Co‑Nerve: A Bicategory Cellular Nerve
- Co‑Nerve is a homotopy coherent cellular nerve that uses normal pseudofunctors to embed bicategories fully into presheaves over Θ₂.
- It encodes bicategorical structure by explicitly capturing objects, 1‑cells, and 2‑cells, along with coherence via pasting compositions.
- The construction establishes Quillen equivalences linking bicategories, 2‑quasi‑categories, Rezk’s Θ‑spaces, and enriched Segal categories.
Co‑Nerve for bicategories, in the sense of Leinster’s construction, is the homotopy coherent cellular nerve: a presheaf on defined by normal pseudofunctors out of Joyal’s two‑dimensional cell category. In the formulation developed in "A homotopy coherent cellular nerve for bicategories" (Campbell, 2019), it defines a fully faithful functor from the category of bicategories and normal pseudofunctors to the category of presheaves over , the nerve of a bicategory is a $2$‑quasi‑category in Ara’s sense, and the resulting adjunctions place bicategories, $2$‑quasi‑categories, Rezk’s ‑‑spaces, and quasi‑category‑enriched Segal categories into a common Quillen‑equivalence framework.
1. Definition on
The category (also denoted ) has objects given by free strict $2$‑categories
0
generated by the 1‑graph with objects 2 and, for 3, a hom‑graph 4 with 5 composable edges. Concretely, 6 has hom‑categories
7
for 8, with 9 the ordinal category. Morphisms $2$0 consist of a simplicial map $2$1 and, for each $2$2, a functor
$2$3
equivalently a family of simplicial maps indexed by the slots between $2$4 and $2$5. The category carries a Reedy structure with degree
$2$6
and representables $2$7 together with their boundaries $2$8 play the role of cells and shells.
The homotopy coherent cellular nerve is the singular functor induced by the full inclusion $2$9. For any bicategory $2$0 and any cell $2$1,
$2$2
that is, the set of normal pseudofunctors $2$3. In this sense, the co‑nerve is not merely a record of objects and arrows; it is a $2$4‑diagram of normal pseudofunctors whose shape already carries the relevant $2$5‑dimensional composition data.
The inclusion $2$6 is dense, and the nerve $2$7 is fully faithful: the natural map
$2$8
is bijective. The paper also proves that $2$9 is determined by its restriction to the subcategory 0 on objects of degree 1 and monomorphisms from degree 2 to degree 3: the truncated nerve 4 is already fully faithful, and
5
via right Kan extension (Campbell, 2019).
2. Low‑dimensional cells and coherence encoding
The low‑dimensional evaluations of 6 make explicit how the construction encodes objects, 7‑cells, 8‑cells, and bicategorical coherence. On 9, 0 is the set of objects of 1. On 2, elements are 3‑morphisms 4 in 5, with faces encoding source and target. On 6, elements are 7‑cells 8 in 9, with faces encoding source and target of 0‑cells and degeneracies giving identities.
On 1, elements are invertible 2‑simplices 3 witnessing a composite 4; the nerve sends them to pasting composites in 5 that combine the 6‑cell data with the composition constraints of the pseudofunctor shape 7. The formula recorded in the paper is that 8 on 9 is given by
0
On 1, elements are commutative pasting equations involving the associator
2
in 3; in the nerve they assert the axioms expressing preservation of associativity pasting by normal pseudofunctors.
The point is that composition and coherence are encoded intrinsically via pseudofunctorial constraints rather than imposed externally. This is the distinguishing feature of the homotopy coherent cellular nerve. A plausible implication is that the co‑nerve is adapted to weak rather than strict 4‑categorical input precisely because the 5‑cells are evaluated by normal pseudofunctors, not by strict 6‑functors.
3. Fibrancy as a 7‑quasi‑category
Ara’s model structure on 8 takes cofibrations to be monomorphisms and generating weak equivalences to be all spine inclusions
9
together with
0
where 1 is the strict nerve of the free 2‑category with a single invertible 3‑cell. Its fibrant objects are called 4‑quasi‑categories; they satisfy inner horn‑filling properties in 5. A horn‑filling style characterization states that 6 is fibrant iff 7 has right lifting for the set consisting of
8
9
and
$2$0
for all $2$1 and $2$2.
The main fibrancy theorem is that the nerve functor $2$3 participates as right adjoint in a Quillen adjunction
$2$4
between bicategories with Lack’s model structure and $2$5 with Ara’s model structure. Since every bicategory is fibrant in Lack’s model, $2$6 preserves fibrant objects; hence $2$7 is a $2$8‑quasi‑category for every bicategory $2$9. The proof strategy described in the paper is to show that 00 sends the generators, including spines and 01, to biequivalences in 02, while 03 preserves trivial fibrations and reflects them via lifting against boundary inclusions (Campbell, 2019).
This result identifies the co‑nerve as a fibrant 04‑cellular model of bicategorical input. It succeeds exactly where the strict cellular nerve fails, because the coherent construction records associators and unitors through pseudofunctorial data.
4. Homotopy bicategory and equivalence criterion
For a 05‑quasi‑category 06, the paper constructs the homotopy bicategory 07. The underlying bisimplicial set
08
obtained by restriction along
09
is a Joyal‑enriched Segal category: 10 is discrete, the Segal maps for 11 are weak categorical equivalences, and the 12 are quasi‑categories. Applying 13 levelwise, where 14 is the left adjoint of 15, yields a simplicial category 16 which is a Tamsamani 17‑category. Bicategory reflection 18 then produces a bicategory 19 with objects the elements of 20 and hom‑categories
21
the homotopy categories of the hom‑quasi‑categories.
The universal property is
22
Moreover, the unit
23
is bijective on objects and an equivalence on hom‑quasi‑categories iff 24 is 25‑truncated. Consequently, 26‑truncated 27‑quasi‑categories are exactly those equivalent to coherent nerves of bicategories.
A central structural theorem gives an intrinsic criterion for weak equivalences of 28‑quasi‑categories: a morphism 29 is a weak equivalence iff it is essentially surjective on objects and fully faithful. Essential surjectivity means that
30
is essentially surjective on objects, equivalently
31
Fully faithfulness means that for all 32, the induced map
33
is an equivalence of Kan complexes, that is, a weak homotopy equivalence between fibrant mapping spaces (Campbell, 2019).
5. Quillen equivalences and model comparisons
Lack’s model structure on 34 has weak equivalences given by biequivalences, fibrations given by equifibrations, and trivial fibrations given by morphisms that are surjective on objects and surjective‑on‑objects equivalences on hom‑categories. Against this background, the adjunction
35
is Quillen, the derived right adjoint is fully faithful, and Lack’s structure is right‑induced along 36.
The Quillen equivalence statement requires a Bousfield localisation on the 37 side. Restricting to 38‑truncated 39‑quasi‑categories, namely fibrant objects local with respect to
40
the adjunction 41 becomes a Quillen equivalence between bicategories and 42. The paper then combines this with Ara’s comparison adjunctions
43
between 44 and simplicial presheaves on 45 to deduce that the composite
46
is a Quillen equivalence. Here 47‑48‑spaces are Rezk 49‑spaces 50 such that each hom complete Segal space 51 is 52‑truncated.
The paper also proves a Quillen equivalence
53
between quasi‑category‑enriched Segal categories, denoted 54 in the Hirschowitz–Simpson–Pellissier model, and Ara’s 55‑quasi‑categories. The left adjoint 56 takes a 57‑set to its underlying bisimplicial set, and the right adjoint 58 is right Kan extension along 59. This validates 60‑quasi‑categories as a robust 61‑cellular model for 62‑categories, compatible with enriched Segal models (Campbell, 2019).
6. Variants, examples, and significance
The paper contrasts the homotopy coherent cellular nerve with the strict cellular nerve 63, defined using strict 64‑functors 65. The strict cellular nerve is fully faithful into presheaves but fails fibrancy in general unless 66 is rigid, meaning it has no non‑identity invertible 67‑cells. Nevertheless, the inclusion
68
is a weak equivalence. The paper also situates the construction relative to Street, Duskin, and Roberts nerves, which are simplicial nerves of bicategories or 69‑categories, and to Lurie’s homotopy coherent nerve of simplicial categories. The analogy with Lurie’s construction is explicit: both encode homotopy coherent composition, but the target and cell shapes differ, since the present construction uses 70‑cells rather than 71‑simplices with enrichment.
A worked example is provided by a monoidal category 72 regarded as a one‑object bicategory 73. The single object is 74; the 75‑morphisms 76 are objects of 77; composition is tensor 78; identity is 79; and 80‑morphisms are morphisms in 81, with associator 82 and unitors 83 as the bicategory constraints. Then
84
85
86
Moreover, 87 consists of coherent triangles 88, and 89 encodes the associativity pentagon. Inner horns for 90 of 91‑dimensional cells are filled in 92 by the associator 93 and unitors 94 of 95, matching Ara’s inner horn‑lifting notion via spines and 96.
The significance of the co‑nerve is therefore twofold. First, it provides a fully faithful embedding of bicategories into 97‑cellular presheaves, landing in Ara’s 98‑quasi‑categories. Second, through the adjunctions and Quillen equivalences above, it connects bicategories to Rezk 99‑spaces and enriched Segal categories as models of $2$00. This suggests that the homotopy coherent cellular nerve is not only a representation theorem for bicategories but also a comparison mechanism across several established higher‑categorical formalisms (Campbell, 2019).