H(T), P) \,,

where TT is an object of the tree category, regarded as a free symmetric operad, and W H(T)W_H(T) is its Boardman-Vogt resolution.


Throughout, let \mathcal{E} be a symmetric monoidal model category equipped with an interval object HH as discussed at model structure on operads and at Boardman-Vogt resolution. We consider multi-coloured symmetric operads (symmetric multicategories) enriched in \mathcal{E}.

Standard examples are =\mathcal{E} = Top, sSet, which yields topological operads and simplicial operads, respectively.

BV resolution of trees

We discuss in detail what the Boardman-Vogt resolution of operads free on an object in the tree category Ω\Omega is like (see dendroidal set for details on trees as operads).


Symm:Operad planarOperad Symm : \mathcal{E} Operad_{planar} \to \mathcal{E} Operad

be the symmetrization functor, the left adjoint to the forgetful functor from symmetric operads to planar operads.


The BV resolution commutes with symmetrization: if T=Symm(T¯)T = Symm(\bar T), then

W(T)=Symm(W(T¯)). W(T) = Symm(W(\bar T)) \,.

Therefore we describe in the following explicitly the BV-resolution of planar trees, that of non-planar trees then being the symmetrization of that construction.


For TΩ planarT \in \Omega_{planar}, and (e 1,,e n;e)(e_1, \cdots, e_n; e) a tuple of colours (edges) of TT, notice that the set of operations T(e 1,,e n,e)T(e_1, \cdots, e_n, e) is the set of those subtrees VTV \subset T such that {e 1,,e n}\{e_1, \cdots, e_n\} is the set of leaves and ee is the root of VV.

First regard TT as a topological operad (with a discrete space of operations in each degree). The corresponding Boardman-Vogt resolution W(T)W(T) of TT is the topological operad whose topological space of operations W(T)(e 1,,e n;e)W(T)(e_1, \cdots, e_n; e) is the space of labeled trees as follows.

A point is a set of lengths (e)[0,1]\ell(e) \in [0,1], one for each inner edge eI(T)e \in I(T) of TT. (…)


W(T)(e 1,,e n;e) VT(e 1,,e n;e)(Δ 1) ×i(V) W(T)(e_1, \cdots, e_n; e) \simeq \coprod_{V \in T(e_1, \cdots, e_n; e)} (\Delta^1)^{\times i(V)}

where the coproduct ranges over the set of subtrees VV, as just discussed (which therefore is either the singleton set or is empty), and where i(V)i(V) is the set of inner edges of VV.

Regard then TT as a simplicial operad. The corresponding Boardman-Vogt resolution W(T)W(T) of TT is the simplicial operad whose simplicial sets of operations are

W(T)(e 1,,e n;e)= VT(e 1,,e n;e)Δ[1] ×i(V). W(T)(e_1, \cdots, e_n; e) = \coprod_{V \in T(e_1, \cdots, e_n; e)} \Delta[1]^{\times i(V)} \,.

In general, when TT is regarded as an \mathcal{E}-operad, we have

W(T)(e 1,,e n;e)= VT(e 1,,e n;e)H i(V), W(T)(e_1, \cdots, e_n; e) = \coprod_{V \in T(e_1, \cdots, e_n; e)} H^{\otimes i(V)} \,,

where HH is the given interval object.


The composition operations in W(T)W(T)

W(T)(e 1,,e n;e 0)W(T)(f 1,,f k;e i) i W(T)(e 1,,e i1,f 1,,f k,e i+1,,e n) \array{ W(T)(e_1, \cdots, e_n; e_0) \otimes W(T)(f_1, \cdots, f_k; e_i) \\ \downarrow^{\mathrlap{\circ_i}} \\ W(T)(e_1, \cdots, e_{i-1}, f_1, \cdots, f_k, e_{i+1}, \cdots, e_n) }

correspond to grafting of trees T σ,T ρTT_\sigma, T_\rho \subset T and “assigning unit length to the new inner edge”. On the components as discussed above it is given by

H i(T σ)H i(T ρ) H i(T σ iT ρ) H i(T σ)i(T ρ)I idi 1 H (i(T σ)i(T ρ))H \array{ H^{\otimes i(T_\sigma)} \otimes H^{\otimes i(T_\rho)} && H^{\otimes i(T_\sigma \circ_i T_\rho)} \\ \downarrow^{\mathrlap{\simeq}} && \uparrow^{\mathrlap{\simeq}} \\ H^{\otimes i(T_\sigma) \cup i(T_\rho)} \otimes I &\stackrel{id \otimes i_1}{\to}& H^{\otimes (i(T_\sigma) \cup i(T_\rho))} \otimes H }

The BV-resolution of trees extends to a functor on the category of simplicial operad

W:ΩsSetOperad W : \Omega \to sSet Operad

as follows

  • on an inner face map δ e: eΩ[T]Ω[T]\delta_e : \partial^e\Omega[T] \to \Omega[T] the component of W(δ)W(\delta) on a subtree VV of TT that contains the edge ee is the product of the inclusion

    i 0:IH i_0 : I \to H

    with the identity on H i(V){e}H^{i(V)-\{e\}}

    (meaning: if the label of an inner edge in a tree is 0, then the operations that it connects may be composed);

  • on a degenracy map σ\sigma that sends two given unary vertices to a single one, the component of W(σ)W(\sigma) on subtrees containing these removes one of the factors HH by the map

    HHH H \otimes H \to H

    given by the interval object HH. For both simplicial operads and topological operads this may be taken to be the map

    max:Δ 1×Δ 1Δ 1 max : \Delta^1 \times \Delta^1 \to \Delta^1

    that sends (x,y)(x,y) to max(x,y)max(x,y).

This is discussed in section 4.2 of (Cisinski-Moerdijk).

The homotopy coherent nerve

By the general discussion at nerve and realization, the functor

W:ΩsSetOperad W : \Omega \to sSet Operad

from prop. induces a nerve functor as follows.


The dendroidal homotopy coherent nerve functor is the functor

hcN d:sSetCatdSet hcN_d : sSet Cat \to dSet

given by

P(TsSetOperad(W(T),P)). P \mapsto ( T \mapsto sSet Operad(W(T), P) ) \,.

Its left adjoint (the corresponding “geometric realization”) we denote

W !:dSetsSetOperad. W_! : dSet \to sSet Operad \,.


Specialization to categories


When restricted to \mathcal{E}-enriched categories, the dendroidal homotopy coherent nerve reproduces the homotopy coherent nerve of enriched categories

Cat Operad hcN hcN d sSet dSet. \array{ \mathcal{E} Cat &\hookrightarrow& \mathcal{E} Operad \\ {}^{\mathllap{hcN}}\downarrow && \downarrow^{\mathrlap{hcN_d}} \\ sSet &\hookrightarrow& dSet } \,.

In particular for =\mathcal{E} = Top / sSet it reproduces the original definition of homotopy coherent nerve.

Dendroidal inner Kan complexes


Let POperadP \in \mathcal{E} Operad be such that each object of operations is fibrant in \mathcal{E}. Then its homotopy coherent nerve hcN d(P)hcN_d(P) is a dendroidal inner Kan complex.

This is (Moerdijk-Weiss, theorem 7.1). This statement will also follow as a corollary from prop. below.


Consider a tree TT and an inner edge ee of it. For each morphism ϕ:Λ e[T]X\phi : \Lambda^e[T] \to X we need to find a filler ψ\psi in

Λ e[T] hcN d(X) ψ Ω[T]. \array{ \Lambda^e[T] &\to& hcN_d(X) \\ \downarrow & \nearrow_{\mathrlap{\psi}} \\ \Omega[T] } \,.

Write Λ e[T]= ie ieΩ[T]\Lambda^e[T] = \cup_{i \neq e}\partial^{i \neq e} \Omega[T].

By the definition of dendroidal nerve, this is equivalently a diagram

ieW( iΩ[T]) X ψ^ W(Ω[T]). \array{ \cup_{i \neq e } W(\partial^i \Omega[T]) &\to& X \\ \downarrow & \nearrow_{\mathrlap{\hat \psi}} \\ W(\Omega[T]) } \,.

The undetermined component to fill is that corresponding to the subtree τ\tau of TT which is TT itself. According to prop. on this the operad W(Ω[T])W(\Omega[T]) has the component

H i(τ)H i(τ){e}H. H^{\otimes i(\tau)} \simeq H^{\otimes i(\tau)- \{e\}}\otimes H \,.

The map ψ^\hat \psi has to send this into XX while being compatible with the given faces. By prop. this means that its precomposition with all the inclusions

(id,,id,i 0,id,,id)id:HHIHHHH i(τ){e}H (id, \cdots, id, i_0, id, \cdots, id) \otimes id : H \otimes \cdots \otimes H \otimes I \otimes H \otimes \cdots \otimes H \otimes H \to H^{\otimes i(\tau)- \{e\}}\otimes H

is fixed. Moreover, the assignment needs to be compatible with the composition operations, which by prop. means that also the precomposition with all the maps

(id,,id,i 1,id,,id):HHIHHH i(τ) (id, \cdots, id, i_1, id, \cdots, id) : H \otimes \cdots \otimes H \otimes I \otimes H \otimes \cdots \otimes H \to H^{\otimes i(\tau)}

is fixed. In total this means that the components of ψ^\hat \psi need to form an extension of the form

(H i(τ){e})HH i(τ){e}I X(τ) H i(τ) \array{ (\partial H^{\otimes i(\tau)- \{e\}}) \otimes H \cup H^{\otimes i(\tau) - \{e\}}\otimes I &\to& X(\tau) \\ \downarrow & \nearrow \\ H^{\otimes i(\tau)} }

in \mathcal{E}, where

H n:=(II)H n1H(II)H n2(i 0,i 1)idH n. \partial H^n := (I \coprod I) \otimes H^{n-1} \coprod H (I \coprod I) \otimes H^{n-2} \coprod \cdots \stackrel{(i_0,i_1)\otimes id \coprod \cdots}{\to} H^n \,.

One sees that the left vertical morphism is an acyclic cofibration, by the pushout-product axiom in the monoidal model category \mathcal{E}. Therefore by the assumption that X(τ)X(\tau) is fibrant, such a lift does exist.

Left adjoint



W !:dSetsSetOperad W_! : dSet \to sSet Operad

for the Yoneda extension of

ΩOperadsSetOperadWsSetOperad; \Omega \hookrightarrow Operad \hookrightarrow sSet Operad \stackrel{W}{\to} sSet Operad \,;

hence for the functor from dendroidal sets to simplicial operads, which

By the general lore of nerve and realization we have


W !W_! is left adjoint to hcN dhcN_d

(W !hcN d):sSetOperadhcN dN ddSet. (W_! \dashv hcN_d) : sSet Operad \stackrel{\overset{N_d}{\leftarrow}}{\underset{hcN_d}{\to}} dSet \,.

For PsSetOperadP \in sSet Operad, the counit

W !hcN d(P)P W_! hcN_d (P) \to P

is essentially the Boardman-Vogt resolution of PP.

For a cofibrant and fibrant XdSetX \in dSet, the unit

XhcN dW !(X) X \to hcN_d W_!(X)

may be viewed as a “strictification” of the (infinity,1)-operad given by XX, in that W !(X)W_!(X), being a simplicial operad, has strictly associative composition.

Relation to ordinary dendroidal nerve

By the general properties of the Boardman-Vogt resolution (but also immediately checked directly) we have


There is a natural transformation

ϵ:WΩ():ΩsSetOperad \epsilon : W \Rightarrow \Omega(-) : \Omega \to sSet Operad
ϵ T:W(T)T \epsilon_T : W(T) \to T

(natural in the tree TΩT \in \Omega), which is a bijection on colors and is on the components of prop. the canonical map

Δ[1] i(V)*. \Delta[1]^{i(V)} \to * \,.

Each ϵ T\epsilon_T is hence a weak equivalence of simplicial operads. In particular

π 0(W(T))T \pi_0(W(T)) \to T

is an isomorphism.

This induces hence a natural transformation

W !τ d:dSetsSetOperad W_! \Rightarrow \tau_d : dSet \to sSet Operad

to the left adjoint τ d\tau_d of the ordinary dendroidal nerve (the “fundamental operad” construction).


For every dendroidal set XX, the natural morphism

π 0W !(X)π 0τ d(X)=τ d(X) \pi_0 W_!(X) \to \pi_0 \tau_d (X) = \tau_d(X)

is an isomorphism of simplicial operads.

This appears as Cisinski-Moerdijk, prop. 4.4.


The functors π 0W !\pi_0 W_! and τ d\tau_d, being left adjoints, both preserve small colimits. Therefore it is sufficient to check the statement for X=Ω[T]X = \Omega[T] a tree. There it is prop. .

Quillen equivalence


The adjunction (W !hcN d)(W_! \dashv hcN_d) from above is a Quillen equivalence between the model structure on operads over Top/sSet and the model structure on dendroidal sets.

We discuss some input to this statement.


The functor W !:dSetsSetOperadW_! : dSet \to sSet Operad sends normal monomorphisms to cofibrations, and inner anodyne extensions to acyclic cofibrations in the model structure on sSet-operads.

This appears as Cisinski-Moerdijk, prop. 4.5.


Observe that the morphism classes in question are, as discussed at dendroidal set, the saturated classes generated by the dendroidal boundary inclusions and by the dendroidal horn inclusions, respectively.

Since W !W_! is left adjoint, it therefore suffices to check the statement on these generating inclusions. Moreover, by construction, on trees W !W_! coincides with the Boardman-Vogt resolution of the operads free on these trees.

It follows that the generating inclusions are sent by W !W_! to morphisms of simplicial operads which are

  • bijective on objects;

  • isomorphisms on all but one simplicial set of operations: that corresponding to the maximal subtree;

  • on this remaining simplicial set of operations a product of identities with cofibrations of simplicial sets (monomorphisms), and following through the combinatorics shows that these are acyclic for the case of anodyne extensions.

It follows that these morphisms of simplicial operads have the left lifting property again operation-object-wise Kan fibrations (there is no further composition to be respected, since the maximal subtree operation has no further non-trivial composites), and hence against the fibrations of the model structure on sSet-operads.


Prop. is, in turn, a direct consequence of this.

Resolution and rectification


Let PP be an operad in Set, regarded as an \mathcal{E}-operad. Then the (W !hcN d)(W_! \dashv hcN_d)-counit

W !N d(P)=W !hcN d(P)P W_! N_d(P) = W_! hcN_d (P) \to P

is isomorphic to the Boardman-Vogt resolution W H(P)W_H(P) of PP.

In particular, therefore, there is a natural isomorphism

Hom Operad(W H(P),Q)Hom dSet(N d(P),hcN d(Q)). Hom_{\mathcal{E}Operad}(W_H(P), Q) \simeq Hom_{dSet}(N_d(P), hcN_d(Q)) \,.

(Here we are using that on a discrete operad PP the homotopy coherent dendroidal nerve trivially coincides with the ordinary dendroidal nerve N dN_d.)


By inspection of the relevant formulas.


For a cofibrant and fibrant dendroidal set XX, the (W !hcN d)(W_! \dashv hcN_d)-unit

XhcN dW !X X \to hcN_d W_! X

is an equivalence.


Since composition of operations in a simplicial operad is strictly associative, this may be understood as producing a semi-strictification of the \infty-operad XX.

The entries of the following table display model categories and Quillen equivalences between these that present the (∞,1)-category of (∞,1)-categories (second table), of (∞,1)-operads (third table) and of 𝒪\mathcal{O}-monoidal (∞,1)-categories (fourth table).

general pattern
strict enrichment(∞,1)-category/(∞,1)-operad
enriched (∞,1)-category\hookrightarrowinternal (∞,1)-category
SimplicialCategories-homotopy coherent nerve\toSimplicialSets/quasi-categoriesRelativeSimplicialSets
\downarrowsimplicial nerve\downarrow
SimplicialOperads-homotopy coherent dendroidal nerve\toDendroidalSetsRelativeDendroidalSets
\downarrowdendroidal nerve\downarrow


The fact that the homotopy coherent nerve if a locally fibrant operad is inner Kan is shown in section 7 of

The Quillen adjunction properties of the homotopy coherent dendroidal nerve are discussed in section 4 of

Lecture notes on these two topics are in section 6 and 9 of

Last revised on March 7, 2012 at 12:26:38. See the history of this page for a list of all contributions to it.