On an extension of the notion of Reedy category
aa r X i v : . [ m a t h . A T ] S e p ON AN EXTENSION OF THE NOTION OF REEDY CATEGORY
CLEMENS BERGER AND IEKE MOERDIJK
Abstract.
We extend the classical notion of a Reedy category so as to allownon-trivial automorphisms. Our extension includes many important examplesoccuring in topology such as Segal’s category Γ, or the total category of acrossed simplicial group such as Connes’ cyclic category Λ. For any generalizedReedy category R and any cofibrantly generated model category E , the functorcategory E R is shown to carry a canonical model structure of Reedy type. Introduction.
A Reedy category is a category R equipped with a structure which makes it pos-sible to prove that, for any Quillen [30] model category E , the functor category E R inherits a model structure, in which the cofibrations, weak equivalences and fibra-tions can all three be described explicitly in terms of those in E . Prime examplesof such Reedy categories are the simplex category ∆ and its dual ∆ op ; the corre-sponding model structure on cosimplicial spaces goes back to Bousfield and Kan[9], while the model structure on simplicial objects in an arbitrary model category E is described in an unpublished manuscript by Reedy [32]. The general result foran arbitrary model category E and Reedy category R is by now a standard andimportant tool in homotopy theory, well explained in several textbooks, see e.g.[18, 20, 21].As is well known, Reedy categories are skeletal, and moreover do not permit non-trivial automorphisms. There are, however, important cases in which it is possibleto establish a Reedy-like model structure on the functor category E R even though R does have non-trivial automorphisms. One example is the strict model structureon Γ-spaces (space-valued presheaves on Segal’s [34] category Γ) established byBousfield-Friedlander [8]. Another example is the case of cyclic spaces (space-valued presheaves on Connes’ category Λ, see [12]). This paper grew out of a thirdexample, namely the category of dendroidal spaces [28, Section 7] which carries aReedy-like model structure, although a dendroidal space is by definition a presheafon a category Ω of trees containing many automorphisms. We expect this Reedy-like model structure on dendroidal spaces (or a localization thereof) to be closelyrelated to a model structure on coloured topological operads, although the preciserelation remains to be worked out.In this paper, we introduce the notion of a generalized Reedy category , and provethat for any such category R and any R -projective (e.g. cofibrantly generated)Quillen model category E , the functor category E R inherits a model structure, inwhich the cofibrations, weak equivalences and fibrations can again be described Date : September 17, 2008.1991
Mathematics Subject Classification.
Primary 18G55, 55U35; Secondary 18G30, 20N99.
Key words and phrases.
Generalized Reedy category, Quillen model category, crossed group,dendroidal set. explicitly in terms of those in E . Any classical Reedy category is a generalizedReedy category in our sense; in fact, a generalized Reedy category is equivalent toa classical one if and only if it has no non-trivial automorphisms. Segal’s category Γ(as well as its dual) and the cyclic category Λ of Connes are examples of generalizedReedy categories, as is any (finite) group or groupoid. The cyclic category is anexample of the total category associated to a crossed simplicial group [16, 24]; wewill show that the total category of any crossed group on a classical Reedy categoryis a generalized Reedy category. This method yields many interesting examples ofgeneralized Reedy categories with non-trivial automorphisms. In particular, thecategory Ω mentioned above is of this type. Other examples of generalized Reedycategories relevant in homotopy theory are the orbit category of a finite or compactLie group, and the total category associated to a complex of groups , see e.g. [19].The results of this paper lead to several interesting questions. We already men-tioned the comparison between dendroidal spaces and coloured topological operads,which we expect to be analogous to the comparison between complete Segal spaces(a localization of the Reedy model structure on simplicial spaces) and topologicallyenriched categories – see [6, 23, 25, 33]. We expect the Reedy model structure onspaces over a complex of groups to be useful in describing the derived categoryof the corresponding orbifold. A precise comparison would refine the weak homo-topy equivalence between (the classifying spaces of) the complex of groups and theproper etale groupoid of the corresponding orbifold, cf. [27]. Another topic to beexplored further is the relation between various models for cyclic homology (see e.g.[31]) and the Reedy model structure on cyclic spaces given by applying our maintheorem to Connes’ category Λ. In this context, we note that it is known [10] thata localization of this model structure is Quillen equivalent to the model structure[13] on cyclic sets.In a recent paper, Angeltveit [1] studies Reedy categories enriched in a monoidalmodel category, and obtains examples of such from non-symmetric operads. Weexpect that a similar enrichment is possible for our generalized Reedy categories,so that Angeltveit’s construction can be applied to symmetric operads as well. Itwould also be of interest to extend the results of Barwick [3] to our context.To conclude this introduction, we describe the contents of the different sectionsof this paper. In Section 1, we present our notion of generalized Reedy category,state the main theorem on the existence of a model structure (Theorem 1.6), andlist some of the main examples. In Section 2, we explain a general method forconstructing generalized Reedy categories out of classical ones by means of crossedgroups. Sections 3 and 4 contain some technical preliminaries for the proof of themain theorem which will be given in Section 5. In Section 6, we give a brief intro-duction into skeleta and coskeleta for functor categories of the form E R . We thendiscuss a special class of dualizable generalized Reedy categories R for which theskeleta of set-valued presheaves on R have a simple, explicit description. In Sec-tion 7, we obtain a refinement of the main theorem (Theorem 7.5) giving sufficientconditions on R and E for the Reedy model structure on E R op to be monoidal. Acknowledgements:
The results of this paper were presented at the CRM inBarcelona in February 2008, in the context of the Program on Homotopy Theoryand Higher Categories. The actual writing of this paper was done while the secondauthor was visiting the University of Nice in May 2008. He would like to express hisgratitude to the CRM and the University of Nice for their hospitality and support.
N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 3 Generalized Reedy categories.
Recall that a subcategory S of R is called wide if S has the same objects as R .An example of a wide subcategory of R is the maximal subgroupoid Iso( R ) of R . Definition 1.1. A generalized Reedy structure on a small category R consists ofwide subcategories R + , R − , and a degree-function d : Ob( R ) → N satisfying thefollowing four axioms: (i) non-invertible morphisms in R + (resp. R − ) raise (resp. lower) the degree;isomorphisms in R preserve the degree; (ii) R + ∩ R − = Iso( R ) ; (iii) every morphism f of R factors as f = gh with g ∈ R + and h ∈ R − , andthis factorization is unique up to isomorphism; (iv) If θf = f for θ ∈ Iso( R ) and f ∈ R − , then θ is an identity.A generalized Reedy structure is dualizable if in addition the following axiom holds: (iv) ′ If f θ = f for θ ∈ Iso( R ) and f ∈ R + , then θ is an identity.A (dualizable) generalized Reedy category is a small category equipped with a(dualizable) generalized Reedy structure.A morphism of generalized Reedy categories R → S is a functor which takes R + (resp. R − ) to S + (resp. S − ) and which preserves the degree.Remark . The inclusion from left to right in axiom (ii) follows from axiom (i).Axiom (iv) says that automorphisms in R consider morphisms of R − as epimor-phisms. This last axiom implies that the isomorphism in (iii) is unique. The axioms(i)-(iii) are self-dual while axiom (iv) is dual to axiom (iv) ′ . A generalized Reedycategory R is thus dualizable if and only if R op is also a generalized Reedy category.Most of the examples that we are aware of are dualizable. The asymmetry in thedefinition is related to the asymmetry of the projective model structure on objectswith a group action, which enters in Theorem 1.6; cf. the proof of Lemma 5.5. Remark . If R is a generalized Reedy category, an equivalence of categories R ′ ∼ −→ R induces a generalized Reedy structure on R ′ . In this sense, the existenceof a generalized Reedy structure is invariant under equivalence of categories. Remark . Recall that in the literature (cf. [18, 20, 21, 32]) a category R , equippedwith R + , R − and d as above, is called a Reedy category if it satisfies the followingtwo axioms:(i) non-identity morphisms in R + (resp. in R − ) raise (resp. lower) degree;(ii) every morphism in R factors uniquely as a morphism in R − followed by onein R + .Any such Reedy category is a dualizable generalized Reedy category in our sense.To emphasize the distinction with generalized Reedy categories we will refer to theclassical ones as strict Reedy categories . The notion of a strict Reedy category isnot invariant under equivalence of categories. In fact, one checks that in a strictReedy category every isomorphism is an identity. A generalized Reedy category isequivalent to a strict one if and only if it has no non-trivial automorphisms, and isitself strict if and only if it is moreover skeletal. Indeed, for an isomorphism f , let f = gh and hf − = g ′ h ′ be the unique factorizations. Then id = ghf − = ( gg ′ ) h ′ , so h ′ = id and gg ′ = id , whence g = id and g ′ = id since g, g ′ ∈ R + . Thus f = h ∈ R − . The same argument applied to f − shows that f preserves the degree, hence f = id . CLEMENS BERGER AND IEKE MOERDIJK
Remark . As for strict Reedy categories, all the results concerning a fixed gen-eralized Reedy category R go through if the degree-function takes values in anarbitrary well-ordered set. (However, with these more general degree-functions, thenotion of a morphism of Reedy categories is more subtle to define).For a generalized Reedy category R , we introduce the following notions, whichare classical in the case of a strict Reedy category. For each object r of R , thecategory R + ( r ) has as objects the non-invertible morphisms in R + with codomain r , and as morphisms from u : s → r to u ′ : s ′ → r all w : s → s ′ such that u = u ′ w .Observe that axiom (iii) implies that w ∈ R + ; moreover, the automorphism groupAut( r ) acts on the category R + ( r ) by composition. For each functor X : R → E and each object r of R , the r -th latching object L r ( X ) of X is defined to be L r ( X ) = lim −→ s → r X s where the colimit is taken over the category R + ( r ). We will always assume E tobe sufficiently cocomplete for this colimit to exist (in many examples this colimitis finite). Note that Aut( r ) acts on L r ( X ).Dually, for each object r of R , the category R − ( r ) has as objects the non-invertible morphisms in R − with domain r , and as morphisms from u : r → s to u ′ : r → s ′ all w : s → s ′ such that u ′ = wu . Observe that axiom (iii) implies that w ∈ R − ; moreover, the automorphism group Aut( r ) acts on the category R − ( r ) byprecomposition. For each object X of E R and each object r of R , the r -th matchingobject M r ( X ) of X is defined to be M r ( X ) = lim ←− r → s X s where the limit is taken over the category R − ( r ). We will always assume E to besufficiently complete for this limit to exist (in many examples this limit is finite).Note that Aut( r ) acts on M r ( X ).Each object X of the functor category E R defines for any object r of R naturalAut( r )-equivariant maps L r ( X ) → X r → M r ( X ). For a map f : X → Y in E R these give rise to relative latching, resp. matching maps X r ∪ L r ( X ) L r ( Y ) −→ Y r , resp. X r −→ M r ( X ) × M r ( Y ) Y r . Recall that for any group (or groupoid) Γ and any cofibrantly generated modelcategory E , the category E Γ of objects of E with right Γ-action carries a projectivemodel structure, in which weak equivalences and fibrations are defined by forgettingthe Γ-action. In general, a Quillen model category E will be called R -projective , iffor each object r of R , the category E Aut( r ) admits a projective model structure.For R -projective model categories E , we introduce the following notions:A map f : X → Y in E R is called a– Reedy cofibration if for each r , the relative latching map X r ∪ L r ( X ) L r ( Y ) → Y r is a cofibration in E Aut( r ) ;– Reedy weak equivalence if for each r , the induced map f r : X r → Y r is a weakequivalence in E Aut( r ) ;– Reedy fibration if for each r , the relative matching map X r → M r ( X ) × M r ( Y ) Y r is a fibration in E Aut( r ) . N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 5
Observe that the automorphism group Aut( r ) really enters only in the definition ofa Reedy cofibration, by definition of the model structure on E Aut( r ) just described. Theorem 1.6.
Let R be a generalized Reedy category and let E be an R -projectiveQuillen model category in which the relevant limits and colimits exist (for instance, E can be any cofibrantly generated model category). With the above classes of Reedycofibrations, Reedy weak equivalences and Reedy fibrations, the functor category E R is a Quillen model category. The proof will be supplied in Section 5. Notice that if R = R + then the con-stant functor E → E R sends weak equivalences and fibrations in E to Reedy weakequivalences and Reedy fibrations in E R . Thus, we obtain the following corollarywhich is well known for strict Reedy categories. Corollary 1.7.
Let E and R be as in Theorem 1.6. If R = R + then lim −→ : E R → E is a left Quillen functor.Examples . We give a list of first examples which, among other things, showthat generalized Reedy categories occur naturally in several different contexts inhomotopy theory. More examples are provided in Section 2.(a) For completeness, we mention (again) that any strict Reedy category is adualizable generalized Reedy category (cf. Remark 1.4). This applies in particularto standard examples of Reedy categories such as the simplex category ∆ and itsdual, as well as to ( N , < ), ← · → , · ⇒ · (relevant for homotopy colimits of sequences,for homotopy pushouts and for homotopy coequalizers). Other examples are Joyal’scategory of finite disks and its dual Θ (cf. [22, 5]).(b) Segal’s category Γ (cf. [34]) is a dualizable generalized Reedy category. Infact, Γ op is equivalent to the category F in ∗ of finite pointed sets, and one cantake F in + ∗ to consist of monomorphisms and F in −∗ of epimorphisms, while thedegree-function is given by cardinality. If E is the category of simplicial sets, theReedy model structure on E Γ op given by Theorem 1.6 was discussed in Bousfield-Friedlander [8] and referred to as the strict model structure on Γ-spaces. Thesimplicial circle ∆[1] /∂ ∆[1], when viewed as a functor ∆ → Γ (cf. [34, 5]), is amorphism of generalized Reedy categories.(c) The category
F in of finite sets carries a dualizable generalized Reedy struc-ture, analogous to the pointed case. A skeleton of
F in is often denoted by ∆ sym ,and E ∆ op sym is referred to as the category of symmetric simplicial objects in E , cf.[2, 10]. The inclusion ∆ ֒ → ∆ sym is a morphism of generalized Reedy categories.(d) Any group(oid) is a generalized Reedy category.(e) Orbit categories . The orbit category O ( G ) of a finite group G has the sub-groups of G as objects, and the G -equivariant maps G/H → G/K as morphisms.This orbit category is a generalized Reedy category with O ( G ) = O ( G ) − and d ( H ) = card( G/H ) (the index of H in G ). There is also a dual generalized Reedystructure on O ( G ) with O ( G ) = O ( G ) + and d ( H ) = card( H ). If G is not finite,the first structure still makes sense for subgroups of finite index, the second onefor finite subgroups. The orbit category O ( G ) of a compact Lie group G is thecategory with closed subgroups of G as objects and G -homotopy classes of G -maps G/H → G/K as morphisms from H to K . This is again a generalized Reedy cate-gory with O ( G ) = O ( G ) + . The degree of an object H now takes values in N × N CLEMENS BERGER AND IEKE MOERDIJK with the lexicographical ordering, and is defined by d ( H ) = (dim( H ) , card( π H )).Notice that this generalized Reedy structure is not in general dualizable like in thecase of finite groups, because there may be infinite increasing sequences of closedsubgroups, e.g. the subgroups Z /p n Z of the circle S .(f) Complexes of groups.
Let X be a simplicial complex. Recall that a complexof groups G over X assigns to each simplex σ ∈ X a group G σ , to each inclusion σ ⊆ τ an injective group homomorphism φ σ,τ : G τ → G σ , and to each sequence ρ ⊆ σ ⊆ τ a specific element g = g ρ,σ,τ ∈ G ρ such that the triangle G τ ✲ G σ G ρ ❄ ✲ commutes up to conjugation by g , i.e. for each x ∈ G τ : gφ ρ,τ ( x ) g − = φ ρ,σ ( φ σ,τ ( x )) . Moreover, for π ⊆ ρ ⊆ σ ⊆ τ , the following coherence condition should be satisfied: φ π,ρ ( g ρ,σ,τ ) g π,ρ,τ = g π,ρ,σ g π,σ,τ . Such complexes of groups can be used to model orbifold structures on a triangulatedspace | X | , see [19, 27]. To each complex of groups G over X is associated a category∆ X ( G ) whose objects are the simplices σ ∈ X ; if σ ⊆ τ then morphisms y : σ → τ in ∆ X ( G ) are given by elements y ∈ G σ . Composition of y : σ → τ and x : ρ → σ is defined to be φ ρ,σ ( y ) x : ρ → σ . The coherence condition implies that thiscomposition is associative. The category ∆ X ( G ) is a generalized Reedy categoryin which the degree of σ is the dimension of the simplex, and for which ∆ X ( G ) =∆ X ( G ) + . This example is a special case of Corollary 1.10.The class of generalized Reedy categories is closed under arbitrary coproductsand under finite products. A more subtle closure property is the following: Proposition 1.9.
Let S → R be a fibered category over R . Suppose that the base R and each of the fibers S r are equipped with generalized Reedy structures. Assumefurthermore that for each morphism α : r → s in the base R , (i) the base change α ∗ : S s → S r preserves the degree; (ii) if α belongs to R + then α ∗ takes S + s to S + r ; (iii) if α belongs to R − then α ∗ has a left adjoint α ! which takes S − r to S − s .Then S can be equipped with a generalized Reedy structure such that the fiber inclu-sions S r ֒ → S and the projection S → R preserve the factorization systems.Proof. Consider a morphism f : x → y in S over α : r → s in R . Say f ∈ S + if α ∈ R + and the unique morphism x → α ∗ ( y ) in S r determined by a cartesianlift α ∗ ( y ) → y of α lies in S + r . Say f ∈ S − if α ∈ R − and the unique morphism α ! ( x ) → y in S y determined by a cocartesian lift x → α ! ( x ) of α lies in S − s . For x ∈ S r , define the degree by d S ( x ) = d R ( r ) + d S r ( x ). With these definitions, it isstraightforward to verify that S is a generalized Reedy category. (cid:3) Corollary 1.10.
Let R be a generalized Reedy category for which R = R + , and let Φ : R op → Cat be a diagram of Reedy categories and morphisms of Reedy categories.Then the Grothendieck construction S = R R Φ is again a generalized Reedy category. N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 7 Crossed groups.
In this section, we introduce the notion of a crossed group G on a category R ,and discuss the construction of the associated total category R G . We will show thatfor any strict Reedy category R and crossed R -group G , the total category R G isa generalized Reedy category, which is no longer strict unless G is trivial. Many ofour examples of generalized Reedy categories are instances of this construction.Crossed groups on the simplex category have been studied in the literature underthe name skew-simplicial groups (see Krasauskas [24]), resp. crossed simplicialgroups (see Fiedorowicz-Loday [16]). Recently, Batanin-Markl [4, 2.2] considered crossed cosimplicial groups which are crossed groups on the dual of the simplexcategory. Feigin-Tsygan already spelled out the axioms of a crossed group in [15,A4.1-4]. Cisinski considers the more general concept of a thickening in [10, 8.5.8]. Definition 2.1.
For any small category R , a crossed R -group G is a set-valuedpresheaf on R , together with, for each object r of R , (i) a group structure on G r , (ii) left G r -actions on the hom-sets Hom R ( s, r ) with codomain r ,such that the following identities hold for all g, h ∈ G r , α : s → r, β : t → s , g ∗ ( α ◦ β ) = g ∗ ( α ) ◦ α ∗ ( g ) ∗ ( β ) , (1) g ∗ (1 r ) = 1 r , (2) α ∗ ( g · h ) = h ∗ ( α ) ∗ ( g ) · α ∗ ( h ) , (3) α ∗ ( e r ) = e s , (4) where the presheaf action of α : s → r is denoted by α ∗ : G r → G s and the groupaction of g ∈ G r is denoted by g ∗ : Hom R ( s, r ) → Hom R ( s, r ) . Moreover, for eachobject r , the identity of r (resp. neutral element of G r ) is denoted by r (resp. e r ).Remark . In what follows we shall make no difference in notation between com-position in R and composition in G r , especially since both structures will agree inthe total category R G . In addition to the four identities spelled out in Definition2.1, the following four identities also hold in any R -crossed group G (by the axiomsfor a presheaf, resp. group action):( αβ ) ∗ ( g ) = β ∗ α ∗ ( g ) , (5) 1 ∗ r ( g ) = g, (6) ( gh ) ∗ ( α ) = g ∗ h ∗ ( α ) , (7) ( e r ) ∗ ( α ) = α. (8)2.3. The total category.
For any small category R and crossed R -group G , thetotal category R G is the category with the same objects as R , and with morphisms r → s the pairs ( α, g ) where α : r → s belongs to R , and g ∈ G r . Composition of( α, g ) : s → t and ( β, h ) : r → s is defined as( α, g ) ◦ ( β, h ) = ( α · g ∗ ( β ) , β ∗ ( g ) · h ) . One easily checks that this composition is associative and has a two-sided unit(1 r , e r ) for each object r of R G . Remark . In the special case where G is a constant presheaf (i.e. G = G r for afixed group G and α ∗ ( g ) = g for all g and all α ), the total category R G reduces tothe familiar Grothendieck construction for a diagram of categories on G . CLEMENS BERGER AND IEKE MOERDIJK
In the special case where the left action of G on R is trivial (i.e. g ∗ ( α ) = α forall g and all α ), the crossed group is actually a presheaf of groups, and the totalcategory R G again reduces to a Grothendieck construction, this time for a diagramof groups on R op .Returning to the general case of a crossed R -group G , notice that we alwayshave a canonical embedding R ֒ → R G which sends α : r → s to ( α, e r ) : r → s ,and identifies R with a wide subcategory of R G . Elements g ∈ G r of the crossedgroup may be identified with special automorphisms (1 r , g ) in the total category R G , and every morphism ( α, g ) in R G factors uniquely as a special automorphism(1 r , g ) followed by a morphism ( α, e r ) in R . This unique factorization property ischaracteristic for total categories of crossed groups as asserted by: Proposition 2.5.
Let R ⊆ S be a wide subcategory and assume that there existsubgroups G s ⊆ Aut S ( s ) of special automorphisms such that each morphism in S factors uniquely as a special automorphism followed by a morphism in R . Then thegroups G s define a crossed R -group, and S is isomorphic to R G (under R ).Proof. For any morphism α : r → s of R and special automorphism g ∈ G s , thepresheaf action of R as well as the group action of G are defined by factoring thecomposite gα : r → s uniquely as in the hypothesis of the proposition, as r α ✲ srα ∗ ( g ) ❄ g ∗ ( α ) ✲ s.g ❄ With this explicit description, the proof of the identities of Definition 2.1 and of theisomorphism S ∼ = R G is a matter of (lengthy but) straightforward verification. (cid:3) Remark . Fiedorowicz-Loday [16] take Proposition 2.5 for R = ∆ as the defi-nition of a crossed simplicial group (with contravariant instead of covariant groupaction), and state Definition 2.1 of a crossed ∆-group as a proposition. Example . The most prominent example of a crossed group is the simplicial circle C = ∆[1] /∂ ∆[1] whose total category ∆ C is isomorphic to the cyclic category Λ ofConnes [12]. It is convenient to embed C in a larger crossed ∆-group Σ, formed bythe permutation groups Σ [ n ] of the sets [ n ] = { , , . . . , n } . The crossed ∆-groupstructure of Σ is defined as follows: given α : [ m ] → [ n ] in ∆ and g : [ n ] → [ n ]in Σ [ n ] , the map α ∗ ( g ) : [ m ] → [ m ] is the unique permutation which is order-preserving on the fibers of α , and for which g ∗ ( α ) = g ◦ α ◦ α ∗ ( g ) − : [ m ] → [ n ] isorder-preserving: [ m ] α ✲ [ n ][ m ] α ∗ ( g ) ❄ g ∗ ( α ) ✲ [ n ] .g ❄ Let C [ n ] ⊂ Σ [ n ] be the subgroup generated by the cycle 0
7→ · · · 7→ n g ∈ C [ n ] then α ∗ ( g ) ∈ C [ m ] for each α : [ m ] → [ n ] in ∆, so that C inherits a crossed ∆-group structure. The total category ∆ C is then isomorphic N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 9 to the cyclic category Λ of Connes [12], and embeds in the total category ∆Σ.The latter has been described in detail by Feigin-Tsygan [15, A10] and plays animportant role in the general classification of crossed ∆-groups, see [16, 24].
Example . One of the examples of a generalized Reedy category which motivatedthis paper is the category Ω of trees introduced by Moerdijk-Weiss in [28]. Theobjects of this category are finite trees with a distinguished output edge and a setof distinguished input edges, as common in the context of operads. Any such tree T freely generates a symmetric coloured operad Ω( T ) whose colour-set is the set E ( T )of edges of T ; the morphisms T → T ′ in Ω are the maps of symmetric colouredoperads Ω( T ) → Ω( T ′ ). For a more precise description, we refer to [28]. Here, it isenough to observe that any such morphism T → T ′ induces a map E ( T ) → E ( T ′ ) ina functorial way, and that this induced map completely determines the morphism.The category Ω carries a natural dualizable generalized Reedy structure, for whichthe degree is given by the number of vertices in the tree, while a morphism belongsto Ω + (resp. Ω − ) when it induces an injection (resp. surjection) between the setsof edges.For such a tree T , one can consider the set of planar structures p on T . Sinceevery tree in Ω carries at least one planar structure, the category Ω is equivalentto the category Ω ′ whose objects are planar trees ( T, p ), and whose morphisms(
T, p ) → ( T ′ , p ′ ) are the morphisms T → T ′ in Ω. For every such morphism, one canpull back the planar structure p ′ on T ′ to one on T , and call the morphism planarif this pulled back structure coincides with p . The planar morphisms form a widesubcategory of Ω ′ , denoted Ω planar ; in this latter category, every automorphism istrivial, and Ω planar is equivalent to a strict Reedy category. Every morphism inΩ ′ factors uniquely as an automorphism followed by a planar map. This shows byProposition 2.5 that the category Ω is equivalent to the total category of a crossedgroup on Ω planar .The embedding i : ∆ ֒ → Ω (cf. [28]) is a morphism of generalized Reedy cate-gories, and Theorem 1.6 gives a Reedy model structure on dendroidal spaces , whichis compatible with the Reedy model structure on simplicial spaces. At the end ofSection 7 (cf. Exampe 7.6(iii)), we will show that the model structure on dendroidalspaces is monoidal (in the sense of Hovey [21]) with respect to the
Boardman-Vogttensor product on dendroidal spaces (cf. [28, appendix]).Consider a crossed R -group G , and suppose that R carries a generalized Reedystructure. We will say that the crossed R -group is compatible with the generalizedReedy structure if the following two conditions hold:(i) the G -action respects R + and R − (i.e. if α : r → s belongs to R ± and g ∈ G s then g ∗ ( α ) : r → s belongs to R ± );(ii) if α : r → s belongs to R − and g ∈ G s is such that α ∗ ( g ) = e r and g ∗ ( α ) = α , then g = e s . Remark . Observe that condition (i) is in particular satisfied if any morphismin R , which in R G is isomorphic to a morphism in R ± , already belongs to R ± .Condition (ii) is equivalent to the condition that R − fulfills axiom (iv) of Definition1.1 with respect to special automorphisms of R G , cf. the proof of Proposition 2.5. Because in the simplex category ∆ the morphisms of ∆ + (resp. of ∆ − ) are themonomorphisms (resp. split epimorphisms) of ∆, any crossed ∆-group is compati-ble with the Reedy structure of ∆. The same property holds for crossed groups onΩ planar , cf. Example 2.8, and in general for crossed groups on strict EZ-categories,cf. Definition 6.6. Proposition 2.10.
Let R be a strict Reedy category, and let G be a compatiblecrossed R -group. Then there is a unique dualizable generalized Reedy structureon R G for which the embedding R ֒ → R G is a morphism of generalized Reedycategories.Proof. Necessarily, ( R G ) ± consists of those morphisms ( α, g ) for which α ∈ R ± .Because of compatibility condition (i), ( R G ) ± is closed under composition. It isnow straightforward to verify that this defines a generalized Reedy structure on R G . In particular, axiom (iv) follows from compatibility condition (ii) and the factthat all automorphisms of R G are special since R is a strict Reedy category. Thedual axiom (iv) ′ holds automatically. (cid:3) Kan extensions and the projection formula.
In this section we recall some basic facts about Kan extensions for diagramcategories. Let φ : D −→ C be a functor between small categories, and write φ ∗ : E C −→ E D for precomposition with φ . The left and right adjoints of φ ∗ areusually called left and right Kan extension along φ .If E is sufficiently cocomplete, the left Kan extension φ ! : E D −→ E C can becomputed pointwise by φ ! ( X ) c = lim −→ φ/c X ◦ π c where φ/c is the comma category with objects ( d, u : φ ( d ) → c ) and morphisms( d, u ) → ( d ′ , u ′ ) given by f : d → d ′ in D such that u ′ ◦ φ ( d ) = u . The functor π c : φ/c −→ D is defined by ( d, u ) d . We will often informally write φ ! ( X ) c = lim −→ φ ( d ) → c X d . The formula for left Kan extension simplifies if the functor φ : D → C is cofibered.Recall (cf. [7]) that for a given functor φ : D → C , a morphism f : d → d ′ in D is cocartesian if for any g : d → d ′′ such that φ ( g ) = hφ ( f ), there is a unique k : d ′ → d ′′ such that g = kf and φ ( k ) = h . The functor φ is called cofibered , ifmorphisms in C have cocartesian lifts in D , and if moreover cocartesian morphismsin D are closed under composition. If φ is cofibered, then for any object c of C ,the embedding of the fiber φ − ( c ) into the comma category φ/c (given on objectsby d ( d, φ ( c ) )) has a left adjoint, so φ − ( c ) is cofinal in φ/c , and hence φ ! ( X ) c = lim −→ φ − ( c ) X is the colimit over the fiber. This implies that for any pullback diagram of categories D ′ β ✲ DC ′ ψ ❄ α ✲ C φ ❄ N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 11 with φ (and hence ψ ) cofibered, the natural transformation of functors ψ ! β ∗ −→ α ∗ φ ! is an isomorphism . This is often called the projection formula , and will be appliedbelow in the special case where D = R C F is the Grothendieck construction of acovariant diagram F : C → Cat.Dually, if E is sufficiently complete, the right Kan extension φ ∗ : E D → E C canbe computed pointwise by φ ∗ ( X ) c = lim ←− c → φ ( d ) X d and this formula simplifies for fibered functors φ : D → C . Recall that a functor φ iscalled fibered , if morphisms in C have cartesian lifts in D , and if moreover cartesianmorphisms in D are closed under composition. If φ is fibered, then for any object c of C , the embedding of the fiber φ − ( c ) into the comma category c/φ (given onobjects by d (1 φ ( c ) , d )) has a right adjoint, so φ − ( c ) is final in c/φ , and hence φ ∗ ( X ) c = lim ←− φ − ( c ) X is the limit over the fiber. This implies that for any pullback diagram of categories D ′ β ✲ DC ′ ψ ❄ α ✲ C φ ❄ with φ (and hence ψ ) fibered , the natural transformation of functors α ∗ φ ∗ −→ ψ ∗ β ∗ is an isomorphism . This dual projection formula will be applied below in the specialcase where D = R C F is the Grothendieck construction of a contravariant diagram F : C op → Cat. 4.
Latching and matching objects.
In this section we give an alternative, more global definition of latching andmatching objects. Throughout, we consider a fixed generalized Reedy category R with wide subcategories R ± and degree-function d as in Definition 1.1, and assumethat E is a sufficiently bicomplete category.4.1. The groupoids of objects of fixed degree.
For each natural number n ,the full subcategory of R of objects of degree ≤ n will be denoted R ≤ n , the full subgroupoid of Iso( R ) spanned by the objects of degree n will be denoted G n ( R ),and the discrete category of objects of R of degree n will be denoted R n . Overcategories.
For each natural number n , the category R + (( n )) has asobjects the non-invertible morphisms u : s → r in R + such that d ( r ) = n , and asmorphisms from u to u ′ the commutative squares s f ✲ s ′ ru ❄ g ✲ r ′ u ′ ❄ such that f ∈ R + and g ∈ G n ( R ).The wide subcategory R + ( n ) of R + (( n )) contains those morphisms for which g is an identity. The category R + ( r ) of Section 1 may thus be identified with the fullsubcategory of R + ( n ) spanned by the objects with codomain r . Notice that R + ( n ) = a d ( r )= n R + ( r ) . The categories introduced so far assemble into the following commutative dia-gram: R ✛ d n R + (( n )) c n ✲ G n ( R ) j n ✲ RR + ( n ) k n ✻ b n ✲ R n i n ✻ where d n denotes the domain-functor, b n and c n denote codomain-functors, and i n , j n and k n are inclusion-functors. Note that c n is cofibered, i.e. R + (( n )) ∼ = Z G n ( R ) R + ( − ) , and that the square is a pullback. In particular, the projection formula yields i ∗ n ( c n ) ! ∼ = ( b n ) ! k ∗ n . Latching objects.
The definition of the latching object L n ( X ) for an object X of E R now takes the following form: L n ( X ) = ( c n ) ! d ∗ n ( X ) ∈ E G n ( R ) . We write X n = j ∗ n ( X ) = X | G n ( R ) , so that we get in each degree n a latching map L n ( X ) −→ X n . Note that, since c n is cofibered, we have more concretely: L n ( X ) r = lim −→ s → r X s , where the colimit is taken over the category R + ( r ) as in Section 1. Accordingly, wewill often simplify notation and write L r ( X ) for L n ( X ) r .Observe that a morphism φ : S → R of generalized Reedy categories induces for k ∈ N and X ∈ E R a natural map L k ( φ ∗ ( X )) −→ φ ∗ k ( L k ( X ))where φ ∗ : E R → E S and φ ∗ k : E G k ( R ) → E G k ( S ) are induced by φ . N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 13
Lemma 4.4.
Let φ : S → R be a morphism of generalized Reedy categories. Supposethat the induced square S + (( k )) ✲ G k ( S ) R + (( k )) φ + k ❄ ✲ G k ( R ) φ k ❄ is a pullback. Then, for each object X of E R , the natural comparison map of latchingobjects L k ( φ ∗ ( X )) → φ ∗ k ( L k ( X )) is an isomorphism.The pullback hypothesis holds in particular in the following two cases: (i) S = R + ( n ) and φ = d n k n : R + ( n ) → R is the domain functor; (ii) S = R ≤ n and φ : R ≤ n → R is the canonical embedding.Proof. The pullback square is part of the commutative diagram S ✛ ¯ d k S + (( k )) ¯ c k ✲ G k ( S ) R φ ❄✛ d k R + (( k )) φ + k ❄ c k ✲ G k ( R ) φ k ❄ whose rows enter into the definition of the latching objects. Together with theprojection formula, this yields canonical isomorphisms: L k ( φ ∗ ( X )) = ¯ c k ! ¯ d ∗ k φ ∗ ( X )= ¯ c k ! ( φ + k ) ∗ d ∗ k ( X ) ∼ = φ ∗ k c k ! d ∗ k ( X )= φ ∗ k ( L k ( X )) . If S = R + ( n ) then S + (( k )) has as objects the composable pairs t → s → r of non-invertible, degree-raising maps such that d ( r ) = n and d ( s ) = k , and asmorphisms those transformations of diagrams which are the identity on the lastobject, an isomorphism on the intermediate object, and degree-raising on the firstobject; this category coincides with the fiber product of φ k : G k ( S ) → G k ( R ) and c k : R + (( k )) → G k ( R ).If S = R ≤ n the pullback hypothesis follows from the fact that an object of R + (( k )) belongs to S + (( k )) if and only if its codomain is of degree ≤ n . (cid:3) Undercategories.
The category R − (( n )) has as objects the non-invertiblemorphisms u : r → s in R − such that d ( r ) = n , and as morphisms from u to u ′ thecommutative squares r g ✲ r ′ su ❄ f ✲ s ′ u ′ ❄ such that f ∈ R − and g ∈ G n ( R ).The wide subcategory R − ( n ) of R − (( n )) contains those morphisms for which g is an identity. The category R − ( r ) of Section 1 may then be identified with the full subcategory of R − ( n ) spanned by the objects with domain r . Notice that R − ( n ) = a d ( r )= n R − ( r ) . The categories introduced so far assemble into the following commutative dia-gram: R ✛ γ n R − (( n )) δ n ✲ G n ( R ) j n ✲ RR − ( n ) κ n ✻ β n ✲ R n i n ✻ where γ n denotes the codomain-functor, β n and δ n denote domain-functors, and i n , j n and κ n are inclusion-functors.Note that δ n is fibered, i.e. R − (( n )) ∼ = Z G n ( R ) R − ( − ) , and that the square is a pullback. In particular, the dual projection formula yields i ∗ n ( δ n ) ∗ ∼ = ( β n ) ∗ κ ∗ n . Matching objects.
The definition of the matching object M n ( X ) of an object X of E R now takes the following form: M n ( X ) = ( δ n ) ∗ γ ∗ n ( X ) ∈ E G n ( R ) . We write X n = j ∗ n ( X ) = X | G n ( R ) , so that we get in each degree n a matching map X n −→ M n ( X ) . Note that, since δ n is fibered, we have more concretely: M n ( X ) r = lim ←− r → s X s , where the limit is taken over the category R − ( r ) as in Section 1. Accordingly, wewill often simplify notation and write M r ( X ) for M n ( X ) r . Lemma 4.7.
Let φ : S → R be a morphism of generalized Reedy categories. Supposethat the induced square S − (( k )) ✲ G k ( S ) R − (( k )) φ − k ❄ ✲ G k ( R ) φ k ❄ is a pullback. Then, for each object X of E R , the natural comparison map of match-ing objects φ ∗ k ( M k ( X )) → M k ( φ ∗ ( X )) is an isomorphism.The pullback hypothesis holds in particular in the following two cases: (i) S = R − ( n ) and φ = γ n κ n : R − ( n ) → R is the codomain functor; (ii) S = R ≤ n and φ : R ≤ n → R is the canonical embedding.Proof. Dual to the proof of Lemma 4.4. (cid:3)
N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 15 The Reedy model structure.
We can reformulate the definition of the classes of maps in Section 1 as follows:
Lemma 5.1.
A map X → Y in E R is a Reedy cofibration (resp. a Reedy weakequivalence, resp. a Reedy fibration) if and only if, for each natural number n , themap X n ∪ L n ( X ) L n ( Y ) → Y n (resp. X n → Y n , resp. X n → M n ( X ) × M n ( Y ) Y n ) isa cofibration (resp. a weak equivalence, resp. a fibration) in E G n ( R ) .Proof. This just follows from the equivalence of categories E G n ( R ) ∼ −→ Y r E Aut( r ) where r runs through a set of representatives for the connected components of thegroupoid G n ( R ). (cid:3) A Reedy (co)fibration which is also a Reedy weak equivalence will be referredto as a trivial Reedy (co)fibration . The following lemmas are preparatory for theproof of Theorem 1.6.
Lemma 5.2.
Let f : A → B be a trivial Reedy cofibration; suppose that, for each n , the induced map L n ( f ) : L n ( A ) → L n ( B ) is a pointwise trivial cofibration (i.e. L n ( f ) r is a trivial cofibration in E for each object r of R ). Then f : A → B hasthe left lifting property with respect to Reedy fibrations.Proof. Consider a commutative square in E R A α ✲ YBf ❄ β ✲ Xg ❄ where f is a trivial Reedy cofibration and g is a Reedy fibration, and furthermore L n ( f ) : L n ( A ) → L n ( B ) is a pointwise trivial cofibration for all n . We constructa diagonal filler γ : B → Y by constructing inductively a filler γ ≤ n : B ≤ n → Y ≤ n on the full subcategory R ≤ n of objects of R of degree ≤ n . For n = 0, we get adiagonal filler γ : B → Y in A α ✲ Y B f ❄ β ✲ X g ❄ since R ≤ is the groupoid G ( R ), and L ( A ) = 0, M ( X ) = 1, so that by hypothesis f is a trivial cofibration in E G ( R ) and g is a fibration in E G ( R ) .Assume by induction that a filler γ ≤ n − : B ≤ n − → Y ≤ n − has been found for A ≤ n − α ≤ n − ✲ Y ≤ n − B ≤ n − f ≤ n − ❄ β ≤ n − ✲ X ≤ n − .g ≤ n − ❄ This yields composite maps L n ( B ) −→ L n ( Y ) −→ Y n and B n −→ M n ( B ) −→ M n ( Y ) as well as a commutative square A n ∪ L n ( A ) L n ( B ) ✲ Y n B n v n ❄ ✲ X n × M n ( X ) M n ( Y ) .w n ❄ A G n ( R )-equivariant filler is exactly what is needed in order to complete the induc-tive step. To see that such a filler exists, note that by hypothesis v n is a cofibrationand w n is a fibration in E G n ( R ) . It is thus enough to check that v n is a weak equiv-alence. For this, consider the following diagram in which the square is a pushout: L n ( A ) r ✲ A r f r ✲ B r L n ( B ) r L n ( f ) r ❄ ✲ ( A ∪ L n ( A ) L n ( B )) r . ❄ ( v n ) r ✲ Since, by hypothesis, L n ( f ) r is a trivial cofibration in E , and f r is a weak equiva-lence, v n is a weak equivalence as required. (cid:3) Lemma 5.3.
Let f : A → B be a Reedy cofibration such that f r : A r → B r isa weak equivalence for all objects r of R of degree < n . Then, the induced map L n ( f ) : L n ( A ) → L n ( B ) is a pointwise trivial cofibration.Proof. For n = 0, there is nothing to prove; therefore, we can assume inductivelythat L k ( f ) : L k ( A ) → L k ( B ) is a pointwise trivial cofibration for k < n . We wantto show that i ∗ n L n ( f ) is a trivial cofibration in E R n . To this end, we have to find afiller for any commutative square i ∗ n L n ( A ) ✲ Yi ∗ n L n ( B ) i ∗ n L n ( f ) ❄ ✲ Xg ❄ in E R n in which g : Y → X is a fibration. Since i ∗ n L n = i ∗ n ( c n ) ! d ∗ n = ( b n ) ! k ∗ n d ∗ n , afiller for the former square is the same as a filler for the following square in E R + ( n ) : k ∗ n d ∗ n ( A ) ✲ b ∗ n ( Y ) k ∗ n d ∗ n ( B ) k ∗ n d ∗ n ( f ) ❄ ✲ b ∗ n ( X ) .b ∗ n ( g ) ❄ In order to finish the proof, we shall apply Lemma 5.2 to this square. The category S = R + ( n ) is a generalized Reedy category for which S = S + . In particular,Reedy fibrations are the same as pointwise fibrations, so b ∗ n ( g ) is a Reedy fibration.Moreover, k ∗ n d ∗ n ( f ) is a Reedy weak equivalence in E S , since the objects of S havedegree < n . It remains to be shown that k ∗ n d ∗ n ( f ) is a Reedy cofibration whoseinduced maps on latching objects of degree < n are pointwise trivial cofibrations.Write φ = d n k n . By Lemma 4.4, the functor φ ∗ k : E G k ( R ) → E G k ( S ) induces acanonical isomorphism L k ( φ ∗ ( A )) ∼ = φ ∗ k ( L k ( A )). Therefore, the relative latchingmap φ ∗ ( A ) ∪ L k ( φ ∗ ( A )) L k ( φ ∗ ( B )) → φ ∗ ( B ) may be identified with φ ∗ k of the relativelatching map A k ∪ L k ( A ) L k ( B ) → B k . Observe that φ k : G k ( S ) → G k ( R ) is a faithfulfunctor between groupoids, so φ ∗ k preserves cofibrations, thus k ∗ n d ∗ n ( f ) is a Reedy N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 17 cofibration. Moreover, L k ( φ ∗ ( A )) → L k ( φ ∗ ( B )) is a pointwise trivial cofibrationfor k < n , since L k ( A ) → L k ( B ) is so by induction hypothesis. (cid:3) Lemma 5.4.
Let g : Y → X be a trivial Reedy fibration; suppose that for each n ,the induced map M n ( g ) : M n ( Y ) → M n ( X ) is a (pointwise) trivial fibration. Then g : Y → X has the right lifting property with respect to Reedy cofibrations.Proof. Dual to the proof of Lemma 5.2. (cid:3)
Lemma 5.5.
Let g : Y → X be a Reedy fibration such that g r : Y r → X r isa weak equivalence for all objects r of R of degree < n . Then, the induced map M n ( g ) : M n ( Y ) → M n ( X ) is a (pointwise) trivial fibration.Proof. For n = 0, there is nothing to prove; therefore, we can assume inductivelythat M k ( g ) : M k ( Y ) → M k ( X ) is a trivial fibration for k < n . We want to showthat i ∗ n M n ( g ) is a trivial fibration in E R n . To this end, we have to find a filler forany commutative square A ✲ i ∗ n M n ( Y ) Bf ❄ ✲ i ∗ n M n ( X ) i ∗ n M n ( g ) ❄ in E R n in which f : A → B is a cofibration. Since i ∗ n M n = i ∗ n ( δ n ) ∗ γ ∗ n = ( β n ) ∗ κ ∗ n γ ∗ n ,a filler for the former square is the same as a filler for the following square in E R − ( n ) : β ∗ n ( A ) ✲ κ ∗ n γ ∗ n ( Y ) β ∗ n ( B ) β ∗ n ( f ) ❄ ✲ κ ∗ n γ ∗ n ( X ) .κ ∗ n γ ∗ n ( g ) ❄ In order to finish the proof, we shall apply Lemma 5.4 to this square. The category S = R − ( n ) is a generalized Reedy category for which S = S − ; notice that S ishas no non-trivial automorphisms in virtue of axiom (iv) of Definition 1.1; in otherwords, S is equivalent to a strict Reedy category. Therefore, Reedy cofibrationsare the same as pointwise cofibrations, so β ∗ n ( f ) is a Reedy cofibration. Moreover, κ ∗ n γ ∗ n ( f ) is a Reedy weak equivalence in E S , since the objects of S have degree < n .It remains to be shown that κ ∗ n γ ∗ n ( f ) is a Reedy fibration whose induced maps onmatching objects of degree < n are trivial fibrations.Write φ = γ n κ n . By Lemma 4.7, the functor φ ∗ k : E G k ( R ) → E G k ( S ) induces acanonical isomorphism φ ∗ k ( M k ( Y )) ∼ = M k ( φ ∗ ( X )). Therefore, the relative matchingmap φ ∗ ( Y ) → φ ∗ ( X ) × M k ( φ ∗ ( X )) M k ( φ ∗ ( Y )) may be identified with φ ∗ k of the relativematching map Y k → X k × M k ( X ) M k ( Y ). Observe that φ ∗ k preserves fibrations, thus κ ∗ n γ ∗ n ( g ) is a Reedy fibration. Moreover, M k ( φ ∗ ( Y )) → M k ( φ ∗ ( X )) is a trivialfibration for k < n , since M k ( Y ) → M k ( X ) is so by induction hypothesis. (cid:3) Proof of Theorem 1.6.
Limits and colimits in E R are constructed pointwise. Theclass of Reedy weak equivalences has the two-out-of-three property. Moreover, allthree classes are closed under retract. It remains to be shown that the lifting andfactorization axioms of a Quillen model category hold.For the lifting axiom, observe that by Lemma 5.3, trivial Reedy cofibrationsfulfill the hypothesis of Lemma 5.2, and therefore have the left lifting property with respect to Reedy fibrations. Dually, Lemmas 5.5 and 5.4 imply that trivial Reedyfibrations have the right lifting property with respect to Reedy cofibrations.We now come to the factorization axiom. Given a map f : X → Y in E R , weshall construct inductively a factorization X → A → Y of f into a trivial Reedycofibration followed by a Reedy fibration.For n = 0, factor f in E G ( R ) as X −→ A −→ Y into a trivial cofibrationfollowed by a fibration. Next, if X ≤ n − → A ≤ n − → Y ≤ n − is a factorizationof f ≤ n − into trivial Reedy cofibration followed by Reedy fibration in E R ≤ n − , weobtain the following commutative diagram in E G n ( R ) : L n ( X ) ✲ L n ( A ) ✲ L n ( Y ) X n ❄ Y n ❄ M n ( X ) ❄ ✲ M n ( A ) ✲ M n ( Y ) . ❄ This diagram induces a map X n ∪ L n ( X ) L n ( A ) → M n ( A ) × M n ( Y ) Y n which we factoras a trivial cofibration followed by a fibration in E G n ( R ) : X n ∪ L n ( X ) L n ( A ) ∼ ✲ A n ✲ M n ( A ) × M n ( Y ) Y n . The object A n of E G n ( R ) together with the maps L n ( A ) → A n → M n ( A ) definean extension of A ≤ n − to an object A ≤ n in E R ≤ n together with a factorization of f ≤ n : X ≤ n → Y ≤ n into a Reedy cofibration X ≤ n → A ≤ n followed by a Reedyfibration A ≤ n → Y ≤ n . The former map is a trivial Reedy cofibration, because themap X n → A n decomposes into two maps X n → X n ∪ L n ( X ) L n ( A ) → A n , the firstone of which is a weak equivalence by Lemma 5.3, the second one by construction.This defines the required factorization of f ≤ n in E R ≤ n .The factorization of f into a Reedy cofibration followed by a trivial Reedy fibra-tion is constructed in a dual manner using Lemma 5.5 instead of Lemma 5.3. (cid:3) The proof of Theorem 1.6 uses implicitly that trivial Reedy (co)fibrations may becharacterized in terms of relative matching (latching) maps. Since this is a pivotalproperty of the Reedy model structure, we state it explicitly:
Proposition 5.6.
A map f : A → B in E R is a trivial Reedy cofibration if andonly if, for each n , the relative latching map A n ∪ L n ( A ) L n ( B ) → B n is a trivialcofibration in E G n ( R ) .A map g : Y → X in E R is a trivial Reedy fibration if and only if, for each n ,the relative matching map Y n → X n × M n ( X ) M n ( Y ) is a trivial fibration in E G n ( R ) .Proof. For each n , the induced map f n : A n → B n in E G n ( R ) factors as A n u n −→ A n ∪ L n ( A ) L n ( B ) v n −→ B n . If f is a trivial Reedy cofibration then f n is a weak equivalence, so that, by Lemma5.3, u n is a weak equivalence, and hence v n is a trivial cofibration. Conversely, ifeach v n is a trivial cofibration then an induction on n , based on Lemma 5.3, showsthat u n is a weak equivalence, and hence f is a trivial Reedy cofibration.The dual proof for a trivial Reedy fibration g : Y → X uses Lemma 5.5 insteadof Lemma 5.3. (cid:3) N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 19 Skeleta and coskeleta.
In this section we define the skeletal filtration and the coskeletal tower of anyfunctor X : R → E on a generalized Reedy category R , and study their interactionwith the Reedy model structure on E R for a Quillen model category E . We thenintroduce a special class of dualizable generalized Reedy categories for which theskeleta in Sets R op are simple to describe.Recall that for any object X of E R , the restriction j ∗ n X : G n ( R ) → E along j n : G n ( R ) → R is denoted X n . We shall write t n : R ≤ n ֒ → R for the fullembedding of the subcategory of objects of degree ≤ n , cf. Section 4.1. Definition 6.1.
The n -skeleton functor is the endofunctor sk n = t n ! t ∗ n . The n -coskeleton functor is the endofunctor cosk n = t n ∗ t ∗ n . Since t n : R ≤ n ֒ → R is a full embedding, the unit of the ( t n ! , t ∗ n )-adjunction(resp. the counit of the ( t ∗ n , t n ∗ )-adjunction) is an isomorphism; in other words, theendofunctor sk n (resp. cosk n ) is an idempotent comonad (resp. monad ) on E R .The counit of the ( t n ! , t ∗ n )-adjunction (resp. unit of the ( t ∗ n , t n ∗ )-adjunction)provides for each object X of E R a map sk n ( X ) → X (resp. X → cosk n ( X )) in E R .Observe however that these maps need not be monic (resp. epic) for general X .For consistency, sk − ( X ) (resp. cosk − ( X )) will denote an initial (resp. termi-nal) object of E R . Lemma 6.2.
For each object X of E R , the n -th latching object L n ( X ) is canoni-cally isomorphic to sk n − ( X ) n , and the n -th matching object M n ( X ) is canonicallyisomorphic to cosk n − ( X ) n .Under these isomorphisms, the n -th latching map L n ( X ) → X n is induced bythe counit sk n − ( X ) → X , and the n -th matching map X n → M n ( X ) is inducedby the unit X → cosk n − ( X ) .Proof. This follows from the explicit formulas for the left and right Kan extensions t n ! and t n ∗ (cf. Section 3), and from axiom (iii) of Definition 1.1. Indeed, the latterimplies that for any object r of R , the category R + ( r ) is cofinal in the commacategory R ≤ n /r , while the category R − ( r ) is final in the comma category r/ R ≤ n .Moreover, the latching map L n ( X ) → X n of Section 4.3 factors canonically throughthe counit sk n − ( X ) n → X n , while the matching map X n → M n ( X ) of Section 4.6factors canonically through the unit X n → cosk n − ( X ) n . (cid:3) Lemma 6.3.
For any natural numbers m ≤ n , there are canonical isomorphismssk n ◦ sk m ∼ = sk m ∼ = sk m ◦ sk n as well as cosk n ◦ cosk m ∼ = cosk m ∼ = cosk m ◦ cosk n .Proof. This follows readily from the fact that sk n (resp. cosk n ) is an idempotentcomonad (resp. monad) on E R . (cid:3) Lemma 6.3 implies in particular the existence of a compatible system of maps sk m → sk n (resp. cosk n → cosk m ) in E R . The colimit sk ∞ (resp. limit cosk ∞ ) ofthis system is isomorphic to the identity functor of E R . We shall now discuss forwhich objects X of E R , this defines a skeletal filtration (resp. coskeletal tower).Recall that a functor between Quillen model categories is called a left (resp. right ) Quillen functor if it preserves cofibrations and trivial cofibrations (resp.fibrations and trivial fibrations).
Lemma 6.4.
Let E be a Quillen model category and let R be a generalized Reedycategory. We endow E R and E R ≤ n with their Reedy model structures. Then, (i) the left Kan extension t n ! is a left Quillen functor; (ii) the right Kan extension t n ∗ is a right Quillen functor; (iii) the restriction functor t ∗ n is simultaneously a left and right Quillen functor.In particular, sk n (resp. cosk n ) is a left (resp. right) Quillen endofunctor of E R .Proof. By adjointness, (iii) is equivalent to the conjunction of (i) and (ii). Property(iii) follows from Proposition 5.6 and Lemmas 4.4 and 4.7. (cid:3)
Proposition 6.5.
Let E be a Quillen model category and let R be a generalizedReedy category. For any m < n ≤ ∞ , and any object X of E R , (i) if X is Reedy cofibrant, the canonical map sk m ( X ) → sk n ( X ) is a Reedycofibration between Reedy cofibrant objects; (ii) if X is Reedy fibrant, the canonical map cosk n ( X ) → cosk m ( X ) is a Reedyfibration between Reedy fibrant objects.Proof. The proofs of (i) and (ii) are dual; we shall establish (i). By Lemma 6.3, wecan stick to the case n = ∞ , i.e. to the case sk m ( X ) → sk ∞ ( X ) = X . For this,consider the commutative square: L k ( sk m ( X )) ✲ L k ( X ) sk m ( X ) k ❄ ✲ X k . ❄ For k ≤ m , the horizontal maps are isomorphisms, thus the relative latching map sk m ( X ) k ∪ L k ( sk m ( X )) L k ( X ) → X k is an isomorphism too. For k > m , the leftvertical map is an isomorphism by Lemmas 6.2 and 6.3, thus the relative latchingmap coincides with L k ( X ) → X k which is a cofibration by hypothesis. Moreover,Lemma 6.4 shows that sk m ( X ) is Reedy cofibrant. (cid:3) We shall now introduce a special class of generalized Reedy categories R forwhich the skeletal filtration in Sets R op admits a particularly simple description, asin Corollary 6.8 below. In the particular case of the simplex category ∆, this propo-sition was first observed by Eilenberg and Zilber (see [14, 17]), and therefore wehave chosen to name these special categories Eilenberg-Zilber categories , or brieflyEZ-categories. Their formal definition is the following:
Definition 6.6. An EZ-category is a small category R , equipped with a degree-function d : Ob( R ) → N , such that (i) monomorphisms preserve (resp. raise) the degree if and only if they areinvertible (resp. non-invertible); (ii) every morphism factors as a split epimorphism followed by a monomorhism; (iii) any pair of split epimorphisms with common domain has an absolute pushout. Any EZ-category is a dualizable generalized Reedy category where R + (resp. R − )is defined to be the wide subcategory containing all monomorphisms (resp. splitepimorphisms). Notice however that, although the dual of an EZ-category is ageneralized Reedy category, it is in general not an EZ-category. We are mostlyinterested in presheaves on R , so that the reader should be aware of the fact thatthe roles of R ± have to be reversed in the definitions of Sections 4-6. N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 21
Recall (cf. [7]) that an absolute pushout is a pushout which is preserved byany functor or, equivalently, just by the
Yoneda-embedding R ֒ → b R = Sets R op .Notice also that any epimorphism between representable presheaves is split. Axiom(ii) expresses thus that the epi-mono factorization system of the presheaf topos b R restricts (under the Yoneda-embedding) to R , while axiom (iii) can be restated asfollows: in b R , the pushout of any pair of representable epimorphisms with commondomain is representable. This in turn means that for any representable presheaf, theequivalence relation generated by two “representable” equivalence relations is again“representable”. For instance, in the simplex category ∆, representable quotientsof ∆[ m ] correspond bijectively to ordered partitions of [ m ]; one checks that therepresentable quotients form a sublattice of the entire quotient-lattice of ∆[ m ].The presheaf R ( − , r ) represented by an object r of R will be denoted R [ r ].The split epimorphisms of an EZ-category will be called degeneracy operators ; themonomorphisms will be called face operators .Examples of EZ-categories include the simplex category ∆, Segal’s category Γ,the category ∆ sym (see examples 1.8.a-c), as well as the total category R G of acrossed group G on a strict EZ-category R (e.g., the category Λ for cyclic sets, resp.the category Ω for dendroidal sets, see examples 2.7 and 2.8). Indeed, Proposition2.10 shows that R G is a dualizable generalized Reedy category in which axiom (ii)of an EZ-category holds; moreover, the restriction functor d R G → b R is monadic andhence creates absolute pushouts, so that axiom (iii) of an EZ-category also holds.Recall that the Yoneda-lemma allows us to identify elements of a set-valuedpresheaf X on R with maps x : R [ r ] → X in b R ; such a map (or element) x will becalled degenerate if x factors through a non-invertible degeneracy R [ r ] → R [ s ], and non-degenerate otherwise. Proposition 6.7.
Let R be an EZ-category and let X be a presheaf on R . Thenany element x : R [ r ] → X factors in an essentially unique way as a degeneracy ρ x : R [ r ] ։ R [ s ] followed by a non-degenerate element σ x : R [ s ] → X . Any such decomposition will be referred to as a standard decomposition of x . Proof.
The existence of a standard decomposition follows from the facts that thedegree-function takes values in N , and that non-invertible degeneracies lower thedegree by 6.6(i). For the essential uniqueness, observe first that there can be at mostone comparison map from a standard decomposition x = σ x ρ x to another x = σ ′ x ρ ′ x ,since degeneracies are epic. It remains to be shown that such a comparison mapalways exists. Take the absolute pushout of ρ x and ρ ′ x , as provided by 6.6(iii): R [ r ] ρ x ✲ R [ s ] R [ s ′ ] ρ ′ x ❄ τ ′ x ✲ R [ t ] τ x ❄ There exists therefore a map φ x : R [ t ] → X such that φ x τ x = σ x and φ x τ ′ x = σ ′ x .Since σ x and σ ′ x are non-degenerate, the split epimorphisms τ x and τ ′ x must preservethe degree. It then follows from 6.6(i) that τ x and τ ′ x are invertible so that ( τ ′ x ) − τ x provides the required comparison map. (cid:3) Corollary 6.8.
Let R be an EZ-category and let X be a set-valued presheaf on R .Then the counit sk n ( X ) → X is monic and its image is the subobject X ( n ) of thoseelements of X which factor through an element R [ s ] → X of degree s ≤ n .Proof. Notice that the counit sk n ( X ) → X factors through X ( n ) since by definition,for each object r of R , we have sk n ( X ) r = t n ! t ∗ n ( X ) r = lim −→ r → t n ( s ) X s = lim −→ r → s,d ( s ) ≤ n X s . The induced map sk n ( X ) → X ( n ) is pointwise surjective. It remains to be shownthat sk n ( X ) → X ( n ) is pointwise injective. Take two elements x, y in sk n ( X ) givingrise to the same element z in X ( n ) . Then, the essential uniqueness of the standarddecomposition of z shows that x and y define the same element in sk n ( X ). (cid:3) Monoidal Reedy model structures
From now on, we shall assume that E = ( E , ⊗ E , I E , τ E ) is a closed symmetricmonoidal category , see e.g. [7]. Observe that if arbitrary (small) coproducts of theunit object I E exist, there is a canonical functor Sets → E given by S ` S I E .The symmetric monoidal structure will be called solid if these coproducts exist, andif moreover the resulting functor from the category of sets to E is faithful . Objectsand morphisms of E which belong to the essential image of this functor will becalled discrete . Likewise, the presheaf topos b R maps to E R op . Observe that bothfunctors have right adjoints, and hence preserve colimits.Recall that, according to Hovey [21], a monoidal model category is a categorywhich is simultaneously a closed symmetric monoidal category and a Quillen modelcategory such that unit and pushout-product axioms hold. For brevity, we shall saythat a monoidal model category E is solid if(i) the symmetric monoidal structure is solid in the sense mentioned above;(ii) the unit I E is cofibrant ;(iii) for any discrete group G , discrete cofibrations in E G are free G -extensions .Observe that condition (ii) makes the unit axiom redundant, and condition (iii)(applied to the trivial group) implies that discrete cofibrations in E are monic. If E is cofibrantly generated (cf. [20, 21]) and discrete cofibrations in E are monic , thencondition (iii) is automatically satisfied, since in this case the discrete cofibrations in E G are generated by free G -extensions. Examples of solid monoidal model categoriesinclude the category of compactly generated spaces, the category of simplicial sets(both equipped with Quillen’s model structure), and the category of differentialgraded R -modules with the projective model structure.7.1. Boundary inclusions and cofibrations.
For each object r of an EZ-category R , the formal boundary ∂ R [ r ] of R [ r ] is defined to be the subobject of those elementsof R [ r ] which factor through a non-invertible face operator s → r . By Corollary6.8, we have ∂ R [ r ] = sk d ( r ) − R [ r ]. Our main purpose here is to single out a classof maps in b R which induce Reedy cofibrations in E R op for any solid monoidal modelcategory E . This class coincides with Cisinski’s class of normal monomorphisms ,see [10, 8.1.30]. A G -equivariant map of G -sets f : A → B is a free G -extension iff f is monic and G actsfreely on the complement B \ f ( A ) N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 23
Proposition 7.2.
For a map φ : X → Y of set-valued presheaves on an EZ-category R , the following three properties are equivalent: (i) for each object r of R , the relative latching map X r ∪ L r ( X ) L r ( Y ) → Y r isa free Aut( r ) -extension; (ii) φ is monic, and for each object r of R and each non-degenerate element y ∈ Y r \ φ ( X ) r , the isotropy group { g ∈ Aut( r ) | g ∗ ( y ) = y } is trivial; (iii) for each n ≥ , the relative n -skeleton sk n ( φ ) = X ∪ sk n ( X ) sk n ( Y ) is ob-tained from the relative ( n − -skeleton sk n − ( φ ) by attaching a coproductof representable presheaves of degree n along their formal boundary.Proof. (ii) = ⇒ (i). By Lemmas 6.8 and 6.2, the latching object L r ( X ) may beidentified with the subobject of degenerate elements of X r . Since φ is monic,the induced map L r ( φ ) : L r ( X ) → L r ( Y ) is monic; moreover, since split epi-morphisms have the left lifting property with respect to monomorphisms, φ takesnon-degenerate elements of X to non-degenerate elements of Y , i.e. the comple-ment X r \ L r ( X ) to the complement Y r \ L r ( Y ). In particular, the relative latchingmap X r ∪ L r ( X ) L r ( Y ) → Y r is monic, and the complement of its image may beidentified with the set of non-degenerate elements of Y r \ φ ( X ) r .(i) = ⇒ (iii). It follows from Lemmas 6.3 and 6.2 that the canonical map sk n − ( φ ) → sk n ( φ ), evaluated at objects r of degree < n , is an isomorphism,while at objects r of degree n , it evaluates to X r ∪ L r ( X ) L r ( Y ) → Y r . The latter isa free Aut( r )-extension by hypothesis. Since neither sk n − ( φ ) nor sk n ( φ ) containnon-degenerate elements of degree > n , this shows that sk n ( φ ) is obtained from sk n − ( φ ) by attaching, for each Aut( r )-orbit in sk n ( φ ) r \ sk n − ( φ ) r , a distinct copyof R [ r ]; since the orbit is free, the complement R [ r ] \ ∂ R [ r ] is freely attached.(iii) = ⇒ (ii). Since property (ii) is stable under pushout and sequential colimit,it suffices to show that the boundary inclusions ∂ R [ r ] ֒ → R [ r ] have property (ii).The only non-degenerate elements of R [ r ] \ ∂ R [ r ] are the automorphisms of r ; thelatter have trivial isotropy groups. (cid:3) Cisinski shows the equivalence of 7.2(ii) and 7.2(iii) in a slightly different setting, cf.[10, 8.1.1, 8.1.29-35]. In the special cases R = Γ and R = Ω, the skeletal filtration7.2(iii) has been described by Lydakis [26] and Moerdijk-Weiss [29].A map φ : X → Y in b R , fulfilling one of the equivalent conditions of Proposition7.2, will be called a cofibration . Condition 7.2(i) readily implies Corollary 7.3.
Let R be an EZ-category and E be a solid monoidal model category.A map of set-valued presheaves on R is a cofibration if and only if the induced mapin E R op is a Reedy cofibration. Definition 7.4.
An EZ-category R is called quasi-monoidal if the presheaf topos b R carries a symmetric monoidal structure ( b R , (cid:3) , I (cid:3) , τ (cid:3) ) such that (i) the bifunctor − (cid:3) − : b R × b R → b R preserves colimits in both variables; (ii) the unit I (cid:3) is cofibrant; (iii) for all objects r, s of R , the boundary inclusions induce a pullback square ∂ R [ r ] (cid:3) ∂ R [ s ] ✲ ∂ R [ r ] (cid:3) R [ s ] R [ r ] (cid:3) ∂ R [ s ] ❄ ✲ R [ r ] (cid:3) R [ s ] ❄ in b R consisting of cofibrations. Since cofibrations are monic, the induced map R [ r ] (cid:3) ∂ R [ s ] ∪ ∂ R [ r ] (cid:3) ∂ R [ s ] ∂ R [ r ] (cid:3) R [ s ] → R [ r ] (cid:3) R [ s ]is also monic, and hence, by 7.2(ii), a cofibration. It then follows from 7.2(iii) and7.4(i) that the class of cofibrations in b R satisfies Hovey’s pushout-product axiom. Theorem 7.5.
Let R be a quasi-monoidal EZ-category and let E be a cofibrantlygenerated, solid monoidal model category. Then the functor category E R op , equippedwith the Reedy model structure of 1.6 and with the symmetric monoidal structureobtained by convolution, is a cofibrantly generated, solid monoidal model category.Proof. We shall first show that E R op is cofibrantly generated, then define the sym-metric monoidal structure (cid:3) E on E R op , and finally check the pushout-product axiomfor the generating (trivial) cofibrations of E R op .The generating (trivial) Reedy cofibrations of E R op are obtained by “twisting”the generating (trivial) cofibrations of E against the boundary inclusions of b R . Tobe more precise, let f : A → B be an arbitrary generating (trivial) cofibration of E and let i r : ∂ R [ r ] → R [ r ] be a boundary inclusion of b R . For brevity, for any object A of E and any set S , the tensor A ⊗ E ( ` S I E ) will be written A ⊗ E S , and similarlyfor set-valued presheaves on R . We thus obtain the following commutative square A ⊗ E ∂ R [ r ] ✲ B ⊗ E ∂ R [ r ] A ⊗ E R [ r ] ❄ ✲ B ⊗ E R [ r ] ❄ in E R op . The induced comparison map A ⊗ E R [ r ] ∪ A ⊗ E ∂ R [ r ] B ⊗ E ∂ R [ r ] → B ⊗ E R [ r ]is a generating (trivial) Reedy cofibration of E R op , and they are all of this form.Indeed, since the Reedy model structure on E R op is well defined by Theorem 1.6,the generating property just expresses that a map X → Y is a trivial Reedy fibration(resp. Reedy fibration) if and only if it has the right lifting property with respectto the generating Reedy cofibrations (resp. trivial Reedy cofibrations). This inturn follows from the fact that, by adjointness, one of the following two squares A ⊗ E R [ r ] ∪ A ⊗ E ∂ R [ r ] B ⊗ E ∂ R [ r ] ✲ X A ✲ X r B ⊗ E R [ r ] ❄ ✲ Y ❄ B ❄ ✲ Y r × M r ( Y ) M r ( X ) ❄ has a diagonal filler if and only if the other has.The symmetric monoidal structure − (cid:3) E − : E R op × E R op → E R op is defined by theuniversal property that for any objects X, Y, Z of E R op , maps X (cid:3) E Y → Z in E R op correspond bijectively to natural systems of maps X r ⊗ E Y s → Z t in E , indexed bymaps R [ t ] → R [ r ] (cid:3) R [ s ] in b R . In other words, we have the formula( X (cid:3) E Y ) t = lim −→ R [ t ] → R [ r ] (cid:3) R [ s ] X r ⊗ E Y s . In particular, the monoidal structure on E R op is closed (cf. [28, appendix]) and ex-tends the given one on b R ; both share the same unit I (cid:3) (which is cofibrant by 7.4(ii) N AN EXTENSION OF THE NOTION OF REEDY CATEGORY 25 and 7.3) so that the canonical map
Sets → E R op factors through b R . Therefore,all axioms of a cofibrantly generated, solid monoidal model category are satisfied,except possibly the pushout-product axiom. In order to establish the latter, taketwo generating cofibrations f : A → B, g : C → D in E as well as two boundaryinclusions i r , i s of b R , and consider the associated generating Reedy cofibrations f /i r : A ⊗ E R [ r ] ∪ A ⊗ E ∂ R [ r ] B ⊗ E ∂ R [ r ] → B ⊗ E R [ r ] ,g/i s : C ⊗ E R [ s ] ∪ A ⊗ E ∂ R [ s ] D ⊗ E ∂ R [ s ] → D ⊗ E R [ s ] . We shall denote them by f /i r : A/i r → B/i r and g/i s : C/i s → D/i s . We have toshow that the pushout-product map( A/i r (cid:3) E D/i s ) ∪ ( A/i r (cid:3) E C/i s ) ( B/i r (cid:3) E C/i s ) → ( B/i r (cid:3) E D/i s )(9)is a Reedy cofibration which is trivial if f or g is trivial.The operation ( f, i r ) f /i r extends in an evident way to a bifunctor − / − : Arr( E ) × Arr( b R ) → Arr( E R op ) , where Arr( C ) denotes the category of arrows in C .It is now straightforward to verify that (9) is isomorphic to h/i t where h : A ⊗ E D ∪ A ⊗ E C B ⊗ E C → B ⊗ E D and i t : R [ r ] (cid:3) ∂ R [ s ] ∪ ∂ R [ r ] (cid:3) ∂ R [ s ] ∂ R [ r ] (cid:3) R [ s ] → R [ r ] (cid:3) R [ s ]are the canonical comparison maps. Since in E and b R the pushout-product axiomholds, it remains to be shown that for a (trivial) cofibration h in E , and cofibration i t in b R , the map h/i t is a (trivial) Reedy cofibration in E R op . By the adjointnessargument given above, this holds whenever h is a generating (trivial) cofibration,and i t a boundary inclusion; the general case reduces to this special case, sincethe operation ( h, i t ) h/i t commutes with sequential colimits and retracts in eachvariable, and takes pushout squares in each variable to pushout squares in E R op . (cid:3) Examples . (a) The simplex category ∆ is a quasi-monoidal EZ-category for thecartesian product on b ∆. Therefore, the category of simplicial spaces is a monoidalmodel category for the cartesian product, where “space” means either compactlygenerated topological space or simplicial set. In the latter case, the Reedy cofi-brations are precisely the monomorphisms, and the result is of course well-known.Notice that in general, even for strict EZ-categories R , the exactness axiom 7.4(iii)may not be true for the cartesian product on b R .(b) Segal’s [34] category Γ is a (quasi-)monoidal EZ-category for the smash prod-uct on b Γ, as can be deduced from the work of Lydakis [26]. This means that the cat-egory of Γ-spaces, equipped with the strict model structure of Bousfield-Friedlander[8], is a monoidal model category.(c) The category Ω for dendroidal sets (see 2.8) is a quasi-monoidal EZ-categoryfor the
Boardman-Vogt tensor product on b Ω, cf. [28, 11]. Therefore, the category ofdendroidal spaces is a monoidal model category in such a way that the embedding i : ∆ ֒ → Ω induces a monoidal Quillen adjunction between simplicial spaces anddendroidal spaces. It can be shown that, in complete analogy to Rezk’s localizationof simplicial spaces (the model structure for complete Segal spaces, cf. [6, 23, 25,33]), there is a localization of the model category of dendroidal spaces which isQuillen equivalent to the model category of quasi-operads introduced in [11].
References [1] V. Angeltveit –
Enriched Reedy categories , Proc. Amer. Math. Soc. (2008), 2323–2332.[2] E.R. Antokoletz –
Nonabelian Algebraic Models for Classical Homotopy Types , PhD thesis,University of California at Berkeley, 2008.[3] C. Barwick –
On Reedy Model Categories , arXiv:math/0708.2832.[4] M. M. Batanin, M. Markl –
Crossed interval groups and operations on the Hochschild coho-mology , arXiv:math/0803.2249.[5] C. Berger –
Iterated wreath product of the simplex category and iterated loop spaces , Adv.Math. (2007), 230–270.[6] J. E. Bergner –
Three models for the homotopy theory of homotopy theories , Topology (2007), 397–436.[7] F. Borceux – Handbook of Categorical Algebra II , Enc. Math. , Camb. Univ. Press 1994.[8] A. K. Bousfield, E. M. Friedlander – Homotopy theory of Γ -spaces, spectra, and bisimplicialsets , Lecture Notes in Math., vol. , Springer Verlag 1978, 80–130.[9] A. K. Bousfield, D. M. Kan – Homotopy limits, completions, and localization , Lecture Notesin Math., vol. , Springer Verlag, 1972.[10] D.-C. Cisinski –
Les pr´efaisceaux comme mod`eles des types d’homotopie , Ast´erisque, vol. , Soc. Math. France, 2006.[11] D.-C. Cisinski and I. Moerdijk –
Dendroidal sets as a model for homotopy operads , in prep.[12] A. Connes –
Cyclic homology and functor
Ext n , C. R. Acad. Sci. Paris (1983), 953–958.[13] W. G. Dwyer, M. J. Hopkins, D. M. Kan – The homotopy theory of cyclic sets , Trans. Amer.Math. Soc. (1985), 281–289.[14] S. Eilenberg, J. A. Zilber –
Semi-simplicial complexes and singular homology , Ann. of Math. (1950), 499–513.[15] B. L. Feigin, B. L. Tsygan – Additive K -theory , Lecture Notes in Math., vol. SpringerVerlag 1987, 67–209.[16] Z. Fiedorowicz, J.-L. Loday –
Crossed simplicial groups and their associated homology , Trans.Amer. Math. Soc. (1991), 57–87.[17] P. Gabriel, M. Zisman –
Calculus of fractions and homotopy theory , Ergebnisse der Mathe-matik und ihrer Grenzgebiete, vol. , Springer Verlag 1967.[18] P. F. Goerss, R. J. Jardine – Simplicial Homotopy Theory , Progress in Mathematics, vol. , Birkh¨auser, 1999.[19] A. Haefliger –
Extension of complexes of groups , Ann. Inst. Fourier, (1992), 275–311.[20] P. Hirschhorn – Model Categories and Their Localizations , Math. Surveys Monogr., vol. ,Amer. Math. Soc. 2003.[21] M. Hovey – Model Categories , Math. Surveys Monogr., vol. , Amer. Math. Soc., 1999.[22] A. Joyal – Disks, duality and θ -categories , preprint (1997).[23] A. Joyal, M. Tierney – Quasi-categories vs Segal spaces , Contemp. Math. (2007), 277–326.[24] R. Krasauskas –
Skew-simplicial groups , Lithuanian Math. J. (1987), 47–54.[25] J. Lurie – Higher Topos Theory , arXiv:math/0608040.[26] M. Lydakis –
Smash products and Γ -spaces , Math. Camb. Philos. Soc. (1999), 311–328.[27] I. Moerdijk – Orbifolds as groupoids: an introduction , Contemp. Math. (2002), 205–221.[28] I. Moerdijk, I. Weiss –
Dendroidal sets , Alg. and Geom. Top. (2007), 1441–1470.[29] I. Moerdijk, I. Weiss – On inner Kan complexes in the category of dendroidal sets ,arXiv:math/0701295.[30] D. G. Quillen –
Homotopical Algebra , Lecture Notes in Math., vol. , Springer Verlag, 1967.[31] D. G. Quillen – Bivariant cyclic cohomology and models for cyclic homology types , J. PureAppl. Algebra (1995), 1–33.[32] C. L. Reedy –
Homotopy theories of model categories , preprint (1973).[33] C. Rezk –
A model for the homotopy theory of homotopy theory , Trans. Amer. Math. Soc. (2003), 973-1007.[34] G. Segal –
Categories and cohomology theories , Topology (1974), 293–312. Universit´e de Nice, Lab. J.-A. Dieudonn´e, Parc Valrose, 06108 Nice, France.
E-mail: [email protected]
Mathematisch Instituut, Postbus 80.010, 3508 TA Utrecht, The Netherlands.