Introduction
Motivation
Our primary motivation to write this paper comes from two questions, which are broadly related under the microcosm principle. The first one concerns the dualities of categorical structures. Consider the op 1 \operatorname{op}_{1} op 1 duality on 1 C a t \mathbf{1Cat} 1Cat , expressed as a 2-functor op 1 : op 2 ( 1 C a t ) → 1 C a t \operatorname{op}_{1}: \operatorname{op}_{2}(\mathbf{1Cat}) \to\mathbf{1Cat} op 1 : op 2 ( 1Cat ) → 1Cat , which satisfy the relation op 1 ∘ op 2 ( op 1 ) = 1 \operatorname{op}_{1} \circ\operatorname{op}_{2}(\operatorname{op}_{1}) = 1 op 1 ∘ op 2 ( op 1 ) = 1 . Clearly, this is not an involution in the traditional sense of satisfying op 1 ∘ op 1 = 1 \operatorname{op}_{1} \circ\operatorname{op}_{1} = 1 op 1 ∘ op 1 = 1 , since this composition is simply not defined. Rather, it is an involution relative to a higher involution op 2 \operatorname{op}_{2} op 2 . This motivates the definition of a relative involution.
Now consider the duality op 1 : op 2 ( M o n C a t c ) → M o n C a t l \operatorname{op}_{1}: \operatorname{op}_{2}(\mathbf{MonCat}_{c}) \to\mathbf{MonCat}_{l} op 1 : op 2 ( MonCat c ) → MonCat l . Here, the domain consists of colax monoidal functors, whereas the codomain consists of lax monoidal functors. This is again a duality, but it is no longer directly expressible even using relative involution as before. Instead, this motivates a further generalization to relative inverses.
As with many algebraic structures, specifying the complete coherence data for these definitions becomes increasingly difficult in higher dimensions. Nevertheless, it is tractable and useful enough to define these for involutive bicategories and biadjunction of bicategories. In particular, this framework is sufficiently general to express the dualities of monoidal structures. Furthermore, it is also natural to consider module structures in the presence of monoidal structures. Thus, one of the main goals of this paper is to extend this framework onto virtual double categories and the corresponding module construction.
The second question concerns slicing and its interaction with these dualities. In particular, we wish to describe a general construction of enriched slice and coslice categories, especially under the consideration of model structures. Although there are some accounts of enriched slice constructions, it is useful to have a systematic treatment that encompasses both slice and coslice constructions. Furthermore, the slice and coslice category construction is related by the op 1 \operatorname{op}_{1} op 1 duality. This suggests that the enriched coslice case should be a dual of the enriched slice case in some appropriate sense. Establishing such a construction is the second main goal of this paper.
Main results and applications
The results can be divided into three major parts. The first part establishes a desired framework of relative involutions and relative inverses. The second part applies this framework to characterize dualities of constructions on virtual double categories. The third part discuss the slicing of module structures and establish a general construction of enriched slice and coslice model categories.
Recall that one can define a 1-category of relative involutions, called OInv ( C , T , α ) \operatorname{OInv}(\mathcal{C}, T, \alpha) OInv ( C , T , α ) , inside an arbitrary lax involutive category ( C , T , α ) (\mathcal{C}, T, \alpha) ( C , T , α ) [34 ] (or see section 3.1). This is an example of the microcosm principle. Motivated by this fact, we show a similar construction for adjunctions by defining relative inverses. The main result in this part is to generalize this framework one dimension higher for weak 2-involutive bicategories and weak 2-inverses of bicategories. This can be summarized by the following proposition.
Proposition. The microcosm principle holds for weak 2-involutive bicategories and weak 2-inverses of bicategories.
In particular, this provides suitable definitions of relative weak 1 1 1 -inverses, relative weak 1 1 1 -involutions, and their 1 1 1 -morphisms. Note that expressing the dualities of strict n n n -categories only require relative involution or relative inverse (see definition 2.1.4). This is in contrast to the monoidal structures, which also require the 1 1 1 -morphisms. The main result here is that the 2 2 2 -functors M o d : v D b l C a t → v D b l C a t \mathbf{Mod}\colon\mathbf{vDblCat} \to\mathbf{vDblCat} Mod : vDblCat → vDblCat and C o M o d : v c o D b l C a t → v c o D b l C a t \mathbf{CoMod}\colon\mathbf{vcoDblCat} \to\mathbf{vcoDblCat} CoMod : vcoDblCat → vcoDblCat are compatible with the relevant dualities.
Theorem. The following extensions hold:
M o d \mathbf{Mod} Mod extends to a 1 1 1 -morphism of involution, respecting the horizontal duality of virtual double categories.
M o d \mathbf{Mod} Mod and C o M o d \mathbf{CoMod} CoMod together extend to a 1 1 1 -morphism of inverses, respecting the vertical duality of virtual double categories.
In the final part, we first briefly review the slice constructions for module categories. Recall that standard definitions of an enriched model category assumes the existence of a two-variable closure adjunction relating tensoring, powering, and enrichment [20] (or see § 5.2 ). Hence, we focus primarily on the behavior of this two-variable adjunction under slicing. We show that it can be sliced in two dual ways, one by slicing the tensoring and the other by slicing the powering. The main result is that these slicing also respect the model structures.
Theorem. Let C \mathcal{C} C be a V \mathcal{V} V -enriched model categories. Let M ∈ V M \in\mathcal{V} M ∈ V and Z ∈ C Z \in\mathcal{C} Z ∈ C be suitable monoid and module object, respectively. Then, both Z / C Z/\mathcal{C} Z / C and C / Z \mathcal{C}/Z C / Z are V / M \mathcal{V}/M V / M -enriched model categories.
Overview
Section 2 2 2 discusses the dualities of strict n n n -categories as a motivating example of nested algebraic structures. To formulate this more precisely, it recall the microcosm principle and define the sliced microcosm principle. A relevant example of these principles is then provided in terms of pseudomonoid and pseudomodule objects.
Section 3 3 3 begins by briefly reviewing the theory of involutions and relative involutions. Correspondingly, it defines the similar theory of inverses and relative inverses. Then, the analogues of these structures are described in one dimension higher.
Section 4 4 4 starts with reformulating the dualities of monoidal categories within the framework of relative inverses and involutions. This framework is then used to describe the corresponding dualities of internal monoid structures. Subsequently, the dualities are generalized to the level of virtual double categories, encoding the action on the module data.
Section 5 5 5 applies the sliced microcosm principle to monoidal and module structures. First, it reviews enrichment through tensoring and states the corresponding slice construction. The results are then dualized for the case of cotensoring. Finally, these constructions are extended to the setting of enriched model categories.
Preliminaries
TEXT
nCat, monoidal n n n -structures, and categorical dualities
Recall that the category 1 C a t = C a t \mathbf{1}\mathrm{Cat}=\mathrm{Cat} 1 Cat = Cat is canonically enriched over itself using the functor categories C a t ( C , D ) \mathrm{Cat}(\mathcal{C},\mathcal{D}) Cat ( C , D ) . It also has a Kelly tensor product given by the cartesian product C × D \mathcal{C}\times\mathcal{D} C × D of two categories, which forms a monoidal structure. Therefore, it is natural to ask whether this enrichment and the monoidal structure are compatible. Indeed, we can enrich the monoidal structure to a monoidal 2 2 2 -structure1 in a canonical way. There exists a C a t \mathrm{Cat} Cat enriched adjunction
C a t ‾ ( C × D , E ) ≃ C a t ‾ ( C , C a t ‾ ( D , E ) ) , \underline{\mathrm{Cat}}(\mathcal{C}\times\mathcal{D},\mathcal{E})\simeq\underline{\mathrm{Cat}}(\mathcal{C},\underline{\mathrm{Cat}}(\mathcal{D},\mathcal{E})), Cat ( C × D , E ) ≃ Cat ( C , Cat ( D , E )) ,
by using a general result which holds for any closed symmetric monoidal category category V \mathcal{V} V [22, Prop. 3.1.11]. In this enriched adjunction, both the product and the internal hom functors are 2 2 2 -functors. Furthermore, the left unitor, the right unitor, the associator, and the symmetry are now all 2 2 2 -natural transformations. This defines, naturally, a symmetric cartesian closed monoidal 2 2 2 -structure.
Lemma 2.1.1. The tuple ( C a t , × , { ∗ } , a C a t , λ C a t , ρ C a t ) (\mathrm{Cat},\times,\{\ast\},a_{\mathrm{Cat}},\lambda_{\mathrm{Cat}},\rho_{\mathrm{Cat}}) ( Cat , × , { ∗ } , a Cat , λ Cat , ρ Cat ) defines a symmetric cartesian closed monoidal structure on C a t \mathrm{Cat} Cat . Furthermore, this monoidal structure can be C a t \mathrm{Cat} Cat -enriched by canonically extending the monoidal data { × , a C a t , λ C a t , ρ C a t } \{\times,a_{\mathrm{Cat}},\lambda_{\mathrm{Cat}},\rho_{\mathrm{Cat}}\} { × , a Cat , λ Cat , ρ Cat } to a strict 2 2 2 -functor and strict 2 2 2 -natural transformations. All the coherence relations remain unchanged. This makes C a t \mathrm{Cat} Cat a monoidal 2 2 2 -category.
As a consequence, it is natural question to ask whether the same result extends to n C a t \mathrm{nCat} nCat or ω C a t \omega\mathrm{Cat} ω Cat . The following general lemma answers this in affirmative.
Lemma 2.1.2. Let V \mathcal{V} V be a locally small, complete, and cocomplete cartesian closed monoidal category. Then, the category V C a t \mathcal{V}\mathrm{Cat} V Cat (of small V \mathcal{V} V -categories) is also locally small, complete, cocomplete, and cartesian closed. Furthermore, the cartesian product of V C a t \mathcal{V}\mathrm{Cat} V Cat is precisely the Kelly tensor product, as given by the cartesian product of sets of objects and the corresponding products of hom-objects.
Proof. Given any symmetric closed monoidal category V \mathcal{V} V , the (co)completeness of V \mathcal{V} V implies the (co)completeness of V C a t \mathcal{V}\mathrm{Cat} V Cat by results of [33 ] [6, 24]. And, the Kelly tensor product of V C a t \mathcal{V}\mathrm{Cat} V Cat forms a symmetric closed monoidal structure by [23, Sec. 2.3]. We are only left to verify that the kelly tensor product is indeed the cartesian product. This last part essentially follows from the monadicity of V C a t \mathcal{V}\mathrm{Cat} V Cat over V G r a p h \mathcal{V}\mathrm{Graph} V Graph .
In particular, this lemma provides the following desired corollary.
Corollary 2.1.3. The tuple ( n C a t , × , { ∗ } , a n C a t , λ n C a t , ρ n C a t ) (\mathrm{nCat},\times,\{\ast\},a_{\mathrm{nCat}},\lambda_{\mathrm{nCat}},\rho_{\mathrm{nCat}}) ( nCat , × , { ∗ } , a nCat , λ nCat , ρ nCat ) canonically defines a symmetric cartesian closed monoidal ( n + 1 ) (n+1) ( n + 1 ) -structure on n C a t \mathrm{nCat} nCat .
Finally, this result also extends to ω C a t \omega\mathrm{Cat} ω Cat , which can be defined as a form of colimit of the enrichment process. In particular, the category of all (small) ω C a t \omega\mathrm{Cat} ω Cat -enriched category is just the category ω C a t \omega\mathrm{Cat} ω Cat itself and is cartesian closed. The interested reader maybe look up the details in [26, 32].
Using suitable universes, it is also clear that each monoidal ( n + 1 ) (n+1) ( n + 1 ) -category n C a t \mathrm{nCat} nCat is equivalently a “monoidal object” inside the monoidal ( n + 2 ) (n+2) ( n + 2 ) -category ( n + 1 ) C a t (n+1)\mathrm{Cat} ( n + 1 ) Cat . This provides our first example of a nested algebraic structure .
Another primary example is given by the duality involution of strict n n n -categories. Consider the 1 1 1 -functor op 1 : 1 C a t → 1 C a t \operatorname{op}_{1}: 1\mathrm{Cat} \to1\mathrm{Cat} op 1 : 1 Cat → 1 Cat which sends each category C C C to its opposite category C o p C^{\mathrm{op}} C op . This is a duality involution 1 1 1 -functor on 1 C a t 1\mathrm{Cat} 1 Cat , that is, op 1 2 = 1 \operatorname{op}_{1}^{2}=1 op 1 2 = 1 . However, as 1 C a t 1\mathrm{Cat} 1 Cat is naturally a strict 2 2 2 -category, we may wish to upgrade this duality involution to a strict 2 2 2 -functor. This is not possible. What we instead have is a strict 2 2 2 -functor op 1 : op 2 ( 1 C a t ) → 1 C a t \operatorname{op}_{1}: \operatorname{op}_{2}(1\mathrm{Cat}) \to1\mathrm{Cat} op 1 : op 2 ( 1 Cat ) → 1 Cat , where op 2 \operatorname{op}_{2} op 2 is a strict 1 1 1 -functor op 2 : 2 C a t ≤ 1 → 2 C a t ≤ 1 \operatorname{op}_{2}: 2\mathrm{Cat}_{\leq1} \to2\mathrm{Cat}_{\leq1} op 2 : 2 Cat ≤ 1 → 2 Cat ≤ 1 and ( − ) ≤ i (-)_{\leq i} ( − ) ≤ i denotes truncation of categorical structures. Moreover, the duality of op 1 \operatorname{op}_{1} op 1 is now expressed by the equation op 1 ∘ op 2 ( op 1 ) = 1 \operatorname{op}_{1} \circ\operatorname{op}_{2}(\operatorname{op}_{1})=1 op 1 ∘ op 2 ( op 1 ) = 1 .
Thus, op 1 \operatorname{op}_{1} op 1 duality is fully defined only in the presence of a higher dimensional op 2 \operatorname{op}_{2} op 2 duality. More generally, let op n , k \operatorname{op}_{n,k} op n , k denote the k k k -cell duality on n n n -categories for 1 ≤ k ≤ n 1 \leq k \leq n 1 ≤ k ≤ n . All the 1 1 1 -functors ( op n , k ) ≤ 1 : n C a t ≤ 1 → n C a t ≤ 1 (\operatorname{op}_{n,k})_{\leq1}: n\mathrm{Cat}_{\leq1} \to n\mathrm{Cat}_{\leq1} ( op n , k ) ≤ 1 : n Cat ≤ 1 → n Cat ≤ 1 are definable readily. See [3, 5] for example. This leads to the following natural proposition.
Proposition 2.1.4. For each 1 ≤ k ≤ n 1 \leq k \leq n 1 ≤ k ≤ n , the 1 1 1 -functor ( op n , k ) ≤ 1 : n C a t ≤ 1 → n C a t ≤ 1 (\operatorname{op}_{n,k})_{\leq1}: n\mathrm{Cat}_{\leq1} \to n\mathrm{Cat}_{\leq1} ( op n , k ) ≤ 1 : n Cat ≤ 1 → n Cat ≤ 1 extends to a corresponding ( n + 1 ) (n+1) ( n + 1 ) -functor op n , k : op n + 1 , k + 1 ( n C a t ) → n C a t \operatorname{op}_{n,k}: \operatorname{op}_{n+1,k+1}(n\mathrm{Cat}) \to n\mathrm{Cat} op n , k : op n + 1 , k + 1 ( n Cat ) → n Cat .
Note that both n C a t n\mathrm{Cat} n Cat and op n + 1 , k + 1 ( n C a t ) \operatorname{op}_{n+1,k+1}(n\mathrm{Cat}) op n + 1 , k + 1 ( n Cat ) are enriched over n C a t n\mathrm{Cat} n Cat by definition as they are ( n + 1 ) (n+1) ( n + 1 ) -categories. This means that our primary goal is to extend the given 1 1 1 -functors to corresponding n C a t n\mathrm{Cat} n Cat -enriched functor. Consequently, we first state the following more general lemma inspired by an example in [31, Ex. 4.6].
Lemma 2.1.5. Let ( V , ⊗ , I ) (\mathcal{V}, \otimes, I) ( V , ⊗ , I ) be a closed symmetric monoidal category equipped with a strict symmetric monoidal functor o : V → V o: \mathcal{V} \to\mathcal{V} o : V → V satisfying o 2 = 1 o^{2}=1 o 2 = 1 . Then, there exists a canonical V \mathcal{V} V -enriched functor o : o ∗ ( V ) → V o: o_{*}(\mathcal{V}) \to\mathcal{V} o : o ∗ ( V ) → V , which extends the 1 1 1 -functor o : V → V o: \mathcal{V} \to\mathcal{V} o : V → V .
Proof. V \mathcal{V} V is enriched, tensored, and cotensored over itself because it is a closed symmetric monoidal category. More precisely, V \mathcal{V} V extends to a symmetric monoidal V \mathcal{V} V -enriched category [22, Thm. 3.3.2]. Thus, we can do a change of base of enrichment to produce a V \mathcal{V} V -enriched category o ∗ ( V ) o_{*}(\mathcal{V}) o ∗ ( V ) . Note that the underlying category of o ∗ ( V ) o_{*}(\mathcal{V}) o ∗ ( V ) is still V \mathcal{V} V itself because o o o is a strict monoidal functor.
The key idea is to note that o ∗ ( V ) o_{*}(\mathcal{V}) o ∗ ( V ) is also tensored over V \mathcal{V} V . We define the tensoring operation using the functor ◃ : V × V → o × 1 V × V → ⊗ V \triangleleft: \mathcal{V} \times\mathcal{V} \xrightarrow{o\times1} \mathcal{V} \times\mathcal{V} \xrightarrow{\otimes} \mathcal{V} ◃ : V × V o × 1 V × V ⊗ V . Our goal is now to show that there is a V \mathcal{V} V -isomorphism
o ∗ ( V ) ( A ◃ X , Y ) ≅ V ( A , o ∗ ( V ) ( X , Y ) ) o_{*}(\mathcal{V})(A \triangleleft X,Y) \cong\mathcal{V}(A,o_{*}(\mathcal{V})(X,Y)) o ∗ ( V ) ( A ◃ X , Y ) ≅ V ( A , o ∗ ( V ) ( X , Y ))
natural in all of A , X , Y A,X,Y A , X , Y . This is a straightforward exercise using the Yoneda lemma. Thus, we have established that ◃ \triangleleft ◃ is indeed a valid tensoring operation.
Finally, we note the following commuting diagram
V × V → 1 × o V × V → X ⊗ o ( Y ) o × 1 ↓ ↓ ⊗ ↓ V × V → ⊗ V → o V → o ( X ◃ Y ) \begin{CD}
\mathcal{V}\times\mathcal{V} @>{1\times o}>> \mathcal{V}\times\mathcal{V} @>>> X\otimes o(Y) \\
@V{o\times1}VV @VV{\otimes}V @VVV \\
\mathcal{V}\times\mathcal{V} @>{\otimes}>> \mathcal{V} @>{o}>> \mathcal{V} @>>> o(X\triangleleft Y)
\end{CD} V × V o × 1 ↓ ⏐ V × V 1 × o ⊗ V × V ↓ ⏐ ⊗ V o X ⊗ o ( Y ) ↓ ⏐ V o ( X ◃ Y )
which provides a canonical identity left tensorial strength for the unenriched functor o o o . This implies the existence of a canonical V \mathcal{V} V -enriched functor o : o ∗ ( V ) → V o:o_*(\mathcal{V})\to\mathcal{V} o : o ∗ ( V ) → V extending the 1-functor o : V → V o:\mathcal{V}\to\mathcal{V} o : V → V . For more details, see [30, Def. 3.1.1 and Prop 3.2.4]. Explicitly, the action on hom objects is given by the adjoint to composite
o ( o ∗ ( V ) ( X , Y ) ◃ X ) → o ( ◃ X Y ) o Y ≅ ↑ ↑ ≅ o ∗ ( V ) ( X , Y ) ⊗ o X → o ( ⊗ ϵ X Y ) o Y \begin{CD}
o(o_*(\mathcal{V})(X,Y)\triangleleft X) @>{o(\triangleleft_X^Y)}>> oY \\
@A{\cong}AA @AA{\cong}A \\
o_*(\mathcal{V})(X,Y)\otimes oX @>{o(\otimes\epsilon_X^Y)}>> oY
\end{CD} o ( o ∗ ( V ) ( X , Y ) ◃ X ) ≅ ⏐ ↑ o ∗ ( V ) ( X , Y ) ⊗ o X o ( ◃ X Y ) o ( ⊗ ϵ X Y ) o Y ⏐ ↑ ≅ o Y
Now, we can use this lemma in our context to obtain the desired result.
Proof of definition 2.1.4. This readily follows once we note that o p n + 1 , k + 1 ( n C a t ) = ( o p n , k ) ∗ ( n C a t ) o_{p_{n+1,k+1}}(\mathsf{nCat})=(o_{p_{n,k}})_*(\mathsf{nCat}) o p n + 1 , k + 1 ( nCat ) = ( o p n , k ) ∗ ( nCat ) and that all the 1-functors ( o p n , k ) ≤ 1 (o_{p_{n,k}})_{\leq1} ( o p n , k ) ≤ 1 (as defined in [5, Sec. 5]) are all strictly symmetric and strictly involutive in the desired sense.
We can also provide a more concrete description of these ( n + 1 ) (n+1) ( n + 1 ) -functors on the hom n n n -categories. Notice that the diagram [ 0 ] ⇉ [ 1 ] [0]\rightrightarrows[1] [ 0 ] ⇉ [ 1 ] defines a cocategory object in C a t \mathsf{Cat} Cat . In particular, this allows one to compute the functor categories D C \mathcal{D}^{\mathcal{C}} D C as ( D C ) 0 ≅ C a t ( [ 0 ] × C , D ) (\mathcal{D}^{\mathcal{C}})_0\cong\mathsf{Cat}([0]\times\mathcal{C},\mathcal{D}) ( D C ) 0 ≅ Cat ([ 0 ] × C , D ) and ( D C ) 1 ≅ C a t ( [ 1 ] × C , C ) (\mathcal{D}^{\mathcal{C}})_1\cong\mathsf{Cat}([1]\times\mathcal{C},\mathcal{C}) ( D C ) 1 ≅ Cat ([ 1 ] × C , C ) . All of the source, target, identity, and composition maps for D C \mathcal{D}^{\mathcal{C}} D C are derived from the cocategory object structure of [ 0 ] ⇉ [ 1 ] [0]\rightrightarrows[1] [ 0 ] ⇉ [ 1 ] .
More generally, the category of all n n n -categories n C a t \mathsf{nCat} nCat has a n n n -cocategory object which can be generated from the diagram G 0 ⇒ G 1 ⇒ G 2 ⋯ ⇒ G n G_0\Rightarrow G_1\Rightarrow G_2\cdots\Rightarrow G_n G 0 ⇒ G 1 ⇒ G 2 ⋯ ⇒ G n , where each G i G_i G i is the i i i -globe. This essentially follows from [2, Prop. 3.14 and Prop. 3.15] by considering the Globe category embedding G n → Θ n \mathbb{G}_n\to\Theta_n G n → Θ n .
For each 1 ≤ k ≤ n 1\leq k\leq n 1 ≤ k ≤ n , the globe G i G_i G i has canonical isomorphisms o p n , k ( G i ) ≅ G i o_{p_{n,k}}(G_i)\cong G_i o p n , k ( G i ) ≅ G i induced by the corresponding duality on the n n n -Globe category o p n , k : G n → G n o_{p_{n,k}}:\mathbb{G}_n\to\mathbb{G}_n o p n , k : G n → G n , which swaps the source map and target map that goes into the k k k th level. Indeed, the corresponding ( o p n , k ) ≤ 1 (o_{p_{n,k}})_{\leq1} ( o p n , k ) ≤ 1 dualities on n n n -globular presheaves or n n n -categories are also precisely induced from the ones on the Globe category. See [5, Sec. 5] for more details. Furthermore, like any arbitrary functor, the globe category dualities are compatible with the diagonal map G n → G n × G n \mathbb{G}_n\to\mathbb{G}_n\times\mathbb{G}_n G n → G n × G n and the swap map G n × G n → s G n × G n \mathbb{G}_n\times\mathbb{G}_n\xrightarrow{s}\mathbb{G}_n\times\mathbb{G}_n G n × G n s G n × G n . Thus, demonstrating the strict symmetric monoidal nature of the induced dualities.
Using this, we can verify that the action of op n , k \operatorname{op}_{n,k} op n , k on the i i i -transfor G i × C → D G_i \times C \to D G i × C → D is given by
G i × C → ( op n , k ) ≤ 1 op n , k G i × op n , k C → ≅ G i × op n , k C ↓ ↓ ↓ D → op n , k op n , k D → op n , k op n , k D \begin{CD}
G_i \times C @>{(\operatorname{op}_{n,k})_{\leq1}}>> \operatorname{op}_{n,k}G_i \times\operatorname{op}_{n,k}C @>{\cong}>> G_i \times\operatorname{op}_{n,k}C \\
@VVV @VVV @VVV \\
D @>{\operatorname{op}_{n,k}}>> \operatorname{op}_{n,k}D @>{\operatorname{op}_{n,k}}>> \operatorname{op}_{n,k}D
\end{CD} G i × C ↓ ⏐ D ( op n , k ) ≤ 1 op n , k op n , k G i × op n , k C ↓ ⏐ op n , k D ≅ op n , k G i × op n , k C ↓ ⏐ op n , k D
This ensures that op n , k \operatorname{op}_{n,k} op n , k changes the direction of only the k k k -transfor. For instance, op 1 , 1 \operatorname{op}_{1,1} op 1 , 1 acting on the 1 1 1 -functor [ 1 ] × C → D [1] \times C \to D [ 1 ] × C → D changes only the direction of [ 1 ] [1] [ 1 ] , which is the direction of natural transformations.
Ultimately, each lower dimensional duality is fully definable only when the higher dimensional duality is defined, at least in a 1 1 1 -truncated form. These nested structures can be better formulated and understood using the microcosm principle
Mircocosm principle
The initial statement of the microcosm principle goes back to Baez and Dolan, abstracting a general guiding principle from the observation that monoids are definable in any monoidal category.
Definition 2.2.1 (Microcosm principle, Baez and Dolan). The microcosm principle suggests that, in order to define a generic ( n − 1 ) (n - 1) ( n − 1 ) -coherent algebraic structure inside an n n n -category, the category itself should be equipped with a categorified n n n -coherent version of the same algebraic structure.
A precise formulation of the microcosm principle is given in [4, Def. 57] using the opetopic foundations of n n n -categories. The key idea is to define internal algebraic structures by using maps out of the terminal categorified algebraic structure. However, this result is not immediately applicable to our usual algebraic foundations. Nor does it address whether slicing with respect to an internal algebraic object is always possible. Indeed, consider a monoid object M M M in a monoidal category V \mathcal{V} V . The slice category V / M \mathcal{V}/M V / M admits a canonical monoidal category structure with the sliced tensor product X ⊗ Y → M ⊗ M → M X \otimes Y \to M \otimes M \to M X ⊗ Y → M ⊗ M → M . This example suggests another general principle governing these slice constructions. Accordingly, we propose the following variant of the microcosm principle.
Definition 2.2.2 (Sliced microcosm principle). Let A A A be a ( n − 1 ) (n - 1) ( n − 1 ) -coherent algebraic structure inside a n n n -category C \mathcal{C} C equipped with the categorified n n n -coherent algebraic structure. The slice n n n -category C / A \mathcal{C}/A C / A admits a canonical sliced n n n -coherent algebraic structure.
We emphasize that both of these are merely guiding principles rather than theorems. For instance, applying the microcosm principle on the Lawvere theory of groups doesn’t yield a meaningful notion of a group object inside a “groupoidal category”. The key obstruction for groups is the necessity of the copy operation Δ : G → G × G \Delta: G \to G \times G Δ : G → G × G and the deletion operation G → 1 G \to1 G → 1 for defining the inverse axioms.
More generally, the microcosm principle is incompatible with algebraic theories that demands some form of copying or deletion of data. In contrast, it is well suited for theories, such as that of monoids, which require no copying or deletion. We may, in general, expect the microcosm principle to be compatible with any “linear” multisorted algebraic theory. Furthermore, one should also expect the existence of sliced algebraic structures in such a case.
Pseudomonoids and pseudomodules
Monoidal categories, module categories, and their dualities are important in the latter part of this paper. Therefore, it is useful to introduce more abstract definitions of pseudomonoid objects and pseudomodule objects, both for conceptual clarity and to provide another useful example of the microcosm principle.
Pseudomonoids and pseudomodules are 1-level higher weak categorification of the idea of a monoid object and a module over a monoid object. These objects can be defined in a general monoidal bicategory V \mathcal{V} V . We shall view V \mathcal{V} V as a one-object tricategory [18, Def. 4.1]. Furthermore, all diagrams are read inside the one object cubical tricategory st V \operatorname{st}\mathcal{V} st V (see [18, Thm. 8.4] and section B for explanation).
Definition 2.3.1. A pseudomonoid in a monoidal bicategory V \mathcal{V} V is given by the data ( M , μ , e , a , λ , ρ ) (M,\mu,e,a,\lambda,\rho) ( M , μ , e , a , λ , ρ ) such that
M ∈ V M \in\mathcal{V} M ∈ V is an object, μ \mu μ is a 1-cell μ : M ⊗ M → M \mu:M\otimes M\to M μ : M ⊗ M → M , and e e e is a 1-cell e : I → M e:I\to M e : I → M ,
a , λ , ρ a,\lambda,\rho a , λ , ρ are invertible 2-cells in V \mathcal{V} V which are defined using the following diagrams:
M ⊗ ( M ⊗ M ) → 1 ⊗ μ M ⊗ M a V ↓ ↓ μ ( M ⊗ M ) ⊗ M → μ ⊗ 1 M ⊗ M → μ M M ⊗ I → ρ V M 1 ⊗ e ↓ ↑ μ M ⊗ M → μ M I ⊗ M → λ V M e ⊗ 1 ↓ ↑ μ M ⊗ M → μ M \begin{CD}
M\otimes(M\otimes M) @>{1\otimes\mu}>> M\otimes M \\
@V{a_{\mathcal{V}}}VV @VV{\mu}V \\
(M\otimes M)\otimes M @>{\mu\otimes1}>> M\otimes M @>{\mu}>> M
\end{CD}
\qquad
\begin{CD}
M\otimes I @>{\rho_{\mathcal{V}}}>> M \\
@V{1\otimes e}VV @AA{\mu}A \\
M\otimes M @>>{\mu}> M
\end{CD}
\qquad
\begin{CD}
I\otimes M @>{\lambda_{\mathcal{V}}}>> M \\
@V{e\otimes1}VV @AA{\mu}A \\
M\otimes M @>>{\mu}> M
\end{CD} M ⊗ ( M ⊗ M ) a V ↓ ⏐ ( M ⊗ M ) ⊗ M 1 ⊗ μ μ ⊗ 1 M ⊗ M ↓ ⏐ μ M ⊗ M μ M M ⊗ I 1 ⊗ e ↓ ⏐ M ⊗ M ρ V μ M ⏐ ↑ μ M I ⊗ M e ⊗ 1 ↓ ⏐ M ⊗ M λ V μ M ⏐ ↑ μ M the associator a a a and the unitors λ , ρ \lambda,\rho λ , ρ are subject to the pentagon axiom (section B.1) and the triangle axiom (section B.2).
Definition 2.3.2. A (lax) 1-morphism ( f , χ , ι ) : ( M , μ , e , a , λ , ρ ) → ( M ′ , μ ′ , e ′ , a ′ , λ ′ , ρ ′ ) (f,\chi,\iota):(M,\mu,e,a,\lambda,\rho)\to(M',\mu',e',a',\lambda',\rho') ( f , χ , ι ) : ( M , μ , e , a , λ , ρ ) → ( M ′ , μ ′ , e ′ , a ′ , λ ′ , ρ ′ ) is given by
a 1-cell f : M → M ′ f:M\to M' f : M → M ′ ,
two 2-cell χ \chi χ and ι \iota ι as depicted by the following diagram:
M ⊗ M → f ⊗ f M ′ ⊗ M ′ μ ↓ ↓ μ ′ M → f M ′ I → e ′ M ′ e ↓ ↑ f M → f M ′ \begin{CD}
M\otimes M @>{f\otimes f}>> M'\otimes M' \\
@V{\mu}VV @VV{\mu'}V \\
M @>>{f}> M'
\end{CD}
\qquad
\begin{CD}
I @>{e'}>> M' \\
@V{e}VV @AA{f}A \\
M @>>{f}> M'
\end{CD} M ⊗ M μ ↓ ⏐ M f ⊗ f f M ′ ⊗ M ′ ↓ ⏐ μ ′ M ′ I e ↓ ⏐ M e ′ f M ′ ⏐ ↑ f M ′ such that the two 2-cells satisfy the morphism hexagon (section B.4), morphism left square (section B.5), and morphism right square (section B.6).
Definition 2.3.3. A 2-morphism r : ( f , χ f , ι f ) ⇒ ( g , χ g , ι g ) r : (f, \chi_f, \iota_f) \Rightarrow(g, \chi_g, \iota_g) r : ( f , χ f , ι f ) ⇒ ( g , χ g , ι g ) is given by
a 2-cell r : f ⇒ g r : f \Rightarrow g r : f ⇒ g ,
such that the following two identities hold:
M ⊗ M → f ⊗ f M ′ ⊗ M ′ ↓ μ ↙ χ g ↓ μ ′ M → g M ′ ≡ M ⊗ M → f ⊗ f M ′ ⊗ M ′ ↓ μ ↙ χ f ↓ μ ′ M → f M ′ ↓ r g \begin{array}{ccccc}
M \otimes M & \xrightarrow{f\otimes f} & M' \otimes M' \\
\downarrow{\mu} & \underset{\chi_g}{\swarrow} & \downarrow{\mu'} \\
M & \xrightarrow{g} & M'
\end{array}
\qquad\equiv\qquad
\begin{array}{ccccc}
M \otimes M & \xrightarrow{f\otimes f} & M' \otimes M' \\
\downarrow{\mu} & \underset{\chi_f}{\swarrow} & \downarrow{\mu'} \\
M & \xrightarrow{f} & M' \\
& \underset{r}{\downarrow} & \\
& g &
\end{array} M ⊗ M ↓ μ M f ⊗ f χ g ↙ g M ′ ⊗ M ′ ↓ μ ′ M ′ ≡ M ⊗ M ↓ μ M f ⊗ f χ f ↙ f r ↓ g M ′ ⊗ M ′ ↓ μ ′ M ′ I ↓ e ↘ ι ′ M → f M ′ ↓ r g ≡ I ↓ e ↘ ι ′ M → g M ′ \begin{array}{ccccc}
I & & \\
\downarrow{e} & \searrow{\iota'} & \\
M & \xrightarrow{f} & M' \\
& \underset{r}{\downarrow} & \\
& g &
\end{array}
\qquad\equiv\qquad
\begin{array}{ccccc}
I & & \\
\downarrow{e} & \searrow{\iota'} & \\
M & \xrightarrow{g} & M'
\end{array} I ↓ e M ↘ ι ′ f r ↓ g M ′ ≡ I ↓ e M ↘ ι ′ g M ′
Lemma 2.3.4. The data of pseudomonoids and their morphisms in V \mathcal{V} V assemble into a bicategory of pseudomonoids called 1 M o n ( V ) 1\mathrm{Mon}(\mathcal{V}) 1 Mon ( V ) .
Proof. This lemma primarily follows from [7, Sec. 3] by considering one object version of iconic weak enrichment over a monoidal bicategory.
Example 2.3.5. As expected, if V \mathcal{V} V is the monoidal bicategory 1 C a t 1\mathrm{Cat} 1 Cat , then the pseudomonoid objects and their morphisms are usual monoidal categories and their lax monoidal functors.
Now, we shall describe braided pseudomonoid objects and their morphisms. For this, we require a braided monoidal bicategory. The details are ultimately irrelevant for us but can be found carefully constructed in [16]. In particular, V \mathcal{V} V is a braided monoidal bicategory if it is a monoidal bicategory equipped with a family of adjoint equivalences b X , Y : X ⊗ Y ↔ Y ⊗ X : b X , Y − 1 b_{X,Y} : X \otimes Y \leftrightarrow Y \otimes X : b^{-1}_{X,Y} b X , Y : X ⊗ Y ↔ Y ⊗ X : b X , Y − 1 that behaves as the braiding.
Definition 2.3.6. A braided pseudomonoid in a braided monoidal bicategory V \mathcal{V} V is a pseudomonoid object M M M together with the additional data of
a braiding 2-cell b b b
M ⊗ M → b M , M M ⊗ M μ ↓ ↓ μ M = M \begin{CD}
M \otimes M @>{b_{M,M}}>> M \otimes M \\
@V{\mu}VV @VV{\mu}V \\
M @= M
\end{CD} M ⊗ M μ ↓ ⏐ M b M , M M ⊗ M ↓ ⏐ μ M
such that the two hexagon axioms (section B.3) hold.
Definition 2.3.7.
A 1-morphism ( f , χ , i ) : ( M , b ) → ( M ′ , b ′ ) (f,\chi,i):(M,b)\to(M',b') ( f , χ , i ) : ( M , b ) → ( M ′ , b ′ ) of braided pseudomonoid is a 1-morphism of pseudomonoid together with the following additional axiom:
M ⊗ M → b M , M M ⊗ M μ ↓ f ⊗ f ↓ M → f M ′ M ⊗ M → f ⊗ f M ′ ⊗ M ′ μ ↓ ↓ μ ′ M → f M ′ \begin{CD}
M \otimes M @>{b_{M,M}}>> M \otimes M \\
@V{\mu}VV @V{f\otimes f}VV \\
M @>{f}>> M'
\end{CD}
\qquad
\begin{CD}
M \otimes M @>{f\otimes f}>> M' \otimes M' \\
@V{\mu}VV @VV{\mu'}V \\
M @>{f}>> M'
\end{CD} M ⊗ M μ ↓ ⏐ M b M , M f M ⊗ M f ⊗ f ↓ ⏐ M ′ M ⊗ M μ ↓ ⏐ M f ⊗ f f M ′ ⊗ M ′ ↓ ⏐ μ ′ M ′ M ⊗ M → b M , M M ⊗ M μ ↓ b ↓ M → μ M M ⊗ M → f ⊗ f M ′ ⊗ M ′ b ↓ ↓ μ ′ M ′ ⊗ M ′ → μ ′ M ′ \begin{CD}
M \otimes M @>{b_{M,M}}>> M \otimes M \\
@V{\mu}VV @V{b}VV \\
M @>{\mu}>> M
\end{CD}
\qquad
\begin{CD}
M \otimes M @>{f\otimes f}>> M' \otimes M' \\
@V{b}VV @VV{\mu'}V \\
M' \otimes M' @>{\mu'}>> M'
\end{CD} M ⊗ M μ ↓ ⏐ M b M , M μ M ⊗ M b ↓ ⏐ M M ⊗ M b ↓ ⏐ M ′ ⊗ M ′ f ⊗ f μ ′ M ′ ⊗ M ′ ↓ ⏐ μ ′ M ′ A 2-morphism of braided pseudomonoid is the same data as that of a pseudomonoid.
Lemma 2.3.8. There is a bicategory of braided pseudomonoids 1 b M o n ( V ) 1\mathrm{bMon}(\mathcal{V}) 1 bMon ( V ) .
Proof. Using that the 2-morphism of braided pseudomonoids are the same as that of pseudomonoid, we define the compositional data and the coherence constraint of the bicategory 1 b M o n ( V ) 1\mathrm{bMon}(\mathcal{V}) 1 bMon ( V ) to be the same as that of 1 M o n ( V ) 1\mathrm{Mon}(\mathcal{V}) 1 Mon ( V ) . Thus, we only need to verify that the composite of 1-morphism of braided pseudomonoid also satisfy the additional braiding axiom. Suppressing notations and labeling, this follows directly from the following pasting equalities:
M M → P P ↑ ↑ M M → P P ↓ ↓ M → N → P M M → N N → P P ↑ ↑ ↑ M → N → P ≡ M M → N N → P P ↑ ↑ ↑ M M → N N → P P ↓ ↓ ↓ M → N → P M M → N N → P P ↑ ↑ ↑ M M → N N → P P ↓ ↓ ↓ M → N → P \begin{array}{c}
\begin{CD}
MM @>>> PP\\
@AAA @AAA\\
MM @>>> PP\\
@VVV @VVV\\
M @>>> N @>>> P
\end{CD}
\qquad
\begin{CD}
MM @>>> NN @>>> PP\\
@AAA @AAA @AAA\\
M @>>> N @>>> P
\end{CD}\\[1em]
\equiv\\[1em]
\begin{CD}
MM @>>> NN @>>> PP\\
@AAA @AAA @AAA\\
MM @>>> NN @>>> PP\\
@VVV @VVV @VVV\\
M @>>> N @>>> P
\end{CD}
\qquad
\begin{CD}
MM @>>> NN @>>> PP\\
@AAA @AAA @AAA\\
MM @>>> NN @>>> PP\\
@VVV @VVV @VVV\\
M @>>> N @>>> P
\end{CD}
\end{array} M M ⏐ ↑ M M ↓ ⏐ M P P ⏐ ↑ P P ↓ ⏐ N P M M ⏐ ↑ M N N ⏐ ↑ N P P ⏐ ↑ P ≡ M M ⏐ ↑ M M ↓ ⏐ M N N ⏐ ↑ N N ↓ ⏐ N P P ⏐ ↑ P P ↓ ⏐ P M M ⏐ ↑ M M ↓ ⏐ M N N ⏐ ↑ N N ↓ ⏐ N P P ⏐ ↑ P P ↓ ⏐ P
Finally, we can also describe symmetric pseudomonoid object inside a sylleptic (or symmetric) monoidal bicategory. This bicategory is properly defined in [19] and is braided monoidal bicategory together with a 2-cell υ X , Y : b Y , X b X , Y ⇒ 1 \upsilon_{X,Y}: b_{Y,X}b_{X,Y} \Rightarrow1 υ X , Y : b Y , X b X , Y ⇒ 1 that essentially identifies b X , Y b_{X,Y} b X , Y with b Y , X − 1 b^{-1}_{Y,X} b Y , X − 1 .
Definition 2.3.9.
A symmetric pseudomonoid is a braided pseudomonoid that satisfy the following additional axiom:
M ⊗ M → ϵ M ⊗ M M ⊗ M μ ↓ ↓ μ M = M ≡ M ⊗ M → ϵ M ⊗ M M ⊗ M μ ↓ ↓ μ M = M \begin{CD}
M \otimes M @>{\epsilon_{M\otimes M}}>> M \otimes M\\
@V{\mu}VV @VV{\mu}V\\
M @= M
\end{CD}
\qquad\equiv\qquad
\begin{CD}
M \otimes M @>{\epsilon_{M\otimes M}}>> M \otimes M\\
@V{\mu}VV @VV{\mu}V\\
M @= M
\end{CD} M ⊗ M μ ↓ ⏐ M ϵ M ⊗ M M ⊗ M ↓ ⏐ μ M ≡ M ⊗ M μ ↓ ⏐ M ϵ M ⊗ M M ⊗ M ↓ ⏐ μ M The 1-morphisms and 2-morphisms of symmetric pseudomonoid object has no additional data.
Lemma 2.3.10. There is a bicategory of symmetric pseudomonoids 1 s M o n ( V ) \mathrm{1sMon}(\mathcal{V}) 1sMon ( V ) , which is formed as a full sub-bicategory of braided pseudomonoids.
For the case of V = 1 C a t \mathcal{V} = \mathrm{1Cat} V = 1Cat , we know that 1 C a t \mathrm{1Cat} 1Cat is a symmetric monoidal 2-category. Thus, as expected, 1 b M o n ( 1 C a t ) = B r M o n C a t l \mathrm{1bMon}(\mathrm{1Cat}) = \mathrm{BrMonCat}_{\mathrm{l}} 1bMon ( 1Cat ) = BrMonCat l and 1 s M o n ( 1 C a t ) = S y m M o n C a t l \mathrm{1sMon}(\mathrm{1Cat}) = \mathrm{SymMonCat}_{\mathrm{l}} 1sMon ( 1Cat ) = SymMonCat l . Similarly, it is possible to consider colax 1-morphisms in the constructions of 1 M o n \mathrm{1Mon} 1Mon and its variants. In this case, we obtain the corresponding 2-category of monoidal categories with colax monoidal functors.
With all the relevant definitions, we can now define pseudomodules over pseudomonoid objects.
Definition 2.3.11. A left pseudomodule object over a pseudomonoid object ( M , μ , e , a M , λ M , ρ M ) (M,\mu,e,a_M,\lambda_M,\rho_M) ( M , μ , e , a M , λ M , ρ M ) in a monoidal bicategory V \mathcal{V} V is given by the collection ( Z , ◃ , a Z , λ Z ) (\mathcal{Z},\triangleleft,a_{\mathcal{Z}},\lambda_{\mathcal{Z}}) ( Z , ◃ , a Z , λ Z ) such that:
Z ∈ V \mathcal{Z} \in\mathcal{V} Z ∈ V is an object, ◃ \triangleleft ◃ defines a left action 1-cell M ⊗ Z → ◃ Z M \otimes\mathcal{Z} \xrightarrow{\triangleleft} \mathcal{Z} M ⊗ Z ◃ Z in V \mathcal{V} V ,
the associator a Z a_{\mathcal{Z}} a Z and left unitor λ Z \lambda_{\mathcal{Z}} λ Z are invertible 2-cells in V \mathcal{V} V which are defined using the following diagrams:
( M ⊗ M ) ⊗ Z → μ ⊗ 1 M ⊗ Z → ◃ Z ↑ a V ↓ ◃ M ⊗ ( M ⊗ Z ) → 1 ⊗ ◃ M ⊗ Z ↗ a Z \begin{aligned}
(M \otimes M) \otimes\mathcal{Z} &\xrightarrow{\mu\otimes1} M \otimes\mathcal{Z} \xrightarrow{\triangleleft} \mathcal{Z} \\
\mathord{\uparrow}^{a_{\mathcal{V}}} &\qquad\qquad\qquad\mathord{\downarrow}^{\triangleleft} \\
M \otimes(M \otimes\mathcal{Z}) &\xrightarrow{1\otimes\triangleleft} M \otimes\mathcal{Z} \\
&\mathord{\nearrow}_{a_{\mathcal{Z}}}
\end{aligned} ( M ⊗ M ) ⊗ Z ↑ a V M ⊗ ( M ⊗ Z ) μ ⊗ 1 M ⊗ Z ◃ Z ↓ ◃ 1 ⊗◃ M ⊗ Z ↗ a Z I ⊗ Z → λ V Z ↑ e ⊗ 1 ↗ λ Z M ⊗ Z → ◃ Z \begin{aligned}
I \otimes\mathcal{Z} &\xrightarrow{\lambda_{\mathcal{V}}} \mathcal{Z} \\
\mathord{\uparrow}^{e\otimes1} &\mathord{\nearrow}_{\lambda_{\mathcal{Z}}} \\
M \otimes\mathcal{Z} &\xrightarrow{\triangleleft} \mathcal{Z}
\end{aligned} I ⊗ Z ↑ e ⊗ 1 M ⊗ Z λ V Z ↗ λ Z ◃ Z the pentagon axiom (section B.1) and the triangle axiom (section B.2) hold with replacing the corresponding copy of M M M with Z \mathcal{Z} Z .
Definition 2.3.12. A right pseudomodule object over a pseudomonoid object ( M , μ , e , a M , λ M , ρ M ) (M,\mu,e,a_M,\lambda_M,\rho_M) ( M , μ , e , a M , λ M , ρ M ) in a monoidal bicategory V \mathcal{V} V is given by the collection ( Z , ▹ , a Z , ρ Z ) (\mathcal{Z},\triangleright,a_{\mathcal{Z}},\rho_{\mathcal{Z}}) ( Z , ▹ , a Z , ρ Z ) such that:
Z ∈ V \mathcal{Z} \in\mathcal{V} Z ∈ V is an object, ▹ \triangleright ▹ defines a right action 1-cell Z ⊗ M → ▹ Z \mathcal{Z} \otimes M \xrightarrow{\triangleright} \mathcal{Z} Z ⊗ M ▹ Z in V \mathcal{V} V ,
the associator a Z a_{\mathcal{Z}} a Z and right unitor ρ Z \rho_{\mathcal{Z}} ρ Z are invertible 2-morphisms in V \mathcal{V} V which are defined using the following diagrams:
( Z ⊗ M ) ⊗ M → ▹ ⊗ 1 Z ⊗ M → ▹ Z ↓ a V ↓ ▹ Z ⊗ ( M ⊗ M ) → 1 ⊗ μ Z ⊗ M ↙ a Z \begin{aligned}
(\mathcal{Z} \otimes M) \otimes M &\xrightarrow{\triangleright\otimes1} \mathcal{Z} \otimes M \xrightarrow{\triangleright} \mathcal{Z} \\
\mathord{\downarrow}^{a_{\mathcal{V}}} &\qquad\qquad\qquad\mathord{\downarrow}^{\triangleright} \\
\mathcal{Z} \otimes(M \otimes M) &\xrightarrow{1\otimes\mu} \mathcal{Z} \otimes M \\
&\mathord{\swarrow}_{a_{\mathcal{Z}}}
\end{aligned} ( Z ⊗ M ) ⊗ M ↓ a V Z ⊗ ( M ⊗ M ) ▹⊗ 1 Z ⊗ M ▹ Z ↓ ▹ 1 ⊗ μ Z ⊗ M ↙ a Z Z ⊗ I → ρ V Z ↑ 1 ⊗ e ↗ ρ Z Z ⊗ M → ▹ Z \begin{aligned}
\mathcal{Z} \otimes I &\xrightarrow{\rho_{\mathcal{V}}} \mathcal{Z} \\
\mathord{\uparrow}^{1\otimes e} &\mathord{\nearrow}_{\rho_{\mathcal{Z}}} \\
\mathcal{Z} \otimes M &\xrightarrow{\triangleright} \mathcal{Z}
\end{aligned} Z ⊗ I ↑ 1 ⊗ e Z ⊗ M ρ V Z ↗ ρ Z ▹ Z
the pentagon axiom (section B.1) and the triangle axiom (section B.2) hold with replacing the corresponding copy of M M M with Z \mathcal{Z} Z .
Definition 2.3.13. A bipseudomodule Z ∈ V \mathcal{Z} \in\mathcal{V} Z ∈ V between two pseudomonoids ( M , μ M , e M , a M , λ M , ρ M ) (M,\mu_M,e_M,a_M,\lambda_M,\rho_M) ( M , μ M , e M , a M , λ M , ρ M ) and ( N , μ N , e N , a N , λ N , ρ N ) (N,\mu_N,e_N,a_N,\lambda_N,\rho_N) ( N , μ N , e N , a N , λ N , ρ N ) is given by
a left pseudomodule ( Z , ◃ , a ( Z , M ) , λ Z ) (\mathcal{Z},\triangleleft,a_{(\mathcal{Z},M)},\lambda_{\mathcal{Z}}) ( Z , ◃ , a ( Z , M ) , λ Z ) over M M M ,
a right pseudomodule ( Z , ▹ , a ( Z , N ) , ρ Z ) (\mathcal{Z},\triangleright,a_{(\mathcal{Z},N)},\rho_{\mathcal{Z}}) ( Z , ▹ , a ( Z , N ) , ρ Z ) over N N N ,
an invertible balancing 2-cell l l l in V \mathcal{V} V which is defined using the following diagram:
( M ⊗ Z ) ⊗ N → a V M ⊗ ( Z ⊗ N ) a Z ⊗ 1 ↓ ↓ 1 ⊗ a Z Z ⊗ N → l M ⊗ Z ▹ Z ↓ ↓ ◃ Z Z = Z \begin{CD}
(M \otimes\mathcal{Z}) \otimes N @>{a_{\mathcal{V}}}>> M \otimes(\mathcal{Z} \otimes N) \\
@V{a_{\mathcal{Z}}\otimes1}VV @VV{1\otimes a_{\mathcal{Z}}}V \\
\mathcal{Z}\otimes N @>{l}>> M\otimes\mathcal{Z} \\
@V{\triangleright_{\mathcal{Z}}}VV @VV{\triangleleft_{\mathcal{Z}}}V \\
\mathcal{Z} @= \mathcal{Z}
\end{CD} ( M ⊗ Z ) ⊗ N a Z ⊗ 1 ↓ ⏐ Z ⊗ N ▹ Z ↓ ⏐ Z a V l M ⊗ ( Z ⊗ N ) ↓ ⏐ 1 ⊗ a Z M ⊗ Z ↓ ⏐ ◃ Z Z two instances of the pentagon axiom (section B.1) hold, one for ( M , Z , N , N ) (M,\mathcal{Z},N,N) ( M , Z , N , N ) and the other for ( M , M , Z , N ) (M,M,\mathcal{Z},N) ( M , M , Z , N ) .
All the axioms for left, right, and bi-pseudomodule are motivated by viewing monoidal categories and module categories as parts of a two objects bicategory K K K . Furthermore, in the case of V = 1 C a t \mathcal{V} = 1\mathrm{Cat} V = 1 Cat , these provide the usual definitions.
Generalized involutions and inverses
We aim to characterize involutions and inverses in a more general setting using the microcosm principle. In a certain sense, the data of a single mutually invertible pair of arrows or a single involution lies within the scope of this principle. This allows one to define relative inverses and involutions which are quite essential in describing dualities of monoidal structures. The first part primarily focus on structures internal to bicategories, whereas the latter part shall explore structures internal to tricategories.
TEXT
We begin by briefly reviewing 0-dimensional involutions for motivation. As this theory is well known, we omit most of the details. The interested reader should refer to [34 ] .
Definition 3.1.1.
A strict involutive object in C C C is given by a morphism f : x → x f:x \to x f : x → x such that f ∘ f = 1 f \circ f = 1 f ∘ f = 1 .
A morphism of involutions ( x , f ) → ( y , g ) (x,f)\to(y,g) ( x , f ) → ( y , g ) is an arrow r : x → y r:x\to y r : x → y such that g r = r f gr=rf g r = r f holds.
Involutive objects and their morphisms define a category of involutions 0 I n v ( C ) \mathbf{0Inv}(\mathcal{C}) 0Inv ( C ) . This construction extends to a 2-functor 0 I n v : 1 C a t → 1 C a t \mathbf{0Inv}:\mathbf{1Cat}\to\mathbf{1Cat} 0Inv : 1Cat → 1Cat , which comes equipped with two canonical natural 2-transformation, U : 0 I n v ⇒ 1 C a t U:\mathbf{0Inv}\Rightarrow\mathbf{1Cat} U : 0Inv ⇒ 1Cat given by the forgetful action, and its section i : 1 C a t ⇒ 0 I n v i:\mathbf{1Cat}\Rightarrow\mathbf{0Inv} i : 1Cat ⇒ 0Inv given by inclusion as identity involution.
Furthermore, the category 0 I n v ( 0 I n v ( C ) ) \mathbf{0Inv}(\mathbf{0Inv}(\mathcal{C})) 0Inv ( 0Inv ( C )) has as objects a pair of involutions f , g : x → x f,g:x\to x f , g : x → x such that f g = g f fg=gf f g = g f . This implies that we have a natural 2-transformation 0 I n v 2 ⇒ 0 I n v \mathbf{0Inv}^{2}\Rightarrow\mathbf{0Inv} 0Inv 2 ⇒ 0Inv given by ( f , g ) ↦ ( f g ) (f,g)\mapsto(fg) ( f , g ) ↦ ( f g ) . In particular, 0 I n v \mathbf{0Inv} 0Inv is a strict 2-Monad on 1 C a t \mathbf{1Cat} 1Cat .
Example 3.1.2. (op n , k \operatorname{op}_{n,k} op n , k duality of n C a t \mathbf{nCat} nCat ). The 1-functors op n , k : n C a t ≤ 1 → n C a t ≤ 1 \operatorname{op}_{n,k}:\mathbf{nCat}_{\leq1}\to\mathbf{nCat}_{\leq1} op n , k : nCat ≤ 1 → nCat ≤ 1 satisfy the desired axiom op n , k 2 = 1 n C a t \operatorname{op}_{n,k}^{2}=1_{\mathbf{nCat}} op n , k 2 = 1 nCat . This is a strict 1-involutive object.
The phrase strict involutive object already suggests the general definition of a weak (1-dimensional) involutive object. In particular, we can define weak (or pseudo-) involutive objects inside an arbitrary bicategory K K K . All the corresponding data and axioms shall be written using pasting diagrams, which are to be interpreted via the evaluation functor e : st K → K e:\operatorname{st}K\to K e : st K → K . See section A for more details.
Definition 3.1.3. (pseudo involutive object). Let K K K be a bicategory. We define a weak/pseudo involutive object by the data of
a 1-cell f : x → x ∈ K f:x\to x\in K f : x → x ∈ K ,
an invertible 2-cell
x → f x f ↓ ↓ f x → ε x x \begin{CD}
x @>{f}>> x \\
@V{f}VV @VV{f}V \\
x @>{\varepsilon_x}>> x
\end{CD} x f ↓ ⏐ x f ε x x ↓ ⏐ f x
such that the 2-cell α \alpha α satisfy an additional axiom as depicted by the following diagram:
x → ε x x f ↓ ↓ α x → ε x x ≡ x → ε x x f ↓ ↓ f x → ε x x \begin{CD}
x @>{\varepsilon_x}>> x \\
@V{f}VV @VV{\alpha}V \\
x @>{\varepsilon_x}>> x
\end{CD}
\qquad\equiv\qquad
\begin{CD}
x @>{\varepsilon_x}>> x \\
@V{f}VV @VV{f}V \\
x @>{\varepsilon_x}>> x
\end{CD} x f ↓ ⏐ x ε x ε x x ↓ ⏐ α x ≡ x f ↓ ⏐ x ε x ε x x ↓ ⏐ f x
Example 3.1.4. Every strict involutive object (where f 2 = 1 f^{2}=1 f 2 = 1 ) in a strict 2-category is trivially a weak involutive object.
One might wonder why do we require the additional identities in the specification of a weak involutive object. The key point is that the structure is incoherent without the additional axiom, analogous to how specifying a monoidal category without its pentagon or triangle identity is meaningless. This extra identity ensures that all reasonable ways of going from f n ⇒ f m f^{n}\Rightarrow f^{m} f n ⇒ f m are equivalent. In fact, there are four equivalent way of expressing this additional axiom. The proof is relatively simple and uses the pasting of the 2-cells α \alpha α , α − 1 \alpha^{-1} α − 1 , or the unitors appropriately.
Lemma 3.1.5. The data of an involution in a bicategory is equivalently expressed as a 1-cell f : x → x f:x\to x f : x → x and an invertible 2-cell α : f 2 ⇒ 1 \alpha:f^{2}\Rightarrow1 α : f 2 ⇒ 1 satisfying any one of the following axioms:
1.
x → ε x x f ↓ ↓ f x → ε x x x → ε x x f ↓ ↓ f x → ε x x ≡ x → ε x x f ↓ ↓ f x → ε x x \begin{CD}
x @>{\varepsilon_x}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD}
\quad
\begin{CD}
x @>{\varepsilon_x}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD}
\quad\equiv\quad
\begin{CD}
x @>{\varepsilon_x}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD} x f ↓ ⏐ x ε x ε x x ↓ ⏐ f x x f ↓ ⏐ x ε x ε x x ↓ ⏐ f x ≡ x f ↓ ⏐ x ε x ε x x ↓ ⏐ f x 2.
x → f x f ↓ ↓ f x → f x x → f x f ↓ ↓ f x → f x ≡ x → f x f ↓ ↓ f x → f x \begin{CD}
x @>{f}>> x\\
@V{f}VV @VV{f}V\\
x @>{f}>> x
\end{CD}
\quad
\begin{CD}
x @>{f}>> x\\
@V{f}VV @VV{f}V\\
x @>{f}>> x
\end{CD}
\quad\equiv\quad
\begin{CD}
x @>{f}>> x\\
@V{f}VV @VV{f}V\\
x @>{f}>> x
\end{CD} x f ↓ ⏐ x f f x ↓ ⏐ f x x f ↓ ⏐ x f f x ↓ ⏐ f x ≡ x f ↓ ⏐ x f f x ↓ ⏐ f x 3.
x → ε x x f ↓ ↓ f x → ε x x x → ε x x f ↓ ↓ f x → ε x x ≡ x → ε x x f ↓ ↓ f x → ε x x \begin{CD}
x @>{\varepsilon_x}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD}
\quad
\begin{CD}
x @>{\varepsilon_x}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD}
\quad\equiv\quad
\begin{CD}
x @>{\varepsilon_x}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD} x f ↓ ⏐ x ε x ε x x ↓ ⏐ f x x f ↓ ⏐ x ε x ε x x ↓ ⏐ f x ≡ x f ↓ ⏐ x ε x ε x x ↓ ⏐ f x 4.
x → f x f ↓ ↓ f x → ε x x x → f x f ↓ ↓ f x → ε x x ≡ x → f x f ↓ ↓ f x → ε x x \begin{CD}
x @>{f}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD}
\quad
\begin{CD}
x @>{f}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD}
\quad\equiv\quad
\begin{CD}
x @>{f}>> x\\
@V{f}VV @VV{f}V\\
x @>{\varepsilon_x}>> x
\end{CD} x f ↓ ⏐ x f ε x x ↓ ⏐ f x x f ↓ ⏐ x f ε x x ↓ ⏐ f x ≡ x f ↓ ⏐ x f ε x x ↓ ⏐ f x
The 2nd and the 4th form of the additional axiom are particularly interesting as they allow one to define a lax/colax variant for involutive object in a (lax/colax) bicategory. Thus, more generally we can work with lax involutive objects. It is now instructive to define 1-morphisms and 2-morphisms of these involutions explicitly.
Definition 3.1.6. Let ( x , f , α ) (x,f,\alpha) ( x , f , α ) and ( y , g , β ) (y,g,\beta) ( y , g , β ) be two (lax) involutive objects in a bicategory K K K . A (lax) 1-morphism of involution ( r , u ) : ( x , f , α ) → ( y , g , β ) (r,u):(x,f,\alpha)\to(y,g,\beta) ( r , u ) : ( x , f , α ) → ( y , g , β ) is
a 1-cell r : x → y ∈ K r:x\to y\in K r : x → y ∈ K
a 2-cell
x → r y f ↓ ↓ g x → r y \begin{CD}
x @>{r}>> y\\
@V{f}VV @VV{g}V\\
x @>{r}>> y
\end{CD} x f ↓ ⏐ x r r y ↓ ⏐ g y
such that the following axiom holds:
x → r y f ↓ ↓ g x → r y ≡ x → r y ε x ↓ ↓ ε y x → r y \begin{CD}
x @>{r}>> y\\
@V{f}VV @VV{g}V\\
x @>{r}>> y
\end{CD}
\quad\equiv\quad
\begin{CD}
x @>{r}>> y\\
@V{\varepsilon_x}VV @VV{\varepsilon_y}V\\
x @>{r}>> y
\end{CD} x f ↓ ⏐ x r r y ↓ ⏐ g y ≡ x ε x ↓ ⏐ x r r y ↓ ⏐ ε y y
Definition 3.1.7. Let ( r , u ) (r,u) ( r , u ) and ( t , v ) (t,v) ( t , v ) be two 1-morphisms of involution between ( x , f , α ) (x,f,\alpha) ( x , f , α ) and ( y , g , β ) (y,g,\beta) ( y , g , β ) . A 2-morphism γ : ( r , u ) ⇒ ( t , v ) \gamma:(r,u)\Rightarrow(t,v) γ : ( r , u ) ⇒ ( t , v ) is
a 2-cell r ⇒ t ∈ K r\Rightarrow t\in K r ⇒ t ∈ K ,
which satisfy the following axiom
x → r y f ↓ ↓ g x → r y ≡ x → t y f ↓ ↓ g x → t y \begin{CD}
x @>{r}>> y\\
@V{f}VV @VV{g}V\\
x @>{r}>> y
\end{CD}
\quad\equiv\quad
\begin{CD}
x @>{t}>> y\\
@V{f}VV @VV{g}V\\
x @>{t}>> y
\end{CD} x f ↓ ⏐ x r r y ↓ ⏐ g y ≡ x f ↓ ⏐ x t t y ↓ ⏐ g y
As usual, all of this data assembles into a bicategory of involutions 1 I n v ( K ) \mathbf{1Inv}(K) 1Inv ( K ) associated to any bicategory K K K . This follows by simply defining all the relevant compositional data strictly in s t K \mathop{\mathrm{st}}K st K and then descending those diagrams onto K K K using the evaluation functor e : s t K → K e:\mathop{\mathrm{st}}K\to K e : st K → K appropriately. We do note that the identity 1-morphisms in 1 I n v ( K ) \mathbf{1Inv}(K) 1Inv ( K ) are given by the 1-cells 1 x : x → x 1_x:x\to x 1 x : x → x , which are not the empty strings.
Lemma 3.1.8. There is a bicategory of involutions 1 I n v ( K ) \mathbf{1Inv}(K) 1Inv ( K ) corresponding to each bicategory K K K .
Remark 3.1.9. With a lengthy computation and using strifications appropriately, one could show that 1 I n v \mathbf{1Inv} 1Inv is a (strict) trifunctor 1 I n v : B i c a t → B i c a t \mathbf{1Inv}: \mathbf{Bicat} \to\mathbf{Bicat} 1Inv : Bicat → Bicat on the tricategory of Bicat. One could also show that there is a (strict) tritransformation U : 1 I n v ⇒ 1 B i c a t U: \mathbf{1Inv} \Rightarrow\mathbf{1Bicat} U : 1Inv ⇒ 1Bicat and a (strict) tritransformation i : 1 B i c a t ⇒ 1 I n v i: \mathbf{1Bicat} \Rightarrow\mathbf{1Inv} i : 1Bicat ⇒ 1Inv . However, we shall not go into the details here.
We have now provided enough discussion about the theory of involutions and want to motivate a more general definition of an involutive object. Consider again the 2-functor op 1 : op 2 ( 1 C a t ) → 1 C a t \operatorname{op}_{1}: \operatorname{op}_{2}(\mathbf{1Cat}) \to\mathbf{1Cat} op 1 : op 2 ( 1Cat ) → 1Cat . Here, the lower dimensional involution has its domain altered by a higher dimensional involution. This suggests the definition of a relative involutive object. A similar account of these can also be found in [34 ] . Hence, we shall be brief in describing the known details and provide a proof only for the missing slice construction.
Definition 3.1.10. Let ( C , T , α ) (\mathcal{C},T,\alpha) ( C , T , α ) be a pseudo (or lax) involutive object in 1 C a t \mathbf{1Cat} 1Cat . We define an involutive object relative to T T T as a morphism f : T x → x ∈ C f:Tx \to x \in\mathcal{C} f : T x → x ∈ C such that the following diagram commutes:
T 2 x → T f T x α x ↓ ↓ f x → 1 x \begin{CD}
T^{2}x @>{Tf}>> Tx \\
@V{\alpha_x}VV @VV{f}V \\
x @>{1}>> x
\end{CD} T 2 x α x ↓ ⏐ x T f 1 T x ↓ ⏐ f x
Remark 3.1.11. To relate with the previous definition, we define the adjunct morphism by f − 1 : x → α x − 1 T 2 x → T f T x f^{-1}:x \xrightarrow{\alpha_x^{-1}} T^{2}x \xrightarrow{T f} Tx f − 1 : x α x − 1 T 2 x T f T x . From this, it follows that both f − 1 ∘ f = 1 f^{-1}\circ f=1 f − 1 ∘ f = 1 and 1 = f ∘ f − 1 1=f\circ f^{-1} 1 = f ∘ f − 1 are satisfied. Furthermore, if one of the equation is satisfied then the other one is automatically satisfied due to the additional axiom of T T T . However, the invertibility of α \alpha α is essential for this, as f − 1 f^{-1} f − 1 is not defined without α − 1 \alpha^{-1} α − 1 .
Lemma 3.1.12. Let ( C , T , α ) (\mathcal{C},T,\alpha) ( C , T , α ) be a lax involutive object in 1 C a t \mathbf{1Cat} 1Cat and ( x , f : T x → x ) (x,f:Tx\to x) ( x , f : T x → x ) be an involutive object relative to T T T . Then C / x \mathcal{C}/x C / x admits a canonical lax involutive structure using:
C / x → T / f C / x ( α l ) c : T 2 c → α c c ↓ ↓ T 2 x → α x x \begin{CD}
\mathcal{C}/x @>{T/f}>> \mathcal{C}/x
\end{CD}
\qquad
(\alpha_l)_c:
\begin{CD}
T^{2}c @>{\alpha_c}>> c \\
@VVV @VVV \\
T^{2}x @>{\alpha_x}>> x
\end{CD} C / x T / f C / x ( α l ) c : T 2 c ↓ ⏐ T 2 x α c α x c ↓ ⏐ x
Proof. The following identity of diagrams implies the result immediately:
T 3 C → T α c T C T 3 C → α T c T C ↓ ↓ ≡ ↓ T 3 x T x T 3 x → α T x T x ↓ T 2 f ↘ T α x ↓ f ↓ T 2 f T 2 x ↘ α x x T 2 x ↘ α x x ↓ T f ↓ T f T x → f x T x → f x \begin{array}{ccccc}
T^{3}C & \xrightarrow{T\alpha_{c}} & TC & & T^{3}C \xrightarrow{\alpha_{Tc}} TC \\
\downarrow& & \downarrow& \equiv& \downarrow\\
T^{3}x & & Tx & & T^{3}x \xrightarrow{\alpha_{Tx}} Tx \\
\downarrow^{T^{2}f} & \searrow^{T\alpha_{x}} & \downarrow^{f} & & \downarrow^{T^{2}f} \\
T^{2}x & \searrow^{\alpha_{x}} & x & & T^{2}x \searrow^{\alpha_{x}} x \\
\downarrow^{Tf} & & & & \downarrow^{Tf} \\
Tx & \xrightarrow{f} & x & & Tx \xrightarrow{f} x
\end{array} T 3 C ↓ T 3 x ↓ T 2 f T 2 x ↓ T f T x T α c ↘ T α x ↘ α x f T C ↓ T x ↓ f x x ≡ T 3 C α T c T C ↓ T 3 x α T x T x ↓ T 2 f T 2 x ↘ α x x ↓ T f T x f x
As before, there is also a 1-category 0 I n v ( C , T , α ) \mathbf{0Inv}(C,T,\alpha) 0Inv ( C , T , α ) , whose objects are relative involutive objects and whose morphisms are maps r : ( x , f ) → ( g , y ) r:(x,f)\to(g,y) r : ( x , f ) → ( g , y ) such that g ∘ T r = r ∘ f g\circ Tr=r\circ f g ∘ T r = r ∘ f . We can also describe this modified 0 I n v \mathbf{0Inv} 0Inv as a 2-functor that extends the original functor.
Lemma 3.1.13.
There exists a 2-functor 0 I n v : 1 I n v ( 1 C a t ) → 1 C a t \mathbf{0Inv}: \mathbf{1Inv}(\mathbf{1Cat})\to\mathbf{1Cat} 0Inv : 1Inv ( 1Cat ) → 1Cat extending the construction of involutive objects along the identity inclusion tritransformation component i 1 C a t : 1 C a t ↪ 1 I n v ( 1 C a t ) i_{\mathbf{1Cat}}:\mathbf{1Cat}\hookrightarrow\mathbf{1Inv}(\mathbf{1Cat}) i 1Cat : 1Cat ↪ 1Inv ( 1Cat ) .
This further extends to the 2-functor 0 I n v : 1 I n v ( 1 C a t ) → 1 I n v ( 1 C a t ) \mathbf{0Inv}: \mathbf{1Inv}(\mathbf{1Cat})\to\mathbf{1Inv}(\mathbf{1Cat}) 0Inv : 1Inv ( 1Cat ) → 1Inv ( 1Cat ) along the forgetful tritransformation component U 1 C a t : 1 I n v ( 1 C a t ) → 1 C a t U_{\mathbf{1Cat}}:\mathbf{1Inv}(\mathbf{1Cat})\to\mathbf{1Cat} U 1Cat : 1Inv ( 1Cat ) → 1Cat .
The natural 2-transformation U U U itself extends to U : 0 I n v ⇒ 1 1 I n v ( 1 C a t ) U:\mathbf{0Inv}\Rightarrow\mathbf{1}_{\mathbf{1Inv}(\mathbf{1Cat})} U : 0Inv ⇒ 1 1Inv ( 1Cat ) . However, it doesn’t admit any section as before. In particular, this 0 I n v \mathbf{0Inv} 0Inv is not a Monad.
Proof. The proof of (1), (2) and (3) mainly follows from a slight modification of the arguments found in [34 Props. 2.6.2, 2.6.3, and 2.6.5] .
We now briefly describe the analogous definitions for inverse structures. Our goal is to provide a better perspective on these definitions that aligns with the microcosm principle and coherence principle.
Definition 3.1.14 (pseudo inverse objects). Let K K K be a bicategory. We define a tuple ( f : x → y , g : y → x , η , ϵ ) (f:x\to y,g:y\to x,\eta,\epsilon) ( f : x → y , g : y → x , η , ϵ ) to be an inverse object whenever it is an internal adjoint equivalence. That is, the
following two triangle identities hold:
x → ε x x f ↓ ↓ ε y → ε y y x → g x f ↓ ↓ f y → ε y y ≡ x → ε x x f ↓ ↓ f y → y \begin{CD}
x @>{\varepsilon_x}>> x \\
@V{f}VV @VV{\varepsilon}V \\
y @>{\varepsilon_y}>> y
\end{CD}
\qquad
\begin{CD}
x @>{g}>> x \\
@V{f}VV @VV{f}V \\
y @>{\varepsilon_y}>> y
\end{CD}
\qquad\equiv\qquad
\begin{CD}
x @>{\varepsilon_x}>> x \\
@V{f}VV @VV{f}V \\
y @>{}>> y
\end{CD} x f ↓ ⏐ y ε x ε y x ↓ ⏐ ε y x f ↓ ⏐ y g ε y x ↓ ⏐ f y ≡ x f ↓ ⏐ y ε x x ↓ ⏐ f y
y → g x ε y ↓ ↓ ε y → f y y → g x g ↓ ↓ g y → ε y y ≡ y → g x g ↓ ↓ g y → x \begin{CD}
y @>{g}>> x \\
@V{\varepsilon_y}VV @VV{\varepsilon}V \\
y @>{f}>> y
\end{CD}
\qquad
\begin{CD}
y @>{g}>> x \\
@V{g}VV @VV{g}V \\
y @>{\varepsilon_y}>> y
\end{CD}
\qquad\equiv\qquad
\begin{CD}
y @>{g}>> x \\
@V{g}VV @VV{g}V \\
y @>{}>> x
\end{CD} y ε y ↓ ⏐ y g f x ↓ ⏐ ε y y g ↓ ⏐ y g ε y x ↓ ⏐ g y ≡ y g ↓ ⏐ y g x ↓ ⏐ g x
Example 3.1.15 (pseudo inverse objects in 1Cat). A pair of functors F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D and G : D → C G : \mathcal{D} \to\mathcal{C} G : D → C specify a weak inverse object whenever they are part of an adjoint equivalence of categories.
We strongly believe that this shift in perspective of viewing an adjoint equivalence as a coherent weak inverse is important and not merely pedantic. Indeed, assume that data of an adjoint equivalence of categories were genuinely different from the data of an equivalence of categories. In this hypothetical situation, working with equivalence of categories would have been no better than working with monoidal categories without the pentagon or triangle axioms.
With this idea, we can now work with general adjunctions instead of adjoint equivalences. One could view the data η : 1 ⇒ G F \eta: 1 \Rightarrow GF η : 1 ⇒ GF as a laxly categorified version of f f f being right inverse of g g g (or g f = 1 gf = 1 g f = 1 ), and the similar goes for ϵ : F G ⇒ 1 \epsilon: FG \Rightarrow1 ϵ : F G ⇒ 1 as laxly categorified left inverse f f f . Using this viewpoint and convention, the data of an adjunction F ⊣ G F \dashv G F ⊣ G simultaneously expresses that F F F is a two-sided lax inverse of G G G and that G G G is a two-sided colax inverse of F F F .
There is also an analogous notion of 1-morphisms and 2-morphisms defining the bicategory of adjunctions in K K K . To avoid notation overload, we will not label the data.
Definition 3.1.16. A (lax) 1-morphism ( x , y , f , g , η , ϵ ) → ( x ′ , y ′ , f ′ , g ′ , η ′ , ϵ ′ ) (x,y,f,g,\eta,\epsilon) \to(x',y',f',g',\eta',\epsilon') ( x , y , f , g , η , ϵ ) → ( x ′ , y ′ , f ′ , g ′ , η ′ , ϵ ′ ) consists of
two 1-cells x → x ′ x \to x' x → x ′ and y → y ′ y \to y' y → y ′ ,
two 2-cells2 as depicted by the following diagram:
x → x ′ f ↓ ↓ f ′ y → y ′ y → y ′ g ↓ ↓ g ′ x → x ′ \begin{CD}
x @>>> x' \\
@V{f}VV @VV{f'}V \\
y @>>> y'
\end{CD}
\qquad
\begin{CD}
y @>>> y' \\
@V{g}VV @VV{g'}V \\
x @>>> x'
\end{CD} x f ↓ ⏐ y x ′ ↓ ⏐ f ′ y ′ y g ↓ ⏐ x y ′ ↓ ⏐ g ′ x ′
such that the following two identities holds:
y ⟶ y ′ ↓ ε y ↙ g ⇓ ↘ g ′ y ← ε x ⟶ x ′ ↘ f ↙ f ′ y ⟶ y ′ y ⟶ y ′ ↓ ε y ↙ ⇓ ↘ g ′ y x ← ε ′ x ′ ↙ ↘ f ′ y ′ ≡ x ⟶ x ′ ↓ f ↙ ↘ f ′ y ⟶ y ′ ← η ′ ↓ g ↙ ↓ ε x ′ x ⟶ x ′ x ⟶ x ′ ↙ f ↓ ε x ↘ ↓ ε x ′ y ← η y ′ ↘ g ↙ x ⟶ x ′ \begin{array}{c c}
\begin{array}{ccccc}
y&\longrightarrow&y'\\
\downarrow^{\varepsilon_y}&\swarrow^{g}&\Downarrow&&\searrow^{g'}\\
y&\xleftarrow{\varepsilon}&x&\longrightarrow&x'\\
&\searrow^{f}&&\swarrow^{f'}&\\
y&\longrightarrow&y'
\end{array}
&
\begin{array}{ccccc}
y&\longrightarrow&y'\\
\downarrow^{\varepsilon_y}&\swarrow&&\Downarrow&\searrow^{g'}\\
y&&x&\xleftarrow{\varepsilon'}&x'\\
&\swarrow&&&\searrow^{f'}\\
&&y'&&
\end{array}
\end{array}
\quad\equiv\quad
\begin{array}{c c}
\begin{array}{ccccc}
x&\longrightarrow&x'\\
\downarrow^{f}&\swarrow&&\searrow^{f'}\\
y&\longrightarrow&y'&\xleftarrow{\eta'}&\\
\downarrow^{g}&&\swarrow&&\downarrow^{\varepsilon_{x'}}\\
x&\longrightarrow&x'
\end{array}
&
\begin{array}{ccccc}
x&\longrightarrow&x'\\
\swarrow^{f}&&\downarrow^{\varepsilon_x}&\searrow&\downarrow^{\varepsilon_{x'}}\\
y&\xleftarrow{\eta}&y'&&\\
\searrow^{g}&&\swarrow&&\\
x&\longrightarrow&x'
\end{array}
\end{array} y ↓ ε y y y ⟶ ↙ g ε ↘ f ⟶ y ′ ⇓ x y ′ ⟶ ↙ f ′ ↘ g ′ x ′ y ↓ ε y y ⟶ ↙ ↙ y ′ x y ′ ⇓ ε ′ ↘ g ′ x ′ ↘ f ′ ≡ x ↓ f y ↓ g x ⟶ ↙ ⟶ ⟶ x ′ y ′ ↙ x ′ ↘ f ′ η ′ ↓ ε x ′ x ↙ f y ↘ g x ⟶ η ⟶ x ′ ↓ ε x y ′ ↙ x ′ ↘ ↓ ε x ′
Definition 3.1.17. A 2-morphism consists of
two 2-cells given as
x ⟶ ⟶ ⇓ x ′ y ⟶ ⟶ ⇓ y ′ , x \underset{\Downarrow}{\substack{\longrightarrow\\[-0.4em]\longrightarrow}} x'
\qquad
y \underset{\Downarrow}{\substack{\longrightarrow\\[-0.4em]\longrightarrow}} y', x ⇓ ⟶ ⟶ x ′ y ⇓ ⟶ ⟶ y ′ ,
such that the following two identities hold:
x ⟶ x ′ ↓ f ↙ ↓ f ′ y ⟶ y ′ ≡ x ⟶ x ′ ↓ f ↙ ↓ f ′ y ⟶ y ′ \begin{array}{ccc}
x&\longrightarrow&x'\\
\downarrow^{f}&\swarrow&\downarrow^{f'}\\
y&\longrightarrow&y'
\end{array}
\quad\equiv\quad
\begin{array}{ccc}
x&\longrightarrow&x'\\
\downarrow^{f}&\swarrow&\downarrow^{f'}\\
y&\longrightarrow&y'
\end{array} x ↓ f y ⟶ ↙ ⟶ x ′ ↓ f ′ y ′ ≡ x ↓ f y ⟶ ↙ ⟶ x ′ ↓ f ′ y ′
y ⟶ y ′ ↓ g ↙ ↓ g ′ x ⟶ x ′ ≡ y ⟶ y ′ ↓ g ↙ ↓ g ′ x ⟶ x ′ \begin{array}{ccc}
y&\longrightarrow&y'\\
\downarrow^{g}&\swarrow&\downarrow^{g'}\\
x&\longrightarrow&x'
\end{array}
\quad\equiv\quad
\begin{array}{ccc}
y&\longrightarrow&y'\\
\downarrow^{g}&\swarrow&\downarrow^{g'}\\
x&\longrightarrow&x'
\end{array} y ↓ g x ⟶ ↙ ⟶ y ′ ↓ g ′ x ′ ≡ y ↓ g x ⟶ ↙ ⟶ y ′ ↓ g ′ x ′
One objection the reader may have is that we never discussed the composites of adjoints in the context of our proposed bicategory of adjunctions . To resolve this discrepancy, we shall illustrate the idea one dimension lower with a straightforward lemma.
Lemma 3.1.18. There exists a double category of inverses in C C C defined by the following structure:
Objects - are precisely the objects of C C C itself,
Horizontal 1-cells - are precisely the morphisms of C C C ,
Vertical 1-cells - are precisely the invertible arrows in C C C ,
Square 2-cells - are commuting squares.
From this double category, one can extract the usual 1-category of inverses, whose objects are invertible pairs of morphisms ( f , f − 1 ) (f,f^{-1}) ( f , f − 1 ) , and whose morphisms are commuting square.
This suggests that our bicategory of adjunction in K K K should itself be a part of some higher double categorical structure, such as a Verity double bicategory [27] or even an intercategory [14]. However, developing this structure is outside the scope of this paper, and we shall not pursue it further here.
Analogous to the involution case, there is also a 1-category of relative inverse objects . Furthermore, adjunctions admit slicing using these inverse objects.
Definition 3.1.19 (inverse objects). Let ( F : C → D , G : D → C , η , ϵ ) (F:C\to D,G:D\to C,\eta,\epsilon) ( F : C → D , G : D → C , η , ϵ ) be an adjunction. We define a relative inverse object to be the pair ( f : F c → d , g : G d → c ) (f:Fc\to d,g:Gd\to c) ( f : F c → d , g : G d → c ) , such that the morphisms satisfy the following identities:
G ( F c ) → G ( f ) G ( d ) η c ↓ ↓ g c → 1 c c F ( G d ) → F ( g ) F ( c ) ϵ d ↓ ↓ f d → 1 d d \begin{CD}
G(Fc) @>{G(f)}>> G(d) \\
@V{\eta_c}VV @VV{g}V \\
c @>{1_c}>> c
\end{CD}
\qquad
\begin{CD}
F(Gd) @>{F(g)}>> F(c) \\
@V{\epsilon_d}VV @VV{f}V \\
d @>{1_d}>> d
\end{CD} G ( F c ) η c ↓ ⏐ c G ( f ) 1 c G ( d ) ↓ ⏐ g c F ( G d ) ϵ d ↓ ⏐ d F ( g ) 1 d F ( c ) ↓ ⏐ f d
Lemma 3.1.20. Let F ⊣ G F\dashv G F ⊣ G be an adjunction (not necessarily equivalence) between C C C and D D D . And let ( f : F c → d , g : G d → c ) (f:Fc\to d,g:Gd\to c) ( f : F c → d , g : G d → c ) be the data of an inverse object relative to this adjunction. Then, there exists a canonical sliced adjunction:
c 1 → F ( c 1 ) ↓ ↓ c → f F ( c ) → d C / c → F / f D / d G / g ↑ ↑ ⊥ G ( d 1 ) ← d 1 ↓ ↓ G ( d ) ← g c \begin{CD}
c_1 @>>> F(c_1) \\
@VVV @VVV \\
c @>{f}>> F(c) @>>> d
\end{CD}
\qquad
\begin{CD}
C/c @>{F/f}>> D/d \\
@A{G/g}AA @AA{\perp}A
\end{CD}
\qquad
\begin{CD}
G(d_1) @<<< d_1 \\
@VVV @VVV \\
G(d) @<{g}<< c
\end{CD} c 1 ↓ ⏐ c f F ( c 1 ) ↓ ⏐ F ( c ) d C / c G / g ⏐ ↑ F / f D / d ⏐ ↑ ⊥ G ( d 1 ) ↓ ⏐ G ( d ) g d 1 ↓ ⏐ c
Here, the action on morphisms is the one induced by F F F and G G G .
Proof. Let us denote the new adjoint pair using the tuple ( F / f , G / g , η / , ϵ / ) (F/f, G/g, \eta/, \epsilon/) ( F / f , G / g , η / , ϵ / ) , where η / \eta/ η / and ϵ / \epsilon/ ϵ / are defined using the following two diagrams:
( η / ) α : c 1 → η c 1 G F ( c 1 ) α ↓ ↓ G F ( α ) c → η c G F ( c ) 1 c ↓ ↓ G f c → g G d ( ϵ / ) β : F G ( d 1 ) → ϵ d 1 d 1 F G ( β ) ↓ ↓ β F G ( d ) → ϵ d d F ( g ) ↓ ↓ 1 d F c → f d \begin{aligned}
(\eta/)_\alpha:\quad&
\begin{CD}
c_1 @>{\eta_{c_1}}>> GF(c_1)\\
@V{\alpha}VV @VV{GF(\alpha)}V\\
c @>{\eta_c}>> GF(c)\\
@V{1_c}VV @VV{Gf}V\\
c @>{g}>> Gd
\end{CD}
\qquad
(\epsilon/)_\beta:\quad
\begin{CD}
FG(d_1) @>{\epsilon_{d_1}}>> d_1\\
@V{FG(\beta)}VV @VV{\beta}V\\
FG(d) @>{\epsilon_d}>> d\\
@V{F(g)}VV @VV{1_d}V\\
Fc @>{f}>> d
\end{CD}
\end{aligned} ( η / ) α : c 1 α ↓ ⏐ c 1 c ↓ ⏐ c η c 1 η c g GF ( c 1 ) ↓ ⏐ GF ( α ) GF ( c ) ↓ ⏐ G f G d ( ϵ / ) β : F G ( d 1 ) F G ( β ) ↓ ⏐ F G ( d ) F ( g ) ↓ ⏐ F c ϵ d 1 ϵ d f d 1 ↓ ⏐ β d ↓ ⏐ 1 d d
The relevant triangle identities are then immediate by pasting these two diagrams appropriately. For instance, one of the triangle identity is given by the following commuting diagram:
\begin{CD}
F(c_1) @>{F\eta_{c_1}}>> FGF(c_1) @>{\epsilon_{Fc_1}}>> F(c_1)\\
@V{F(\alpha)}VV @VV{FGF(\alpha)}V @VV{F(\alpha)}V\\
F(c) @>{F\eta_c}>> FGF(c) @>{\epsilon_{Fc}}>> F(c)\\
@V{1_{F(c)}}VV @VV{FG(f)}V @VV{f}V\\
F(c) @>{F(g)}>> FGd @>{\epsilon_d}>> d\\
@. @V{}VV @A{1_d}A{}\\
@. d
\end{CD}
At this stage, we have provided ample motivation and theory to view dualities and inverses through the lens of the microcosm principle. However, the original statement holds between two arbitrary dimensions n n n and n + 1 n+1 n + 1 . Therefore, a similar theory of dualities and inverses should also hold in one dimension higher. In particular, we want to define relative pseudoinvolutions and pseudoinverses.
2-dimensional variants
To obtain the definition of 2-dimensional variant of involutive objects, it is best to utilize a categorification of the lax compatible form of 1-dimensional involutive objects. A somewhat similar definition can be found in [31, Def. 2.1] for the special case of T = B i c a t T = \mathbf{Bicat} T = Bicat .
Definition 3.2.1. A weak 2-involution in a tricategory T T T consists of
a 1-cell H : K → K H: K \to K H : K → K ,
an adjoint equivalence α : H ⊗ H ⟺ 1 K : α ′ \alpha: H \otimes H \Longleftrightarrow1_K: \alpha' α : H ⊗ H ⟺ 1 K : α ′ ,
a 3-cell ζ \zeta ζ as depicted by the following diagram:
K → H K K → H K H ↑ ↓ H ⟹ ζ H ↑ ↓ H K → H K K → H K \begin{array}{ccccc}
K & \xrightarrow{\ H\ } & K & & K \xrightarrow{\ H\ } K \\
{\scriptstyle H}\uparrow& & \downarrow{\scriptstyle H} & \Longrightarrow_{\zeta} & {\scriptstyle H}\uparrow\qquad\qquad\downarrow{\scriptstyle H} \\
K & \xrightarrow{\ H\ } & K & & K \xrightarrow{\ H\ } K
\end{array} K H ↑ K H H K ↓ H K ⟹ ζ K H K H ↑ ↓ H K H K
( H ⊗ H ) ⊗ H → a H ⊗ ( H ⊗ H ) α ⊗ 1 ↓ ↗ ζ ↓ 1 ⊗ α 1 K ⊗ H → l H ← r H ⊗ 1 K \begin{array}{ccccc}
(H \otimes H) \otimes H & \xrightarrow{\ a\ } & H \otimes(H \otimes H) \\
{\scriptstyle\alpha\otimes1}\downarrow& \nearrow{\scriptstyle\zeta} & \downarrow{\scriptstyle1 \otimes\alpha} \\
1_K \otimes H & \xrightarrow{\ l\ } & H & \xleftarrow{\ r\ } & H \otimes1_K
\end{array} ( H ⊗ H ) ⊗ H α ⊗ 1 ↓ 1 K ⊗ H a ↗ ζ l H ⊗ ( H ⊗ H ) ↓ 1 ⊗ α H r H ⊗ 1 K
the 3-cell ζ \zeta ζ satisfy the weak 2-involution axiom (section D.1).
Although the morphisms of weak 2-involutions are not directly relevant for our purposes, we record their definitions in section C.1. We now define the relative pseudo involutive objects as suggested by the microcosm principle.
Definition 3.2.2. Let ( K , H , α , ζ ) (K, H, \alpha, \zeta) ( K , H , α , ζ ) be a weak 2-involution in the tricategory Bicat . We define a relative lax involutive object by
a 1-cell f : H x → x ∈ B f: Hx \to x \in B f : H x → x ∈ B ,
a 2-cell depicted by the following diagram:
H 2 x → H f H x α x ↓ ↙ ϕ ↓ f x → ε x x \begin{array}{ccccc}
H^2x & \xrightarrow{\ Hf\ } & Hx \\
{\scriptstyle\alpha_x}\downarrow& \swarrow{\scriptstyle\phi} & \downarrow{\scriptstyle f} \\
x & \xrightarrow{\ \varepsilon_x\ } & x
\end{array} H 2 x α x ↓ x H f ↙ ϕ ε x H x ↓ f x
such that the 2-cell satisfy the following identity:
H 3 x → H 2 f H 2 x → H f H x H α x ↓ α x ↓ f ↓ H x → f x → ε x x ≡ H 3 x → H 2 f H 2 x H α x ↓ H f ↓ H x → ε H x H x → f x \begin{CD}
H^{3}x @>{H^{2}f}>> H^{2}x @>{Hf}>> Hx\\
@V{H\alpha_x}VV @V{\alpha_x}VV @V{f}VV\\
Hx @>{f}>> x @>{\varepsilon_x}>> x
\end{CD}
\quad\equiv\quad
\begin{CD}
H^{3}x @>{H^{2}f}>> H^{2}x\\
@V{H\alpha_x}VV @V{Hf}VV\\
Hx @>{\varepsilon_{Hx}}>> Hx @>{f}>> x
\end{CD} H 3 x H α x ↓ ⏐ H x H 2 f f H 2 x α x ↓ ⏐ x H f ε x H x f ↓ ⏐ x ≡ H 3 x H α x ↓ ⏐ H x H 2 f ε H x H 2 x H f ↓ ⏐ H x f x
Example 3.2.3. The functor op 2 : 2 C a t ≤ 2 → 2 C a t ≤ 2 \operatorname{op}_{2}: 2\mathrm{Cat}_{\leq2} \to2\mathrm{Cat}_{\leq2} op 2 : 2 Cat ≤ 2 → 2 Cat ≤ 2 defines a strict 2-involutive object. Thus, op 1 : op 2 ( 1 C a t ) → 1 C a t \operatorname{op}_{1}: \operatorname{op}_{2}(1\mathrm{Cat}) \to1\mathrm{Cat} op 1 : op 2 ( 1 Cat ) → 1 Cat is a relative strict 1-involutive object.
One can similarly define the 1-morphisms and the 2-morphisms of these relative involutions.
Definition 3.2.4.
A 1-morphism ( r , u ) (r,u) ( r , u ) is given by r : x → y r:x\to y r : x → y , and a corresponding 2-cell u u u that satisfy the following identity:
H 2 x → H 2 r H 2 y α x ↓ ↓ α y x → r y ≡ H 2 x → H 2 r H 2 y α x ↓ ↓ H g x → r y \begin{CD}
H^{2}x @>{H^{2}r}>> H^{2}y\\
@V{\alpha_x}VV @VV{\alpha_y}V\\
x @>{r}>> y
\end{CD}
\quad\equiv\quad
\begin{CD}
H^{2}x @>{H^{2}r}>> H^{2}y\\
@V{\alpha_x}VV @VV{Hg}V\\
x @>{r}>> y
\end{CD} H 2 x α x ↓ ⏐ x H 2 r r H 2 y ↓ ⏐ α y y ≡ H 2 x α x ↓ ⏐ x H 2 r r H 2 y ↓ ⏐ H g y
A 2-morphism is a 2-cell γ : r ⇒ t \gamma:r\Rightarrow t γ : r ⇒ t that satisfy the following identity:
H x → H r H y f ↓ ↓ g x → r y ≡ H x → H r H y f ↓ ↓ g x → t y \begin{CD}
Hx @>{Hr}>> Hy\\
@V{f}VV @VV{g}V\\
x @>{r}>> y
\end{CD}
\quad\equiv\quad
\begin{CD}
Hx @>{Hr}>> Hy\\
@V{f}VV @VV{g}V\\
x @>{t}>> y
\end{CD} H x f ↓ ⏐ x H r r H y ↓ ⏐ g y ≡ H x f ↓ ⏐ x H r t H y ↓ ⏐ g y
Therefore, we obtain the bicategory 1 I n v ( K , H , α , ζ ) 1\mathrm{Inv}(K,H,\alpha,\zeta) 1 Inv ( K , H , α , ζ ) . The axioms can be verified similar as before by using the strictification to encode the desired compositional data.
Proposition 3.2.5. The microcosm principle holds for weak 2-involutive bicategories.
Finally, in parallel, we can also write these higher dimensional definitions in the context of inverses.
Definition 3.2.6 ([17, Def. 2.1]). A weak 2-adjunction in a tricategory T T T is
two 1-cell F : K 1 → K 2 F: K_1 \to K_2 F : K 1 → K 2 and G : K 2 → K 1 G: K_2 \to K_1 G : K 2 → K 1 ,
two 2-cells ϵ : F G ⇒ 1 K 2 \epsilon: FG \Rightarrow1_{K_2} ϵ : F G ⇒ 1 K 2 and η : 1 K 1 ⇒ G F \eta: 1_{K_1} \Rightarrow GF η : 1 K 1 ⇒ GF ,
two 3-cells ϕ \phi ϕ and ψ \psi ψ as depicted by the following diagrams:
K 1 → F K 1 → F K 2 F ↓ η ↑ ↑ ϵ K 2 → G K 1 → F K 2 F ↓ ↑ 1 F K 1 → 1 K 1 K 1 ⟹ ϕ K 1 → F K 2 F ↓ ↑ 1 F K 2 K 2 \begin{CD}
K_1 @>{F}>> K_1 @>{F}>> K_2 \\
@V{F}VV @A{\eta}AA @AA{\epsilon}A \\
K_2 @>{G}>> K_1 @>{F}>> K_2 \\
@V{F}VV @. @AA{1_F}A \\
K_1 @>{1_{K_1}}>> K_1
\end{CD}
\qquad\Longrightarrow^{\phi}\qquad
\begin{CD}
K_1 @>{F}>> K_2 \\
@V{F}VV @AA{1_F}A \\
K_2 @. K_2
\end{CD} K 1 F ↓ ⏐ K 2 F ↓ ⏐ K 1 F G 1 K 1 K 1 η ⏐ ↑ K 1 K 1 F F K 2 ⏐ ↑ ϵ K 2 ⏐ ↑ 1 F ⟹ ϕ K 1 F ↓ ⏐ K 2 F K 2 ⏐ ↑ 1 F K 2 F ⊗ ( G ⊗ F ) → a ( F ⊗ G ) ⊗ F 1 ⊗ η ↓ ↓ ϵ ⊗ 1 F ⊗ 1 K 1 1 K 2 ⊗ F r ↑ ↑ l F → 1 F F \begin{CD}
F \otimes(G \otimes F) @>{a}>> (F \otimes G) \otimes F \\
@V{1\otimes\eta}VV @VV{\epsilon\otimes1}V \\
F \otimes1_{K_1} @. 1_{K_2} \otimes F \\
@A{r}AA @AA{l}A \\
F @>{1_F}>> F
\end{CD} F ⊗ ( G ⊗ F ) 1 ⊗ η ↓ ⏐ F ⊗ 1 K 1 r ⏐ ↑ F a 1 F ( F ⊗ G ) ⊗ F ↓ ⏐ ϵ ⊗ 1 1 K 2 ⊗ F ⏐ ↑ l F
G → K 1 ↓ ↓ K 2 → K 1 K 1 → K 1 ↓ ↓ K 2 → K 2 K 2 → K 2 ↓ ↓ G → G \begin{CD}
G @>>> K_1 \\
@VVV @VVV \\
K_2 @>>> K_1
\end{CD}
\qquad
\begin{CD}
K_1 @>>> K_1 \\
@VVV @VVV \\
K_2 @>>> K_2
\end{CD}
\qquad
\begin{CD}
K_2 @>>> K_2 \\
@VVV @VVV \\
G @>>> G
\end{CD} G ↓ ⏐ K 2 K 1 ↓ ⏐ K 1 K 1 ↓ ⏐ K 2 K 1 ↓ ⏐ K 2 K 2 ↓ ⏐ G K 2 ↓ ⏐ G
( G ⊗ F ) ⊗ G → a G ⊗ ( F ⊗ G ) η ⊗ 1 ↑ ↓ 1 ⊗ ϵ 1 K 1 ⊗ G G ⊗ 1 K 2 l ′ ↑ ↓ r G → 1 G G \begin{CD}
(G \otimes F) \otimes G @>{a}>> G \otimes(F \otimes G) \\
@A{\eta\otimes1}AA @VV{1\otimes\epsilon}V \\
1_{K_1}\otimes G @. G\otimes1_{K_2} \\
@A{l^{\prime}}AA @VV{r}V \\
G @>{1_G}>> G
\end{CD} ( G ⊗ F ) ⊗ G η ⊗ 1 ⏐ ↑ 1 K 1 ⊗ G l ′ ⏐ ↑ G a 1 G G ⊗ ( F ⊗ G ) ↓ ⏐ 1 ⊗ ϵ G ⊗ 1 K 2 ↓ ⏐ r G
such that two pasting diagrams as given in [17, Def. 2.1] or in section D.4 hold.
As before, the morphisms of weak 2-adjunctions can be found in section C.2.
Definition 3.2.7. Let ( F : K 1 → K 2 , G : K 2 → K 1 , ϵ , η , ϕ , ψ ) (F: K_1 \to K_2, G: K_2 \to K_1, \epsilon, \eta, \phi, \psi) ( F : K 1 → K 2 , G : K 2 → K 1 , ϵ , η , ϕ , ψ ) be the data of a weak 2-adjunction in the tricategory Bicat . We define a relative (lax) weak 1-inverse object by
two 1-cells f : F k 1 → k 2 ∈ K 2 f: Fk_1 \to k_2 \in K_2 f : F k 1 → k 2 ∈ K 2 and g : G k 2 → k 1 ∈ K 1 g: Gk_2 \to k_1 \in K_1 g : G k 2 → k 1 ∈ K 1 ,
two 2-cells
F G k 2 → F ( g ) F k 1 ϵ k 2 ↓ ↓ f k 2 → ϵ k 2 k 2 G F k 1 → G ( f ) G k 2 η k 1 ↑ ↓ g k 1 → ϵ k 1 k 1 \begin{CD}
FGk_2 @>{F(g)}>> Fk_1 \\
@V{\epsilon_{k_2}}VV @VV{f}V \\
k_2 @>{\epsilon_{k_2}}>> k_2
\end{CD}
\qquad
\begin{CD}
GFk_1 @>{G(f)}>> Gk_2 \\
@A{\eta_{k_1}}AA @VV{g}V \\
k_1 @>{\epsilon_{k_1}}>> k_1
\end{CD} F G k 2 ϵ k 2 ↓ ⏐ k 2 F ( g ) ϵ k 2 F k 1 ↓ ⏐ f k 2 GF k 1 η k 1 ⏐ ↑ k 1 G ( f ) ϵ k 1 G k 2 ↓ ⏐ g k 1
such that the 2-cells satisfy the following identities:
F k 1 → f k 2 , F k 1 → ε F k 1 F k 1 , F k 1 → F η k 1 F G F k 1 , F G F k 1 → F G ( f ) F G k 2 , F G F k 1 ← ϕ F G F k 1 , F G F k 1 → F G k 2 F G k 2 , F G k 2 → F g F k 1 , F G k 2 → ε k 2 k 2 , F k 1 → ε F k 1 F k 1 , F k 1 → f k 2 , F k 1 → F β F G F k 1 , F G k 2 → α k 2 , G k 2 → g k 1 , G k 2 → η G k 2 G F G k 2 , G F G k 2 → G F ( g ) G F k 1 , G F k 1 → G f G k 2 , G k 2 → ε G k 2 G k 2 , G k 2 → g k 1 , G k 1 → η k 1 G F k 1 , G F k 1 → β G k 2 , G k 2 → G ε k 2 G k 2 , G F G k 2 → G α G k 2 , G F G k 2 → G ε k 2 G k 2 \begin{gathered}
F k_1 \xrightarrow{f} k_2,\qquad F k_1 \xrightarrow{\varepsilon_{F k_1}} F k_1,\qquad F k_1 \xrightarrow{F\eta_{k_1}} F G F k_1,\qquad F G F k_1 \xrightarrow{F G(f)} F G k_2,\\
F G F k_1 \xleftarrow{\phi} F G F k_1,\qquad F G F k_1 \xrightarrow{F G k_2} F G k_2,\qquad F G k_2 \xrightarrow{F g} F k_1,\qquad F G k_2 \xrightarrow{\varepsilon_{k_2}} k_2,\\
F k_1 \xrightarrow{\varepsilon_{F k_1}} F k_1,\qquad F k_1 \xrightarrow{f} k_2,\qquad F k_1 \xrightarrow{F\beta} F G F k_1,\qquad F G k_2 \xrightarrow{\alpha} k_2,\\
G k_2 \xrightarrow{g} k_1,\qquad G k_2 \xrightarrow{\eta_{G k_2}} G F G k_2,\qquad G F G k_2 \xrightarrow{G F(g)} G F k_1,\qquad G F k_1 \xrightarrow{G f} G k_2,\\
G k_2 \xrightarrow{\varepsilon_{G k_2}} G k_2,\qquad G k_2 \xrightarrow{g} k_1,\qquad G k_1 \xrightarrow{\eta_{k_1}} G F k_1,\qquad G F k_1 \xrightarrow{\beta} G k_2,\\
G k_2 \xrightarrow{G\varepsilon_{k_2}} G k_2,\qquad G F G k_2 \xrightarrow{G\alpha} G k_2,\qquad G F G k_2 \xrightarrow{G\varepsilon_{k_2}} G k_2
\end{gathered} F k 1 f k 2 , F k 1 ε F k 1 F k 1 , F k 1 F η k 1 F GF k 1 , F GF k 1 F G ( f ) F G k 2 , F GF k 1 ϕ F GF k 1 , F GF k 1 F G k 2 F G k 2 , F G k 2 F g F k 1 , F G k 2 ε k 2 k 2 , F k 1 ε F k 1 F k 1 , F k 1 f k 2 , F k 1 F β F GF k 1 , F G k 2 α k 2 , G k 2 g k 1 , G k 2 η G k 2 GF G k 2 , GF G k 2 GF ( g ) GF k 1 , GF k 1 G f G k 2 , G k 2 ε G k 2 G k 2 , G k 2 g k 1 , G k 1 η k 1 GF k 1 , GF k 1 β G k 2 , G k 2 G ε k 2 G k 2 , GF G k 2 G α G k 2 , GF G k 2 G ε k 2 G k 2
F k 1 → f k 2 , F k 1 → ε F k 1 F k 1 , F k 1 → ε F k 1 F k 1 , F k 1 → f k 2 , F k 1 → ε F k 1 F k 1 , F k 1 → ε F k 1 F k 1 , k 2 → ε k 2 k 2 \begin{gathered}
F k_1 \xrightarrow{f} k_2,\qquad F k_1 \xrightarrow{\varepsilon_{F k_1}} F k_1,\qquad F k_1 \xrightarrow{\varepsilon_{F k_1}} F k_1,\qquad F k_1 \xrightarrow{f} k_2,\\
F k_1 \xrightarrow{\varepsilon_{F k_1}} F k_1,\qquad F k_1 \xrightarrow{\varepsilon_{F k_1}} F k_1,\qquad k_2 \xrightarrow{\varepsilon_{k_2}} k_2
\end{gathered} F k 1 f k 2 , F k 1 ε F k 1 F k 1 , F k 1 ε F k 1 F k 1 , F k 1 f k 2 , F k 1 ε F k 1 F k 1 , F k 1 ε F k 1 F k 1 , k 2 ε k 2 k 2
G k 2 → g k 1 , G k 2 → η G k 2 G F G k 2 , G F G k 2 → G F ( g ) G F k 1 , G F k 1 → G f G k 2 , G k 2 → G ε k 2 G k 2 , G k 2 → ε G k 2 G k 2 , G k 2 → g k 1 , k 1 → η k 1 G F k 1 , G F k 1 → β G k 2 , G α → G k 2 , G k 2 → ε G k 2 G k 2 \begin{gathered}
G k_2 \xrightarrow{g} k_1,\qquad G k_2 \xrightarrow{\eta_{G k_2}} G F G k_2,\qquad G F G k_2 \xrightarrow{G F(g)} G F k_1,\qquad G F k_1 \xrightarrow{G f} G k_2,\\
G k_2 \xrightarrow{G\varepsilon_{k_2}} G k_2,\qquad G k_2 \xrightarrow{\varepsilon_{G k_2}} G k_2,\qquad G k_2 \xrightarrow{g} k_1,\qquad k_1 \xrightarrow{\eta_{k_1}} G F k_1,\\
G F k_1 \xrightarrow{\beta} G k_2,\qquad G\alpha\xrightarrow{} G k_2,\qquad G k_2 \xrightarrow{\varepsilon_{G k_2}} G k_2
\end{gathered} G k 2 g k 1 , G k 2 η G k 2 GF G k 2 , GF G k 2 GF ( g ) GF k 1 , GF k 1 G f G k 2 , G k 2 G ε k 2 G k 2 , G k 2 ε G k 2 G k 2 , G k 2 g k 1 , k 1 η k 1 GF k 1 , GF k 1 β G k 2 , G α G k 2 , G k 2 ε G k 2 G k 2
Definition 3.2.8. Let ( f : F k 1 → k 2 , g : G k 2 → k 1 , α , β ) (f : F k_1 \to k_2, g : G k_2 \to k_1, \alpha, \beta) ( f : F k 1 → k 2 , g : G k 2 → k 1 , α , β ) and ( f ′ : F k 1 ′ → k 2 ′ , g ′ : G k 2 ′ → k 1 ′ , α ′ , β ′ ) (f' : F k'_1 \to k'_2, g' : G k'_2 \to k'_1, \alpha', \beta') ( f ′ : F k 1 ′ → k 2 ′ , g ′ : G k 2 ′ → k 1 ′ , α ′ , β ′ ) be two relative inverse objects. The data of a (lax) 1-morphism is given by
two 1-cells k 1 → k 1 ′ k_1 \to k'_1 k 1 → k 1 ′ and k 2 → k 2 ′ k_2 \to k'_2 k 2 → k 2 ′ ,
two 2-cells
F k 1 ⟶ F k 1 ′ G k 2 ⟶ G k 2 ′ ↓ f ↙ ↓ f ′ ↓ g ↓ g ′ k 2 ⟶ k 2 ′ k 1 ⟶ k 1 ′ \begin{array}{ccccc}
Fk_1 & \longrightarrow& Fk'_1 & \qquad& Gk_2 \longrightarrow Gk'_2 \\
\downarrow^{f} & \swarrow& \downarrow^{f'} & & \downarrow^{g}\qquad\qquad\downarrow^{g'} \\
k_2 & \longrightarrow& k'_2 & \qquad& k_1 \longrightarrow k'_1
\end{array} F k 1 ↓ f k 2 ⟶ ↙ ⟶ F k 1 ′ ↓ f ′ k 2 ′ G k 2 ⟶ G k 2 ′ ↓ g ↓ g ′ k 1 ⟶ k 1 ′
such that the following two identities hold:
k 1 ⟶ k 1 ′ ↓ η k 1 ↙ ↙ ↓ η k 1 ′ G F k 1 ⟶ G F k 1 ′ ↓ G f ↙ ↙ ↓ G f ′ ↓ ϵ k 1 ′ G k 2 ⟶ G k 2 ′ ↙ g ↘ g ′ ↙ ↙ ↓ k 1 ⟶ k 1 ′ k 1 ⟶ k 1 ′ ↓ η k 1 ↘ ↓ ϵ k 1 ′ G F k 1 G F k 1 ′ ↓ G f ⇐ β = ↓ G k 2 ↘ g ↓ k 1 ⟶ k 1 ′ \begin{array}{ccccc}
k_1 & \longrightarrow& k'_1 & & \\
\downarrow^{\eta_{k_1}} & \swarrow\!\!\swarrow& \downarrow^{\eta_{k'_1}} & & \\
GFk_1 & \longrightarrow& GFk'_1 & & \\
\downarrow^{Gf} & \swarrow\!\!\swarrow& \downarrow^{Gf'} & & \downarrow^{\epsilon_{k'_1}} \\
Gk_2 & \longrightarrow& Gk'_2 & & \\
\swarrow^{g} & \searrow_{g'} & \swarrow\!\!\swarrow& & \downarrow\\
k_1 & \longrightarrow& k'_1 & &
\end{array}
\qquad
\begin{array}{c}
k_1 \longrightarrow k'_1\\
\downarrow^{\eta_{k_1}}\quad\searrow\quad\downarrow^{\epsilon_{k'_1}}\\
GFk_1 \qquad\qquad GFk'_1\\
\downarrow^{Gf}\qquad\Leftarrow\beta=\qquad\downarrow\\
Gk_2\\
\searrow^{g}\qquad\downarrow\\
k_1 \longrightarrow k'_1
\end{array} k 1 ↓ η k 1 GF k 1 ↓ G f G k 2 ↙ g k 1 ⟶ ↙ ↙ ⟶ ↙ ↙ ⟶ ↘ g ′ ⟶ k 1 ′ ↓ η k 1 ′ GF k 1 ′ ↓ G f ′ G k 2 ′ ↙ ↙ k 1 ′ ↓ ϵ k 1 ′ ↓ k 1 ⟶ k 1 ′ ↓ η k 1 ↘ ↓ ϵ k 1 ′ GF k 1 GF k 1 ′ ↓ G f ⇐ β = ↓ G k 2 ↘ g ↓ k 1 ⟶ k 1 ′
F G k 2 ⟶ F G k 2 ′ ↓ ϵ k 2 ↙ ↙ ↓ ϵ k 2 ′ F k 1 ⟶ F k 1 ′ ↓ f ↙ ↙ ↓ f ′ k 2 ⟶ k 2 ′ F G k 2 ⟶ F G k 2 ′ ↓ ϵ k 1 ↙ ↙ ↓ ϵ k 1 k 2 ⟶ k 2 ′ ⇐ α ′ = F k 1 ′ ↓ ϵ k 2 ↓ ϵ k 2 ′ ↓ f ′ k 2 ⟶ k 2 ′ \begin{array}{ccccc}
FGk_2 & \longrightarrow& FGk'_2 & & \\
\downarrow^{\epsilon_{k_2}} & \swarrow\!\!\swarrow& \downarrow^{\epsilon_{k'_2}} & & \\
Fk_1 & \longrightarrow& Fk'_1 & & \\
\downarrow^{f} & \swarrow\!\!\swarrow& \downarrow^{f'} & & \\
k_2 & \longrightarrow& k'_2 & &
\end{array}
\qquad
\begin{array}{c}
FGk_2 \longrightarrow FGk'_2\\
\downarrow^{\epsilon_{k_1}}\quad\swarrow\!\!\swarrow\quad\downarrow^{\epsilon_{k_1}}\\
k_2 \longrightarrow k'_2\qquad\Leftarrow\alpha'=\qquad Fk'_1\\
\downarrow^{\epsilon_{k_2}}\qquad\downarrow^{\epsilon_{k'_2}}\qquad\downarrow^{f'}\\
k_2 \longrightarrow k'_2
\end{array} F G k 2 ↓ ϵ k 2 F k 1 ↓ f k 2 ⟶ ↙ ↙ ⟶ ↙ ↙ ⟶ F G k 2 ′ ↓ ϵ k 2 ′ F k 1 ′ ↓ f ′ k 2 ′ F G k 2 ⟶ F G k 2 ′ ↓ ϵ k 1 ↙ ↙ ↓ ϵ k 1 k 2 ⟶ k 2 ′ ⇐ α ′ = F k 1 ′ ↓ ϵ k 2 ↓ ϵ k 2 ′ ↓ f ′ k 2 ⟶ k 2 ′
Definition 3.2.9. The data of a 2-morphism between two 1-morphisms is given
two 2-cells
k 1 ⟹ k 1 ′ k 2 ⟹ k 2 ′ \begin{array}{ccccc}
k_1 & \Longrightarrow& k'_1 & \qquad& k_2 \Longrightarrow k'_2
\end{array} k 1 ⟹ k 1 ′ k 2 ⟹ k 2 ′
such that the following two identities hold:
F k 1 → F k 1 ′ f ↓ ↓ f ′ k 2 → k 2 ′ ≡ F k 1 → F k 1 ′ f ↓ ↓ f ′ k 2 → k 2 ′ G k 2 → G k 2 ′ g ↓ ↓ g ′ k 1 → k 1 ′ ≡ G k 2 → G k 2 ′ g ↓ ↓ g ′ k 1 → k 1 ′ \begin{aligned}
&\begin{CD}
Fk_1 @>{}>> Fk'_1\\
@V{f}VV @VV{f'}V\\
k_2 @>{}>> k'_2
\end{CD}
\quad\equiv\quad
\begin{CD}
Fk_1 @>{}>> Fk'_1\\
@V{f}VV @VV{f'}V\\
k_2 @>{}>> k'_2
\end{CD}\\[2em]
&\begin{CD}
Gk_2 @>{}>> Gk'_2\\
@V{g}VV @VV{g'}V\\
k_1 @>{}>> k'_1
\end{CD}
\quad\equiv\quad
\begin{CD}
Gk_2 @>{}>> Gk'_2\\
@V{g}VV @VV{g'}V\\
k_1 @>{}>> k'_1
\end{CD}
\end{aligned} F k 1 f ↓ ⏐ k 2 F k 1 ′ ↓ ⏐ f ′ k 2 ′ ≡ F k 1 f ↓ ⏐ k 2 F k 1 ′ ↓ ⏐ f ′ k 2 ′ G k 2 g ↓ ⏐ k 1 G k 2 ′ ↓ ⏐ g ′ k 1 ′ ≡ G k 2 g ↓ ⏐ k 1 G k 2 ′ ↓ ⏐ g ′ k 1 ′
Proposition 3.2.10. The microcosm principle holds for weak 2-adjunction of bicategories.
We have now developed all the desired definitions to fully express the dualities of monoidal structures, albeit at a greater level of generality than necessary. For all our purposes, we shall only require T = G r a y T = \mathbf{Gray} T = Gray .
As a closing remark, we note that so far we have only employed the microcosm principle in doubly nested structures. This is in accordance with how the original statement is written and demonstrated. However, there are instances where we can achieve triply/higher nested structures.
Example 3.2.11. The dual vector space 1-functor ( − ) ∗ : op 1 ( V e c t k f . d . ) → V e c t k f . d . ( - )^{*}\colon\operatorname{op}_{1}\left(\mathbf{Vect}_{k}^{f.d.}\right) \to\mathbf{Vect}_{k}^{f.d.} ( − ) ∗ : op 1 ( Vect k f . d . ) → Vect k f . d . defines an involutive object relative to the strict involutive object op 1 : op 2 ( 1 C a t ) → 1 C a t \operatorname{op}_{1}\colon\operatorname{op}_{2}(\mathbf{1Cat}) \to\mathbf{1Cat} op 1 : op 2 ( 1Cat ) → 1Cat . The structure cells are simply η V : V → V ∗ ∗ ∈ V e c t k f . d . \eta_V\colon V \to V^{**} \in\mathbf{Vect}_{k}^{f.d.} η V : V → V ∗∗ ∈ Vect k f . d . as the unit and its opposite cell op 1 ( η V ) \operatorname{op}_{1}(\eta_V) op 1 ( η V ) as the counit. One could view ( − ) ∗ ( - )^{*} ( − ) ∗ as a contravariant autoequivalence on V e c t k f . d . \mathbf{Vect}_{k}^{f.d.} Vect k f . d . .
In this example, we have a top (2-truncated) 3-dimensional involution op 2 : 2 C a t ≤ 2 → 2 C a t ≤ 2 \operatorname{op}_{2}\colon\mathbf{2Cat}_{\leq2} \to\mathbf{2Cat}_{\leq2} op 2 : 2Cat ≤ 2 → 2Cat ≤ 2 , inside which we define a (1-dimensional promoted to) 2-dimensional involution op 1 \operatorname{op}_{1} op 1 , and finally inside which we define a 1-dimensional involution ( − ) ∗ ( - )^{*} ( − ) ∗ . In general, it is not possible to obtain such triply (or higher) nested structures just by simply considering the microcosm principle.
Dualities of monoidal structures
We wish to describe various “transport” constructions for monoids and modules across different duals of a monoidal category V \mathcal{V} V . To do this, we first briefly describe these relevant dualities on M o n C a t x \mathbf{MonCat}_{x} MonCat x , where x x x is used as a placeholder for lax, colax, or pseudo. Next, we discuss how morphisms of involutions and inverses provide a natural framework to encode this dualities. For encoding the dualities of modules, virtual double categories shall play an essential role. Moreover, these dualities themselves generalize appropriately to virtual double categories.
Dualities of MonCat
The op 1 \operatorname{op}_{1} op 1 duality (and its generalized inverse op 1 t \operatorname{op}_{1}^{t} op 1 t ) can be extended to the 2 2 2 -category of monoidal categories and provides the following relation between the lax and colax variants.
Lemma 4.1.1. There exists two functors
op 1 : op 2 ( M o n C a t c ) → M o n C a t l \operatorname{op}_{1}\colon\operatorname{op}_{2}(\mathbf{MonCat}_{c}) \to\mathbf{MonCat}_{l} op 1 : op 2 ( MonCat c ) → MonCat l and
op 1 t : op 2 ( M o n C a t l ) → M o n C a t c \operatorname{op}_{1}^{t}\colon\operatorname{op}_{2}(\mathbf{MonCat}_{l}) \to\mathbf{MonCat}_{c} op 1 t : op 2 ( MonCat l ) → MonCat c extending the involution op 1 : op 2 ( 1 C a t ) → 1 C a t \operatorname{op}_{1}\colon\operatorname{op}_{2}(\mathbf{1Cat}) \to\mathbf{1Cat} op 1 : op 2 ( 1Cat ) → 1Cat along the forgetful 2 2 2 -functors U x : M o n C a t x → C a t U_{x}\colon\mathbf{MonCat}_{x} \to\mathbf{Cat} U x : MonCat x → Cat . These functors are strict inverses of each other relative to the op 2 \operatorname{op}_{2} op 2 involution. That is, they satisfy the constraints as defined by the following diagrams:
M o n C a t c → op 2 ( op 1 ) op 2 ( M o n C a t l ) 1 ↓ ↓ op 1 t M o n C a t c ← op 1 t op 2 ( M o n C a t l ) M o n C a t l → op 2 ( op 1 t ) op 2 ( M o n C a t c ) 1 ↓ ↓ op 1 M o n C a t l ← op 1 op 2 ( M o n C a t c ) \begin{CD}
\mathbf{MonCat}_{c} @>{\operatorname{op}_{2}(\operatorname{op}_{1})}>> \operatorname{op}_{2}(\mathbf{MonCat}_{l}) \\
@V{1}VV @VV{\operatorname{op}_{1}^{t}}V \\
\mathbf{MonCat}_{c} @<<{\operatorname{op}_{1}^{t}}< \operatorname{op}_{2}(\mathbf{MonCat}_{l})
\end{CD}
\qquad
\begin{CD}
\mathbf{MonCat}_{l} @>{\operatorname{op}_{2}(\operatorname{op}_{1}^{t})}>> \operatorname{op}_{2}(\mathbf{MonCat}_{c}) \\
@V{1}VV @VV{\operatorname{op}_{1}}V \\
\mathbf{MonCat}_{l} @<<{\operatorname{op}_{1}}< \operatorname{op}_{2}(\mathbf{MonCat}_{c})
\end{CD} MonCat c 1 ↓ ⏐ MonCat c op 2 ( op 1 ) op 1 t op 2 ( MonCat l ) ↓ ⏐ op 1 t op 2 ( MonCat l ) MonCat l 1 ↓ ⏐ MonCat l op 2 ( op 1 t ) op 1 op 2 ( MonCat c ) ↓ ⏐ op 1 op 2 ( MonCat c ) Furthermore, U x U_{x} U x can be extended to a (strict) morphism of inverses and the 2 2 2 -functors op 1 \operatorname{op}_{1} op 1 and op 1 t \operatorname{op}_{1}^{t} op 1 t are isomorphic when restricted onto the variant M o n C a t p \mathbf{MonCat}_{p} MonCat p .
Proof. For each diagram defining some structural monoidal data or coherence constraint, apply the functor op 1 : op 2 ( 1 C a t ) → 1 C a t \operatorname{op}_{1}\colon\operatorname{op}_{2}(\mathbf{1Cat}) \to\mathbf{1Cat} op 1 : op 2 ( 1Cat ) → 1Cat to obtain the required dual diagram. More precisely, op 1 \operatorname{op}_{1} op 1 is a monoidal 1 C a t \mathbf{1Cat} 1Cat -enriched functor of monoidal 1 C a t \mathbf{1Cat} 1Cat -enriched categories. Thus, one could also adapt [34 Thm. 4.5.1] in our situation by replacing the strict structures with pseudo structures. □
The same duality holds for B r M o n C a t x \mathbf{BrMonCat}_{x} BrMonCat x and S y m M o n C a t x \mathbf{SymMonCat}_{x} SymMonCat x . For instance, the restriction of op 1 \operatorname{op}_{1} op 1 functor yields the commuting diagram
op 2 ( S y m M o n C a t c ) → op 2 ( U ) op 2 ( B r M o n C a t c ) → op 2 ( U ) op 2 ( M o n C a t c ) op 1 ↓ op 1 ↓ op 1 ↓ S y m M o n C a t l → U B r M o n C a t l → U M o n C a t l \begin{CD}
\operatorname{op}_{2}(\mathbf{SymMonCat}_{c}) @>{\operatorname{op}_{2}(U)}>> \operatorname{op}_{2}(\mathbf{BrMonCat}_{c}) @>{\operatorname{op}_{2}(U)}>> \operatorname{op}_{2}(\mathbf{MonCat}_{c}) \\
@V{\operatorname{op}_{1}}VV @V{\operatorname{op}_{1}}VV @V{\operatorname{op}_{1}}VV \\
\mathbf{SymMonCat}_{l} @>>{U}> \mathbf{BrMonCat}_{l} @>>{U}> \mathbf{MonCat}_{l}
\end{CD} op 2 ( SymMonCat c ) op 1 ↓ ⏐ SymMonCat l op 2 ( U ) U op 2 ( BrMonCat c ) op 1 ↓ ⏐ BrMonCat l op 2 ( U ) U op 2 ( MonCat c ) op 1 ↓ ⏐ MonCat l
Here, op 1 \operatorname{op}_{1} op 1 maps a braided monoidal category as ( V , b x , y ) ↦ ( op 1 ( V ) , b x , y − 1 ) (\mathcal{V}, b_{x,y}) \mapsto(\operatorname{op}_{1}(\mathcal{V}), b^{-1}_{x,y}) ( V , b x , y ) ↦ ( op 1 ( V ) , b x , y − 1 ) . And, each square in the diagram represents a part of the corresponding morphism of inverses.
Remark 4.1.2. We can combine both op 1 \operatorname{op}_{1} op 1 and op 1 t \operatorname{op}_{1}^{t} op 1 t as a single involution on the double category of monoidal category ([13, Sec. 4.2.6]), when considered relative to the transpose involution ([13, Sec. 3.2.2]) of the 1 1 1 -category of (strict) double categories. In particular, there exists a double functor op 1 : t ( M o n C a t ) → M o n C a t \operatorname{op}_{1}:t(\mathrm{MonCat})\to\mathrm{MonCat} op 1 : t ( MonCat ) → MonCat that satisfies t ( op 1 ) ∘ op 1 = 1 t(\operatorname{op}_{1})\circ\operatorname{op}_{1}=1 t ( op 1 ) ∘ op 1 = 1 or equivalently op 1 ∘ t ( op 1 ) = 1 \operatorname{op}_{1}\circ t(\operatorname{op}_{1})=1 op 1 ∘ t ( op 1 ) = 1 . In particular, we have the following equivalent diagram of morphisms of involutions:
t ( S y m M o n C a t ) → t ( U ) t ( B r M o n C a t ) → t ( U ) t ( M o n C a t ) ↓ op 1 ↓ op 1 ↓ op 1 S y m M o n C a t → U B r M o n C a t → U M o n C a t . \begin{aligned}
t(\mathrm{SymMonCat})&\xrightarrow{t(U)}t(\mathrm{BrMonCat})\xrightarrow{t(U)}t(\mathrm{MonCat})\\
\mathop{\downarrow}\limits^{\operatorname{op}_{1}}&&\mathop{\downarrow}\limits^{\operatorname{op}_{1}}&&\mathop{\downarrow}\limits^{\operatorname{op}_{1}}\\
\mathrm{SymMonCat}&\xrightarrow{U}\mathrm{BrMonCat}\xrightarrow{U}\mathrm{MonCat}.
\end{aligned} t ( SymMonCat ) ↓ op 1 SymMonCat t ( U ) t ( BrMonCat ) t ( U ) t ( MonCat ) U BrMonCat U MonCat . ↓ op 1 ↓ op 1
The other important duality is that of reversal of a monoidal category. In particular, for each monoidal category ( V , ⊗ , I ) (\mathcal{V},\otimes,I) ( V , ⊗ , I ) , there exists a corresponding reversed monoidal category ( V , ⊗ r , I ) (\mathcal{V},\otimes_{r},I) ( V , ⊗ r , I ) , which is denoted by r ( V ) r(\mathcal{V}) r ( V ) . The monoidal product of r ( V ) r(\mathcal{V}) r ( V ) is defined using the composite V × V → s V × V → ⊗ V \mathcal{V}\times\mathcal{V}\xrightarrow{s}\mathcal{V}\times\mathcal{V}\xrightarrow{\otimes}\mathcal{V} V × V s V × V ⊗ V , and the monoidal unit remains the same. The associator is given by ( a r ( V ) ) x , y , z = ( a V ) z , y , x − 1 (a_{r(\mathcal{V})})_{x,y,z}=(a_{\mathcal{V}})^{-1}_{z,y,x} ( a r ( V ) ) x , y , z = ( a V ) z , y , x − 1 , and the unitors by λ r ( V ) = ρ V \lambda_{r(\mathcal{V})}=\rho_{\mathcal{V}} λ r ( V ) = ρ V , ρ r ( V ) = λ V \rho_{r(\mathcal{V})}=\lambda_{\mathcal{V}} ρ r ( V ) = λ V .
The reversal construction r r r also extends to an involution. That is, r : M o n C a t x → M o n C a t x r:\mathrm{MonCat}_{x}\to\mathrm{MonCat}_{x} r : MonCat x → MonCat x is a 2 2 2 -functor which satisfy r 2 = 1 r^{2}=1 r 2 = 1 . Furthermore, this duality restricts to the 2 2 2 -categories B r M o n C a t x \mathrm{BrMonCat}_{x} BrMonCat x and S y m M o n C a t x \mathrm{SymMonCat}_{x} SymMonCat x appropriately, as before. That is, there are commuting squares
S y m M o n C a t x → U B r M o n C a t x → U M o n C a t x ↓ r ↓ r ↓ r S y m M o n C a t x → U B r M o n C a t x → U M o n C a t x , \begin{aligned}
\mathrm{SymMonCat}_{x}&\xrightarrow{U}\mathrm{BrMonCat}_{x}\xrightarrow{U}\mathrm{MonCat}_{x}\\
\mathop{\downarrow}\limits^{r}&&\mathop{\downarrow}\limits^{r}&&\mathop{\downarrow}\limits^{r}\\
\mathrm{SymMonCat}_{x}&\xrightarrow{U}\mathrm{BrMonCat}_{x}\xrightarrow{U}\mathrm{MonCat}_{x},
\end{aligned} SymMonCat x ↓ r SymMonCat x U BrMonCat x U MonCat x U BrMonCat x U MonCat x , ↓ r ↓ r
representing the corresponding morphisms of involutions. Here, r r r sends a braiding b x , y b_{x,y} b x , y to the reversed braiding b y , x b_{y,x} b y , x .
Remark 4.1.3. Analogous double categorical extension to MonCat (and variants) can also be formulated for the reversal duality as relative to the identity involution of strict double categories.
Transport of monoids in virtual double categories
It is well known that the category of monoids in a monoidal category can be identified using the isomorphisms M o n ( V ) ≅ M o n C a t l ( ∗ , V ) \mathrm{Mon}(\mathcal{V})\cong\mathrm{MonCat}_{l}(\ast,\mathcal{V}) Mon ( V ) ≅ MonCat l ( ∗ , V ) . Similarly, commutative monoids in a braided monoidal category can be identified using c M o n ( V ) ≅ B r M o n C a t l ( ∗ , V ) \mathrm{cMon}(\mathcal{V})\cong\mathrm{BrMonCat}_{l}(\ast,\mathcal{V}) cMon ( V ) ≅ BrMonCat l ( ∗ , V ) . And if V \mathcal{V} V is symmetric, we have further isomorphisms B r M o n C a t l ( ∗ , V ) ≅ S y m M o n C a t l ( ∗ , V ) \mathrm{BrMonCat}_{l}(\ast,\mathcal{V})\cong\mathrm{SymMonCat}_{l}(\ast,\mathcal{V}) BrMonCat l ( ∗ , V ) ≅ SymMonCat l ( ∗ , V ) . Our goal is to understand various ways to transport monoids and comonoids using these identifications. As mentioned before, we also wish to describe these constructions within the framework of generalized involutions/inverses, with the main motivation being the intuition for what to expect in higher dimensions.
We start with defining the op duality functors between monoids and comonoids for a given V \mathcal{V} V .
Definition 4.2.1. For each monoidal category V \mathcal{V} V , define a 1-functor op V : op 1 ( CoMon ( V ) ) → Mon ( op 1 ( V ) ) \operatorname{op}_{\mathcal{V}} \colon\operatorname{op}_{1}(\operatorname{CoMon}(\mathcal{V})) \to\operatorname{Mon}(\operatorname{op}_{1}(\mathcal{V})) op V : op 1 ( CoMon ( V )) → Mon ( op 1 ( V )) using the commutative diagram
op 1 ( CoMon ( V ) ) → op V Mon ( op 1 ( V ) ) ≅ ↓ ↓ ≅ op 1 ( MonCat c ( ∗ , V ) ) → op 1 ( ≅ ) MonCat l ( ∗ , op 1 ( V ) ) . \begin{CD}
\operatorname{op}_{1}(\operatorname{CoMon}(\mathcal{V})) @>{\operatorname{op}_{\mathcal{V}}}>> \operatorname{Mon}(\operatorname{op}_{1}(\mathcal{V})) \\
@V{\cong}VV @VV{\cong}V \\
\operatorname{op}_{1}(\operatorname{MonCat}_{c}(*,\mathcal{V})) @>{\operatorname{op}_{1}(\cong)}>> \operatorname{MonCat}_{l}(*,\operatorname{op}_{1}(\mathcal{V})) .
\end{CD} op 1 ( CoMon ( V )) ≅ ↓ ⏐ op 1 ( MonCat c ( ∗ , V )) op V op 1 ( ≅ ) Mon ( op 1 ( V )) ↓ ⏐ ≅ MonCat l ( ∗ , op 1 ( V )) . The dual version op V t : op 1 ( Mon ( V ) ) → CoMon ( op 1 ( V ) ) \operatorname{op}_{\mathcal{V}}^{t} \colon\operatorname{op}_{1}(\operatorname{Mon}(\mathcal{V})) \to\operatorname{CoMon}(\operatorname{op}_{1}(\mathcal{V})) op V t : op 1 ( Mon ( V )) → CoMon ( op 1 ( V )) is defined similarly.
We already observe that there is no obvious way to compose op V \operatorname{op}_{\mathcal{V}} op V and op V t \operatorname{op}_{\mathcal{V}}^{t} op V t to produce the identity map as all inverses do. At the same time, it is clear that this is a form of duality. The primary issue is that we are not incorporating the functoriality in V \mathcal{V} V . In particular, both CoMon ( − ) \operatorname{CoMon}(-) CoMon ( − ) and Mon ( − ) \operatorname{Mon}(-) Mon ( − ) are strict 2-functors as they are defined by a corepresentation. Therefore, we ought to extend both op V \operatorname{op}_{\mathcal{V}} op V and op V t \operatorname{op}_{\mathcal{V}}^{t} op V t to strict 2-natural transformations.
Lemma 4.2.2. The functors op V \operatorname{op}_{\mathcal{V}} op V and op V t \operatorname{op}_{\mathcal{V}}^{t} op V t form the components of strict 2-natural transformations
op 2 ( MonCat c ) → op 2 ( CoMon ( − ) ) op 2 ( Cat ) op 1 ↓ ↓ op 1 MonCat l → Mon ( − ) Cat op 2 ( MonCat l ) → op 2 ( Mon ( − ) ) op 2 ( Cat ) op 1 t ↓ ↓ op 1 MonCat c → CoMon ( − ) Cat . \begin{CD}
\operatorname{op}_{2}(\operatorname{MonCat}_{c}) @>{\operatorname{op}_{2}(\operatorname{CoMon}(-))}>> \operatorname{op}_{2}(\operatorname{Cat}) \\
@V{\operatorname{op}_{1}}VV @VV{\operatorname{op}_{1}}V \\
\operatorname{MonCat}_{l} @>{\operatorname{Mon}(-)}>> \operatorname{Cat}
\end{CD}
\qquad
\begin{CD}
\operatorname{op}_{2}(\operatorname{MonCat}_{l}) @>{\operatorname{op}_{2}(\operatorname{Mon}(-))}>> \operatorname{op}_{2}(\operatorname{Cat}) \\
@V{\operatorname{op}_{1}^{t}}VV @VV{\operatorname{op}_{1}}V \\
\operatorname{MonCat}_{c} @>{\operatorname{CoMon}(-)}>> \operatorname{Cat} .
\end{CD} op 2 ( MonCat c ) op 1 ↓ ⏐ MonCat l op 2 ( CoMon ( − )) Mon ( − ) op 2 ( Cat ) ↓ ⏐ op 1 Cat op 2 ( MonCat l ) op 1 t ↓ ⏐ MonCat c op 2 ( Mon ( − )) CoMon ( − ) op 2 ( Cat ) ↓ ⏐ op 1 Cat . Furthermore, these two transformations form a morphism of inverses relative to the strict 2-involution op 2 : 2 Cat ≤ 2 → 2 Cat ≤ 2 \operatorname{op}_{2} \colon2\operatorname{Cat}_{\leq2} \to2\operatorname{Cat}_{\leq2} op 2 : 2 Cat ≤ 2 → 2 Cat ≤ 2 .
Proof. The 2-naturality and strictness of the two collection of maps op V \operatorname{op}_{\mathcal{V}} op V and op V t \operatorname{op}_{\mathcal{V}}^{t} op V t is a direct consequence of their definition in terms of the corepresentation by the terminal monoidal category.
The two axioms for morphism of inverses are similarly straightforward. For instance, one of these axioms can be verified as indicated below:
M o n C a t c → C o M o n ( − ) C a t M o n C a t c ↓ ↓ C o M o n ( − ) op 2 2 ( M o n C a t c ) → op 2 2 ( C o M o n ( − ) ) op 2 2 ( C a t ) C a t ↓ op 2 ( op 1 ) ↓ 1 C a t op 2 ( M o n C a t l ) → op 2 ( M o n ( − ) ) op 2 ( C a t ) ≡ ↓ op 1 t ↓ op 1 M o n C a t c → C o M o n ( − ) C a t C a t op 1 ( M o n op 1 ( V ) ) → op op 1 t V C o M o n ( op 1 2 ( V ) ) ↑ op V ↓ C o M o n ( 1 V ) C o M o n ( V ) → 1 C o M o n ( V ) C o M o n ( V ) \begin{aligned}
&\mathrm{MonCat}_{c}\xrightarrow{\mathrm{CoMon}(-)}\mathrm{Cat}
&&\mathrm{MonCat}_{c}\\
&\quad\downarrow&&\quad\downarrow\scriptstyle{\mathrm{CoMon}(-)}\\
&\operatorname{op}_{2}^{2}(\mathrm{MonCat}_{c})\xrightarrow{\operatorname{op}_{2}^{2}(\mathrm{CoMon}(-))}\operatorname{op}_{2}^{2}(\mathrm{Cat})
&&\mathrm{Cat}\\
&\quad\downarrow\scriptstyle{\operatorname{op}_{2}(\operatorname{op}_{1})}
&&\quad\downarrow\scriptstyle{1_{\mathrm{Cat}}}\\
&\operatorname{op}_{2}(\mathrm{MonCat}_{l})\xrightarrow{\operatorname{op}_{2}(\mathrm{Mon}(-))}\operatorname{op}_{2}(\mathrm{Cat})
&&\equiv\\
&\quad\downarrow\scriptstyle{\operatorname{op}_{1}^{t}}
&&\quad\downarrow\scriptstyle{\operatorname{op}_{1}}\\
&\mathrm{MonCat}_{c}\xrightarrow{\mathrm{CoMon}(-)}\mathrm{Cat}
&&\mathrm{Cat}\\
&\operatorname{op}_{1}(\mathrm{Mon}_{\operatorname{op}_{1}(\mathcal V)})\xrightarrow{\operatorname{op}_{\operatorname{op}_{1}^{t}}\mathcal V}\mathrm{CoMon}(\operatorname{op}_{1}^{2}(\mathcal V))\\
&\quad\uparrow\scriptstyle{\operatorname{op}\mathcal V}
&&\quad\downarrow\scriptstyle{\mathrm{CoMon}(1_{\mathcal V})}\\
&\mathrm{CoMon}(\mathcal V)\xrightarrow{1_{\mathrm{CoMon}(\mathcal V})}\mathrm{CoMon}(\mathcal V)
\end{aligned} MonCat c CoMon ( − ) Cat ↓ op 2 2 ( MonCat c ) op 2 2 ( CoMon ( − )) op 2 2 ( Cat ) ↓ op 2 ( op 1 ) op 2 ( MonCat l ) op 2 ( Mon ( − )) op 2 ( Cat ) ↓ op 1 t MonCat c CoMon ( − ) Cat op 1 ( Mon op 1 ( V ) ) op op 1 t V CoMon ( op 1 2 ( V )) ↑ op V CoMon ( V ) 1 CoMon ( V ) CoMon ( V ) MonCat c ↓ CoMon ( − ) Cat ↓ 1 Cat ≡ ↓ op 1 Cat ↓ CoMon ( 1 V )
The phenomenon of transport across r r r and i i i is more interesting as considering the braiding or symmetry provides additional transport structure unlike op 1 \operatorname{op}_{1} op 1 . First, consider that V \mathcal V V is just an arbitrary monoidal category. In this case, we only have the expected reversal duality.
Lemma 4.2.3. There exists a strict 2-natural transformation
M o n C a t l → M o n ( − ) C a t r ↓ ↓ 1 M o n C a t l → M o n ( − ) C a t \begin{CD}
\mathrm{MonCat}_{l} @>{\mathrm{Mon}(-)}>> \mathrm{Cat}\\
@V{r}VV @VV{1}V\\
\mathrm{MonCat}_{l} @>{\mathrm{Mon}(-)}>> \mathrm{Cat}
\end{CD} MonCat l r ↓ ⏐ MonCat l Mon ( − ) Mon ( − ) Cat ↓ ⏐ 1 Cat which defines a morphism of involutions relative to the identity functor 1 : 2 C a t ≤ 2 → 2 C a t ≤ 2 1:2\mathrm{Cat}_{\leq2}\to2\mathrm{Cat}_{\leq2} 1 : 2 Cat ≤ 2 → 2 Cat ≤ 2 .
Now take V \mathcal V V to be a braided monoidal category. The key distinction from op 1 \operatorname{op}_{1} op 1 duality is that B r M o n C a t \mathrm{BrMonCat} BrMonCat admits two distinct non-trivial 2-cells
B r M o n C a t l → 1 B r M o n C a t l b ↓ ↑ r \begin{CD}
\mathrm{BrMonCat}_{l} @>{1}>> \mathrm{BrMonCat}_{l}\\
@V{b}VV @AA{r}A
\end{CD} BrMonCat l b ↓ ⏐ 1 BrMonCat l ⏐ ↑ r
b V : ( r ( V ) , b y , x ) → ( 1 , b x , y , 1 ) ( V , b x , y ) b_{\mathcal V}\colon(r(\mathcal V),b_{y,x})\xrightarrow{(1,b_{x,y},1)}(\mathcal V,b_{x,y}) b V : ( r ( V ) , b y , x ) ( 1 , b x , y , 1 ) ( V , b x , y )
B r M o n C a t l → 1 B r M o n C a t l b − 1 ↓ ↑ r \begin{CD}
\mathrm{BrMonCat}_{l} @>{1}>> \mathrm{BrMonCat}_{l}\\
@V{b^{-1}}VV @AA{r}A
\end{CD} BrMonCat l b − 1 ↓ ⏐ 1 BrMonCat l ⏐ ↑ r
b V − 1 : ( r ( V ) , b y , x ) → ( 1 , b y , x − 1 , 1 ) ( V , b x , y ) b_{\mathcal V}^{-1}\colon(r(\mathcal V),b_{y,x})\xrightarrow{(1,b^{-1}_{y,x},1)}(\mathcal V,b_{x,y}) b V − 1 : ( r ( V ) , b y , x ) ( 1 , b y , x − 1 , 1 ) ( V , b x , y )
between the identity functor and the reversal functor. And, these 2-cell b b b and b − 1 b^{-1} b − 1 are part of a morphism of inverses, albeit degenerately,
( V , b x , y ) B r M o n C a t l r ( b V ) : ( 1 , b y , x , 1 ) ↓ ↑ 1 ( r ( V ) , b y , x ) B r M o n C a t l b V − 1 : ( 1 , b y , x − 1 , 1 ) ↓ ↑ b − 1 ( V , b x , y ) B r M o n C a t l \begin{CD}
(\mathcal V,b_{x,y}) @. \mathrm{BrMonCat}_{l}\\
@V{r(b_{\mathcal V})\colon(1,b_{y,x},1)}VV @AA{1}A\\
(r(\mathcal V),b_{y,x}) @. \mathrm{BrMonCat}_{l}\\
@V{b_{\mathcal V}^{-1}\colon(1,b^{-1}_{y,x},1)}VV @AA{b^{-1}}A\\
(\mathcal V,b_{x,y}) @. \mathrm{BrMonCat}_{l}
\end{CD} ( V , b x , y ) r ( b V ) : ( 1 , b y , x , 1 ) ↓ ⏐ ( r ( V ) , b y , x ) b V − 1 : ( 1 , b y , x − 1 , 1 ) ↓ ⏐ ( V , b x , y ) BrMonCat l ⏐ ↑ 1 BrMonCat l ⏐ ↑ b − 1 BrMonCat l
Expectedly, if V \mathcal V V is a symmetric monoidal category then both the 2-cells b b b and b − 1 b^{-1} b − 1 are identical and become part of a morphism of involutions. In general, we can compose this morphism of inverses ( b , b − 1 ) (b,b^{-1}) ( b , b − 1 ) to define other variants of the reversal duality 2-cell, such as:
B r M o n C a t l → U M o n C a t l → Mon ( − ) C a t 1 ↓ r ↓ ↑ r B r M o n C a t l → U M o n C a t l → Mon ( − ) C a t ↑ 1 \begin{CD}
\mathrm{BrMonCat}_{l} @>{U}>> \mathrm{MonCat}_{l} @>{\operatorname{Mon}(-)}>> \mathrm{Cat}\\
@V{1}VV @V{r}VV @AA{r}A\\
\mathrm{BrMonCat}_{l} @>{U}>> \mathrm{MonCat}_{l} @>{\operatorname{Mon}(-)}>> \mathrm{Cat}\\
@. @. @AA{1}A
\end{CD} BrMonCat l 1 ↓ ⏐ BrMonCat l U U MonCat l r ↓ ⏐ MonCat l Mon ( − ) Mon ( − ) Cat ⏐ ↑ r Cat ⏐ ↑ 1
Mon ( V ) → r V Mon ( r ( V ) ) → Mon ( b V ) Mon ( V ) . \operatorname{Mon}(\mathcal V)\xrightarrow{r_{\mathcal V}}\operatorname{Mon}(r(\mathcal V))\xrightarrow{\operatorname{Mon}(b_{\mathcal V})}\operatorname{Mon}(\mathcal V). Mon ( V ) r V Mon ( r ( V )) Mon ( b V ) Mon ( V ) .
To illustrate the distinction between these variants of reversal, we provide the following example.
Example 4.2.4. Let V \mathcal V V be the symmetric monoidal category S e t \mathrm{Set} Set with the symmetry given by the swap function s x , y : x × y → y × x s_{x,y}\colon x\times y\to y\times x s x , y : x × y → y × x . In this case, the component b S e t b_{\mathrm{Set}} b Set is given by the braided monoidal functor ( 1 S e t , s x , y , 1 ) : ( r ( S e t ) , s y , x ) → ( S e t , s x , y ) (1_{\mathrm{Set}},s_{x,y},1)\colon(r(\mathrm{Set}),s_{y,x})\to(\mathrm{Set},s_{x,y}) ( 1 Set , s x , y , 1 ) : ( r ( Set ) , s y , x ) → ( Set , s x , y ) . The morphism r S e t r_{\mathrm{Set}} r Set merely reinterprets a given monoid M ∈ Mon ( S e t ) M\in\operatorname{Mon}(\mathrm{Set}) M ∈ Mon ( Set ) as a monoid in Mon ( r ( S e t ) ) \operatorname{Mon}(r(\mathrm{Set})) Mon ( r ( Set )) in the sense depicted by
M × M M × r M μ ↓ ↓ μ M = M \begin{CD}
M\times M @. M\times_{r}M\\
@V{\mu}VV @VV{\mu}V\\
M @= M
\end{CD} M × M μ ↓ ⏐ M M × r M ↓ ⏐ μ M
Thus, we should view the morphism r S e t r_{\mathrm{Set}} r Set as a comprehension duality. In contrast, the morphism M o n ( b S e t ) \mathbf{Mon}(\mathbf{bSet}) Mon ( bSet ) then transfers this monoid back to M o n ( S e t ) \mathbf{Mon}(\mathbf{Set}) Mon ( Set ) by precomposing with the symmetry as in
M × M → s M , M M × r M μ ↓ M × r M → M \begin{CD}
M \times M @>{s_{M,M}}>> M \times_r M \\
@. @V{\mu}VV \\
M \times_r M @>{}>> M
\end{CD} M × M M × r M s M , M M × r M μ ↓ ⏐ M
The key takeaway is that one could replace M o n \mathbf{Mon} Mon with P s M o n \mathbf{PsMon} PsMon and S e t \mathbf{Set} Set with C a t \mathbf{Cat} Cat to draw similar observation about the duality r : M o n C a t → M o n C a t r : \mathbf{MonCat} \to\mathbf{MonCat} r : MonCat → MonCat itself. We discuss this later in section 4.3.
A similar theory of dualities can be developed for arbitrary bimodules in a monoidal category. In particular, we could define a 2-functor M o d ( − ) \mathbf{Mod}(-) Mod ( − ) that takes the input as a monoidal category and outputs the bicategory of bimodules over arbitrary monoids (as long as we restrict our attention to monoidally cocomplete categories). However, this is rather unwieldy and introduces unnecessary assumptions. A more natural framework to encode these dualities is that of virtual double categories. See [9, Defs. 2.1, 3.1, and 7.6] for a complete account and also see [26, Ch. 5] for our conventions on horizontal 1-cells. In particular, we shall generally read horizontal 1-cells M : A → B M : A \to B M : A → B as ( B , A ) (B,A) ( B , A ) -bimodule.
Note that in the original definition, one can vertically reverse the direction of the square multicell to obtain the dual notion of a “co”-virtual double category. As the corresponding definitions are apparent in this dual case, we shall not write them explicitly.
The only non-trivial result which we requires is that v c o D b l C a t \mathbf{vcoDblCat} vcoDblCat and v D b l C a t \mathbf{vDblCat} vDblCat exhibits a vertical duality involution in the following sense.
Lemma 4.2.5. There exist two 2-functors
op V : op 2 ( v c o D b l C a t ) → v D b l C a t \operatorname{op}_{V} : \operatorname{op}_{2}(\mathbf{vcoDblCat}) \to\mathbf{vDblCat} op V : op 2 ( vcoDblCat ) → vDblCat and
op V t : op 2 ( v D b l C a t ) → v c o D b l C a t , \operatorname{op}_{V}^{t} : \operatorname{op}_{2}(\mathbf{vDblCat}) \to\mathbf{vcoDblCat}, op V t : op 2 ( vDblCat ) → vcoDblCat , which form a morphism of inverses relative to the op 2 \operatorname{op}_{2} op 2 involution.
Proof. we shall mainly define the action of op V \operatorname{op}_{V} op V on each co-virtual double categories V \mathcal{V} V . The rest is technical but straightforward. The data of op V ( V ) \operatorname{op}_{V}(\mathcal{V}) op V ( V ) is given by
Objects
Table 1.
V 1-cell
V op V ( V ) X ↑ X Y Y \begin{array}{ccc}
\mathcal{V} && \operatorname{op}_{\mathcal{V}}(\mathcal{V})\\
X & \uparrow& X\\
Y && Y
\end{array} V X Y ↑ op V ( V ) X Y
H 1-cell
V op V ( V ) X ⟶ Y X ⟶ Y \begin{array}{ccc}
\mathcal{V} && \operatorname{op}_{\mathcal{V}}(\mathcal{V})\\
X & \longrightarrow& Y\\
X & \longrightarrow& Y
\end{array} V X X ⟶ ⟶ op V ( V ) Y Y
Sq multicell
V op V ( V ) X 0 ⟶ ⋯ ⟶ X n X 0 ⟶ ⋯ ⟶ X n Y 0 ⟶ Y 1 Y 0 ⟶ Y 1 \begin{array}{ccc}
\mathcal{V} && \operatorname{op}_{\mathcal{V}}(\mathcal{V})\\
X_{0}\longrightarrow\cdots\longrightarrow X_{n} &&
X_{0}\longrightarrow\cdots\longrightarrow X_{n}\\
Y_{0}\longrightarrow Y_{1} &&
Y_{0}\longrightarrow Y_{1}
\end{array} V X 0 ⟶ ⋯ ⟶ X n Y 0 ⟶ Y 1 op V ( V ) X 0 ⟶ ⋯ ⟶ X n Y 0 ⟶ Y 1
All of the data on the right is formal data similar to how the category C op \mathcal{C}^{\operatorname{op}} C op is defined when given a category C \mathcal{C} C . Therefore, we simply induce the compositions and units of op V ( V ) \operatorname{op}_{\mathcal{V}}(\mathcal{V}) op V ( V ) from V \mathcal{V} V itself in an analogous manner.
More technically, if the underlying fc-graph of a virtual double category V \mathcal{V} V is given by T ( C 0 ) ← s C 1 → t C 0 T(\mathcal{C}_{0}) \xleftarrow{s} \mathcal{C}_{1} \xrightarrow{t} \mathcal{C}_{0} T ( C 0 ) s C 1 t C 0 , we can obtain a dual co-fc-graph given by T ( C 0 ) ← t ′ C 1 → s ′ C 0 T(\mathcal{C}_{0}) \xleftarrow{t'} \mathcal{C}_{1} \xrightarrow{s'} \mathcal{C}_{0} T ( C 0 ) t ′ C 1 s ′ C 0 with t ′ = s t'=s t ′ = s and s ′ = t s'=t s ′ = t . □ \square □
We briefly recall that there exists a chain of embeddings MonCat l ↪ ( Bicat 2 ) l ↪ vDblCat \operatorname{MonCat}_{l} \hookrightarrow(\operatorname{Bicat}_{2})_{l} \hookrightarrow\operatorname{vDblCat} MonCat l ↪ ( Bicat 2 ) l ↪ vDblCat . The first one is well known and simply identifies one object bicategories with monoidal categories [25]. The second one, we can describe as follows:
B ⟶ B \mathbb{B}\longrightarrow\mathbb{B} B ⟶ B
x 0 → f 1 x 1 → f 2 x 2 ⟶ ⋯ ⟶ x n ↓ ε x 0 ↓ ε x n x 0 ⟶ x n α : ( f n ⋯ f 2 ) f 1 ⇒ g 1 ⇓ g 1 \begin{array}{c}
x_{0}\xrightarrow{f_{1}}x_{1}\xrightarrow{f_{2}}x_{2}\longrightarrow\cdots\longrightarrow x_{n}\\
\downarrow\varepsilon_{x_{0}}\qquad\qquad\qquad\qquad\qquad\qquad\downarrow\varepsilon_{x_{n}}\\
x_{0}\longrightarrow\qquad\qquad\qquad\qquad\qquad\qquad x_{n}\\
\qquad\qquad\alpha:(f_{n}\cdots f_{2})f_{1}\Rightarrow g_{1}\\
\qquad\qquad\qquad\qquad\Downarrow\\
\qquad\qquad\qquad\qquad g_{1}
\end{array} x 0 f 1 x 1 f 2 x 2 ⟶ ⋯ ⟶ x n ↓ ε x 0 ↓ ε x n x 0 ⟶ x n α : ( f n ⋯ f 2 ) f 1 ⇒ g 1 ⇓ g 1
e ( f n , f n − 1 , … , f 1 ) e(f_n,f_{n-1},\ldots,f_1) e ( f n , f n − 1 , … , f 1 )
α : ( f n ⋯ f 2 ) f 1 ⇒ g 1 \alpha:(f_n\cdots f_2)f_1\Rightarrow g_1 α : ( f n ⋯ f 2 ) f 1 ⇒ g 1
An analogous embedding M o n C a t c ↪ ( B i c a t 2 ) c ↪ v c o D b l C a t \mathbf{MonCat}_{c}\hookrightarrow(\mathbf{Bicat}_{2})_{c}\hookrightarrow\mathbf{vcoDblCat} MonCat c ↪ ( Bicat 2 ) c ↪ vcoDblCat holds for the colax case. The first one is again clear. For the second one, we have the similar following description:
B ⟶ B x 0 → f 1 x 1 → f 2 x 2 ⟶ ⋯ ⟶ x n x 0 ⟶ g 1 g 1 x n , α : g 1 ⇒ ( f n ⋯ f 2 ) f 1 , ε x 0 , ε x n \begin{aligned}
&B \longrightarrow\mathbb{B}\\
&x_{0}\xrightarrow{f_{1}}x_{1}\xrightarrow{f_{2}}x_{2}\longrightarrow\cdots\longrightarrow x_{n}\\
&\phantom{x_{0}}\underset{g_{1}}{\underset{\phantom{g_{1}}}{\longrightarrow}}x_{n},\qquad
\alpha:g_{1}\Rightarrow(f_{n}\cdots f_{2})f_{1},\qquad
\varepsilon_{x_{0}},\varepsilon_{x_{n}}
\end{aligned} B ⟶ B x 0 f 1 x 1 f 2 x 2 ⟶ ⋯ ⟶ x n x 0 g 1 g 1 ⟶ x n , α : g 1 ⇒ ( f n ⋯ f 2 ) f 1 , ε x 0 , ε x n
The important fact about these two embeddings is that they canonically extend to morphisms of involutions and morphisms of inverses.
Lemma 4.2.6.
There exist morphisms of inverses whose one direction is given by
op 2 ( M o n C a t c ) ⟶ op 2 ( B i c a t 2 ) c ⟶ op 2 ( v c o D b l C a t ) op 2 ( M o n C a t c ) ↓ op 1 ↓ op 2 ↓ op V M o n C a t l ⟶ ( B i c a t 2 ) l ⟶ v D b l C a t \begin{aligned}
\operatorname{op}_{2}(\mathbf{MonCat}_{c})&\longrightarrow\operatorname{op}_{2}(\mathbf{Bicat}_{2})_{c}\longrightarrow\operatorname{op}_{2}(\mathbf{vcoDblCat})\\
\mathrel{\phantom{\operatorname{op}_{2}(\mathbf{MonCat}_{c})}}\downarrow\operatorname{op}_{1}\qquad\qquad\qquad\qquad\downarrow\operatorname{op}_{2}\qquad\qquad\qquad\qquad\downarrow\operatorname{op}_{V}\\
\mathbf{MonCat}_{l}&\longrightarrow(\mathbf{Bicat}_{2})_{l}\longrightarrow\mathbf{vDblCat}
\end{aligned} op 2 ( MonCat c ) op 2 ( MonCat c ) ↓ op 1 ↓ op 2 ↓ op V MonCat l ⟶ op 2 ( Bicat 2 ) c ⟶ op 2 ( vcoDblCat ) ⟶ ( Bicat 2 ) l ⟶ vDblCat and whose dual generalized inverse is given similarly.
There exist morphisms of involution as given by
M o n C a t l ⟶ ( B i c a t 2 ) l ⟶ v D b l C a t M o n C a t l ↓ r ↓ op 1 ↓ op H M o n C a t l ⟶ ( B i c a t 2 ) l ⟶ v D b l C a t \begin{aligned}
\mathbf{MonCat}_{l}&\longrightarrow(\mathbf{Bicat}_{2})_{l}\longrightarrow\mathbf{vDblCat}\\
\mathrel{\phantom{\mathbf{MonCat}_{l}}}\downarrow r\qquad\qquad\qquad\qquad\downarrow\operatorname{op}_{1}\qquad\qquad\qquad\qquad\downarrow\operatorname{op}_{H}\\
\mathbf{MonCat}_{l}&\longrightarrow(\mathbf{Bicat}_{2})_{l}\longrightarrow\mathbf{vDblCat}
\end{aligned} MonCat l MonCat l ↓ r ↓ op 1 ↓ op H MonCat l ⟶ ( Bicat 2 ) l ⟶ vDblCat ⟶ ( Bicat 2 ) l ⟶ vDblCat
Note that the 2-cell appearing in the involution case primarily arises from the fact that ( ( f n ∘ f n − 1 ) ∘ ⋯ ) ∘ f 1 ≠ f n ∘ ( f n − 1 ∘ ( ⋯ ∘ f 1 ) ) ((f_{n}\circ f_{n-1})\circ\cdots)\circ f_{1}\ne f_{n}\circ(f_{n-1}\circ(\cdots\circ f_{1})) (( f n ∘ f n − 1 ) ∘ ⋯ ) ∘ f 1 = f n ∘ ( f n − 1 ∘ ( ⋯ ∘ f 1 )) in an arbitrary bicategory. To illustrate this issue further, consider a left-biased lax monoidal category V \mathcal{V} V , that is, the associator need not an invertible 2-cell. Applying r r r to this lax monoidal category doesn’t yield a left-biased lax monoidal category, as there is no α − 1 \alpha^{-1} α − 1 available. At best, we obtain a canonical right-biased lax monoidal category. All of this suggests that a more thorough approach would require generalized inverses r : M o n C a t l L b → M o n C a t l R b r:\mathbf{MonCat}_{l}^{Lb}\to\mathbf{MonCat}_{l}^{Rb} r : MonCat l L b → MonCat l R b and r t : M o n C a t l R b → M o n C a t l L b r^{t}:\mathbf{MonCat}_{l}^{Rb}\to\mathbf{MonCat}_{l}^{Lb} r t : MonCat l R b → MonCat l L b . However, we are content with our approach here.
We have now established that monoidal categories are a special case of a virtual double categories, as well as co-virtual double categories. This suggests that the dualities of monoidal categories should also extend to the virtual double categorical case. We begin by characterization the op 1 \operatorname{op}_{1} op 1 duality.
Theorem 4.2.7. The 2-functors M o d : v D b l C a t → v D b l C a t \mathrm{Mod}\colon\mathrm{vDblCat}\to\mathrm{vDblCat} Mod : vDblCat → vDblCat and C o M o d : v c o D b l C a t → v c o D b l C a t \mathrm{CoMod}\colon\mathrm{vcoDblCat}\to\mathrm{vcoDblCat} CoMod : vcoDblCat → vcoDblCat extend to a 1-morphism of relative inverses by the strict 2-natural transformations
op 2 ( v c o D b l C a t ) → op 2 ( C o M o d ( − ) ) op 2 ( v c o D b l C a t ) op V ↓ ↓ op V v D b l C a t → M o d ( − ) v D b l C a t \begin{CD}
\operatorname{op}_{2}(\mathrm{vcoDblCat}) @>{\operatorname{op}_{2}(\mathrm{CoMod}(-))}>> \operatorname{op}_{2}(\mathrm{vcoDblCat}) \\
@V{\operatorname{op}_{V}}VV @VV{\operatorname{op}_{V}}V \\
\mathrm{vDblCat} @>{\mathrm{Mod}(-)}>> \mathrm{vDblCat}
\end{CD} op 2 ( vcoDblCat ) op V ↓ ⏐ vDblCat op 2 ( CoMod ( − )) Mod ( − ) op 2 ( vcoDblCat ) ↓ ⏐ op V vDblCat and
op 2 ( v D b l C a t ) → op 2 ( M o d ( − ) ) op 2 ( v D b l C a t ) op V t ↓ ↓ op V t v c o D b l C a t → C o M o d ( − ) v c o D b l C a t . \begin{CD}
\operatorname{op}_{2}(\mathrm{vDblCat}) @>{\operatorname{op}_{2}(\mathrm{Mod}(-))}>> \operatorname{op}_{2}(\mathrm{vDblCat}) \\
@V{\operatorname{op}_{V}^{t}}VV @VV{\operatorname{op}_{V}^{t}}V \\
\mathrm{vcoDblCat} @>{\mathrm{CoMod}(-)}>> \mathrm{vcoDblCat}.
\end{CD} op 2 ( vDblCat ) op V t ↓ ⏐ vcoDblCat op 2 ( Mod ( − )) CoMod ( − ) op 2 ( vDblCat ) ↓ ⏐ op V t vcoDblCat .
Proof. As usual, we need to define the components op V : op V ( C o M o d ( V ) ) → M o d ( op V ( V ) ) \operatorname{op}_{V}\colon\operatorname{op}_{V}(\mathrm{CoMod}(\mathcal{V}))\to\mathrm{Mod}(\operatorname{op}_{V}(\mathcal{V})) op V : op V ( CoMod ( V )) → Mod ( op V ( V )) .
Objects
C : = ( ∙ , C , μ , e ) ⟶ C o p : = ( ∙ , C , μ o p , e o p ) C := (\bullet,C,\mu,e)\longrightarrow C^{\mathrm{op}} := (\bullet,C,\mu^{\mathrm{op}},e^{\mathrm{op}}) C := ( ∙ , C , μ , e ) ⟶ C op := ( ∙ , C , μ op , e op )
\begin{CD}
\bullet@>{C}>> \bullet\\
@V{}VV @VV{}V \\
\bullet@>{C}>> \bullet
\end{CD}
\qquad
\begin{CD}
\bullet@>{C}>> \bullet\\
@V{}VV @VV{}V \\
\bullet@>{C}>> \bullet
\end{CD}
\begin{CD}
\bullet@>>> \bullet\\
@. @A{\mu}AA @. \\
\bullet@>>> \bullet
\end{CD}
\qquad
\begin{CD}
\bullet@>>> \bullet\\
@. @V{\mu^{\mathrm{op}}}VV @. \\
\bullet@>>> \bullet
\end{CD}
\begin{CD}
\bullet@>>> \bullet\\
@. @A{e}AA @. \\
\bullet@>>> \bullet
\end{CD}
\qquad
\begin{CD}
\bullet@>>> \bullet\\
@. @V{e^{\mathrm{op}}}VV @. \\
\bullet@>>> \bullet
\end{CD}
V 1-cell
\begin{CD}
\bullet@>{C}>> \bullet\\
@A{f_{0}}AA @AA{f_{0}}A \\
\star@>{D}>> \star
\end{CD}
\longrightarrow
\begin{CD}
\bullet@>{C^{\mathrm{op}}}>> \bullet\\
@V{f_{0}^{\mathrm{op}}}VV @VV{f_{0}^{\mathrm{op}}}V \\
\star@>{D^{\mathrm{op}}}>> \star
\end{CD}
H 1-cell
( C , ∙ ) → X : = ( X , λ , ρ ) ( D , ⋆ ) ⟼ ( C o p , ∙ ) → X o p : = ( X , λ o p , ρ o p ) ( D o p , ⋆ ) (C,\mathbin{\bullet}) \xrightarrow{X:=(X,\lambda,\rho)} (D,\mathbin{\star})
\quad\longmapsto\quad
(C^{\mathrm{op}},\mathbin{\bullet}) \xrightarrow{X^{\mathrm{op}}:=(X,\lambda^{\mathrm{op}},\rho^{\mathrm{op}})} (D^{\mathrm{op}},\mathbin{\star}) ( C , ∙ ) X := ( X , λ , ρ ) ( D , ⋆ ) ⟼ ( C op , ∙ ) X op := ( X , λ op , ρ op ) ( D op , ⋆ )
\begin{aligned}
&\bullet\xrightarrow{X} \star
\qquad\qquad
\bullet\xrightarrow{X} \star\\[1ex]
&\begin{array}{ccccc}
\bullet& \xrightarrow{C} & \bullet& \xrightarrow{X} & \star\\
\vert& & \mathord{\uparrow}\limits_{\rho} & & \vert\\
\bullet& \xrightarrow{X} & \bullet& \xrightarrow{} & \star
\end{array}
\qquad
\begin{array}{ccccc}
\bullet& \xrightarrow{C^{\mathrm{op}}} & \bullet& \xrightarrow{X^{\mathrm{op}}} & \star\\
\vert& & \mathord{\downarrow}\limits_{\rho^{\mathrm{op}}} & & \vert\\
\bullet& \xrightarrow{X^{\mathrm{op}}} & \bullet& \xrightarrow{} & \star
\end{array} \\[2ex]
&\begin{array}{ccccc}
\bullet& \xrightarrow{X} & \star& \xrightarrow{D} & \star\\
\vert& & \mathord{\uparrow}\limits_{\lambda} & & \vert\\
\bullet& \xrightarrow{X} & \star& \xrightarrow{} & \star
\end{array}
\qquad
\begin{array}{ccccc}
\bullet& \xrightarrow{X^{\mathrm{op}}} & \star& \xrightarrow{D^{\mathrm{op}}} & \star\\
\vert& & \mathord{\downarrow}\limits_{\lambda^{\mathrm{op}}} & & \vert\\
\bullet& \xrightarrow{X^{\mathrm{op}}} & \star& \xrightarrow{} & \star
\end{array}
\end{aligned}
Sq multicell
\begin{array}{ccccccccc}
\bullet_1 & \xrightarrow{X_1} & \cdots& \xrightarrow{X_n} & \bullet_n \\
\mathord{\uparrow}\limits_f & & & \mathord{\uparrow}\limits_{\alpha} & & & \mathord{\uparrow}\limits_g \\
\bullet& \xrightarrow{Y} & \star
\end{array}
\qquad\longmapsto\qquad
\begin{array}{ccccccccc}
\bullet_1 & \xrightarrow{X_1^{\mathrm{op}}} & \cdots& \xrightarrow{X_n^{\mathrm{op}}} & \bullet_n \\
\mathord{\downarrow}\limits_{f^{\mathrm{op}}} & & & \mathord{\downarrow}\limits_{\alpha^{\mathrm{op}}} & & & \mathord{\downarrow}\limits_{g^{\mathrm{op}}} \\
\bullet& \xrightarrow{Y^{\mathrm{op}}} & \star
\end{array}
The other family of components can be defined similarly.
The case for reversal duality is exactly the same as above.
Theorem 4.2.8. The 2 2 2 -functor Mod : v D b l C a t → v D b l C a t \operatorname{Mod}\colon\mathrm{vDblCat}\to\mathrm{vDblCat} Mod : vDblCat → vDblCat extends to a 1 1 1 -morphism of involutions by the strict 2 2 2 -natural transformation
\begin{array}{ccccc}
\mathrm{vDblCat} & \xrightarrow{\operatorname{Mod}(-)} & \mathrm{vDblCat} \\
\mathord{\downarrow}\limits_{\operatorname{op}_{H}} & & \mathord{\downarrow}\limits_{\operatorname{op}_{H}} \\
\mathrm{vDblCat} & \xrightarrow{\operatorname{Mod}(-)} & \mathrm{vDblCat}
\end{array}
\qquad
r
Proof. The components r V : op H ( Mod ( V ) ) → Mod ( op H ( V ) ) r_{\mathbb{V}}\colon\operatorname{op}_{H}(\operatorname{Mod}(\mathbb{V}))\to\operatorname{Mod}(\operatorname{op}_{H}(\mathbb{V})) r V : op H ( Mod ( V )) → Mod ( op H ( V )) are defined by
Objects
M : = ( ∙ , M , μ , e ) M := (\bullet,M,\mu,e) M := ( ∙ , M , μ , e ) \hspace{4em} ⟶ \longrightarrow ⟶ \hspace{4em} M r : = ( ∙ , M r , μ , e ) M^{r} := (\bullet,M^{r},\mu,e) M r := ( ∙ , M r , μ , e )
∙ ← M ∙ ← M ∙ ↓ ↓ ∙ ← M ∙ ∙ → M r ∙ → M r ∙ ↓ ↓ ∙ → M r ∙ ∙ ╲ ↓ e ╲ ∙ ← M ∙ ∙ ╲ ↓ e ╲ ∙ → M r ∙ \begin{array}{cc}
\begin{array}{ccc}
\bullet& \xleftarrow{\ M\ } & \bullet\xleftarrow{\ M\ } \bullet\\
\downarrow&& \downarrow\\
\bullet& \xleftarrow{\ M\ } & \bullet
\end{array}
&\qquad\qquad\begin{array}{ccc}
\bullet& \xrightarrow{\ M^{r}\ } & \bullet\xrightarrow{\ M^{r}\ } \bullet\\
\downarrow&& \downarrow\\
\bullet& \xrightarrow{\ M^{r}\ } & \bullet
\end{array}\\[2em]
\begin{array}{ccc}
& \bullet&\\
{}^{\diagdown} & \downarrow^{e} & {}^{\diagdown}\\
\bullet& \xleftarrow{\ M\ } & \bullet
\end{array}
&\qquad\qquad\begin{array}{ccc}
& \bullet&\\
{}^{\diagdown} & \downarrow^{e} & {}^{\diagdown}\\
\bullet& \xrightarrow{\ M^{r}\ } & \bullet
\end{array}
\end{array} ∙ ↓ ∙ M M ∙ M ∙ ↓ ∙ ╲ ∙ ∙ ↓ e M ╲ ∙ ∙ ↓ ∙ M r M r ∙ M r ∙ ↓ ∙ ╲ ∙ ∙ ↓ e M r ╲ ∙
V 1-cell
∙ ← M ∙ ↓ f 0 ↓ f ↓ f 0 ⋆ ← N ⋆ ∙ → M r ∙ ↓ f 0 ↓ f ↓ f 0 ⋆ → N r ⋆ ⟶ \begin{array}{cc}
\begin{array}{ccc}
\bullet& \xleftarrow{\ M\ } & \bullet\\
\downarrow^{f_{0}} & \downarrow^{f} & \downarrow^{f_{0}}\\
\star& \xleftarrow{\ N\ } & \star
\end{array}
&\qquad\qquad\begin{array}{ccc}
\bullet& \xrightarrow{\ M^{r}\ } & \bullet\\
\downarrow^{f_{0}} & \downarrow^{f} & \downarrow^{f_{0}}\\
\star& \xrightarrow{\ N^{r}\ } & \star
\end{array}
\end{array}
\qquad\longrightarrow ∙ ↓ f 0 ⋆ M ↓ f N ∙ ↓ f 0 ⋆ ∙ ↓ f 0 ⋆ M r ↓ f N r ∙ ↓ f 0 ⋆ ⟶
H 1-cell
X : = ( X , λ , ρ ) X := (X,\lambda,\rho) X := ( X , λ , ρ )
( N , ⋆ ) ← X ( M , ∙ ) ⟶ ( N r , ⋆ ) → X r ( M r , ∙ ) (N,\star)\xleftarrow{\ X\ }(M,\bullet)
\qquad\longrightarrow\qquad
(N^{r},\star)\xrightarrow{\ X^{r}\ }(M^{r},\bullet) ( N , ⋆ ) X ( M , ∙ ) ⟶ ( N r , ⋆ ) X r ( M r , ∙ )
⋆ ← X ∙ ⋆ ↓ ← X ∙ ← M ∙ ↓ ρ ⋆ ← X ∙ ⋆ → X r ∙ ⋆ ↓ → X r ∙ → M r ∙ ↓ ρ ⋆ → X r ∙ ⋆ ← N ⋆ ← X ∙ ↓ ↓ λ ⋆ ← X ∙ ⋆ → N r ⋆ → X r ∙ ↓ ↓ λ ⋆ → X r ∙ \begin{array}{cc}
\begin{array}{ccc}
\star& \xleftarrow{\ X\ } & \bullet\\[0.5em]
\begin{array}{c}\star\\[-0.2em]\downarrow\end{array}
&
\begin{array}{c}\xleftarrow{\ X\ }\bullet\xleftarrow{\ M\ }\bullet\\[-0.2em]\downarrow^{\rho}\end{array}
\\
\star& \xleftarrow{\ X\ } & \bullet
\end{array}
&\qquad\qquad\begin{array}{ccc}
\star& \xrightarrow{\ X^{r}\ } & \bullet\\[0.5em]
\begin{array}{c}\star\\[-0.2em]\downarrow\end{array}
&
\begin{array}{c}\xrightarrow{\ X^{r}\ }\bullet\xrightarrow{\ M^{r}\ }\bullet\\[-0.2em]\downarrow^{\rho}\end{array}
\\
\star& \xrightarrow{\ X^{r}\ } & \bullet
\end{array}\\[2em]
\begin{array}{ccc}
\star& \xleftarrow{\ N\ } & \star\xleftarrow{\ X\ } \bullet\\
\downarrow&& \downarrow^{\lambda}\\
\star& \xleftarrow{\ X\ } & \bullet
\end{array}
&\qquad\qquad\begin{array}{ccc}
\star& \xrightarrow{\ N^{r}\ } & \star\xrightarrow{\ X^{r}\ } \bullet\\
\downarrow&& \downarrow^{\lambda}\\
\star& \xrightarrow{\ X^{r}\ } & \bullet
\end{array}
\end{array} ⋆ ⋆ ↓ ⋆ X X ∙ M ∙ ↓ ρ X ∙ ∙ ⋆ ↓ ⋆ N X ⋆ X ∙ ↓ λ ∙ ⋆ ⋆ ↓ ⋆ X r X r ∙ M r ∙ ↓ ρ X r ∙ ∙ ⋆ ↓ ⋆ N r X r ⋆ X r ∙ ↓ λ ∙
Sq multicell
∙ n ← X n ⋯ ← X 1 ∙ 1 ↓ g ↓ α ↓ f ⋆ ← Y ∙ ∙ n → X n r ⋯ → X 1 r ∙ 1 ↓ g ↓ α ↓ f ⋆ → Y r ∙ ⟶ \begin{array}{cc}
\begin{array}{ccccccccc}
\bullet_{n} & \xleftarrow{\ X_{n}\ } & \cdots& \xleftarrow{\ X_{1}\ } & \bullet_{1}\\
\downarrow^{g} && \downarrow^{\alpha} && \downarrow^{f}\\
\star& \xleftarrow{\qquad Y\qquad} & & & \bullet
\end{array}
&\qquad\qquad\begin{array}{ccccccccc}
\bullet_{n} & \xrightarrow{\ X_{n}^{r}\ } & \cdots& \xrightarrow{\ X_{1}^{r}\ } & \bullet_{1}\\
\downarrow^{g} && \downarrow^{\alpha} && \downarrow^{f}\\
\star& \xrightarrow{\qquad Y^{r}\qquad} & & & \bullet
\end{array}
\end{array}
\qquad\longrightarrow ∙ n ↓ g ⋆ X n Y ⋯ ↓ α X 1 ∙ 1 ↓ f ∙ ∙ n ↓ g ⋆ X n r Y r ⋯ ↓ α X 1 r ∙ 1 ↓ f ∙ ⟶
We shall now specialize back to the case of monoidal categories. In particular, we shall restrict Mod ( − ) \operatorname{Mod}(-) Mod ( − ) and CoMod ( − ) \operatorname{CoMod}(-) CoMod ( − ) to monoidal categories. In what follows, we write our diagrams in strict monoidal categories st V \operatorname{st}\mathcal{V} st V and interpret the 2-cells using the canonical evaluation map.
Proposition 4.2.9. The ( op , op t ) (\operatorname{op},\operatorname{op}^{t}) ( op , op t ) morphism of inverses on monoidal categories extends to virtual double categories through the following strict 2-transformation cells:
op 2 ( M o n C a t c ) → op 2 ( CoMod ( − ) ) op 2 ( v c o D b l C a t ) op 1 ↓ ↓ op V M o n C a t l → Mod ( − ) v D b l C a t \begin{CD}
\operatorname{op}_{2}(\mathsf{MonCat}_{c}) @>{\operatorname{op}_{2}(\operatorname{CoMod}(-))}>> \operatorname{op}_{2}(\mathsf{vcoDblCat}) \\
@V{\operatorname{op}_{1}}VV @VV{\operatorname{op}_{V}}V \\
\mathsf{MonCat}_{l} @>{\operatorname{Mod}(-)}>> \mathsf{vDblCat}
\end{CD} op 2 ( MonCat c ) op 1 ↓ ⏐ MonCat l op 2 ( CoMod ( − )) Mod ( − ) op 2 ( vcoDblCat ) ↓ ⏐ op V vDblCat op 2 ( M o n C a t l ) → op 2 ( Mod ( − ) ) op 2 ( v D b l C a t ) op 1 t ↓ ↓ op V t M o n C a t c → CoMod ( − ) v c o D b l C a t \begin{CD}
\operatorname{op}_{2}(\mathsf{MonCat}_{l}) @>{\operatorname{op}_{2}(\operatorname{Mod}(-))}>> \operatorname{op}_{2}(\mathsf{vDblCat}) \\
@V{\operatorname{op}_{1}^{t}}VV @VV{\operatorname{op}_{V}^{t}}V \\
\mathsf{MonCat}_{c} @>{\operatorname{CoMod}(-)}>> \mathsf{vcoDblCat}
\end{CD} op 2 ( MonCat l ) op 1 t ↓ ⏐ MonCat c op 2 ( Mod ( − )) CoMod ( − ) op 2 ( vDblCat ) ↓ ⏐ op V t vcoDblCat
Proof. We primarily need to compute the components op V \operatorname{op}_{V} op V and op V t \operatorname{op}_{V}^{t} op V t . For the component op V \operatorname{op}_{V} op V , this means that we need to obtain a virtual double functor op V : op V ( CoMod ( V ) ) → Mod ( op 1 ( V ) ) \operatorname{op}_{V}\colon\operatorname{op}_{V}(\operatorname{CoMod}(\mathcal{V})) \to\operatorname{Mod}(\operatorname{op}_{1}(\mathcal{V})) op V : op V ( CoMod ( V )) → Mod ( op 1 ( V )) . This is given by the following data:
Objects
C : = ( C , μ , e ) ⟶ C op : = ( C , μ op , e op ) C := (C,\mu,e) \longrightarrow C^{\operatorname{op}} := (C,\mu^{\operatorname{op}},e^{\operatorname{op}}) C := ( C , μ , e ) ⟶ C op := ( C , μ op , e op )
V 1-cell
C ( C ′ ) op ↑ ↓ C ′ → C op \begin{CD}
C @. (C')^{\operatorname{op}} \\
@A{}AA @VV{}V \\
C' @>{}>> C^{\operatorname{op}}
\end{CD} C ⏐ ↑ C ′ ( C ′ ) op ↓ ⏐ C op
H 1-cell
C 1 → X : = ( X , λ , ρ ) C 2 ⟶ C 1 op → X op : = ( X , λ op , ρ op ) C 2 op C_{1} \xrightarrow{X:=(X,\lambda,\rho)} C_{2}
\longrightarrow
C_{1}^{\operatorname{op}} \xrightarrow{X^{\operatorname{op}}:=(X,\lambda^{\operatorname{op}},\rho^{\operatorname{op}})} C_{2}^{\operatorname{op}} C 1 X := ( X , λ , ρ ) C 2 ⟶ C 1 op X op := ( X , λ op , ρ op ) C 2 op
Sq multicell
C 1 ′ → Y ⋯ → Z C 2 ′ ( C 1 ′ ) o p → Y o p ⋯ → Z o p ( C 2 ′ ) o p ↑ ↑ ↓ ↓ C 1 → X C 2 C 1 o p → X o p C 2 o p f : X → Z ⊗ ⋯ ⊗ Y f o p : Z ⊗ ⋯ ⊗ Y → X \begin{aligned}
C'_1 &\xrightarrow{Y} \cdots\xrightarrow{Z} C'_2
&
(C'_1)^{\mathrm{op}} &\xrightarrow{Y^{\mathrm{op}}} \cdots\xrightarrow{Z^{\mathrm{op}}} (C'_2)^{\mathrm{op}}\\
\uparrow&& \uparrow&& \downarrow&& \downarrow\\
C_1 &\xrightarrow{X} C_2
&
C_1^{\mathrm{op}} &\xrightarrow{X^{\mathrm{op}}} C_2^{\mathrm{op}}\\
&f:X\to Z\otimes\cdots\otimes Y&&
&f^{\mathrm{op}}:Z\otimes\cdots\otimes Y\to X&
\end{aligned} C 1 ′ ↑ C 1 Y ⋯ Z C 2 ′ X C 2 f : X → Z ⊗ ⋯ ⊗ Y ( C 1 ′ ) op ↑ C 1 op Y op ⋯ Z op ( C 2 ′ ) op X op C 2 op ↓ f op : Z ⊗ ⋯ ⊗ Y → X ↓
The virtual double functor op V t : op V t ( Mod ( V ) ) → CoMod ( op 1 ( V ) ) \operatorname{op}_{\mathcal{V}}^{t}:\operatorname{op}_{\mathcal{V}}^{t}(\operatorname{Mod}(\mathcal{V}))\to\operatorname{CoMod}(\operatorname{op}_{1}(\mathcal{V})) op V t : op V t ( Mod ( V )) → CoMod ( op 1 ( V )) can be obtained similarly.
Finally, we shall similarly extend the reversal duality.
Proposition 4.2.10. The r r r morphism of involution on monoidal categories extends to the virtual double categories case through the following strict 2 2 2 -transformation cells:
MonCat l → Mod ( − ) vDblCat ↓ r ↓ op H MonCat l → Mod ( − ) vDblCat \begin{aligned}
\operatorname{MonCat}_{l}&\xrightarrow{\operatorname{Mod}(-)}\operatorname{vDblCat}\\
\downarrow r&\qquad\qquad\qquad\downarrow\operatorname{op}_{H}\\
\operatorname{MonCat}_{l}&\xrightarrow{\operatorname{Mod}(-)}\operatorname{vDblCat}
\end{aligned} MonCat l ↓ r MonCat l Mod ( − ) vDblCat ↓ op H Mod ( − ) vDblCat
Proof. As before, we specify the virtual double functor r V : op H ( Mod ( V ) ) → Mod ( r ( V ) ) r_{\mathcal{V}}:\operatorname{op}_{H}(\operatorname{Mod}(\mathcal{V}))\to\operatorname{Mod}(r(\mathcal{V})) r V : op H ( Mod ( V )) → Mod ( r ( V )) .
Objects
M : = ( M , μ , e ) ⟶ M r : = ( M , μ , e ) \mathcal{M}:=(M,\mu,e)\longrightarrow\mathcal{M}^{r}:=(M,\mu,e) M := ( M , μ , e ) ⟶ M r := ( M , μ , e )
V 1-cell
M ⟶ M r ↓ ↓ M ′ ⟶ ( M ′ ) r \begin{aligned}
M&\longrightarrow M^{r}\\
\downarrow&&\downarrow\\
M'&\longrightarrow(M')^{r}
\end{aligned} M ↓ M ′ ⟶ M r ⟶ ( M ′ ) r ↓
H 1-cell
M 2 ← X : = ( X , λ , ρ ) M 1 ⟶ ( M 2 ) r → X r : = ( X , ρ , λ ) ( M 1 ) r M_{2}\xleftarrow{X:=(X,\lambda,\rho)}M_{1}
\longrightarrow
(M_{2})^{r}\xrightarrow{X^{r}:=(X,\rho,\lambda)}(M_{1})^{r} M 2 X := ( X , λ , ρ ) M 1 ⟶ ( M 2 ) r X r := ( X , ρ , λ ) ( M 1 ) r
Sq multicell
M 2 ′ ← Z ⋯ ← Y M 1 ′ ( M 2 ′ ) r → Z r ⋯ → Y r ( M 1 ′ ) r ↓ ↓ ↓ ↓ M 2 ← X M 1 M 2 r → X r M 1 r f : Z ⊗ ⋯ ⊗ Y → X f : Y r ⊗ ⋯ ⊗ Z r → X r \begin{aligned}
M'_{2}&\xleftarrow{Z}\cdots\xleftarrow{Y}M'_{1}
&
(M'_{2})^{r}&\xrightarrow{Z^{r}}\cdots\xrightarrow{Y^{r}}(M'_{1})^{r}\\
\downarrow&&\downarrow&&\downarrow&&\downarrow\\
M_{2}&\xleftarrow{\phantom{X}}M_{1}
&
M_{2}^{r}&\xrightarrow{\phantom{X^{r}}}M_{1}^{r}\\
&f:Z\otimes\cdots\otimes Y\to X&&
&f:Y^{r}\otimes\cdots\otimes Z^{r}\to X^{r}
\end{aligned} M 2 ′ ↓ M 2 Z ⋯ Y M 1 ′ X M 1 f : Z ⊗ ⋯ ⊗ Y → X ( M 2 ′ ) r ↓ M 2 r Z r ⋯ Y r ( M 1 ′ ) r X r M 1 r ↓ f : Y r ⊗ ⋯ ⊗ Z r → X r ↓
There are many other variants to consider, such as the pseudo case or the commutative case of each of these. However, the respective results are mostly analogous.
Remark 4.2.11. In the case of a braided or symmetric monoidal category V \mathcal{V} V , we can again compose with the canonical braiding functor b V b_{\mathcal{V}} b V to obtain the following inverse form of reversal component at V \mathcal{V} V :
op H ( Mod ( V ) ) → r V Mod ( r ( V ) ) → Mod ( b V ) Mod ( V ) . \operatorname{op}_{H}(\operatorname{Mod}(\mathcal{V})) \xrightarrow{r_{\mathcal{V}}} \operatorname{Mod}(r(\mathcal{V})) \xrightarrow{\operatorname{Mod}(b_{\mathcal{V}})} \operatorname{Mod}(\mathcal{V}). op H ( Mod ( V )) r V Mod ( r ( V )) Mod ( b V ) Mod ( V ) .
Another well known construction on virtual double categories is the 2-functor Mat : MonCat l → vDblCat \operatorname{Mat}\colon\operatorname{MonCat}_{l} \to\operatorname{vDblCat} Mat : MonCat l → vDblCat . Although this functor is not our central focus, it is worth discussing the dualities of this functor. We state the following lemma on reversal duality of Mat \operatorname{Mat} Mat .
Lemma 4.2.12. The 2-functor Mat \operatorname{Mat} Mat extends to a 1-morphism of involutions by the strict 2-natural transformation
MonCat l → Mat ( − ) vDblCat r ↓ ↓ op H MonCat l → Mat ( − ) vDblCat . \begin{CD}
\operatorname{MonCat}_{l} @>{\operatorname{Mat}(-)}>> \operatorname{vDblCat}\\
@V{r}VV @VV{\operatorname{op}_{H}}V\\
\operatorname{MonCat}_{l} @>>{\operatorname{Mat}(-)}> \operatorname{vDblCat}.
\end{CD} MonCat l r ↓ ⏐ MonCat l Mat ( − ) Mat ( − ) vDblCat ↓ ⏐ op H vDblCat .
Proof. Let ( − ) t (-)^{t} ( − ) t here denote the transpose of a matrix. The components r V : op H ( Mat ( V ) ) → Mat ( r ( V ) ) r_{\mathcal{V}}\colon\operatorname{op}_{H}(\operatorname{Mat}(\mathcal{V})) \to\operatorname{Mat}(r(\mathcal{V})) r V : op H ( Mat ( V )) → Mat ( r ( V )) are defined by
Objects
X ⟶ X X \longrightarrow X X ⟶ X
V 1-cell
X → X ↓ ↓ Y → Y \begin{CD}
X @>>> X\\
@VVV @VVV\\
Y @>>> Y
\end{CD} X ↓ ⏐ Y X ↓ ⏐ Y
H 1-cell
Y ← M X ⟶ Y → M t X Y \xleftarrow{\ M\ } X \longrightarrow Y \xrightarrow{\ M^{t}\ } X Y M X ⟶ Y M t X
Sq multicell
\begin{CD}
X_{n} @<{M_{n}}<< \cdots@<<< \cdots@<{M_{1}}<< X_{0}\\
@VVV @VVV @VVV\\
Y_{1} @<{N}<< Y_{0}
\end{CD}
\qquad\longrightarrow\qquad
\begin{CD}
X_{n} @>{M_{n}^{t}}>> \cdots@>>> \cdots@>{M_{1}^{t}}>> X_{0}\\
@VVV @VVV @VVV\\
Y_{1} @>{N^{t}}>> M_{1}^{r}
\end{CD}
Remark 4.2.13. More generally, [11] propose a generalized matrices functor Mat : vDblCat → vDblCat \operatorname{Mat}: \operatorname{vDblCat} \to\operatorname{vDblCat} Mat : vDblCat → vDblCat , which arises primarily from the discrete opfibration Set ∗ → Set \operatorname{Set}_* \to\operatorname{Set} Set ∗ → Set . In particular, they define it as a composite functor
vDblCat → ( Set ∗ ) h c × ( − ) vDblCat / ( Set ∗ ) h c Mat ↓ ↓ Π h c vDblCat ← U vDblCat / ( Set ) h c \begin{CD}
\operatorname{vDblCat} @>{(\operatorname{Set}_*)_{\mathrm{hc}}\times(-)}>> \operatorname{vDblCat}/(\operatorname{Set}_*)_{\mathrm{hc}} \\
@V{\operatorname{Mat}}VV @VV{\Pi_{\mathrm{hc}}}V \\
\operatorname{vDblCat} @<{U}<< \operatorname{vDblCat}/(\operatorname{Set})_{\mathrm{hc}}
\end{CD} vDblCat Mat ↓ ⏐ vDblCat ( Set ∗ ) hc × ( − ) U vDblCat / ( Set ∗ ) hc ↓ ⏐ Π hc vDblCat / ( Set ) hc
One could show that each of the individual 2-functor appearing in this decomposition is compatible with the op H \operatorname{op}_{H} op H duality. In particular, this would generalize the previous lemma to vDblCat \operatorname{vDblCat} vDblCat .
Lastly, in contrast to the horizontal duality, we note that the 2-functor Mat \operatorname{Mat} Mat doesn’t respect the vertical duality of virtual double categories in the expected way. Although one could define a suitable 2-functor CoMat : vcoDblCat → vcoDblCat \operatorname{CoMat}: \operatorname{vcoDblCat} \to\operatorname{vcoDblCat} CoMat : vcoDblCat → vcoDblCat , applying CoMod \operatorname{CoMod} CoMod to this construction would not yield the desired variance for morphisms between cocategories. This is more readily observed by considering the category V op Cat \mathcal{V}^{\operatorname{op}}\operatorname{Cat} V op Cat of op 1 ( V ) \operatorname{op}_{1}(\mathcal{V}) op 1 ( V ) -enriched categories. In particular, the action of functors on the hom-objects and on the set of objects are of opposite variance. If one is content with this disparity, we believe that one could dualize the Mat \operatorname{Mat} Mat construction.
Higher Dualities
We briefly describe the analogous transport phenomenon in the case of 1 Cat 1\operatorname{Cat} 1 Cat -enriched symmetric monoidal category 1 Cat 1\operatorname{Cat} 1 Cat .
Definition 4.3.1. Let ( V , ⊗ , I ) (\mathcal{V},\otimes,I) ( V , ⊗ , I ) be any monoidal category. We define a left V \mathcal{V} V -module category C \mathcal{C} C as a left pseudomodule object in 1 Cat 1\operatorname{Cat} 1 Cat . The definitions for right module or bimodule categories are analogous.
To describe the op 1 \operatorname{op}_{1} op 1 duality of monoidal categories, we need to have the variants 1 Mon c 1\operatorname{Mon}_{c} 1 Mon c and 1 Mon l 1\operatorname{Mon}_{l} 1 Mon l corresponding the construction 1 Mon 1\operatorname{Mon} 1 Mon . This is similar to how one dimension lower, we required separate Mon \operatorname{Mon} Mon and CoMon \operatorname{CoMon} CoMon constructions. In particular, 1 Mon l 1\operatorname{Mon}_{l} 1 Mon l is simply the construction we have already provided involving lax 1-morphisms. Whereas, 1 Mon c 1\operatorname{Mon}_{c} 1 Mon c is the colax 1-morphism variant.
Lemma 4.3.2. The op 1 \operatorname{op}_{1} op 1 and op 1 t \operatorname{op}_{1}^{t} op 1 t duality inverses of monoidal categories are equivalently expressed as the two maps:
op 1 t : op 2 ( 1 Mon l ( 1 Cat ) ) ⟶ 1 Mon c ( op 2 ( 1 Cat ) ) → 1 Mon c ( op 1 ) 1 Mon c ( 1 Cat ) op 1 : op 2 ( 1 Mon c ( 1 Cat ) ) ⟶ 1 Mon l ( op 2 ( 1 Cat ) ) → 1 Mon l ( op 1 ) 1 Mon l ( 1 Cat ) \begin{aligned}
\operatorname{op}_{1}^{t}: \quad&\operatorname{op}_{2}(1\operatorname{Mon}_{l}(1\operatorname{Cat})) \longrightarrow1\operatorname{Mon}_{c}(\operatorname{op}_{2}(1\operatorname{Cat})) \xrightarrow{1\operatorname{Mon}_{c}(\operatorname{op}_{1})} 1\operatorname{Mon}_{c}(1\operatorname{Cat}) \\
\operatorname{op}_{1}: \quad&\operatorname{op}_{2}(1\operatorname{Mon}_{c}(1\operatorname{Cat})) \longrightarrow1\operatorname{Mon}_{l}(\operatorname{op}_{2}(1\operatorname{Cat})) \xrightarrow{1\operatorname{Mon}_{l}(\operatorname{op}_{1})} 1\operatorname{Mon}_{l}(1\operatorname{Cat})
\end{aligned} op 1 t : op 1 : op 2 ( 1 Mon l ( 1 Cat )) ⟶ 1 Mon c ( op 2 ( 1 Cat )) 1 Mon c ( op 1 ) 1 Mon c ( 1 Cat ) op 2 ( 1 Mon c ( 1 Cat )) ⟶ 1 Mon l ( op 2 ( 1 Cat )) 1 Mon l ( op 1 ) 1 Mon l ( 1 Cat )
Proof. The identifications 1 M o n l ( o p 2 ( 1 C a t ) ) ≅ o p 2 ( M o n C a t c ) 1\mathrm{Mon}_{l}(\mathrm{op}_{2}(1\mathrm{Cat})) \cong\mathrm{op}_{2}(\mathrm{MonCat}_{c}) 1 Mon l ( op 2 ( 1 Cat )) ≅ op 2 ( MonCat c ) and 1 M o n c ( o p 2 ( 1 C a t ) ) ≅ o p 2 ( M o n C a t l ) 1\mathrm{Mon}_{c}(\mathrm{op}_{2}(1\mathrm{Cat})) \cong\mathrm{op}_{2}(\mathrm{MonCat}_{l}) 1 Mon c ( op 2 ( 1 Cat )) ≅ op 2 ( MonCat l ) follows directly by the definitions. One should view these identifications as merely comprehension identifications. Whereas, o p 1 : o p 2 ( 1 C a t ) → 1 C a t \mathrm{op}_{1}:\mathrm{op}_{2}(1\mathrm{Cat}) \to1\mathrm{Cat} op 1 : op 2 ( 1 Cat ) → 1 Cat is a strict monoidal 2-functor. Thus, we have transport morphisms 1 M o n c ( o p 1 ) 1\mathrm{Mon}_{c}(\mathrm{op}_{1}) 1 Mon c ( op 1 ) and 1 M o n l ( o p 1 ) 1\mathrm{Mon}_{l}(\mathrm{op}_{1}) 1 Mon l ( op 1 ) given by applying the o p 1 \mathrm{op}_{1} op 1 functor directly on data of the objects and the morphisms. This is exactly the description we provided at the beginning of this section. Now, we are stating the same description in a more formalized way.
In other words, we have fully expressed the o p 1 \mathrm{op}_{1} op 1 and o p 1 t \mathrm{op}_{1}^{t} op 1 t duality of monoidal categories in terms of a hypothetical higher duality o p 2 : o p 3 ( M o n B i c a t x ) → M o n B i c a t y \mathrm{op}_{2}:\mathrm{op}_{3}(\mathrm{MonBicat}_{x}) \to\mathrm{MonBicat}_{y} op 2 : op 3 ( MonBicat x ) → MonBicat y , where x x x and y y y denote some suitable variants. More precisely, assuming we had all the relevant definitions, the duality can be described by the following diagram
∗ → C a t o p 3 ( M o n B i c a t x ) → 1 M o n l o p 3 ( B i c a t x ) ↓ ↓ o p 2 ↓ o p 2 ∗ → C a t M o n B i c a t y → 1 M o n l B i c a t y \begin{aligned}
&\ast\xrightarrow{\mathrm{Cat}} \mathrm{op}_{3}(\mathrm{MonBicat}_{x}) \xrightarrow{1\mathrm{Mon}_{l}} \mathrm{op}_{3}(\mathrm{Bicat}_{x}) \\
&\downarrow\qquad\qquad\quad\downarrow\mathrm{op}_{2}\qquad\qquad\quad\downarrow\mathrm{op}_{2} \\
&\ast\xrightarrow{\mathrm{Cat}} \mathrm{MonBicat}_{y} \xrightarrow{1\mathrm{Mon}_{l}} \mathrm{Bicat}_{y}
\end{aligned} ∗ Cat op 3 ( MonBicat x ) 1 Mon l op 3 ( Bicat x ) ↓ ↓ op 2 ↓ op 2 ∗ Cat MonBicat y 1 Mon l Bicat y
It is now also straightforward to guess how pseudomodules in 1 C a t 1\mathrm{Cat} 1 Cat ought to transport under the dualities.
Lemma 4.3.3. Let V 1 \mathcal{V}_{1} V 1 and V 2 \mathcal{V}_{2} V 2 be any two monoidal categories and let B i m o d ( V 1 , V 2 ) \mathrm{Bimod}(\mathcal{V}_{1},\mathcal{V}_{2}) Bimod ( V 1 , V 2 ) be 2-category of modules between V 1 \mathcal{V}_{1} V 1 and V 2 \mathcal{V}_{2} V 2 . There exists the following involutive isomorphism induced by the action of o p 1 : o p 2 ( 1 C a t ) → 1 C a t \mathrm{op}_{1}:\mathrm{op}_{2}(1\mathrm{Cat}) \to1\mathrm{Cat} op 1 : op 2 ( 1 Cat ) → 1 Cat :
o p 1 : o p 2 ( B i m o d ( V 1 , V 2 ) ) ⟶ B i m o d ( o p 1 ( V 1 ) , o p 1 ( V 2 ) ) . \mathrm{op}_{1}:\quad\mathrm{op}_{2}(\mathrm{Bimod}(\mathcal{V}_{1},\mathcal{V}_{2})) \longrightarrow\mathrm{Bimod}(\mathrm{op}_{1}(\mathcal{V}_{1}),\mathrm{op}_{1}(\mathcal{V}_{2})). op 1 : op 2 ( Bimod ( V 1 , V 2 )) ⟶ Bimod ( op 1 ( V 1 ) , op 1 ( V 2 )) . Simply note that all the defining axioms are comprised of natural isomorphisms which can be inverted. Finally, the duality for reversal is expressible similarly.
Lemma 4.3.4.
The reversal duality r : M o n C a t x → M o n C a t x r : \mathrm{MonCat}_{x} \to\mathrm{MonCat}_{x} r : MonCat x → MonCat x is given by the map
r : 1 M o n ( 1 C a t ) ⟶ 1 M o n ( r ( 1 C a t ) ) → 1 M o n ( ( 1 , s 1 C a t , 1 ) ) 1 M o n ( 1 C a t ) , r:\quad1\mathrm{Mon}(1\mathrm{Cat}) \longrightarrow1\mathrm{Mon}(r(1\mathrm{Cat})) \xrightarrow{1\mathrm{Mon}((1,s_{1\mathrm{Cat}},1))} 1\mathrm{Mon}(1\mathrm{Cat}), r : 1 Mon ( 1 Cat ) ⟶ 1 Mon ( r ( 1 Cat )) 1 Mon (( 1 , s 1 Cat , 1 )) 1 Mon ( 1 Cat ) , where ( 1 , s 1 C a t , 1 ) (1,s_{1\mathrm{Cat}},1) ( 1 , s 1 Cat , 1 ) is the canonical identity symmetric monoidal strict 2-functor whose compositor constraint is the symmetry of 1 C a t 1\mathrm{Cat} 1 Cat .
There exists the following involutive isomorphism on the bimodule 2-categories induced by the action of r r r :
r : B i m o d ( V 1 , V 2 ) ⟶ B i m o d ( r ( V 2 ) , r ( V 1 ) ) . r:\quad\mathrm{Bimod}(\mathcal{V}_{1},\mathcal{V}_{2}) \longrightarrow\mathrm{Bimod}(r(\mathcal{V}_{2}),r(\mathcal{V}_{1})). r : Bimod ( V 1 , V 2 ) ⟶ Bimod ( r ( V 2 ) , r ( V 1 )) .
At this point, we could also obtain similar results relating monoidal and comonoidal categories using the o p 1 \mathrm{op}_{1} op 1 involution of bicategories, that is, o p 1 ( 1 C o M o n l ( 1 C a t ) ) ≅ 1 M o n l ( o p 1 ( 1 C a t ) ) \mathrm{op}_{1}(1\mathrm{CoMon}_{l}(1\mathrm{Cat})) \cong1\mathrm{Mon}_{l}(\mathrm{op}_{1}(1\mathrm{Cat})) op 1 ( 1 CoMon l ( 1 Cat )) ≅ 1 Mon l ( op 1 ( 1 Cat )) . However, these results are not essential for us, so we do not investigate them here.
Sliced modules and enriched categories
The goal of this section is to describe how the microcosm principle allows one to define general enriched slice categories, especially under the consideration of model structures. Some special cases of the following results have already appeared in the literature for slice-over categories. We provide a general characterization encompassing both slice-over and slice-under categories.
Slice monoidal categories and module
The definition of a module category is again within the scope of the microcosm principle. Consequently, we start with the definition of a module object inside a module category.
Definition 5.1.1. Let ( C , ◃ C , a C , λ C ) (\mathcal{C},\triangleleft_{\mathcal{C}},a_{\mathcal{C}},\lambda_{\mathcal{C}}) ( C , ◃ C , a C , λ C ) be a left V \mathcal{V} V -module category over some monoidal category V \mathcal{V} V . Let ( M , μ , e ) ∈ V (M,\mu,e)\in\mathcal{V} ( M , μ , e ) ∈ V be a monoid object. We say that an object Z ∈ C Z\in\mathcal{C} Z ∈ C is a left module object over M M M if it has following data:
a left action morphism ◃ Z : M ◃ C Z → Z \triangleleft_Z:M\triangleleft_{\mathcal{C}}Z\to Z ◃ Z : M ◃ C Z → Z ,
the associativity and unitality axioms hold
( M ⊗ M ) ◃ C Z → μ ◃ 1 M ◃ C Z → ◃ Z Z , ( M ⊗ M ) ◃ C Z → a C M ◃ C ( M ◃ C Z ) → 1 ◃ ( ◃ Z ) M ◃ C Z → ◃ Z Z , I ◃ C Z → e ◃ C Z M ◃ C Z → ◃ Z Z , I ◃ C Z → λ C Z . \begin{aligned}
(M\otimes M)\triangleleft_{\mathcal{C}}Z &\xrightarrow{\mu\triangleleft1} M\triangleleft_{\mathcal{C}}Z \xrightarrow{\triangleleft_Z} Z,\\
(M\otimes M)\triangleleft_{\mathcal{C}}Z &\xrightarrow{a_{\mathcal{C}}} M\triangleleft_{\mathcal{C}}(M\triangleleft_{\mathcal{C}}Z) \xrightarrow{1\triangleleft(\triangleleft_Z)} M\triangleleft_{\mathcal{C}}Z \xrightarrow{\triangleleft_Z} Z,\\
I\triangleleft_{\mathcal{C}}Z &\xrightarrow{e\triangleleft_{\mathcal{C}}Z} M\triangleleft_{\mathcal{C}}Z \xrightarrow{\triangleleft_Z} Z,\\
I\triangleleft_{\mathcal{C}}Z &\xrightarrow{\lambda_{\mathcal{C}}} Z.
\end{aligned} ( M ⊗ M ) ◃ C Z ( M ⊗ M ) ◃ C Z I ◃ C Z I ◃ C Z μ ◃ 1 M ◃ C Z ◃ Z Z , a C M ◃ C ( M ◃ C Z ) 1 ◃ ( ◃ Z ) M ◃ C Z ◃ Z Z , e ◃ C Z M ◃ C Z ◃ Z Z , λ C Z .
Definition 5.1.2. Let ( C , ▹ C , a C , ρ C ) (\mathcal{C},\triangleright_{\mathcal{C}},a_{\mathcal{C}},\rho_{\mathcal{C}}) ( C , ▹ C , a C , ρ C ) be a right V \mathcal{V} V -module category over some monoidal category V \mathcal{V} V . Let ( M , μ , e ) ∈ V (M,\mu,e)\in\mathcal{V} ( M , μ , e ) ∈ V be a monoid object. We say that an object Z ∈ C Z\in\mathcal{C} Z ∈ C is a right module object over M M M if it has following data:
a right action morphism ▹ Z : Z ▹ C M → Z \triangleright_Z:Z\triangleright_{\mathcal{C}}M\to Z ▹ Z : Z ▹ C M → Z ,
the associativity and unitality axioms hold
( Z ▹ C M ) ▹ C M → ( ▹ Z ) ▹ c 1 Z ▹ C M ↓ a C ↓ ▹ Z Z ▹ C ( M ⊗ M ) → 1 ▹ c μ Z ▹ C M → ▹ Z Z \begin{aligned}
(Z \triangleright_{\mathcal C} M)\triangleright_{\mathcal C} M &\xrightarrow{(\triangleright Z)\triangleright c1} Z\triangleright_{\mathcal C} M \\
\downarrow a_{\mathcal C} &\qquad\qquad\qquad\qquad\downarrow\triangleright Z \\
Z\triangleright_{\mathcal C}(M\otimes M) &\xrightarrow{1\triangleright c\mu} Z\triangleright_{\mathcal C}M \xrightarrow{\triangleright Z} Z
\end{aligned} ( Z ▹ C M ) ▹ C M ↓ a C Z ▹ C ( M ⊗ M ) ( ▹ Z ) ▹ c 1 Z ▹ C M ↓ ▹ Z 1 ▹ c μ Z ▹ C M ▹ Z Z
Z ▹ C I → Z ▹ c e Z ▹ C M ↓ ρ C ↓ ▹ Z Z \begin{aligned}
Z\triangleright_{\mathcal C}I &\xrightarrow{Z\triangleright c e} Z\triangleright_{\mathcal C}M \\
\downarrow\rho_{\mathcal C} &\qquad\downarrow\triangleright Z \\
Z &
\end{aligned} Z ▹ C I ↓ ρ C Z Z ▹ ce Z ▹ C M ↓ ▹ Z
A similar definition can be stated for bimodule object in a bimodule category. One only needs to add a balancing axiom for the bimodule object relative to the balancing 2-cell of bimodule category.
As expected, these definitions also provides the slice constructions. Let ( V , ⊗ , I , a V , λ V , ρ V ) (\mathcal V,\otimes,I,a_{\mathcal V},\lambda_{\mathcal V},\rho_{\mathcal V}) ( V , ⊗ , I , a V , λ V , ρ V ) be a monoidal category in what follows.
Lemma 5.1.3. Let ( C , ◃ C , a C , λ C ) (\mathcal C,\triangleleft_{\mathcal C},a_{\mathcal C},\lambda_{\mathcal C}) ( C , ◃ C , a C , λ C ) be a left V \mathcal V V -module. Let ( M , μ , e ) (M,\mu,e) ( M , μ , e ) be a monoid object in V \mathcal V V , and let ( Z , ◃ Z ) (Z,\triangleleft_Z) ( Z , ◃ Z ) be a left M M M -module.
The slice category V / M \mathcal V/M V / M admits a monoidal structure denoted by ( V / M , ⊗ / M , I / M ) (\mathcal V/M,\otimes/M,I/M) ( V / M , ⊗ / M , I / M ) . If V \mathcal V V is braided or symmetric and M M M is commutative then V / M \mathcal V/M V / M is also braided or symmetric.
There is a canonical left V / M \mathcal V/M V / M -module structure on C / Z \mathcal C/Z C / Z denoted by ( C / Z , ◃ / ) (\mathcal C/Z,\triangleleft/) ( C / Z , ◃ / ) .
Proof. For the first statement, the sliced tensoring operation is defined by
f ⊗ / M g : = A ⊗ B → f ⊗ g M ⊗ M → μ M f\otimes/M g := A\otimes B \xrightarrow{f\otimes g} M\otimes M \xrightarrow{\mu} M f ⊗ / M g := A ⊗ B f ⊗ g M ⊗ M μ M
and the unit by
I / M : = I → η M . I/M := I \xrightarrow{\eta} M. I / M := I η M .
The naturality of the associator and the unitors of V \mathcal V V together with the monoid axioms of M M M allows us to extend these given associator and unitors to those of V / M \mathcal V/M V / M directly.
For the second statement, the sliced left action operation is defined exactly the same. Let α : X → M ∈ V / M \alpha:X\to M\in\mathcal V/M α : X → M ∈ V / M and f : A → Z ∈ C / Z f:A\to Z\in\mathcal C/Z f : A → Z ∈ C / Z . The sliced left action is defined by
α ◃ / f : X ◃ A → α ◃ f M ◃ Z → ◃ Z Z . \alpha\triangleleft/ f:X\triangleleft A\xrightarrow{\alpha\triangleleft f}M\triangleleft Z\xrightarrow{\triangleleft_Z}Z. α ◃ / f : X ◃ A α ◃ f M ◃ Z ◃ Z Z .
Thereafter, the extension of coherence 2-cells follow again by the naturality and module axioms. □
This lemma can be dualized in two ways: one by using the o p 1 \mathrm{op}_1 op 1 duality and the other by r r r duality. The reversal duality, however, provides nothing new.
For the o p 1 \mathrm{op}_1 op 1 duality, we consider a monoid object N N N in the monoidal category V o p \mathcal V^{\mathrm{op}} V op and a module object Z Z Z in C o p \mathcal C^{\mathrm{op}} C op . Applying the previous lemma in this situation and transporting structure back using o p 1 \mathrm{op}_1 op 1 duality, we obtain the following.
Lemma 5.1.4. Let ( C , ◃ C , a C , λ C ) (\mathcal{C},\triangleleft_{\mathcal{C}},a_{\mathcal{C}},\lambda_{\mathcal{C}}) ( C , ◃ C , a C , λ C ) be a left V \mathcal{V} V -module. Let ( N , μ , η ) (N,\mu,\eta) ( N , μ , η ) be a comonoid object and ( Z , λ ) (Z,\lambda) ( Z , λ ) be a left N N N -comodule.
The coslice category N / V N/\mathcal{V} N / V admits a monoidal structure denoted by ( N / V , ⊗ N / , I N / ) (N/\mathcal{V},\otimes_{N/},I_{N/}) ( N / V , ⊗ N / , I N / ) . If V \mathcal{V} V is braided or symmetric and N N N is commutative then N / V N/\mathcal{V} N / V is also braided or symmetric.
There is a canonical left N / V N/\mathcal{V} N / V -module structure on Z / V Z/\mathcal{V} Z / V denoted by ( Z / C , ◃ Z / C ) (Z/\mathcal{C},\triangleleft_{Z/\mathcal{C}}) ( Z / C , ◃ Z / C ) .
We now investigate the presence of closures on left V \mathcal{V} V -module structure. That is, we investigate enrichment through tensoring or powering. Consequentially, we shall now refer to the action functor ◃ : V × C → C \triangleleft:\mathcal{V}\times\mathcal{C}\to\mathcal{C} ◃ : V × C → C as tensoring .
Definition 5.1.5. A left V \mathcal{V} V -module category ( C , ◃ ) (\mathcal{C},\triangleleft) ( C , ◃ ) is called left closed if for every c ∈ C c\in\mathcal{C} c ∈ C , the functor − ◃ c : V → V -\triangleleft c:\mathcal{V}\to\mathcal{V} − ◃ c : V → V has a right adjoint C ( c , − ) : C → V \mathcal{C}(c,-):\mathcal{C}\to\mathcal{V} C ( c , − ) : C → V . Thus, we have a functor C ( − , − ) : C o p × C → V \mathcal{C}(-,-):\mathcal{C}^{\mathrm{op}}\times\mathcal{C}\to\mathcal{V} C ( − , − ) : C op × C → V , which is called the internal hom of C \mathcal{C} C .
Definition 5.1.6. A left V \mathcal{V} V -module category ( C , ◃ ) (\mathcal{C},\triangleleft) ( C , ◃ ) is called right closed if for every x ∈ V x\in\mathcal{V} x ∈ V , the functor x ◃ − : C → C x\triangleleft-:\mathcal{C}\to\mathcal{C} x ◃ − : C → C has a right adjoint − ⋔ x : C → C -\pitchfork x:\mathcal{C}\to\mathcal{C} − ⋔ x : C → C . Thus, we have a functor − ⋔ − : C × V o p → C -\pitchfork-:\mathcal{C}\times\mathcal{V}^{\mathrm{op}}\to\mathcal{C} − ⋔ − : C × V op → C , which is called powering . This is usually also denoted by b x b^x b x or { x , b } \{x,b\} { x , b } .
Definition 5.1.7. A left V \mathcal{V} V -module category ( C , ◃ ) (\mathcal{C},\triangleleft) ( C , ◃ ) is closed if it is both left closed and right closed.
Lemma 5.1.8. A closed left V \mathcal{V} V -module category ( C , ◃ ) (\mathcal{C},\triangleleft) ( C , ◃ ) is equivalently specified by the data of a two variable left adjunction
( − ◃ − ) o p : V o p × C o p ⟶ C o p (-\triangleleft-)^{\mathrm{op}}:\mathcal{V}^{\mathrm{op}}\times\mathcal{C}^{\mathrm{op}}\longrightarrow\mathcal{C}^{\mathrm{op}} ( − ◃ − ) op : V op × C op ⟶ C op C ( − , − ) : C o p × C ⟶ V \mathcal{C}(-,-):\mathcal{C}^{\mathrm{op}}\times\mathcal{C}\longrightarrow\mathcal{V} C ( − , − ) : C op × C ⟶ V − ⋔ − : C × V o p ⟶ C . -\pitchfork-:\mathcal{C}\times\mathcal{V}^{\mathrm{op}}\longrightarrow\mathcal{C}. − ⋔ − : C × V op ⟶ C .
Proof. See [8, Def. 2.1 and Sec. 2.2] for the definition and the respective proof. In simpler terms, this is equivalent to the natural isomorphisms C ( x ◃ a , b ) ≅ V ( x , C ( a , b ) ) ≅ C ( a , b ⋔ x ) \mathcal{C}(x\triangleleft a,b)\cong\mathcal{V}(x,\mathcal{C}(a,b))\cong\mathcal{C}(a,b\pitchfork x) C ( x ◃ a , b ) ≅ V ( x , C ( a , b )) ≅ C ( a , b ⋔ x ) .
Example 5.1.9. If the canonical left V \mathcal{V} V -module ( V , ⊗ ) (\mathcal{V},\otimes) ( V , ⊗ ) is left closed then it is a left closed monoidal category in the usual sense. The same is true for the right closure of this left module structure.
Lemma 5.1.10. There is a canonical enrichment of C \mathcal{C} C over V \mathcal{V} V using the internal homs C ( − , − ) \mathcal{C}(-,-) C ( − , − ) , which we call C ‾ \underline{\mathcal{C}} C . Furthermore, the underlying category of C ‾ \underline{\mathcal{C}} C is isomorphic to C \mathcal{C} C itself.
Proof. One can mimic the proof provided in [22, Vol. III, Sec. 3.1] by defining eval d c : C ( c , d ) ◃ c → d \operatorname{eval}^{c}_{d}:\mathcal{C}(c,d)\triangleleft c\to d eval d c : C ( c , d ) ◃ c → d through the counit and coeval x d : x → C ( d , x ◃ d ) \operatorname{coeval}^{d}_{x}:x\to\mathcal{C}(d,x\triangleleft d) coeval x d : x → C ( d , x ◃ d ) through the unit. Thus, compositions are given by the adjoint to the composites
( C ( d , e ) ⊗ C ( c , d ) ) ◃ c → e ↓ a eval e d ↑ C ( d , e ) ◃ ( C ( c , d ) ◃ c ) → 1 ◃ eval d c C ( d , e ) ◃ d \begin{CD}
(\mathcal{C}(d,e)\otimes\mathcal{C}(c,d))\triangleleft c @>>> e \\
@VV a V @A \operatorname{eval}^{d}_{e} AA \\
\mathcal{C}(d,e)\triangleleft(\mathcal{C}(c,d)\triangleleft c) @>{1\triangleleft\operatorname{eval}^{c}_{d}}>> \mathcal{C}(d,e)\triangleleft d
\end{CD} ( C ( d , e ) ⊗ C ( c , d )) ◃ c ↓ ⏐ a C ( d , e ) ◃ ( C ( c , d ) ◃ c ) 1 ◃ eval d c e eval e d ⏐ ↑ C ( d , e ) ◃ d
and unitors are given by the adjoint to the morphisms I ◃ x → λ x x I \triangleleft x \xrightarrow{\lambda_x} x I ◃ x λ x x . Thereafter, the coherence follows from the 2-cells of the left V \mathcal{V} V -module structure.
Let i : I → C ( c , c ) i\colon I \to\mathcal{C}(c,c) i : I → C ( c , c ) be the unitors and let m : C ( d , e ) ⊗ C ( c , d ) → C ( c , e ) m\colon\mathcal{C}(d,e) \otimes\mathcal{C}(c,d) \to\mathcal{C}(c,e) m : C ( d , e ) ⊗ C ( c , d ) → C ( c , e ) be the compositors. In particular, we obtain the following two commuting diagrams
( C ( d , e ) ⊗ C ( c , d ) ) ◃ c → m ◃ 1 C ( c , e ) ◃ c a ↓ ↓ eval e c C ( d , e ) ◃ ( C ( c , d ) ◃ c ) e 1 ◃ eval d c ↓ C ( d , e ) ◃ d → eval e d e I ◃ c → i ◃ 1 C ( c , c ) ◃ c λ ↓ ↓ eval c c c = c \begin{CD}
(\mathcal{C}(d,e)\otimes\mathcal{C}(c,d))\triangleleft c @>{m\triangleleft1}>> \mathcal{C}(c,e)\triangleleft c \\
@V{a}VV @VV{\operatorname{eval}^{c}_{e}}V \\
\mathcal{C}(d,e)\triangleleft(\mathcal{C}(c,d)\triangleleft c) @. e \\
@V{1\triangleleft\operatorname{eval}^{c}_{d}}VV @. \\
\mathcal{C}(d,e)\triangleleft d @>{\operatorname{eval}^{d}_{e}}>> e
\end{CD}
\qquad
\begin{CD}
I\triangleleft c @>{i\triangleleft1}>> \mathcal{C}(c,c)\triangleleft c \\
@V{\lambda}VV @VV{\operatorname{eval}^{c}_{c}}V \\
c @= c
\end{CD} ( C ( d , e ) ⊗ C ( c , d )) ◃ c a ↓ ⏐ C ( d , e ) ◃ ( C ( c , d ) ◃ c ) 1 ◃ eval d c ↓ ⏐ C ( d , e ) ◃ d m ◃ 1 eval e d C ( c , e ) ◃ c ↓ ⏐ eval e c e e I ◃ c λ ↓ ⏐ c i ◃ 1 C ( c , c ) ◃ c ↓ ⏐ eval c c c
The converse of this lemma is not true in general. That is, we can’t obtain left module structures using left adjoints to internal homs. However, when V \mathcal{V} V is a closed symmetric monoidal category, the converse does hold, and the closed module structure lifts from 1 C a t \mathbf{1Cat} 1Cat to V C a t \mathcal{V}\mathbf{Cat} V Cat . In particular, this defines a enriched, tensored, and cotensored category [28, Prop. 10.1.14].
Lemma 5.1.11. The powering functor − ⋔ − : C × V o p → C -\pitchfork- \colon\mathcal{C}\times\mathcal{V}^{\mathrm{op}}\to\mathcal{C} − ⋔ − : C × V op → C defines a canonical right V o p \mathcal{V}^{\mathrm{op}} V op -module structure.
Proof. A left V \mathcal{V} V -module structure ( C , ◃ ) (\mathcal{C},\triangleleft) ( C , ◃ ) is equivalent to providing a strong monoidal functor − ◃ − : V → 1 C a t ( C , C ) -\triangleleft-\colon\mathcal{V}\to\mathbf{1Cat}(\mathcal{C},\mathcal{C}) − ◃ − : V → 1Cat ( C , C ) . The requirement of x ◃ − x\triangleleft- x ◃ − admitting right adjoint is then equivalent to x ◃ − x\triangleleft- x ◃ − admitting a right dual in the endofunctor monoidal category 1 C a t ( C , C ) \mathbf{1Cat}(\mathcal{C},\mathcal{C}) 1Cat ( C , C ) . The right dual construction then defines a “partial” normal strong monoidal functor r ( ( 1 C a t ( C , C ) ) o p ) → 1 C a t ( C , C ) r((\mathbf{1Cat}(\mathcal{C},\mathcal{C}))^{\mathrm{op}})\to\mathbf{1Cat}(\mathcal{C},\mathcal{C}) r (( 1Cat ( C , C ) ) op ) → 1Cat ( C , C ) out of the right dualizable endofunctors (see [10, Sec. 2.10]). In particular, we obtain a canonical left r ( V o p ) r(\mathcal{V}^{\mathrm{op}}) r ( V op ) -module structure on C \mathcal{C} C by postcomposing with this partial right dual functor. Finally, this is equivalent to a canonical right V o p \mathcal{V}^{\mathrm{op}} V op -module structure on C \mathcal{C} C or a canonical right V \mathcal{V} V -module structure on C o p \mathcal{C}^{\mathrm{op}} C op .
In concrete terms, the right V o p \mathcal{V}^{\mathrm{op}} V op -module structure on C \mathcal{C} C has a right unitor given by the diagram
c ⋔ I → λ c ⋔ I − 1 I ◃ ( c ⋔ I ) ↓ ϵ c I c \begin{CD}
c\pitchfork I @>{\lambda^{-1}_{c\pitchfork I}}>> I\triangleleft(c\pitchfork I) \\
@. @VV{\epsilon^{I}_{c}}V \\
@. c
\end{CD} c ⋔ I λ c ⋔ I − 1 I ◃ ( c ⋔ I ) ↓ ⏐ ϵ c I c
where the inverse of this right unitor is given by a morphism c → ( c ⋔ I ) c\to(c\pitchfork I) c → ( c ⋔ I ) that is adjoint to the left unitor I ◃ c → c I\triangleleft c\to c I ◃ c → c . In particular, this inverse right unitor fits in the diagram
I ◃ c → I ◃ ( c ⋔ I ) λ ↓ ↓ ϵ c I c = c \begin{CD}
I\triangleleft c @>>> I\triangleleft(c\pitchfork I) \\
@V{\lambda}VV @VV{\epsilon^{I}_{c}}V \\
c @= c
\end{CD} I ◃ c λ ↓ ⏐ c I ◃ ( c ⋔ I ) ↓ ⏐ ϵ c I c
Similarly, the associator of the right module structure on C \mathcal{C} C is given by a map ( c ◃ x ) ◃ y → c ◃ ( x ⊗ y ) ∈ C (c \triangleleft x) \triangleleft y \to c \triangleleft(x \otimes y) \in\mathcal{C} ( c ◃ x ) ◃ y → c ◃ ( x ⊗ y ) ∈ C which fits into the diagram
( x ⊗ y ) ◃ ( ( c ◃ x ) ◃ y ) → ( x ⊗ y ) ◃ ( c ◃ ( x ⊗ y ) ) a ↓ ↓ ε c x ⊗ y x ◃ ( y ◃ ( ( c ◃ x ) ◃ y ) ) C 1 ◃ ε c ◃ x y ↓ x ◃ ( c ◃ x ) → ε c x c \begin{CD}
(x \otimes y) \triangleleft((c \triangleleft x) \triangleleft y) @>>> (x \otimes y) \triangleleft(c \triangleleft(x \otimes y)) \\
@V{a}VV @VV{\varepsilon_{c}^{x \otimes y}}V \\
x \triangleleft(y \triangleleft((c \triangleleft x) \triangleleft y)) @. \mathcal{C} \\
@V{1 \triangleleft\varepsilon_{c \triangleleft x}^{y}}VV @. \\
x \triangleleft(c \triangleleft x) @>{\varepsilon_{c}^{x}}>> c
\end{CD} ( x ⊗ y ) ◃ (( c ◃ x ) ◃ y ) a ↓ ⏐ x ◃ ( y ◃ (( c ◃ x ) ◃ y )) 1 ◃ ε c ◃ x y ↓ ⏐ x ◃ ( c ◃ x ) ε c x ( x ⊗ y ) ◃ ( c ◃ ( x ⊗ y )) ↓ ⏐ ε c x ⊗ y C c
This lemma can be readily dualized to provide a left V \mathcal{V} V -module structure by tensoring given a right V o p \mathcal{V}^{\mathrm{op}} V op -module structure by powering. Indeed, a right V o p \mathcal{V}^{\mathrm{op}} V op -module C \mathcal{C} C is equivalent to specifying a left τ ( V ) \tau(\mathcal{V}) τ ( V ) -module C o p \mathcal{C}^{\mathrm{op}} C op . Now, we just apply the previous lemma to this equivalent specification to obtain the desired result. Expectedly, it is also true that an enrichment can be defined solely in terms of the powering without requiring the existence of a tensoring operation.
We can now begin discussing the slicing of closed module structures.
Lemma 5.1.12. Let C \mathcal{C} C be a left V \mathcal{V} V -module, let ( M , μ , e ) ∈ V (M,\mu,e) \in\mathcal{V} ( M , μ , e ) ∈ V be a monoid object, and let ( Z , ◃ Z ) ∈ C (Z,\triangleleft_Z) \in\mathcal{C} ( Z , ◃ Z ) ∈ C be a left module object over the monoid M M M . And assume that V \mathcal{V} V has pullbacks.
If the category V \mathcal{V} V is left closed, then the category V / M \mathcal{V}/M V / M is also left closed using the sliced monoidal product. The same is true for right closure.
If C \mathcal{C} C is a left closed left V \mathcal{V} V -module then C / Z \mathcal{C}/Z C / Z is a left closed left-V / M \mathcal{V}/M V / M module. The same is true for right closure as long as C \mathcal{C} C has pullbacks.
Proof.
We define the left closed internal hom V / M ( α : X → M , β : Y → M ) \mathcal{V}/M(\alpha:X \to M,\beta:Y \to M) V / M ( α : X → M , β : Y → M ) and right closed powering β ◃ / α \beta\triangleleft/ \alpha β ◃ / α using the pullback diagrams
\begin{CD}
\mathcal{V}/M(\alpha,\beta) @>{\Gamma}>> \mathcal{V}(X,Y) \\
@VVV @VV{\beta_*}V \\
M @>{\mu}>> \mathcal{V}(M,M) @>{\alpha_*}>> \mathcal{V}(X,M)
\end{CD}
\qquad
\begin{CD}
\beta\triangleleft/ \alpha@>{\Gamma}>> Y \triangleleft X \\
@VVV @VV{\beta\triangleleft X}V \\
M @>{\mu}>> M \triangleleft M @>{M \triangleleft\alpha}>> M \triangleleft X
\end{CD}
Thereafter, maps γ : Z → M \gamma:Z \to M γ : Z → M into V / M ( α , β ) → M \mathcal{V}/M(\alpha,\beta) \to M V / M ( α , β ) → M be can identified with maps γ ⊗ / M α → β \gamma\otimes/M \alpha\to\beta γ ⊗ / M α → β by using the left closure adjunction of V \mathcal{V} V . The right closure is same.
Let α : X → M \alpha: X \to M α : X → M , and f , g : A , B → Z f,g : A,B \to Z f , g : A , B → Z . The definition of internal hom and powering is similarly given pullback diagrams
\begin{CD}
\mathcal{C}/Z(f,g) @>>> \mathcal{C}(A,B) \\
@VVV @VV{g_*}V \\
M @>{\triangleleft_Z}>> \mathcal{C}(Z,Z) @>{f_*}>> \mathcal{C}(A,Z)
\end{CD}
\qquad
\begin{CD}
g \mathbin{\pitchfork} \alpha@>>> B \mathbin{\pitchfork} X \\
@VVV @VV{g\mathbin{\pitchfork}X}V \\
Z @>{\triangleleft_Z}>> Z\mathbin{\pitchfork}M @>{Z\mathbin{\pitchfork}\alpha}>> Z\mathbin{\pitchfork}X
\end{CD}
The verification of adjunction is similar by using the closure adjunctions of left V \mathcal{V} V -module C \mathcal{C} C .
A dual version of the above lemma is not possible for a comonoid N N N as there is no transpose for the comultiplication N → N ⊗ N N \to N \otimes N N → N ⊗ N . However, we can still pursue a form of duality, if we instead start with a right V op \mathcal{V}^{\operatorname{op}} V op -module ( C , ⋔ ) (\mathcal{C},\mathbin{\pitchfork}) ( C , ⋔ ) . This right module structure is meant to encode powering. Thus, we merely ask whether this powering admits tensors using left adjoints or internal homs using mutual right adjoints. More precisely, we use the op 1 \operatorname{op}_1 op 1 duality to encode this as a right V \mathcal{V} V -module C op \mathcal{C}^{\operatorname{op}} C op . Then, equivalently, we ask if this right module is left closed for admitting tensors or right closed for admitting internal homs.
Lemma 5.1.13. Let C op \mathcal{C}^{\operatorname{op}} C op be a right V \mathcal{V} V -module using ( − ⋔ − ) op (-\mathbin{\pitchfork}-)^{\operatorname{op}} ( − ⋔ − ) op , let ( M , μ , e ) ∈ V (M,\mu,e)\in\mathcal{V} ( M , μ , e ) ∈ V be a monoid object, and let ( Z , ▹ Z ) ∈ C op (Z,\triangleright_Z)\in\mathcal{C}^{\operatorname{op}} ( Z , ▹ Z ) ∈ C op be a right module object over the monoid M M M (equivalently, a right comodule object ( Z , ▹ Z ) ∈ C (Z,\triangleright_Z)\in\mathcal{C} ( Z , ▹ Z ) ∈ C over comonoid M ∈ V op M\in\mathcal{V}^{\operatorname{op}} M ∈ V op ). And assume that V \mathcal{V} V has pullbacks.
There is a canonical right V / M \mathcal{V}/M V / M -module structure on C op / Z \mathcal{C}^{\operatorname{op}}/Z C op / Z , equivalently on ( Z / C ) op (Z/\mathcal{C})^{\operatorname{op}} ( Z / C ) op . Under the op 1 \operatorname{op}_1 op 1 duality, this is a right ( V / M ) op (\mathcal{V}/M)^{\operatorname{op}} ( V / M ) op -module structure on Z / C Z/\mathcal{C} Z / C and defines sliced powering.
If C op \mathcal{C}^{\operatorname{op}} C op is a right closed right V \mathcal{V} V -module then ( Z / C ) op (Z/\mathcal{C})^{\operatorname{op}} ( Z / C ) op is a right closed right V / M \mathcal{V}/M V / M -module. This defines a sliced internal hom Z / C ( − , − ) : ( Z / C ) op × ( Z / C ) → V / M Z/\mathcal{C}(-,-):(Z/\mathcal{C})^{\operatorname{op}}\times(Z/\mathcal{C})\to\mathcal{V}/M Z / C ( − , − ) : ( Z / C ) op × ( Z / C ) → V / M .
If C op \mathcal{C}^{\operatorname{op}} C op is a left closed right V \mathcal{V} V module then ( Z / C ) op (Z/\mathcal{C})^{\operatorname{op}} ( Z / C ) op is a left closed right V / M \mathcal{V}/M V / M -module as long as C \mathcal{C} C has pushouts (equivalently C op \mathcal{C}^{\operatorname{op}} C op has pullbacks). This defines a sliced left action functor ( − ◃ − ) : V / M × Z / C → Z / C (-\triangleleft-):\mathcal{V}/M\times Z/\mathcal{C}\to Z/\mathcal{C} ( − ◃ − ) : V / M × Z / C → Z / C .
Proof. This whole lemma is merely a direct application of the definition 5.1.12 under appropriate dualities. Thus, we only define the functors for clarity. Let α : X → M ∈ V / M \alpha:X\to M\in\mathcal{V}/M α : X → M ∈ V / M and f , g : Z → A , B ∈ Z / C f,g:Z\to A,B\in Z/\mathcal{C} f , g : Z → A , B ∈ Z / C .
The sliced powering g α g^\alpha g α is defined by the composite
Z → g ⋔ / α B ⋔ X ▹ Z ↓ ↑ g ⋔ X Z ⋔ M → Z ⋔ α Z ⋔ X \begin{CD}
Z @>{g\mathbin{\pitchfork}/\alpha}>> B\mathbin{\pitchfork}X \\
@V{\triangleright_Z}VV @AA{g\mathbin{\pitchfork}X}A \\
Z\mathbin{\pitchfork}M @>{Z\mathbin{\pitchfork}\alpha}>> Z\mathbin{\pitchfork}X
\end{CD} Z ▹ Z ↓ ⏐ Z ⋔ M g ⋔ / α Z ⋔ α B ⋔ X ⏐ ↑ g ⋔ X Z ⋔ X
The internal hom Z / C ( f , g ) Z/\mathcal{C}(f,g) Z / C ( f , g ) is defined by the pullback
Z / C ( f , g ) → C ( A , B ) ↓ ↓ f ∗ M → ▹ Z C ( Z , Z ) → g ∗ C ( Z , B ) \begin{CD}
Z/\mathcal{C}(f,g) @>>> \mathcal{C}(A,B) \\
@VVV @VV{f^*}V \\
M @>{\triangleright_Z}>> \mathcal{C}(Z,Z) @>{g_*}>> \mathcal{C}(Z,B)
\end{CD} Z / C ( f , g ) ↓ ⏐ M ▹ Z C ( A , B ) ↓ ⏐ f ∗ C ( Z , Z ) g ∗ C ( Z , B )
The sliced tensoring α ◃ Z f \alpha\triangleleft_Z f α ◃ Z f is defined by the pushout
X ◃ Z → 1 ◃ f X ◃ A α ◃ 1 ↓ ↓ M ◃ Z ↓ ▹ Z → α ◃ Z f \begin{CD}
X\triangleleft Z @>{1\triangleleft f}>> X\triangleleft A \\
@V{\alpha\triangleleft1}VV @VVV \\
M\triangleleft Z @VV{\triangleright_Z}V @. \\
Z @>{\alpha\triangleleft_Z f}>> \alpha\triangleleft_Z f
\end{CD} X ◃ Z α ◃ 1 ↓ ⏐ M ◃ Z α ◃ Z f 1 ◃ f ↓ ⏐ ▹ Z X ◃ A ↓ ⏐
In the previous two lemma, we obtained the sliced enrichment structures by slicing the tensoring and by slicing the powering, respectively. Therefore, a natural question that remains is if there exists a slicing of the enrichment itself. We motivate this through the following lemma.
Lemma 5.1.14. Let C \mathcal{C} C be a closed left V \mathcal{V} V -module and ( M , μ , e ) (M,\mu,e) ( M , μ , e ) a monoid object in V \mathcal{V} V . The following data are equivalent.
A left module object ( Z , M ◃ Z → Z ) (Z,M\triangleleft Z\to Z) ( Z , M ◃ Z → Z ) over the monoid object M ∈ V M\in\mathcal{V} M ∈ V .
A morphism of monoid objects M → C ( Z , Z ) M\to\mathcal{C}(Z,Z) M → C ( Z , Z ) .
A right comodule object ( Z , Z → Z ⋔ M ) (Z,Z\to Z\pitchfork M) ( Z , Z → Z ⋔ M ) over the comonoid object M ∈ V o p M\in\mathcal{V}^{\mathrm{op}} M ∈ V op .
Proof. We define the monoid morphism M → C ( Z , Z ) M\to\mathcal{C}(Z,Z) M → C ( Z , Z ) as adjoint to the morphism M ◃ Z → Z M\triangleleft Z\to Z M ◃ Z → Z . In particular, we have a commuting diagram
M ◃ Z → C ( Z , Z ) ◃ Z ↓ ↓ eval Z Z Z \begin{CD}
M\triangleleft Z @>>> \mathcal{C}(Z,Z)\triangleleft Z \\
@VVV @VV{\operatorname{eval}^{Z}_{Z}}V \\
@. Z
\end{CD} M ◃ Z ↓ ⏐ C ( Z , Z ) ◃ Z ↓ ⏐ eval Z Z Z
Similarly, we define the comodule map as an adjoint to the morphism M ◃ Z → Z M\triangleleft Z\to Z M ◃ Z → Z , which fits into the following commuting diagram
M ◃ Z → M ◃ ( Z ⋔ M ) ↓ ↓ ϵ Z M Z \begin{CD}
M\triangleleft Z @>>> M\triangleleft(Z\pitchfork M) \\
@VVV @VV{\epsilon^{M}_{Z}}V \\
@. Z
\end{CD} M ◃ Z ↓ ⏐ M ◃ ( Z ⋔ M ) ↓ ⏐ ϵ Z M Z
The rest is simply an exercise in taking transposes of the defining axioms of the left module structure of ( Z , M ◃ Z → Z ) (Z,M \triangleleft Z \to Z) ( Z , M ◃ Z → Z ) and using the above two diagrams to simplify.
In particular, this suggests that we should be able to perform slice construction on arbitrary enriched categories C \mathcal{C} C over some base V \mathcal{V} V , as long as we have a morphism of monoids M → C ( Z , Z ) M \to\mathcal{C}(Z,Z) M → C ( Z , Z ) . This is indeed true.
Proposition 5.1.15. Consider a monoidal category V \mathcal{V} V with pullbacks. Let C \mathcal{C} C be any enriched category and assume that there exists a morphism of monoid objects r : M → C ( Z , Z ) r:M \to\mathcal{C}(Z,Z) r : M → C ( Z , Z ) for some Z ∈ C Z \in\mathcal{C} Z ∈ C . There exists canonical slice enriched categories C / Z \mathcal{C}/Z C / Z and Z / C Z/\mathcal{C} Z / C enriched over the slice monoidal category V / M \mathcal{V}/M V / M . Furthermore, the underlying category of both C / Z \mathcal{C}/Z C / Z and Z / C Z/\mathcal{C} Z / C over V / M \mathcal{V}/M V / M (and not V \mathcal{V} V ) is isomorphic to the corresponding slices categories C / Z \mathcal{C}/Z C / Z and Z / C Z/\mathcal{C} Z / C .
Proof. Let the maps f ∗ f_\ast f ∗ and f ∗ f^\ast f ∗ be formed by taking the unenriched representables. We only define the compositor and the unitor for the case of C / Z \mathcal{C}/Z C / Z enrichment over V / M \mathcal{V}/M V / M . The case for Z / C Z/\mathcal{C} Z / C is identical.
The sliced internal homs are defined by the pullbacks
C / Z ( f , g ) → C ( A , B ) ↓ ↓ g ∗ M → r C ( Z , Z ) → f ∗ C ( A , Z ) \begin{CD}
\mathcal{C}/Z(f,g) @>{}>> \mathcal{C}(A,B) \\
@V{}VV @VV{g_\ast}V \\
M @>{r}>> \mathcal{C}(Z,Z) @>{f^\ast}>> \mathcal{C}(A,Z)
\end{CD} C / Z ( f , g ) ↓ ⏐ M r C ( A , B ) ↓ ⏐ g ∗ C ( Z , Z ) f ∗ C ( A , Z )
The unitors are defined by the unique universal maps i : I → C ( A , A ) i:I \to\mathcal{C}(A,A) i : I → C ( A , A ) as given in the diagrams
I → 1 A C ( A , A ) i ↓ ↓ f ∗ C / Z ( f , f ) → C ( A , A ) ↓ ↓ f ∗ M → r C ( Z , Z ) → f ∗ C ( A , Z ) \begin{CD}
I @>{1_A}>> \mathcal{C}(A,A) \\
@V{i}VV @VV{f_\ast}V \\
\mathcal{C}/Z(f,f) @>{}>> \mathcal{C}(A,A) \\
@V{}VV @VV{f_\ast}V \\
M @>{r}>> \mathcal{C}(Z,Z) @>{f^\ast}>> \mathcal{C}(A,Z)
\end{CD} I i ↓ ⏐ C / Z ( f , f ) ↓ ⏐ M 1 A r C ( A , A ) ↓ ⏐ f ∗ C ( A , A ) ↓ ⏐ f ∗ C ( Z , Z ) f ∗ C ( A , Z )
where the outer rectangle commutes because r r r is a morphism of monoids.
Finally, the compositors C / Z ( g , h ) ⊗ C / Z ( f , g ) → C / Z ( f , h ) \mathcal{C}/Z(g,h) \otimes\mathcal{C}/Z(f,g) \to\mathcal{C}/Z(f,h) C / Z ( g , h ) ⊗ C / Z ( f , g ) → C / Z ( f , h ) are the unique universal maps induced by the following commuting diagrams
C / Z ( g , h ) ⊗ C / Z ( f , g ) → C / Z ( g , h ) ⊗ C ( A , B ) → C ( B , C ) ⊗ C ( A , B ) → ∘ C ( A , C ) ↓ ↓ h ∗ ⊗ 1 ↑ ↑ h ∗ M ⊗ C / Z ( f , g ) → M ⊗ C ( A , B ) → r ⊗ 1 C ( Z , Z ) ⊗ C ( A , B ) → C ( A , Z ) ↓ 1 ⊗ g ∗ ↓ 1 ⊗ g ∗ ↓ ↑ f ∗ M ⊗ M → 1 ⊗ r M ⊗ C ( Z , Z ) → r ⊗ 1 C ( Z , Z ) ⊗ C ( Z , Z ) → C ( Z , Z ) μ ↓ M → r C ( Z , Z ) \begin{CD}
\mathcal{C}/Z(g,h)\otimes\mathcal{C}/Z(f,g) @>>> \mathcal{C}/Z(g,h)\otimes\mathcal{C}(A,B) @>>> \mathcal{C}(B,C)\otimes\mathcal{C}(A,B) @>{\circ}>> \mathcal{C}(A,C) \\
@VVV @VVV @A{h_*\otimes1}AA @AA{h_*}A \\
M\otimes\mathcal{C}/Z(f,g) @>>> M\otimes\mathcal{C}(A,B) @>{r\otimes1}>> \mathcal{C}(Z,Z)\otimes\mathcal{C}(A,B) @>>> \mathcal{C}(A,Z) \\
@VVV @V{1\otimes g_*}VV @V{1\otimes g_*}VV @AA{f_*}A \\
M\otimes M @>{1\otimes r}>> M\otimes\mathcal{C}(Z,Z) @>{r\otimes1}>> \mathcal{C}(Z,Z)\otimes\mathcal{C}(Z,Z) @>>> \mathcal{C}(Z,Z) \\
@V{\mu}VV @. @. @. \\
M @>{r}>> \mathcal{C}(Z,Z)
\end{CD} C / Z ( g , h ) ⊗ C / Z ( f , g ) ↓ ⏐ M ⊗ C / Z ( f , g ) ↓ ⏐ M ⊗ M μ ↓ ⏐ M 1 ⊗ r r C / Z ( g , h ) ⊗ C ( A , B ) ↓ ⏐ M ⊗ C ( A , B ) 1 ⊗ g ∗ ↓ ⏐ M ⊗ C ( Z , Z ) C ( Z , Z ) r ⊗ 1 r ⊗ 1 C ( B , C ) ⊗ C ( A , B ) h ∗ ⊗ 1 ⏐ ↑ C ( Z , Z ) ⊗ C ( A , B ) 1 ⊗ g ∗ ↓ ⏐ C ( Z , Z ) ⊗ C ( Z , Z ) ∘ C ( A , C ) ⏐ ↑ h ∗ C ( A , Z ) ⏐ ↑ f ∗ C ( Z , Z )
For the associativity and unitality, first note that U : V / M → V U:\mathcal{V}/M\to\mathcal{V} U : V / M → V is a faithful strict monoidal functor. Thus, f = g f=g f = g in V / M \mathcal{V}/M V / M iff U ( f ) = U ( g ) U(f)=U(g) U ( f ) = U ( g ) in V \mathcal{V} V . Second, universal maps defined through pullbacks are uniquely characterized by the two projection maps.
Using these two facts we shall, for instance, demonstrate the left unitality. The associativity is quite similar but only technically more involved. Now, note the commuting diagram
I ⊗ C / Z ( f , g ) → λ C / Z ( f , g ) ↓ ↓ C / Z ( g , g ) ⊗ C / Z ( f , g ) → C / Z ( f , g ) C ( A , B ) C / Z ( g , g ) ⊗ C ( A , B ) → C ( B , B ) ⊗ C ( A , B ) → C ( A , B ) ↑ ↑ 1 B ⊗ 1 I ⊗ C ( A , B ) → C ( B , B ) ⊗ C ( A , B ) C / Z ( g , g ) ⊗ C / Z ( f , g ) → C / Z ( f , g ) → C ( A , B ) \begin{CD}
I\otimes\mathcal{C}/Z(f,g) @>{\lambda}>> \mathcal{C}/Z(f,g) \\
@VVV @VVV \\
\mathcal{C}/Z(g,g)\otimes\mathcal{C}/Z(f,g) @>>> \mathcal{C}/Z(f,g) \\
@. @. \\
@. \mathcal{C}(A,B) \\
@. @. \\
\mathcal{C}/Z(g,g)\otimes\mathcal{C}(A,B) @>>> \mathcal{C}(B,B)\otimes\mathcal{C}(A,B) @>>> \mathcal{C}(A,B) \\
@. @AAA @AA{1_B\otimes1}A \\
@. I\otimes\mathcal{C}(A,B) @>>> \mathcal{C}(B,B)\otimes\mathcal{C}(A,B) \\
@. @. @. \\
\mathcal{C}/Z(g,g)\otimes\mathcal{C}/Z(f,g) @>>> \mathcal{C}/Z(f,g) @>>> \mathcal{C}(A,B)
\end{CD} I ⊗ C / Z ( f , g ) ↓ ⏐ C / Z ( g , g ) ⊗ C / Z ( f , g ) C / Z ( g , g ) ⊗ C ( A , B ) C / Z ( g , g ) ⊗ C / Z ( f , g ) λ C / Z ( f , g ) ↓ ⏐ C / Z ( f , g ) C ( A , B ) C ( B , B ) ⊗ C ( A , B ) ⏐ ↑ I ⊗ C ( A , B ) C / Z ( f , g ) C ( A , B ) ⏐ ↑ 1 B ⊗ 1 C ( B , B ) ⊗ C ( A , B ) C ( A , B )
implies that the left unitality holds under taking postcomposite with one of the projection maps.
Similarly, we have the diagram
I ⊗ C / Z ( f , g ) ⟶ C / Z ( g , g ) ⊗ C / Z ( f , g ) ↓ λ ↘ e ⊗ 1 ↙ ↓ C / Z ( f , g ) M ⊗ C / Z ( f , g ) C / Z ( f , g ) ↘ I ⊗ M → e ⊗ 1 M ⊗ M ↙ ↓ λ M \begin{array}{ccccc}
I \otimes\mathcal{C}/Z(f,g) & \longrightarrow& \mathcal{C}/Z(g,g)\otimes\mathcal{C}/Z(f,g) \\
\downarrow\lambda& \searrow e\otimes1 & \swarrow& \downarrow\\
\mathcal{C}/Z(f,g) & & M\otimes\mathcal{C}/Z(f,g) & & \mathcal{C}/Z(f,g) \\
& \searrow& I\otimes M \xrightarrow{e\otimes1} M\otimes M & \swarrow& \\
& & \downarrow\lambda& & \\
& & M & &
\end{array} I ⊗ C / Z ( f , g ) ↓ λ C / Z ( f , g ) ⟶ ↘ e ⊗ 1 ↘ C / Z ( g , g ) ⊗ C / Z ( f , g ) ↙ M ⊗ C / Z ( f , g ) I ⊗ M e ⊗ 1 M ⊗ M ↓ λ M ↓ ↙ C / Z ( f , g )
for compatibility with the other projection map. Finally, these two commuting diagrams jointly imply the desired commuting diagram
C / Z ( g , g ) ⊗ C / Z ( f , g ) ↙ ↘ I ⊗ C / Z ( f , g ) → λ C / Z ( f , g ) \begin{array}{ccc}
& \mathcal{C}/Z(g,g)\otimes\mathcal{C}/Z(f,g) & \\
\swarrow& & \searrow\\
I\otimes\mathcal{C}/Z(f,g) & \xrightarrow{\lambda} & \mathcal{C}/Z(f,g)
\end{array} ↙ I ⊗ C / Z ( f , g ) C / Z ( g , g ) ⊗ C / Z ( f , g ) λ ↘ C / Z ( f , g )
for left unitality. As already mentioned, the verification for associativity and the underlying category is mostly similar.
Let C \mathcal{C} C be any V \mathcal{V} V enriched category. There is a canonical monoid structure on the monoidal unit I I I of V \mathcal{V} V given by μ : I ⊗ I → λ I = ρ I I \mu: I\otimes I \xrightarrow{\lambda_I=\rho_I} I μ : I ⊗ I λ I = ρ I I and η : I → 1 I I \eta: I \xrightarrow{1_I} I η : I 1 I I . In particular, V / I \mathcal{V}/I V / I is a semicartesian monoidal category. That is, the induced monoidal unit I → I I\to I I → I is a terminal object of V / I \mathcal{V}/I V / I . Furthermore, for any Z ∈ C Z\in\mathcal{C} Z ∈ C , the unitor map I → C ( Z , Z ) I\to\mathcal{C}(Z,Z) I → C ( Z , Z ) is a morphism of monoid objects. This readily provides the following corollary.
Corollary 5.1.16. There exists canonical enriched slice categories C / Z \mathcal{C}/Z C / Z , Z / C Z/\mathcal{C} Z / C over V / I \mathcal{V}/I V / I for every Z ∈ C Z\in\mathcal{C} Z ∈ C .
Corollary 5.1.17. Let V \mathcal{V} V be a closed symmetric monoidal category, and let C \mathcal{C} C be enriched, tensored, and cotensored over V \mathcal{V} V . Assume the relevant pullbacks and pushouts exist. Then, both the categories C / Z \mathcal{C}/Z C / Z and Z / C Z/\mathcal{C} Z / C are enriched, tensored, and cotensored over V / I \mathcal{V}/I V / I for every Z ∈ C Z\in\mathcal{C} Z ∈ C . More generally, this is also true for a commutative monoid object M M M .
Model structures and slices
Monoidal model categories and enriched model categories are an important tool in abstract homotopy theory. Their defining axioms are closely linked with the two variable adjunction defining the closure of left V \mathcal{V} V -modules. Hence, we can also analyze the slicing of enriched model structures.
Definition 5.2.1. A monoidal model category is a closed monoidal category ( V , ⊗ , I ) (\mathcal{V}, \otimes, I) ( V , ⊗ , I ) together with a model structure on its underlying category V \mathcal{V} V that satisfies:
Unit axiom: Let Q I QI Q I be any cofibrant replacement of I I I . The induced maps Q I ⊗ X → I ⊗ X QI \otimes X \to I \otimes X Q I ⊗ X → I ⊗ X and X ⊗ Q I → X ⊗ I X \otimes QI \to X \otimes I X ⊗ Q I → X ⊗ I are weak equivalences, whenever X X X is cofibrant in V \mathcal{V} V .
Pushout product axiom: Let α : X 1 → Y 1 \alpha: X_1 \to Y_1 α : X 1 → Y 1 and β : X 2 → Y 2 \beta: X_2 \to Y_2 β : X 2 → Y 2 be two arbitrary maps in V \mathcal{V} V . The pushout-product map α □ β \alpha\mathbin{\Box} \beta α □ β is a cofibration whenever both α \alpha α and β \beta β are cofibrations. Furthermore, if one of α \alpha α or β \beta β is also a weak equivalence then so is α □ β \alpha\mathbin{\Box} \beta α □ β .
Remark 5.2.2. The closedness assumption of the monoidal structure on V \mathcal{V} V is not entirely redundant. One of its important consequences is that the monoidal product ⊗ \otimes ⊗ preserves the initial object in both of its arguments. In fact, it is sufficient to assume only this additional property for constructing the derived monoidal product using a left deformation. See [28, Sec. 10.2] or [20, Prop. 4.3.1] for more details.
We now verify that this extends to a natural monoidal model structure for V / M \mathcal{V}/M V / M .
Proposition 5.2.3. Let ( V , ⊗ , I ) (\mathcal{V}, \otimes, I) ( V , ⊗ , I ) be a monoidal model category , ( M , μ , η ) (M, \mu, \eta) ( M , μ , η ) be a monoid object inside of this monoidal category. Then, ( V / M , ⊗ / M , I / M ) (\mathcal{V}/M, \otimes/M, I/M) ( V / M , ⊗ / M , I / M ) is also a monoidal model category equipped with the slice model structure.
Proof. The slice model structure is defined by the property that all of its cofibrations, fibrations, and weak equivalences are created under the forgetful functor U : V / M → V U : \mathcal{V}/M \to\mathcal{V} U : V / M → V .
Unit Axiom: Any cofibrant replacement of η : I → M \eta: I \to M η : I → M must also be created under the forgetful functor U U U . In particular, it must be of the form Q I → I → M QI \to I \to M Q I → I → M , where we take the functorial cofibrant replacement in V \mathcal{V} V itself. Now let X → M ∈ V / M X \to M \in\mathcal{V}/M X → M ∈ V / M be any cofibrant object. We obtain that the following diagram which demonstrates the required unit axiom.
Q I ⊗ X → I ⊗ X ↓ ↓ M ⊗ X ↓ ↓ M ⊗ M ↓ M \begin{CD}
QI \otimes X @>>> I \otimes X \\
@VVV @VVV \\
M \otimes X @VVV \\
@VVV \\
M \otimes M \\
@VVV \\
M
\end{CD} Q I ⊗ X ↓ ⏐ M ⊗ X ↓ ⏐ M ⊗ M ↓ ⏐ M ↓ ⏐ I ⊗ X ↓ ⏐
Pushout product axiom: The functor U U U is strict monoidal and creates all colimits. Thus, the pushout product axiom of V \mathcal{V} V lifts to that of V / M \mathcal{V}/M V / M directly. □ \square □
:::
Having obtained the monoidal model category structure on V / M \mathcal{V}/M V / M , we can now analyze the slicing of closed left V \mathcal{V} V -modules under consideration of model structures. In particular, we start with a more general definition of a V \mathcal{V} V -module model category instead of a enriched model category.
Definition 5.2.4. Let ( V , ⊗ , I ) (\mathcal{V},\otimes,I) ( V , ⊗ , I ) be a monoidal model category. A left V \mathcal{V} V -module model category C \mathcal{C} C is a closed left V \mathcal{V} V -module ( C , ◃ ) (\mathcal{C},\triangleleft) ( C , ◃ ) together with a model structure on C \mathcal{C} C that satisfies:
Tensoring unit axiom : Let Q I QI Q I be any cofibrant replacement of I I I . The induced map Q I ◃ A → I ◃ A QI\triangleleft A\to I\triangleleft A Q I ◃ A → I ◃ A is a weak equivalences, whenever A A A is cofibrant in C \mathcal{C} C .
Pushout product axiom : Let α : X → Y ∈ V \alpha:X\to Y\in\mathcal{V} α : X → Y ∈ V and f : A → B ∈ C f:A\to B\in\mathcal{C} f : A → B ∈ C be two arbitrary maps. The pushout-product map α ◃ ˉ f \alpha\mathbin{\bar{\triangleleft}}f α ◃ ˉ f is a cofibration, whenever both α \alpha α and f f f are cofibrations. Furthermore, if one of α \alpha α or f f f is a also weak equivalence then so is α ◃ ˉ f \alpha\mathbin{\bar{\triangleleft}}f α ◃ ˉ f .
Motivated by the definition 5.1.13, we also have a dual definition.
Definition 5.2.5. Let ( V , ⊗ , I ) (\mathcal{V},\otimes,I) ( V , ⊗ , I ) be a monoidal model category. A right V o p \mathcal{V}^{\mathrm{op}} V op -module model category C \mathcal{C} C is a closed right V \mathcal{V} V -module ( C o p , ⋔ o p ) (\mathcal{C}^{\mathrm{op}},\pitchfork^{\mathrm{op}}) ( C op , ⋔ op ) together with a model structure on C \mathcal{C} C that satisfies:
Powering unit axiom : Let Q I QI Q I be any cofibrant replacement of I I I . The induced map A ⋔ I → A ⋔ Q I A\pitchfork I\to A\pitchfork QI A ⋔ I → A ⋔ Q I is a weak equivalences, whenever A A A is fibrant in C \mathcal{C} C .
Pullback power axiom : Let α : X → Y ∈ V \alpha:X\to Y\in\mathcal{V} α : X → Y ∈ V and f : A → B ∈ C f:A\to B\in\mathcal{C} f : A → B ∈ C be two arbitrary maps. The pullback-power map f ⋔ ˉ α f\mathbin{\bar{\pitchfork}}\alpha f ⋔ ˉ α is a fibration, whenever α \alpha α is a cofibration and f f f is a fibrations. Furthermore, if one of α \alpha α or f f f is a also weak equivalence then so is f ⋔ ˉ α f\mathbin{\bar{\pitchfork}}\alpha f ⋔ ˉ α .
Remark 5.2.6. Under the application of both o p 1 \mathrm{op}_{1} op 1 and r r r dualities, the above definition specify precisely the tensoring unit axiom and the pushout-product axioms for the corresponding left r ( V ) r(\mathcal{V}) r ( V ) -module structure on C o p \mathcal{C}^{\mathrm{op}} C op . In particular, the above definition is equivalent to specifying a left r ( V ) r(\mathcal{V}) r ( V ) -module model category C o p \mathcal{C}^{\mathrm{op}} C op . The primary reason for having distinct axioms is that while r ( V ) r(\mathcal{V}) r ( V ) is again a monoidal model category, o p 1 ( V ) \mathrm{op}_{1}(\mathcal{V}) op 1 ( V ) is not, in general.
As before, there is also a correspondence between left V \mathcal{V} V -module model categories and right V o p \mathcal{V}^{\mathrm{op}} V op -module model categories. This primarily follows from definition 5.1.11 together with its dual and [20, Lemmas 4.2.2 and 4.2.7].
Theorem 5.2.7. Let ( V , ⊗ , I ) (\mathcal{V},\otimes,I) ( V , ⊗ , I ) be a monoidal model category and let ( C , ◃ ) (\mathcal{C},\triangleleft) ( C , ◃ ) be a left V \mathcal{V} V -module model category. Let ( M , μ , η ) (M,\mu,\eta) ( M , μ , η ) be a monoid object in V \mathcal{V} V and let ( Z , λ ) (Z,\lambda) ( Z , λ ) be a left M M M -module in C \mathcal{C} C . Then, it follows that ( C / Z , ◃ / Z ) (\mathcal{C}/Z,\triangleleft/Z) ( C / Z , ◃ / Z ) is a V / M \mathcal{V}/M V / M -model category.
Proof. The slice model structure on C / Z \mathcal{C}/Z C / Z is again defined by the property that all of the cofibrations, fibrations, and weak equivalences are created under the forgetful functor U : C / Z → C U:\mathcal{C}/Z\to\mathcal{C} U : C / Z → C .
The rest of the proof is also similar to the case of a monoidal model category. Indeed, the idea of forgetful functors being “strict monoidal” is now encoded by the commuting diagram
V / M × C / Z → ◃ Z C / Z U × U ↓ ↓ U V × C → ◃ C \begin{CD}
\mathcal{V}/M \times\mathcal{C}/Z @>{\triangleleft_Z}>> \mathcal{C}/Z \\
@V{U \times U}VV @VV{U}V \\
\mathcal{V} \times\mathcal{C} @>{\triangleleft}>> \mathcal{C}
\end{CD} V / M × C / Z U × U ↓ ⏐ V × C ◃ Z ◃ C / Z ↓ ⏐ U C
It is not possible to use the previous technique in the case of under categories Z / C Z/\mathcal{C} Z / C . In particular, we only have the commuting diagram
V / M × Z / C → ◃ Z Z / C U × U ↓ ↓ U V × C → ◃ C \begin{CD}
\mathcal{V}/M \times Z/\mathcal{C} @>{\triangleleft_Z}>> Z/\mathcal{C} \\
@V{U \times U}VV @VV{U}V \\
\mathcal{V} \times\mathcal{C} @>{\triangleleft}>> \mathcal{C}
\end{CD} V / M × Z / C U × U ↓ ⏐ V × C ◃ Z ◃ Z / C ↓ ⏐ U C
up to a 2 2 2 -cell l : X ◃ A → α ◃ Z f l\colon X \triangleleft A \to\alpha\triangleleft_Z f l : X ◃ A → α ◃ Z f , encoding a “lax monoidal” functor.
We can attempt to remedy this situation by demanding that the canonical map l l l be a natural weak equivalence. While this may potentially solve the issue, it is not a desirable assumption in practice. Instead, as before, the solution is to dualize the results with right V o p \mathcal{V}^{\mathrm{op}} V op -module structures.
Theorem 5.2.8. Let ( V , ⊗ , I ) (\mathcal{V}, \otimes, I) ( V , ⊗ , I ) be a monoidal model category and let ( C , ⊲ ) (\mathcal{C}, \mathbin{\lhd}) ( C , ⊲ ) be a right V o p \mathcal{V}^{\mathrm{op}} V op -module model category. Let ( M , μ , η ) (M, \mu, \eta) ( M , μ , η ) be a monoid object in V \mathcal{V} V and let ( Z , ρ ) (Z, \rho) ( Z , ρ ) be a right M M M -module in C o p \mathcal{C}^{\mathrm{op}} C op . Then, it follows that ( Z / C , ⊲ / Z ) (Z/\mathcal{C}, \mathbin{\lhd}/Z) ( Z / C , ⊲ / Z ) is a right ( V / M ) o p (\mathcal{V}/M)^{\mathrm{op}} ( V / M ) op -module model category. Consequentially, this automatically implies ( Z / C , ◃ Z ) (Z/\mathcal{C}, \triangleleft_Z) ( Z / C , ◃ Z ) is a left V / M \mathcal{V}/M V / M -module model category.
Proof. The idea of forgetful functors being “strict monoidal” is now encoded by the following commuting diagram
Z / C × ( V / M ) o p → ⊲ / Z Z / C U × U o p ↓ ↓ U C × V o p → ⊲ C \begin{CD}
Z/\mathcal{C} \times(\mathcal{V}/M)^{\mathrm{op}} @>{\mathbin{\lhd}/Z}>> Z/\mathcal{C} \\
@V{U \times U^{\mathrm{op}}}VV @VV{U}V \\
\mathcal{C} \times\mathcal{V}^{\mathrm{op}} @>{\mathbin{\lhd}}>> \mathcal{C}
\end{CD} Z / C × ( V / M ) op U × U op ↓ ⏐ C × V op ⊲ / Z ⊲ Z / C ↓ ⏐ U C
The functor U : Z / C → C U\colon Z/\mathcal{C} \to\mathcal{C} U : Z / C → C creates all limit and lifts the pullback power axiom directly.
Finally, we specialize to the case of a symmetric closed monoidal category for characterizing all enriched slice model categories.
Theorem 5.2.9. Let ( V , ⊗ , I ) (\mathcal{V}, \otimes, I) ( V , ⊗ , I ) be a closed symmetric monoidal model category and let C \mathcal{C} C be a V \mathcal{V} V -enriched model category with the underlying category C \mathcal{C} C . Let ( M , μ , η ) (M, \mu, \eta) ( M , μ , η ) be a commutative monoid object in V \mathcal{V} V and let ( Z , λ ) (Z, \lambda) ( Z , λ ) be a left M M M -module in C \mathcal{C} C relative to tensoring. Then, it follows that both Z / C Z/\mathcal{C} Z / C and C / Z \mathcal{C}/Z C / Z are V / M \mathcal{V}/M V / M -enriched model categories.
Strictification functor of bicategories
There are many sources (such as [12, 18, 21]) that discuss the strictification trifunctor s t : B i c a t → 2 C a t s t , p s \mathrm{st}: \mathbf{Bicat} \to\mathbf{2Cat}_{\mathrm{st,ps}} st : Bicat → 2Cat st , ps and its supposed right triadjoint i : 2 C a t s t , p s → B i c a t i: \mathbf{2Cat}_{\mathrm{st,ps}} \to\mathbf{Bicat} i : 2Cat st , ps → Bicat . Here, 2 C a t s t , p s \mathbf{2Cat}_{\mathrm{st,ps}} 2Cat st , ps is also known as Gray, and is the tricategory of strict 2-categories, strict 2-functors, pseudonatural transformations, and modifications. We aim to expand upon these descriptions by providing a more thorough account of this functor using concrete diagrams to fully specify the relevant trifunctorial data. Then, we shall briefly describe its usage to encode axioms of bicategorical structures using pasting diagrams.
Mapping of Bicategories
Each bicategory B ∈ B i c a t B \in\mathbf{Bicat} B ∈ Bicat maps to a strict 2-category called s t B \mathrm{st}B st B , which is essentially determined by a pair of maps e : i ∘ s t B → B e: i \circ\mathrm{st}B \to B e : i ∘ st B → B and s : B → i ∘ s t B s: B \to i \circ\mathrm{st}B s : B → i ∘ st B . For simplicity, we shall drop the i i i , unless required for better clarity. We first define the action of e e e on the objects and 1-cells of s t B \mathrm{st}B st B .
Definition A.1.1. The map e e e is defined on 0-cells and 1-cells as follows:
0-cells - for each b ∈ s t B b \in\mathrm{st}B b ∈ st B , e ( b ) = b ∈ B e(b)=b \in B e ( b ) = b ∈ B ,
1-cells - for each non-empty string ( f n , f n − 1 , ⋯ , f 1 ) (f_n,f_{n-1},\cdots,f_1) ( f n , f n − 1 , ⋯ , f 1 ) representing a 1-cell in s t B \mathrm{st}B st B , written right to left, e e e is defined recursively as shown below:
b → f 1 b 1 → f 2 b 2 → f 3 ⋯ ⟶ b n ↦ b → e ( f n , ⋯ , f 2 ) f 1 b n b 1 → f 2 b 2 → f 3 ⋯ ⟶ b n ↦ b 1 → e ( f n , ⋯ , f 3 ) f 2 b n ⋮ b n − 1 → f n b n ↦ b n − 1 → f n b n , \begin{aligned}
b \xrightarrow{f_1} b_1 \xrightarrow{f_2} b_2 \xrightarrow{f_3} \cdots\longrightarrow b_n &\mapsto b \xrightarrow{e(f_n,\cdots,f_2)f_1} b_n \\
b_1 \xrightarrow{f_2} b_2 \xrightarrow{f_3} \cdots\longrightarrow b_n &\mapsto b_1 \xrightarrow{e(f_n,\cdots,f_3)f_2} b_n \\
&\vdots\\
b_{n-1} \xrightarrow{f_n} b_n &\mapsto b_{n-1} \xrightarrow{f_n} b_n,
\end{aligned} b f 1 b 1 f 2 b 2 f 3 ⋯ ⟶ b n b 1 f 2 b 2 f 3 ⋯ ⟶ b n b n − 1 f n b n ↦ b e ( f n , ⋯ , f 2 ) f 1 b n ↦ b 1 e ( f n , ⋯ , f 3 ) f 2 b n ⋮ ↦ b n − 1 f n b n , and for the empty string, it is defined seperately by setting ( b → ε b b ) ↦ ( 1 b : b → b ) (b \xrightarrow{\varepsilon_b} b) \mapsto(1_b:b\to b) ( b ε b b ) ↦ ( 1 b : b → b ) .
Having partially defined e e e , we can now fully specify the 2-category s t B \mathrm{st}B st B in terms of its 2-cells, vertical composition, and horizontal composition.
Definition A.1.2. The 2-category s t B \mathrm{st}B st B has the following data:
Objects - exactly the objects of B B B ,
1-cells - free strings of arrows b → f 1 b 1 → f 2 b 2 → f 2 b 3 → f 3 ⋯ b \xrightarrow{f_1} b_1 \xrightarrow{f_2} b_2 \xrightarrow{f_2} b_3 \xrightarrow{f_3}\cdots b f 1 b 1 f 2 b 2 f 2 b 3 f 3 ⋯ , where the empty string is allowed and is labeled as b → ε b b b \xrightarrow{\varepsilon_b} b b ε b b ,
Identity 1-cells - for each b b b , the empty string b → ε b b b \xrightarrow{\varepsilon_b} b b ε b b is the horizontal identity
Hcomp 1-cells - two strings of arrows are composed by concatenation of strings,
2-cells - the 2-cells between two strings f ⃗ \vec{f} f and g ⃗ \vec{g} g are in bijection with the 2-cells α : e ( f ⃗ ) ⇒ e ( g ⃗ ) \alpha:e(\vec{f})\Rightarrow e(\vec{g}) α : e ( f ) ⇒ e ( g ) ,
s t B → e B \begin{CD}
\mathrm{st}B @>{e}>> B
\end{CD} st B e B
\begin{CD}
b @>{f_1}>> b_1 @>{f_2}>> b_2 @>{f_3}>> \cdots@>>> b_n\\
@. @. @. @. @.\\
b @>{g_1}>> b'_1 @>{g_2}>> b'_2 @>{g_3}>> \cdots@>>> b'_m=b_n
\end{CD}
b 1 → e ( f ⃗ ) b 2 α ↓ ↑ b 1 → e ( g ⃗ ) b 2 \begin{CD}
b_1 @>{e(\vec{f})}>> b_2\\
@V{\alpha}VV @AA{}A\\
b_1 @>{e(\vec{g})}>> b_2
\end{CD} b 1 α ↓ ⏐ b 1 e ( f ) e ( g ) b 2 ⏐ ↑ b 2
this bijection of 2-cells also defines the action of e e e on the 2-cells of s t B \mathrm{st}B st B ,
Identity 2-cells - for each string f ⃗ \vec{f} f , the unique 2-cell 1 e ( f ⃗ ) : e ( f ⃗ ) ⇒ e ( f ⃗ ) 1_{e(\vec{f})}:e(\vec{f})\Rightarrow e(\vec{f}) 1 e ( f ) : e ( f ) ⇒ e ( f ) is the identity,
Vcomp 2-cells - the vertical composition of 2-cells is directly lifted from the map e e e as shown below
s t B → e B \begin{CD}
\mathrm{st}B @>{e}>> B
\end{CD} st B e B
\begin{CD}
b @>{f_1}>> b_1 @>{f_2}>> b_2 @>{f_3}>> \cdots@>>> b_n\\
@V{\alpha:\vec{f}\Rightarrow\vec{g}\in B}VV @. @. @. @.\\
b @>{g_1}>> b'_1 @>{g_2}>> b'_2 @>{g_3}>> \cdots@>>> b'_m=b_n\\
@V{\beta:\vec{g}\Rightarrow\vec{h}\in B}VV @. @. @. @.\\
b @>{h_1}>> b''_1 @>{h_2}>> b''_2 @>{h_3}>> \cdots@>>> b''_p=b_n
\end{CD}
b 1 → e ( f ⃗ ) b 2 α ↓ ↑ β α b 1 → e ( g ⃗ ) b 2 β ↓ ↑ b 1 → e ( h ⃗ ) b 2 \begin{CD}
b_1 @>{e(\vec{f})}>> b_2\\
@V{\alpha}VV @AA{\beta\alpha}A\\
b_1 @>{e(\vec{g})}>> b_2\\
@V{\beta}VV @AA{}A\\
b_1 @>{e(\vec{h})}>> b_2
\end{CD} b 1 α ↓ ⏐ b 1 β ↓ ⏐ b 1 e ( f ) e ( g ) e ( h ) b 2 ⏐ ↑ β α b 2 ⏐ ↑ b 2
Hcomp 2-cells - the horizontal composite of two 2-cells α : e ( f ⃗ ) ⇒ e ( g ⃗ ) \alpha:e(\vec{f})\Rightarrow e(\vec{g}) α : e ( f ) ⇒ e ( g ) and β : e ( h ⃗ ) ⇒ e ( k ⃗ ) \beta:e(\vec{h})\Rightarrow e(\vec{k}) β : e ( h ) ⇒ e ( k ) is given as follows:
b → f 1 b 1 → f 2 ⋯ → b n = c → h 1 c 1 → h 2 ⋯ → c p = d ⇓ ( α : f ⃗ ⇒ g ⃗ ∈ B ) ⇓ ( β : h ⃗ ⇒ k ⃗ ∈ B ) b → g 1 b 1 ′ → g 2 ⋯ → b m ′ = b n = c → k 1 c 1 ′ → k 2 ⋯ → c q ′ = c p = d \begin{array}{ccccccccccccc}
b & \xrightarrow{f_1} & b_1 & \xrightarrow{f_2} & \cdots& \xrightarrow{} & b_n=c & \xrightarrow{h_1} & c_1 & \xrightarrow{h_2} & \cdots& \xrightarrow{} & c_p=d \\
& \underset{(\alpha:\vec f\Rightarrow\vec g\in B)}{\Downarrow} & & & & & & \underset{(\beta:\vec h\Rightarrow\vec k\in B)}{\Downarrow} & & & & & \\
b & \xrightarrow{g_1} & b'_1 & \xrightarrow{g_2} & \cdots& \xrightarrow{} & b'_m=b_n=c & \xrightarrow{k_1} & c'_1 & \xrightarrow{k_2} & \cdots& \xrightarrow{} & c'_q=c_p=d
\end{array} b b f 1 ( α : f ⇒ g ∈ B ) ⇓ g 1 b 1 b 1 ′ f 2 g 2 ⋯ ⋯ b n = c b m ′ = b n = c h 1 ( β : h ⇒ k ∈ B ) ⇓ k 1 c 1 c 1 ′ h 2 k 2 ⋯ ⋯ c p = d c q ′ = c p = d
b → f 1 b 1 → f 2 ⋯ → b n = c → h 1 c 1 → h 2 ⋯ → c p = d ⇓ e ( h ⃗ , f ⃗ ) ⇒ e ( h ⃗ ) e ( f ⃗ ) b → e ( f ⃗ ) c → e ( h ⃗ ) d ⇓ β ⋆ α b → e ( g ⃗ ) c → e ( k ⃗ ) d ⇓ e ( k ⃗ ) e ( g ⃗ ) ⇒ e ( k ⃗ , g ⃗ ) b → g 1 b 1 ′ → g 2 ⋯ → b m ′ = b n = c → k 1 c 1 ′ → k 2 ⋯ → c q ′ = c p = d \begin{array}{ccccccccccccc}
b & \xrightarrow{f_1} & b_1 & \xrightarrow{f_2} & \cdots& \xrightarrow{} & b_n=c & \xrightarrow{h_1} & c_1 & \xrightarrow{h_2} & \cdots& \xrightarrow{} & c_p=d \\
& & & & & & \Downarrow_{e(\vec h,\vec f)\Rightarrow e(\vec h)e(\vec f)} & & & & & & \\
b & \xrightarrow{\hspace{1cm}e(\vec f)\hspace{1cm}} & c & \xrightarrow{\hspace{1cm}e(\vec h)\hspace{1cm}} & d \\
& & & & \Downarrow_{\beta\star\alpha} & & & & \\
b & \xrightarrow{\hspace{1cm}e(\vec g)\hspace{1cm}} & c & \xrightarrow{\hspace{1cm}e(\vec k)\hspace{1cm}} & d \\
& & & & \Downarrow_{e(\vec k)e(\vec g)\Rightarrow e(\vec k,\vec g)} & & & & \\
b & \xrightarrow{g_1} & b'_1 & \xrightarrow{g_2} & \cdots& \xrightarrow{} & b'_m=b_n=c & \xrightarrow{k_1} & c'_1 & \xrightarrow{k_2} & \cdots& \xrightarrow{} & c'_q=c_p=d
\end{array} b b b b f 1 e ( f ) e ( g ) g 1 b 1 c c b 1 ′ f 2 e ( h ) e ( k ) g 2 ⋯ d ⇓ β ⋆ α d ⇓ e ( k ) e ( g ) ⇒ e ( k , g ) ⋯ b n = c ⇓ e ( h , f ) ⇒ e ( h ) e ( f ) b m ′ = b n = c h 1 k 1 c 1 c 1 ′ h 2 k 2 ⋯ ⋯ c p = d c q ′ = c p = d
where the 2-cells e ( k ⃗ ) e ( g ⃗ ) ⇒ e ( k ⃗ , g ⃗ ) e(\vec k)e(\vec g)\Rightarrow e(\vec k,\vec g) e ( k ) e ( g ) ⇒ e ( k , g ) and e ( h ⃗ , f ⃗ ) ⇒ e ( h ⃗ ) e ( f ⃗ ) e(\vec h,\vec f)\Rightarrow e(\vec h)e(\vec f) e ( h , f ) ⇒ e ( h ) e ( f ) are induced uniquely from the coherence constraints, and β ⋆ α \beta\star\alpha β ⋆ α is the horizontal composite in the bicategory B B B itself.
Now, we wish to specify e : i ∘ s t B → B e:i\circ\mathrm{st}B\to B e : i ∘ st B → B and s : B → i ∘ s t B s:B\to i\circ\mathrm{st}B s : B → i ∘ st B as pseudofunctors.
Definition A.1.3. The pseudofunctor e : s t B → B e:\mathrm{st}B\to B e : st B → B is defined by the following data:
Objects - for each b ∈ s t B b\in\mathrm{st}B b ∈ st B , the pseudofunctor e e e acts as b ↦ b b\mapsto b b ↦ b ,
1-cells - for each string f ⃗ ∈ s t B \vec f\in\mathrm{st}B f ∈ st B , the pseudofunctor e e e acts as f ⃗ ↦ e ( f ⃗ ) \vec f\mapsto e(\vec f) f ↦ e ( f ) ,
2-cells - each 2-cell in s t B \mathrm{st}B st B is already defined as a 2-cell in B B B ,
Composition constraint - the 2-cell e ( g ⃗ ) e ( f ⃗ ) ⇒ e ( g ⃗ , f ⃗ ) e(\vec g)e(\vec f)\Rightarrow e(\vec g,\vec f) e ( g ) e ( f ) ⇒ e ( g , f ) is the unique coherence constraint 2-cell,
s t B → e B b → f ⃗ c ↘ ⇓ ↙ d ⟺ b → e ( f ⃗ ) c ↘ ⇓ ↙ d \begin{array}{ccccc}
\mathrm{st}B & \xrightarrow{e} & B \\
b & \xrightarrow{\vec f} & c \\
& \searrow& \Downarrow& \swarrow\\
& & d &
\end{array}
\qquad\Longleftrightarrow\qquad
\begin{array}{ccccc}
b & \xrightarrow{e(\vec f)} & c \\
& \searrow& \Downarrow& \swarrow\\
& & d &
\end{array} st B b e f ↘ B c ⇓ d ↙ ⟺ b e ( f ) ↘ c ⇓ d ↙
Identity constraint - the 2-cell 1 b ⇒ e ( ε b → b ) 1_b \Rightarrow e(\varepsilon_b \to b) 1 b ⇒ e ( ε b → b ) induced by the identity 2-cell at 1 b : b → b 1_b:b \to b 1 b : b → b ,
s t B → e B → ε b ⇝ b → 1 b d 1 1 b ↓ ↑ 1 b = e ( ε b ) \begin{CD}
\mathrm{st}B @>e>> B\\
b @>{\varepsilon_b}>> b
\end{CD}
\qquad\leadsto\qquad
\begin{CD}
b @>{1_b}>> d\\
@V{1_{1_b}}VV @AA{1_b=e(\varepsilon_b)}A
\end{CD} st B ε b e B ⇝ b 1 1 b ↓ ⏐ 1 b d ⏐ ↑ 1 b = e ( ε b )
The pseudofunctor s : B → s t B s:B \to\mathrm{st}B s : B → st B is supposed to be a strict section of the pseudofunctor e : s t B → B e:\mathrm{st}B \to B e : st B → B , that is, e ∘ s = 1 B e \circ s=1_B e ∘ s = 1 B . We shall specify s s s with this view in mind.
Definition A.1.4. The pseudofunctor s : B → s t B s:B \to\mathrm{st}B s : B → st B is defined as follows:
Objects - for each b ∈ B b \in B b ∈ B , the pseudofunctor s s s acts as b ↦ b b \mapsto b b ↦ b .
1-cells - for each f : b → c ∈ B f:b \to c \in B f : b → c ∈ B , the pseudofunctor f f f acts as f ↦ ( f ) f \mapsto(f) f ↦ ( f ) , where ( f ) (f) ( f ) is string of length one, for simplicity this will now be labeled as f f f itself,
2-cells - the 2-cell α : f ⇒ g \alpha:f \Rightarrow g α : f ⇒ g is mapped precisely to the 2-cell α : f ⇒ g \alpha:f \Rightarrow g α : f ⇒ g under inverse image in e e e ,
Composition constraint - the 2-cell ( g , f ) → g f (g,f) \to gf ( g , f ) → g f is induced by the identity cell 1 g f : e ( g , f ) ⇒ e ( g f ) 1_{gf}:e(g,f) \Rightarrow e(gf) 1 g f : e ( g , f ) ⇒ e ( g f ) ,
B → f s t B → f ↓ ↓ → g f ⇝ b → f c ↓ ↑ 1 g f d → g f d \begin{CD}
B @>f>> \mathrm{st}B\\
b @>{f}>> c\\
@V{}VV @VV{}V\\
d @>{gf}>> d
\end{CD}
\qquad\leadsto\qquad
\begin{CD}
b @>{f}>> c\\
@V{}VV @AA{1_{gf}}A\\
d @>{gf}>> d
\end{CD} B f g f f ↓ ⏐ st B ↓ ⏐ ⇝ b ↓ ⏐ d f g f c ⏐ ↑ 1 g f d
Identity constraint - the 2-cell ε b ⇒ 1 b \varepsilon_b \Rightarrow1_b ε b ⇒ 1 b is induced by the unique identity cell 1 1 b : 1 b ⇒ 1 b 1_{1_b}:1_b \Rightarrow1_b 1 1 b : 1 b ⇒ 1 b ,
B → f s t B → 1 b ⇝ b → ε b b 1 1 b ↓ ↑ 1 b \begin{CD}
B @>f>> \mathrm{st}B\\
b @>{1_b}>> b
\end{CD}
\qquad\leadsto\qquad
\begin{CD}
b @>{\varepsilon_b}>> b\\
@V{1_{1_b}}VV @AA{1_b}A
\end{CD} B 1 b f st B ⇝ b 1 1 b ↓ ⏐ ε b b ⏐ ↑ 1 b
It is highly important to note that the constraints cells of s s s are non-trivial, even if both of them are represented by the identity 2-cells under e e e . Furthermore, the equality e ∘ s = 1 B e \circ s=1_B e ∘ s = 1 B is meant as equality of pseudofunctors, which includes their coherence constraints. We will not verify the pseudo-functor axioms for both e e e and s s s , as they are routine technical checks.
TEXT
Mapping of Bicat(A A A , B B B )
Our goal now is determine the action of st on the bicategory of pseudofunctors, pseudotransformations, and modification. To this end, the following lemma shall be essential.
Lemma A.2.1. The dotted composite Φ \Phi Φ defined in the diagram
Gray ( s t A , B ) → i Bicat ( i ( s t A ) , i B ) → ≅ Bicat ( i ( s t A ) , i B ) × ∗ Φ ↓ ( s A ) ∗ ↓ ↓ 1 × s A Bicat ( A , i B ) ← ∘ Bicat ( i ( s t A ) , i B ) × Bicat ( A , i ( s t A ) ) \begin{CD}
\operatorname{Gray}(\mathrm{st}A,B) @>{i}>> \operatorname{Bicat}(i(\mathrm{st}A),iB) @>{\cong}>> \operatorname{Bicat}(i(\mathrm{st}A),iB)\times* \\
@V{\Phi}VV @V{(s_A)^*}VV @VV{1\times s_A}V \\
\operatorname{Bicat}(A,iB) @<{\circ}<< \operatorname{Bicat}(i(\mathrm{st}A),iB)\times\operatorname{Bicat}(A,i(\mathrm{st}A))
\end{CD} Gray ( st A , B ) Φ ↓ ⏐ Bicat ( A , i B ) i ∘ Bicat ( i ( st A ) , i B ) ( s A ) ∗ ↓ ⏐ Bicat ( i ( st A ) , i B ) × Bicat ( A , i ( st A )) ≅ Bicat ( i ( st A ) , i B ) × ∗ ↓ ⏐ 1 × s A is a strict 2-functor of strict 2-categories. Here, s A s_A s A is the constant pseudofunctor out of the terminal bicategory which picks s A ∈ Bicat ( A , i ( s t A ) ) s_A\in\operatorname{Bicat}(A,i(\mathrm{st}A)) s A ∈ Bicat ( A , i ( st A )) , and the functor ∘ \circ ∘ is the compositor of Bicat.
Proof. We first note that Bicat ( A , Y ) \operatorname{Bicat}(A,Y) Bicat ( A , Y ) is a strict 2-category, whenever Y Y Y is a strict 2-category. Thus, all of the bicategories involved in defining the above composite are all strict 2-categories. Next, we verify that Φ \Phi Φ is a strict 2-functor of strict 2-categories. Note that Φ ( γ : F ⇒ G ) \Phi(\gamma:F\Rightarrow G) Φ ( γ : F ⇒ G ) is given by γ ⋆ 1 S A \gamma\star1_{S_A} γ ⋆ 1 S A . Using the standard definition of the compositor in Bicat, we can represent this using the following transformation cell in B B B :
Φ ( γ ) ≡ F a → F f F b γ a ↓ ↓ γ b G a → G f G b 1 G a ↓ ↓ 1 G a G a → G f G b ≡ F a → F f F b γ a ↓ ↓ γ b G a → G f G b \Phi(\gamma)\equiv
\begin{CD}
Fa @>{Ff}>> Fb\\
@V{\gamma_a}VV @VV{\gamma_b}V\\
Ga @>{Gf}>> Gb\\
@V{1_{Ga}}VV @VV{1_{Ga}}V\\
Ga @>{Gf}>> Gb
\end{CD}
\equiv
\begin{CD}
Fa @>{Ff}>> Fb\\
@V{\gamma_a}VV @VV{\gamma_b}V\\
Ga @>{Gf}>> Gb
\end{CD} Φ ( γ ) ≡ F a γ a ↓ ⏐ G a 1 G a ↓ ⏐ G a F f G f G f F b ↓ ⏐ γ b G b ↓ ⏐ 1 G a G b ≡ F a γ a ↓ ⏐ G a F f G f F b ↓ ⏐ γ b G b
Crucially, we emphasize that this equality is possible only because both F , G : s t A → B F,G:\mathrm{st}A\to B F , G : st A → B are strict 2-functors of strict 2-categories. With this description, it is now straightforward to verify both ( γ 2 γ 1 ) ⋆ 1 S A = ( γ 2 ⋆ 1 S A ) ( γ 1 ⋆ 1 S A ) (\gamma_2\gamma_1)\star1_{S_A}=(\gamma_2\star1_{S_A})(\gamma_1\star1_{S_A}) ( γ 2 γ 1 ) ⋆ 1 S A = ( γ 2 ⋆ 1 S A ) ( γ 1 ⋆ 1 S A ) and 1 F ⋆ 1 S A = 1 F ⋆ S A 1_F\star1_{S_A}=1_{F\star S_A} 1 F ⋆ 1 S A = 1 F ⋆ S A . Thus, the pseudofunctor Φ \Phi Φ is indeed strict 2-functor.
It turns out that Φ \Phi Φ is actually a strict 2-isomorphism of 2-categories. For this, we only need to show that it is a bijection on the underlying 2-graphs. Thus, we state the following lemma.
For every pseudofunctor U : A → Y U:A\to Y U : A → Y with Y ∈ G r a y Y\in\mathbf{Gray} Y ∈ Gray , there exists a unique strict functor U ~ : st A → Y \widetilde{U}:\operatorname{st}A\to Y U : st A → Y such that U ~ ∘ s A = U \widetilde{U}\circ s_A=U U ∘ s A = U .
For every pseudotransformation γ : U ⇒ V : A → Y \gamma:U\Rightarrow V:A\to Y γ : U ⇒ V : A → Y with Y ∈ G r a y Y\in\mathbf{Gray} Y ∈ Gray , there exists a unique pseudotransformation γ ~ \widetilde{\gamma} γ such that γ ~ ⋆ 1 s A = ρ \widetilde{\gamma}\star1_{s_A}=\rho γ ⋆ 1 s A = ρ .
For every modification θ : γ ⇛ δ : U ⇒ V : A → Y \theta:\gamma\Rrightarrow\delta:U\Rightarrow V:A\to Y θ : γ ⇛ δ : U ⇒ V : A → Y with Y ∈ G r a y Y\in\mathbf{Gray} Y ∈ Gray , there exists a unique modification θ ~ \widetilde{\theta} θ such that θ ~ ⋅ 1 1 s A = ω \widetilde{\theta}\cdot1_{1_{s_A}}=\omega θ ⋅ 1 1 s A = ω .
Proof. The only non-trivial part of the proof is defining each of the unique lifts, the rest is a routine.
Define U ~ \widetilde{U} U by the following data:
Objects - for any a ∈ st A a\in\operatorname{st}A a ∈ st A , U U U maps a ↦ U a ∈ Y a\mapsto Ua\in Y a ↦ U a ∈ Y ,
1-cells - the non-identity 1-cells f ⃗ \vec{f} f are mapped as
a → f 1 a 1 → f 2 ⋯ ↦ U ~ U a → U f 1 U a 1 → U f 2 ⋯ \begin{aligned}
a\xrightarrow{f_1}a_1\xrightarrow{f_2}\cdots
&\mapsto^{\widetilde{U}}
Ua\xrightarrow{Uf_1}Ua_1\xrightarrow{Uf_2}\cdots
\end{aligned} a f 1 a 1 f 2 ⋯ ↦ U U a U f 1 U a 1 U f 2 ⋯
and the identity 1-cell is mapped to the identity 1-cell,
2-cells - the action on 2-cells is given as
a → f 1 a 1 → f 2 ⋯ → b ⇓ ( α : e ( f ⃗ ) ⇒ e ( g ⃗ ) ) a → g 1 a 1 ′ → g 2 ⋯ → b \begin{aligned}
a\xrightarrow{f_1}a_1\xrightarrow{f_2}\cdots\xrightarrow{}b
&\Downarrow^{(\alpha:e(\vec f)\Rightarrow e(\vec g))}\\
a\xrightarrow{g_1}a'_1\xrightarrow{g_2}\cdots\xrightarrow{}b
\end{aligned} a f 1 a 1 f 2 ⋯ b a g 1 a 1 ′ g 2 ⋯ b ⇓ ( α : e ( f ) ⇒ e ( g ))
a → f 1 a 1 → f 2 ⋯ → b ⇒ Coh s A a → e ( f ⃗ ) → b ⇒ α a → e ( g ⃗ ) → b ⇒ Coh s A − 1 a → g 1 a 1 ′ → g 2 ⋯ → b ⇓ U ~ U a → U f 1 U a 1 → U f 2 ⋯ → U b ⇒ Coh U U a → U ( e ( f ⃗ ) ) → U b ⇒ U ( α ) U a → U ( e ( g ⃗ ) ) → U b ⇒ Coh U − 1 U a → U g 1 U a 1 ′ → U g 2 ⋯ → U b . \begin{aligned}
a\xrightarrow{f_1}a_1\xrightarrow{f_2}\cdots\xrightarrow{}b
&\xRightarrow{\operatorname{Coh}_{s_A}}
a\xrightarrow{}e(\vec f)\xrightarrow{}b
\xRightarrow{\alpha}
a\xrightarrow{}e(\vec g)\xrightarrow{}b
\xRightarrow{\operatorname{Coh}_{s_A}^{-1}}
a\xrightarrow{g_1}a'_1\xrightarrow{g_2}\cdots\xrightarrow{}b\\
\Downarrow\widetilde{U}\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\\
Ua\xrightarrow{Uf_1}Ua_1\xrightarrow{Uf_2}\cdots\xrightarrow{}Ub
&\xRightarrow{\operatorname{Coh}_{U}}
Ua\xrightarrow{}U(e(\vec f))\xrightarrow{}Ub
\xRightarrow{U(\alpha)}
Ua\xrightarrow{}U(e(\vec g))\xrightarrow{}Ub
\xRightarrow{\operatorname{Coh}_{U}^{-1}}
Ua\xrightarrow{Ug_1}Ua'_1\xrightarrow{Ug_2}\cdots\xrightarrow{}Ub.
\end{aligned} a f 1 a 1 f 2 ⋯ b ⇓ U U a U f 1 U a 1 U f 2 ⋯ U b Coh s A a e ( f ) b α a e ( g ) b Coh s A − 1 a g 1 a 1 ′ g 2 ⋯ b Coh U U a U ( e ( f )) U b U ( α ) U a U ( e ( g )) U b Coh U − 1 U a U g 1 U a 1 ′ U g 2 ⋯ U b .
where Coh U \operatorname{Coh}_{U} Coh U and Coh s A \operatorname{Coh}_{s_A} Coh s A represent the unique coherence constraint cells.
By choosing 1 f ⃗ : f ⃗ ⇒ e ( f ⃗ ) 1_{\vec f}:\vec f\Rightarrow e(\vec f) 1 f : f ⇒ e ( f ) , the definition readily implies that U ~ ( Coh s A ) = Coh U \widetilde{U}(\operatorname{Coh}_{s_A})=\operatorname{Coh}_{U} U ( Coh s A ) = Coh U . Thus, U ~ ∘ s A = U \widetilde{U}\circ s_A=U U ∘ s A = U holds as desired, including the equality of coherence data.
Define γ ~ \widetilde{\gamma} γ by the following data:
Component 1-cell for each a ∈ st A a \in\operatorname{st}A a ∈ st A , the component ( st γ ) a (\operatorname{st}\gamma)_a ( st γ ) a is defined as γ a \gamma_a γ a itself,
Component 2-cell for each string f ⃗ ∈ st A \vec{f} \in\operatorname{st}A f ∈ st A , the component ( st γ ) f ⃗ (\operatorname{st}\gamma)_{\vec{f}} ( st γ ) f is defined by the following composite 2-cell in Y Y Y
U a → U f 1 U a 1 → U f 2 U a 2 → U f 3 ⋯ ⟶ U b ↓ γ a ↗ γ f 1 ↓ γ a 1 ↗ γ f 2 ↓ γ a 2 ⋯ ↓ γ b V a → V f 1 V a 1 → V f 2 V a 2 → V f 3 ⋯ ⟶ V b \begin{array}{ccccc}
U_a & \xrightarrow{U f_1} & U_{a_1} & \xrightarrow{U f_2} & U_{a_2} \xrightarrow{U f_3} \cdots\longrightarrow U_b \\
\downarrow{\gamma_a} & \nearrow{\gamma_{f_1}} & \downarrow{\gamma_{a_1}} & \nearrow{\gamma_{f_2}} & \downarrow{\gamma_{a_2}} \hspace{1em} \cdots\hspace{1em} \downarrow{\gamma_b} \\
V_a & \xrightarrow{V f_1} & V_{a_1} & \xrightarrow{V f_2} & V_{a_2} \xrightarrow{V f_3} \cdots\longrightarrow V_b
\end{array} U a ↓ γ a V a U f 1 ↗ γ f 1 V f 1 U a 1 ↓ γ a 1 V a 1 U f 2 ↗ γ f 2 V f 2 U a 2 U f 3 ⋯ ⟶ U b ↓ γ a 2 ⋯ ↓ γ b V a 2 V f 3 ⋯ ⟶ V b
Define θ ~ : γ ~ ⇛ δ ~ \widetilde{\theta}:\widetilde{\gamma}\Rrightarrow\widetilde{\delta} θ : γ ⇛ δ by setting ( θ ~ ) a = θ a : γ a ⇒ δ a (\widetilde{\theta})_a=\theta_a:\gamma_a\Rightarrow\delta_a ( θ ) a = θ a : γ a ⇒ δ a
As a consequence, we obtain the desired proposition.
Proposition A.2.3. The dotted composite Φ \Phi Φ defined in the diagram
Gray ( st A , B ) → i Bicat ( i ( st A ) , i B ) → ≅ Bicat ( i ( st A ) , i B ) × ∗ ↘ Φ ↓ ( s A ) ∗ ↓ 1 × s A Bicat ( A , i B ) ← ∘ Bicat ( i ( st A ) , i B ) × Bicat ( A , i ( st A ) ) \begin{array}{ccccc}
\operatorname{Gray}(\operatorname{st}A,B) & \xrightarrow{i} & \operatorname{Bicat}(i(\operatorname{st}A),iB) & \xrightarrow{\cong} & \operatorname{Bicat}(i(\operatorname{st}A),iB)\times* \\
& \searrow\Phi& \downarrow{(s_A)^*} & & \downarrow{1\times s_A} \\
& & \operatorname{Bicat}(A,iB) & \xleftarrow{\ \circ\ } & \operatorname{Bicat}(i(\operatorname{st}A),iB)\times\operatorname{Bicat}(A,i(\operatorname{st}A))
\end{array} Gray ( st A , B ) i ↘ Φ Bicat ( i ( st A ) , i B ) ↓ ( s A ) ∗ Bicat ( A , i B ) ≅ ∘ Bicat ( i ( st A ) , i B ) × ∗ ↓ 1 × s A Bicat ( i ( st A ) , i B ) × Bicat ( A , i ( st A )) is a strict 2-isomorphism of strict 2-categories.
Remark A.2.4. One may visualize this universal property using the diagram
U ~ i ∘ st A ⟶ γ ~ ⇛ δ ~ ⟶ i Y ↑ s A A → U , V i Y \begin{array}{ccccc}
& & \widetilde{U} & & \\
i\circ\operatorname{st}A & \longrightarrow& \widetilde{\gamma}\Rrightarrow\widetilde{\delta} & \longrightarrow& iY \\
\uparrow{s_A} & & & & \\
A & \xrightarrow{U,V} & iY
\end{array} i ∘ st A ↑ s A A ⟶ U , V U γ ⇛ δ iY ⟶ iY
As with any ordinary adjunction, we can utilize this universal property to extend st \operatorname{st} st to a trifunctor. We first extend it on Bicat ( A , B ) \operatorname{Bicat}(A,B) Bicat ( A , B ) .
Definition A.2.5. Define the post composition pseudofunctor ( s B ) ∗ (s_B)_* ( s B ) ∗ , corresponding to s B : B → s t B s_B:B\to\mathrm{st}B s B : B → st B , as the composite pseudofunctor
Bicat ( A , B ) → ≅ ∗ × Bicat ( A , B ) ( s B ) ∗ ↓ ↓ s B × 1 Bicat ( A , i ∘ s t B ) ← ∘ Bicat ( B , i ∘ s t B ) × Bicat ( A , B ) \begin{CD}
\operatorname{Bicat}(A,B) @>{\cong}>> {*} \times\operatorname{Bicat}(A,B) \\
@V{(s_B)_*}VV @VV{s_B\times1}V \\
\operatorname{Bicat}(A,i\circ\mathrm{st}B) @<{\circ}<< \operatorname{Bicat}(B,i\circ\mathrm{st}B)\times\operatorname{Bicat}(A,B)
\end{CD} Bicat ( A , B ) ( s B ) ∗ ↓ ⏐ Bicat ( A , i ∘ st B ) ≅ ∘ ∗ × Bicat ( A , B ) ↓ ⏐ s B × 1 Bicat ( B , i ∘ st B ) × Bicat ( A , B )
Note that ( s B ) ∗ (s_B)_* ( s B ) ∗ is not a strict 2-functor. The compositor is a modification given by unique coherence cell ( 1 s B ⋆ γ 2 ) ( 1 s B ⋆ γ 1 ) ⇒ ( 1 s B ⋆ γ 2 γ 1 ) (1_{s_B}\star\gamma_2)(1_{s_B}\star\gamma_1)\Rightarrow(1_{s_B}\star\gamma_2\gamma_1) ( 1 s B ⋆ γ 2 ) ( 1 s B ⋆ γ 1 ) ⇒ ( 1 s B ⋆ γ 2 γ 1 ) . Similarly, the identity coherence modification is given by the unique coherence cell 1 s B ∘ F ⇒ 1 s B ⋆ 1 F 1_{s_B\circ F}\Rightarrow1_{s_B}\star1_F 1 s B ∘ F ⇒ 1 s B ⋆ 1 F .
Definition A.2.6. Define s t A , B : Bicat ( A , B ) → Gray ( s t A , s t B ) \mathrm{st}_{A,B}:\operatorname{Bicat}(A,B)\to\operatorname{Gray}(\mathrm{st}A,\mathrm{st}B) st A , B : Bicat ( A , B ) → Gray ( st A , st B ) as the composite pseudofunctor
Bicat ( A , B ) → ( s B ) ∗ Bicat ( A , i ∘ s t B ) → Φ − 1 Gray ( s t A , s t B ) . \operatorname{Bicat}(A,B)\xrightarrow{(s_B)_*}\operatorname{Bicat}(A,i\circ\mathrm{st}B)\xrightarrow{\Phi^{-1}}\operatorname{Gray}(\mathrm{st}A,\mathrm{st}B). Bicat ( A , B ) ( s B ) ∗ Bicat ( A , i ∘ st B ) Φ − 1 Gray ( st A , st B ) .
With these definitions, one can readily check that the following diagram commutes strictly
Bicat ( A , B ) → i s t A , s t B ∘ s t A , B Bicat ( i ∘ s t A , i ∘ s t B ) 1 ↓ ↓ ( s A ) ∗ Bicat ( A , B ) → ( s B ) ∗ Bicat ( A , i ∘ s t B ) \begin{CD}
\operatorname{Bicat}(A,B) @>{i_{\mathrm{st}A,\mathrm{st}B}\circ\mathrm{st}_{A,B}}>> \operatorname{Bicat}(i\circ\mathrm{st}A,i\circ\mathrm{st}B) \\
@V{1}VV @VV{(s_A)_*}V \\
\operatorname{Bicat}(A,B) @>{(s_B)_*}>> \operatorname{Bicat}(A,i\circ\mathrm{st}B)
\end{CD} Bicat ( A , B ) 1 ↓ ⏐ Bicat ( A , B ) i st A , st B ∘ st A , B ( s B ) ∗ Bicat ( i ∘ st A , i ∘ st B ) ↓ ⏐ ( s A ) ∗ Bicat ( A , i ∘ st B )
Finally, the coherence data of s t A , B \mathrm{st}_{A,B} st A , B is simply the image under Φ − 1 \Phi^{-1} Φ − 1 of the coherence constraint cells of ( s B ) ∗ (s_B)_* ( s B ) ∗ . Thus, we obtain the following concrete description of s t A , B \mathrm{st}_{A,B} st A , B .
Lemma A.2.7. The coherence data of s t A , B : Bicat ( A , B ) → Gray ( s t A , s t B ) \mathrm{st}_{A,B}:\operatorname{Bicat}(A,B)\to\operatorname{Gray}(\mathrm{st}A,\mathrm{st}B) st A , B : Bicat ( A , B ) → Gray ( st A , st B ) is given by
Composition constraint - for every γ 1 : F ⇒ G \gamma_1:F\Rightarrow G γ 1 : F ⇒ G and γ 2 : G ⇒ H \gamma_2:G\Rightarrow H γ 2 : G ⇒ H , s t ( γ 2 ) s t ( γ 1 ) ⇒ s t ( γ 2 γ 1 ) \mathrm{st}(\gamma_2)\mathrm{st}(\gamma_1)\Rightarrow\mathrm{st}(\gamma_2\gamma_1) st ( γ 2 ) st ( γ 1 ) ⇒ st ( γ 2 γ 1 ) is given by a 2-cell/modification in Gray ( s t A , s t B ) \operatorname{Gray}(\mathrm{st}A,\mathrm{st}B) Gray ( st A , st B ) , whose component at a ∈ s t A a\in\mathrm{st}A a ∈ st A is as depicted in:
F a → ( γ 1 ) a G a → ( γ 2 ) a H a ⟹ ↓ ↑ ( γ 2 ) a ( γ 1 ) a \begin{CD}
F a @>{(\gamma_1)_a}>> G a @>{(\gamma_2)_a}>> H a \\
@. @V{\Longrightarrow}VV @AA{}A \\
@. (\gamma_2)_a(\gamma_1)_a @.
\end{CD} F a ( γ 1 ) a G a ⟹ ↓ ⏐ ( γ 2 ) a ( γ 1 ) a ( γ 2 ) a H a ⏐ ↑
Identity constraint - for every 1 F : F ⇒ F : A → B 1_F:F\Rightarrow F:A\to B 1 F : F ⇒ F : A → B , 1 s t F ⇒ s t ( 1 F ) 1_{\mathrm{st}F}\Rightarrow\mathrm{st}(1_F) 1 st F ⇒ st ( 1 F ) is given by a 2-cell/modification in Gray ( s t A , s t B ) \operatorname{Gray}(\mathrm{st}A,\mathrm{st}B) Gray ( st A , st B ) , whose component at a ∈ s t A a\in\mathrm{st}A a ∈ st A is as depicted in:
\begin{CD}
& Fa & \\
{}_{\varepsilon_{Fa}}\swarrow@. \searrow^{1_{F_a}} \\
& Fa &
\end{CD}
TEXT
Coherence data of st as trifunctor
To specify the functor s t \mathrm{st} st as a trifunctor, we also need to specify its identity and compositor adjoint equivalences. Note that we have not yet imposed any naturality condition on the isomorphisms Φ A , Y : Gray ( s t A , Y ) → Bicat ( A , i Y ) \Phi_{A,Y}:\operatorname{Gray}(\mathrm{st}A,Y)\to\operatorname{Bicat}(A,\mathrm{i}Y) Φ A , Y : Gray ( st A , Y ) → Bicat ( A , i Y ) . Indeed, the optimal condition is to have Φ A , Y \Phi_{A,Y} Φ A , Y be a strictly natural family of isomorphisms. Thus, we begin by noting that there are canonical icons
Bicat ( B , i Y ) → F ∗ Bicat ( A , i Y ) G ∗ ↓ ( F G ) ∗ ↑ Bicat ( C , i Y ) → Bicat ( A , i Y ) Bicat ( A , i Y ) → 1 A ∗ Bicat ( A , i Y ) 1 ↓ ↑ Bicat ( A , i Y ) → Bicat ( A , i Y ) \begin{CD}
\operatorname{Bicat}(B,\mathrm{i}Y) @>{F^*}>> \operatorname{Bicat}(A,\mathrm{i}Y) \\
@V{G^*}VV @A{(FG)^*}AA \\
\operatorname{Bicat}(C,\mathrm{i}Y) @>>{}> \operatorname{Bicat}(A,\mathrm{i}Y)
\end{CD}
\qquad
\begin{CD}
\operatorname{Bicat}(A,\mathrm{i}Y) @>{1_A^*}>> \operatorname{Bicat}(A,\mathrm{i}Y) \\
@V{1}VV @AA{}A \\
\operatorname{Bicat}(A,\mathrm{i}Y) @>>{}> \operatorname{Bicat}(A,\mathrm{i}Y)
\end{CD} Bicat ( B , i Y ) G ∗ ↓ ⏐ Bicat ( C , i Y ) F ∗ Bicat ( A , i Y ) ( F G ) ∗ ⏐ ↑ Bicat ( A , i Y ) Bicat ( A , i Y ) 1 ↓ ⏐ Bicat ( A , i Y ) 1 A ∗ Bicat ( A , i Y ) ⏐ ↑ Bicat ( A , i Y )
If we assume our family of 2-isomorphisms Φ A , Y \Phi_{A,Y} Φ A , Y is strictly natural, this yields similar icons on Gray ( s t ( − ) , Y ) \operatorname{Gray}(\mathrm{st}(-),Y) Gray ( st ( − ) , Y ) by transporting along Φ \Phi Φ . Finally, within the context of Gray-enriched Yoneda lemma, this naturally motivates specific definitions of the identity and compositor adjoint equivalences (or icons). The identity adjoint equivalence is relatively straight forward.
Definition A.3.1. The identity adjoint equivalence ι \iota ι is defined by the following composite of pseudo natural transformation:
∗ → s A Bicat ( A , i s t A ) → Φ − 1 Gray ( s t A , s t A ) 1 A ↓ 1 s t A ↑ ↑ 1 s t A Bicat ( A , A ) → ( s A ) ∗ Bicat ( A , i s t A ) → Φ − 1 Gray ( s t A , s t A ) \begin{CD}
* @>{s_A}>> \operatorname{Bicat}(A,\mathrm{ist}A) @>{\Phi^{-1}}>> \operatorname{Gray}(\mathrm{st}A,\mathrm{st}A) \\
@V{1_A}VV @A{1_{\mathrm{st}A}}AA @AA{1_{\mathrm{st}A}}A \\
\operatorname{Bicat}(A,A) @>{(s_A)_*}>> \operatorname{Bicat}(A,\mathrm{ist}A) @>{\Phi^{-1}}>> \operatorname{Gray}(\mathrm{st}A,\mathrm{st}A)
\end{CD} ∗ 1 A ↓ ⏐ Bicat ( A , A ) s A ( s A ) ∗ Bicat ( A , ist A ) 1 st A ⏐ ↑ Bicat ( A , ist A ) Φ − 1 Φ − 1 Gray ( st A , st A ) ⏐ ↑ 1 st A Gray ( st A , st A )
Φ − 1 s A → 1 ∗ Coh B i c a t Φ − 1 ( ( S A ) ∗ 1 A ) → a ′ ( Φ − 1 ( S A ) ∗ ) 1 A = ↑ = ↑ 1 s t A → ι s t A , A 1 A \begin{CD}
\Phi^{-1}s_A @>{1*\operatorname{Coh}_{\mathrm{Bicat}}}>> \Phi^{-1}((S_A)_*1_A) @>{a'}>> (\Phi^{-1}(S_A)_*)1_A \\
@A{=}AA @A{=}AA \\
1_{\mathrm{st}A} @>>{\iota}> \mathrm{st}_{A,A}1_A
\end{CD} Φ − 1 s A = ⏐ ↑ 1 st A 1 ∗ Coh Bicat ι Φ − 1 (( S A ) ∗ 1 A ) = ⏐ ↑ st A , A 1 A a ′ ( Φ − 1 ( S A ) ∗ ) 1 A
We note that C o h B i c a t \mathrm{CohBicat} CohBicat is an invertible icon whose component at the unique 1-cell 1 ∗ : ∗ → ∗ 1_*: * \to* 1 ∗ : ∗ → ∗ is depicted by the following diagram in B i c a t ( A , i s t A ) \mathrm{Bicat}(A,\mathrm{ist}A) Bicat ( A , ist A ) :
s A → 1 s A s A 1 s A ↓ ↓ 1 s A s A → 1 s A ∗ 1 A s A \begin{CD}
s_A @>{1_{s_A}}>> s_A \\
@V{1_{s_A}}VV @VV{1_{s_A}}V \\
s_A @>{1_{s_A}*1_A}>> s_A
\end{CD} s A 1 s A ↓ ⏐ s A 1 s A 1 s A ∗ 1 A s A ↓ ⏐ 1 s A s A
Thus, the composite invertible icon 1 ⋆ C o h B i c a t 1 \star\mathrm{Coh}_{\mathrm{Bicat}} 1 ⋆ Coh Bicat is depicted in G r a y ( s t A , s t A ) \mathrm{Gray}(\mathrm{st}A,\mathrm{st}A) Gray ( st A , st A ) by
s t ( 1 A ) → 1 s t 1 A s t ( 1 A ) 1 s t 1 A ↓ ↓ 1 s t 1 A s t ( 1 A ) → s t ( 1 1 A ) s t ( 1 A ) a → ε a a 1 a ↑ ↓ a a ← l a \begin{CD}
\mathrm{st}(1_A) @>{1_{\mathrm{st}1_A}}>> \mathrm{st}(1_A) \\
@V{1_{\mathrm{st}1_A}}VV @VV{1_{\mathrm{st}1_A}}V \\
\mathrm{st}(1_A) @>{\mathrm{st}(1_{1_A})}>> \mathrm{st}(1_A)
\end{CD}
\qquad
\begin{CD}
a @>{\varepsilon_a}>> a \\
@A{1_a}AA @VV{a}V \\
a @<{l}<< a
\end{CD} st ( 1 A ) 1 st 1 A ↓ ⏐ st ( 1 A ) 1 st 1 A st ( 1 1 A ) st ( 1 A ) ↓ ⏐ 1 st 1 A st ( 1 A ) a 1 a ⏐ ↑ a ε a l a ↓ ⏐ a a
Unfortunately, the compositor adjoint equivalence χ \chi χ is for more complicated.
Definition A.3.2. The compositor adjoint equivalence χ \chi χ is defined by the following composite of pseudonatural transformations:
χ : ⊗ G r a y ∘ ( s t × s t ) → Φ − 1 ∘ ( s A ) ∗ ∘ ⊗ B i c a t ∘ ( i × i ) ∘ ( s t × s t ) Φ − 1 ∘ ⊗ B i c a t ∘ ( 1 × ( s A ) ∗ ) ∘ ( i × i ) ∘ ( s t × s t ) Φ − 1 ∘ ⊗ B i c a t ∘ ( 1 × ( s B ) ∗ ) ∘ ( i × 1 ) ∘ ( s t × 1 ) Φ − 1 ∘ ⊗ B i c a t ∘ ( ( s B ) ∗ × 1 ) ∘ ( i × 1 ) ∘ ( s t × 1 ) Φ − 1 ∘ ⊗ B i c a t ∘ ( ( s C ) ∗ × 1 ) s t ∘ ⊗ B i c a t → Φ − 1 ∘ ( s C ) ∗ ∘ ⊗ B i c a t s t ∘ ⊗ B i c a t . \begin{aligned}
\chi:\quad
&\otimes_{\mathrm{Gray}}\circ(\mathrm{st}\times\mathrm{st})
\xrightarrow{\ \Phi^{-1}\circ(s_A)^*\circ\otimes_{\mathrm{Bicat}}\circ(i\times i)\circ(\mathrm{st}\times\mathrm{st})\ } \\
&\Phi^{-1}\circ\otimes_{\mathrm{Bicat}}\circ(1\times(s_A)^*)\circ(i\times i)\circ(\mathrm{st}\times\mathrm{st}) \\
&\Phi^{-1}\circ\otimes_{\mathrm{Bicat}}\circ(1\times(s_B)^*)\circ(i\times1)\circ(\mathrm{st}\times1) \\
&\Phi^{-1}\circ\otimes_{\mathrm{Bicat}}\circ((s_B)^*\times1)\circ(i\times1)\circ(\mathrm{st}\times1) \\
&\Phi^{-1}\circ\otimes_{\mathrm{Bicat}}\circ((s_C)^*\times1) \\
&\mathrm{st}\circ\otimes_{\mathrm{Bicat}}
\xrightarrow{\ \Phi^{-1}\circ(s_C)^*\circ\otimes_{\mathrm{Bicat}}\ } \mathrm{st}\circ\otimes_{\mathrm{Bicat}}.
\end{aligned} χ : ⊗ Gray ∘ ( st × st ) Φ − 1 ∘ ( s A ) ∗ ∘ ⊗ Bicat ∘ ( i × i ) ∘ ( st × st ) Φ − 1 ∘ ⊗ Bicat ∘ ( 1 × ( s A ) ∗ ) ∘ ( i × i ) ∘ ( st × st ) Φ − 1 ∘ ⊗ Bicat ∘ ( 1 × ( s B ) ∗ ) ∘ ( i × 1 ) ∘ ( st × 1 ) Φ − 1 ∘ ⊗ Bicat ∘ (( s B ) ∗ × 1 ) ∘ ( i × 1 ) ∘ ( st × 1 ) Φ − 1 ∘ ⊗ Bicat ∘ (( s C ) ∗ × 1 ) st ∘ ⊗ Bicat Φ − 1 ∘ ( s C ) ∗ ∘ ⊗ Bicat st ∘ ⊗ Bicat .
Here, we have suppressed the associators or the unitors of B i c a t \mathrm{Bicat} Bicat for clarity.
The action of χ \chi χ becomes more apparent when we note that it is also an invertible icon whose components are represented by the following chain of modifications in Gray ( s t A , s t C ) \operatorname{Gray}(\mathrm{st}A,\mathrm{st}C) Gray ( st A , st C ) :
χ α G , α F : s t ( α G ) ⋆ s t ( α F ) ⇓ Φ − 1 ( ( i ∘ s t α G ⋆ i ∘ s t α F ) ⋆ 1 s A ) ≡ Φ − 1 ( ( i ∘ ( s t α G ⋆ s t α F ) ) ⋆ 1 s A ) ⇑ ⇑ Φ − 1 ( i ∘ s t α G ⋆ ( i ∘ s t α F ⋆ 1 s A ) ) ≡ Φ − 1 ( i ∘ s t α G ⋆ ( 1 s B ⋆ α F ) ) ⇑ ⇑ Φ − 1 ( ( 1 s C ⋆ α G ) ⋆ α F ) ≡ Φ − 1 ( ( i ∘ s t α G ⋆ 1 s B ) ⋆ α F ) ⇑ ⇑ Φ − 1 ( 1 s C ⋆ ( α G ⋆ α F ) ) ≡ s t ( α G ⋆ α F ) \begin{aligned}
\chi_{\alpha_G,\alpha_F}: \quad&\mathrm{st}(\alpha_G)\star\mathrm{st}(\alpha_F) \\
&\Downarrow\\
&\Phi^{-1}\bigl((i\circ\mathrm{st}\alpha_G\star i\circ\mathrm{st}\alpha_F)\star1_{s_A}\bigr)
\equiv\Phi^{-1}\bigl((i\circ(\mathrm{st}\alpha_G\star\mathrm{st}\alpha_F))\star1_{s_A}\bigr) \\
&\Uparrow\hspace{4em} \Uparrow\\
&\Phi^{-1}\bigl(i\circ\mathrm{st}\alpha_G\star(i\circ\mathrm{st}\alpha_F\star1_{s_A})\bigr)
\equiv\Phi^{-1}\bigl(i\circ\mathrm{st}\alpha_G\star(1_{s_B}\star\alpha_F)\bigr) \\
&\Uparrow\hspace{4em} \Uparrow\\
&\Phi^{-1}\bigl((1_{s_C}\star\alpha_G)\star\alpha_F\bigr)
\equiv\Phi^{-1}\bigl((i\circ\mathrm{st}\alpha_G\star1_{s_B})\star\alpha_F\bigr) \\
&\Uparrow\hspace{4em} \Uparrow\\
&\Phi^{-1}\bigl(1_{s_C}\star(\alpha_G\star\alpha_F)\bigr)
\equiv\mathrm{st}(\alpha_G\star\alpha_F)
\end{aligned} χ α G , α F : st ( α G ) ⋆ st ( α F ) ⇓ Φ − 1 ( ( i ∘ st α G ⋆ i ∘ st α F ) ⋆ 1 s A ) ≡ Φ − 1 ( ( i ∘ ( st α G ⋆ st α F )) ⋆ 1 s A ) ⇑ ⇑ Φ − 1 ( i ∘ st α G ⋆ ( i ∘ st α F ⋆ 1 s A ) ) ≡ Φ − 1 ( i ∘ st α G ⋆ ( 1 s B ⋆ α F ) ) ⇑ ⇑ Φ − 1 ( ( 1 s C ⋆ α G ) ⋆ α F ) ≡ Φ − 1 ( ( i ∘ st α G ⋆ 1 s B ) ⋆ α F ) ⇑ ⇑ Φ − 1 ( 1 s C ⋆ ( α G ⋆ α F ) ) ≡ st ( α G ⋆ α F )
Both the tricategories Bicat and Gray can be considered Bicat 2 _2 2 -enriched bicategories as deduced from [18, Rem. 5.4], where Bicat 2 _2 2 is the strict 2-category of bicategories, pseudofunctors, and Icons. Thus, the iconic structure of st implies that the trifunctor st can itself be considered as a Bicat 2 _2 2 -enriched functor of Bicat 2 _2 2 -enriched bicategories. In particular, any 3-dimensional coherence data of the trifunctor st is merely induced by the coherence constraint cells. Hence, we have fully specified st as a trifunctor.
Encoding bicategorical pasting diagrams
The primary usage of the strictification functor st is that it allows us to encode bicategorical pasting diagram in arbitrary bicategory K K K using the canonical evaluation pseudofunctor e : s t K → K e:\mathrm{st}K\to K e : st K → K . We motivate this with an example.
Suppose we wish to specify the data of an adjunction in a bicategory K K K . Algebraically, it can be expressed using the following data:
f : a → b g : b → a η : 1 a ⇒ g f ϵ : f g ⇒ 1 b \begin{aligned}
f &: a\to b & g &: b\to a \\
\eta&: 1_a\Rightarrow gf & \epsilon&: fg\Rightarrow1_b
\end{aligned} f η : a → b : 1 a ⇒ g f g ϵ : b → a : f g ⇒ 1 b
f 1 a → 1 η f ( g f ) → a − 1 ( f g ) f → ϵ 1 1 b f ↗ r − 1 ↘ l f → 1 f f \begin{aligned}
f1_a &\xrightarrow{1\eta} f(gf)\xrightarrow{a^{-1}}(fg)f\xrightarrow{\epsilon1}1_bf \\
&\mathrel{\substack{\nearrow\\[-0.4em] r^{-1}}} \hspace{12em} \mathrel{\substack{\searrow\\[-0.4em] l}} \\
f &\xrightarrow{1_f}f
\end{aligned} f 1 a f 1 η f ( g f ) a − 1 ( f g ) f ϵ 1 1 b f ↗ r − 1 ↘ l 1 f f
1 a g → η 1 ( g f ) g → a g ( f g ) → 1 ϵ g 1 b ↗ l − 1 ↘ r g → 1 g g \begin{aligned}
1_ag &\xrightarrow{\eta1} (gf)g\xrightarrow{a}g(fg)\xrightarrow{1\epsilon}g1_b \\
&\mathrel{\substack{\nearrow\\[-0.4em] l^{-1}}} \hspace{12em} \mathrel{\substack{\searrow\\[-0.4em] r}} \\
g &\xrightarrow{1_g}g
\end{aligned} 1 a g g η 1 ( g f ) g a g ( f g ) 1 ϵ g 1 b ↗ l − 1 ↘ r 1 g g
In a strict 2-category such as 1 C a t 1\mathrm{Cat} 1 Cat , we could express one of the triangle identity using the following pasting diagram:
A → F B → 1 B B 1 A ↓ η ↑ ↑ ϵ A → F A ≡ B → 1 B B F ↓ ↓ F A → 1 A A \begin{CD}
A @>{F}>> B @>{1_{B}}>> B \\
@V{1_{A}}VV @A{\eta}AA @AA{\epsilon}A \\
A @>{F}>> A
\end{CD}
\qquad\equiv\qquad
\begin{CD}
B @>{1_{B}}>> B \\
@V{F}VV @VV{F}V \\
A @>{1_{A}}>> A
\end{CD} A 1 A ↓ ⏐ A F F B η ⏐ ↑ A 1 B B ⏐ ↑ ϵ ≡ B F ↓ ⏐ A 1 B 1 A B ↓ ⏐ F A
This motivates us to consider encoding such identities inside of s t K \mathrm{st}K st K instead of K K K itself, where such identities are expressible using pasting diagram. To this end, we note the following relations under the evaluation map e e e expressing both the cells η \eta η and ϵ \epsilon ϵ inside s t K \mathrm{st}K st K :
s t K → e K \mathrm{st}K \xrightarrow{\ e\ } K st K e K
a → f b → g a η ↑ ↑ ϵ a a → g f a η ↑ ↑ 1 a \begin{CD}
a @>{f}>> b @>{g}>> a \\
@. @A{\eta}AA @AA{\epsilon_{a}}A
\end{CD}
\qquad
\begin{CD}
a @>{gf}>> a \\
@A{\eta}AA @AA{1_{a}}A
\end{CD} a f b η ⏐ ↑ g a ⏐ ↑ ϵ a a η ⏐ ↑ g f a ⏐ ↑ 1 a
b → g a → f b ϵ ↓ ↓ ϵ b b → f g b ϵ ↓ ↓ 1 b \begin{CD}
b @>{g}>> a @>{f}>> b \\
@. @V{\epsilon}VV @VV{\epsilon_{b}}V
\end{CD}
\qquad
\begin{CD}
b @>{fg}>> b \\
@V{\epsilon}VV @VV{1_{b}}V
\end{CD} b g a ϵ ↓ ⏐ f b ↓ ⏐ ϵ b b ϵ ↓ ⏐ f g b ↓ ⏐ 1 b
In general, let γ \gamma γ be some 2-cell between e ( f ⃗ , b ) e(\vec{f},b) e ( f , b ) and e ( g ⃗ , b ′ ) e(\vec{g},b') e ( g , b ′ ) , where b b b and b ′ b' b ′ signify any arbitrary choice of bracketing in the evaluation. This 2-cell can be sent inside s t K \mathrm{st}K st K using the embedding s K : K → s t K s_K:K\to\mathrm{st}K s K : K → st K . We do this by mapping γ ∈ K \gamma\in K γ ∈ K to the 2-cell f ⃗ ⟹ C o h e ( f ⃗ , b ) ⟹ e ( g ⃗ , b ′ ) ⟹ C o h − 1 g ⃗ \vec{f}\mathrel{\overset{\mathrm{Coh}}{\Longrightarrow}}e(\vec{f},b)\mathrel{\Longrightarrow}e(\vec{g},b')\mathrel{\overset{\mathrm{Coh}^{-1}}{\Longrightarrow}}\vec{g} f ⟹ Coh e ( f , b ) ⟹ e ( g , b ′ ) ⟹ Coh − 1 g in s t K \mathrm{st}K st K , whenever f ⃗ \vec{f} f and g ⃗ \vec{g} g contains no identity 1-cells. If they do, we also remove them using coherence.
With this idea, we now embed one of the triangle identity of ( f , g , η , ϵ ) (f,g,\eta,\epsilon) ( f , g , η , ϵ ) inside s t K \mathrm{st}K st K :
a → ϵ a a → f f f ↓ η ↑ ↓ 1 f f → f b ≡ a → f b 1 f ↑ a → f b \begin{CD}
a @>{\epsilon_{a}}>> a @>{f}>> f \\
@V{f}VV @A{\eta}AA @VV{1_{f}}V \\
f @>{f}>> b
\end{CD}
\qquad\equiv\qquad
\begin{CD}
a @>{f}>> b \\
@. @A{1_{f}}AA \\
a @>{f}>> b
\end{CD} a f ↓ ⏐ f ϵ a f a η ⏐ ↑ b f f ↓ ⏐ 1 f ≡ a a f f b 1 f ⏐ ↑ b
Of course, ε a \varepsilon_a ε a and ε b \varepsilon_b ε b are strict identities inside st K \operatorname{st}K st K . Thus, we could equivalently write
a → ε a a f ↓ ↓ f b → ε b b ≡ a → ε a a f ↓ ↓ f b → b \begin{CD}
a @>{\varepsilon_a}>> a \\
@V{f}VV @VV{f}V \\
b @>{\varepsilon_b}>> b
\end{CD}
\quad\equiv\quad
\begin{CD}
a @>{\varepsilon_a}>> a \\
@V{f}VV @VV{f}V \\
b @>>> b
\end{CD} a f ↓ ⏐ b ε a ε b a ↓ ⏐ f b ≡ a f ↓ ⏐ b ε a a ↓ ⏐ f b
to express the same triangle identity. Finally, Using the evaluation map e : st B → B e\colon\operatorname{st}B \to B e : st B → B , it is relatively straightforward to verify that this pasting diagram equality correspond to the following equality of 2-cells inside B B B :
f 1 a ⇒ η f ( g f ) ⇒ a − 1 ( f g ) f ⇒ ε 1 1 b f ⇒ η ╱ r − 1 ╲ l f ⇒ 1 f f \begin{aligned}
f1_a &\xRightarrow{\eta} f(gf) \xRightarrow{a^{-1}} (fg)f \xRightarrow{\varepsilon1} 1_bf \\
&\phantom{\xRightarrow{\eta}}\diagup^{r^{-1}} \hspace{5.5cm}\diagdown^{l} \\
f &\xRightarrow{1_f} f
\end{aligned} f 1 a f η f ( g f ) a − 1 ( f g ) f ε 1 1 b f η ╱ r − 1 ╲ l 1 f f
Therefore, as a general note, whenever we write bicategorical pasting diagram inside a given bicategory K K K , we always mean to read such diagrams inside st K \operatorname{st}K st K with the use of the evaluation functor e : st K → K e\colon\operatorname{st}K \to K e : st K → K . A more formal treatment of this idea is given using pasting schemes. See [21, Ch. 3] for more details.
Coherence diagrams for pseudomonoid objects
Throughout this section, let V \mathcal{V} V be a monoidal bicategory . In particular, we can choose V \mathcal{V} V to be a one-object algebraic tricategory in the sense of [15, 18], and shall denote the relevant data of the monoidal bicategory V \mathcal{V} V as ( V , ⊗ , I , a V , a V − 1 , λ V , λ V − 1 , ρ V , ρ V − 1 ) (\mathcal{V},\otimes,I,a_{\mathcal{V}},a_{\mathcal{V}}^{-1},\lambda_{\mathcal{V}},\lambda_{\mathcal{V}}^{-1},\rho_{\mathcal{V}},\rho_{\mathcal{V}}^{-1}) ( V , ⊗ , I , a V , a V − 1 , λ V , λ V − 1 , ρ V , ρ V − 1 ) , omitting any labeling of the 2-cell data. Thus, for simplicity, one may also consider V \mathcal{V} V as a 1Cat-enriched monoidal category. Under this consideration, all of the forthcoming coherence diagram are supposed to be read inside the cubical tricategory st V \operatorname{st}\mathcal{V} st V . In particular, we must incorporate the fact that ⊗ \otimes ⊗ is a pseudofunctor with non-trivial coherence constraints. Thus, for instance, a ⊗ 1 a\otimes1 a ⊗ 1 (when seen in st V \operatorname{st}\mathcal{V} st V ) is supposed to be the image of a a a combined with appropriate coherence data under the pseudofunctor V → V × ∗ → 1 × M V × V → ⊗ V \mathcal{V}\to\mathcal{V}\times*\xrightarrow{1\times M}\mathcal{V}\times\mathcal{V}\xrightarrow{\otimes}\mathcal{V} V → V × ∗ 1 × M V × V ⊗ V . Some similar or same diagrams are also available in [22] for the case of 1Cat-enriched categories. More generally, some of these diagrams can be seen as one object versions of [7, Defs. 3.2, 3.3, and 3.4].
Pentagon identity
M ⊗ ( M ⊗ ( M ⊗ M ) ) → 1 ⊗ ( 1 ⊗ μ ) M ⊗ ( M ⊗ M ) ↑ 1 ⊗ a V ↑ 1 ⊗ a V ↓ 1 ⊗ μ ( M ⊗ M ) ⊗ ( M ⊗ M ) M ⊗ ( ( M ⊗ M ) ⊗ M ) → 1 ⊗ ( μ ⊗ 1 ) M ⊗ ( M ⊗ M ) → 1 ⊗ μ M ⊗ M ↑ a V ↑ a V ↑ a V ↓ μ ( ( M ⊗ M ) ⊗ M ) ⊗ M → ( μ ⊗ 1 ) ⊗ 1 ( M ⊗ M ) ⊗ M → μ ⊗ 1 M ⊗ M → μ M \begin{aligned}
& M\otimes(M\otimes(M\otimes M))
\xrightarrow{1\otimes(1\otimes\mu)}
M\otimes(M\otimes M) \\
& \quad\uparrow^{1\otimes a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\uparrow^{1\otimes a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\downarrow^{1\otimes\mu} \\
& (M\otimes M)\otimes(M\otimes M)
\qquad\qquad
M\otimes((M\otimes M)\otimes M)
\xrightarrow{1\otimes(\mu\otimes1)}
M\otimes(M\otimes M)
\xrightarrow{1\otimes\mu}
M\otimes M \\
& \quad\uparrow^{a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\uparrow^{a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\uparrow^{a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\downarrow^{\mu} \\
& ((M\otimes M)\otimes M)\otimes M
\xrightarrow{(\mu\otimes1)\otimes1}
(M\otimes M)\otimes M
\xrightarrow{\mu\otimes1}
M\otimes M
\xrightarrow{\mu}
M
\end{aligned} M ⊗ ( M ⊗ ( M ⊗ M )) 1 ⊗ ( 1 ⊗ μ ) M ⊗ ( M ⊗ M ) ↑ 1 ⊗ a V ↑ 1 ⊗ a V ↓ 1 ⊗ μ ( M ⊗ M ) ⊗ ( M ⊗ M ) M ⊗ (( M ⊗ M ) ⊗ M ) 1 ⊗ ( μ ⊗ 1 ) M ⊗ ( M ⊗ M ) 1 ⊗ μ M ⊗ M ↑ a V ↑ a V ↑ a V ↓ μ (( M ⊗ M ) ⊗ M ) ⊗ M ( μ ⊗ 1 ) ⊗ 1 ( M ⊗ M ) ⊗ M μ ⊗ 1 M ⊗ M μ M
M ⊗ ( M ⊗ ( M ⊗ M ) ) → 1 ⊗ ( 1 ⊗ μ ) M ⊗ ( M ⊗ M ) → 1 ⊗ μ M ⊗ M ↑ a V ↑ a V ↑ a ( M ⊗ M ) ⊗ ( M ⊗ M ) → ( 1 ⊗ 1 ) ⊗ μ ( M ⊗ M ) ⊗ M → μ ⊗ 1 M ⊗ M ↑ a V ∥ ↓ μ ( ( M ⊗ M ) ⊗ M ) ⊗ M → ( μ ⊗ 1 ) ⊗ 1 ( M ⊗ M ) ⊗ M → μ ⊗ 1 M ⊗ M → μ M \begin{aligned}
& M\otimes(M\otimes(M\otimes M))
\xrightarrow{1\otimes(1\otimes\mu)}
M\otimes(M\otimes M)
\xrightarrow{1\otimes\mu}
M\otimes M \\
& \quad\uparrow^{a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\uparrow^{a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\uparrow^{a} \\
& (M\otimes M)\otimes(M\otimes M)
\xrightarrow{(1\otimes1)\otimes\mu}
(M\otimes M)\otimes M
\xrightarrow{\mu\otimes1}
M\otimes M \\
& \quad\uparrow^{a_{\mathcal V}}
\qquad\qquad\qquad\qquad\qquad
\Vert
\qquad\qquad\qquad\qquad\qquad
\downarrow^{\mu} \\
& ((M\otimes M)\otimes M)\otimes M
\xrightarrow{(\mu\otimes1)\otimes1}
(M\otimes M)\otimes M
\xrightarrow{\mu\otimes1}
M\otimes M
\xrightarrow{\mu}
M
\end{aligned} M ⊗ ( M ⊗ ( M ⊗ M )) 1 ⊗ ( 1 ⊗ μ ) M ⊗ ( M ⊗ M ) 1 ⊗ μ M ⊗ M ↑ a V ↑ a V ↑ a ( M ⊗ M ) ⊗ ( M ⊗ M ) ( 1 ⊗ 1 ) ⊗ μ ( M ⊗ M ) ⊗ M μ ⊗ 1 M ⊗ M ↑ a V ∥ ↓ μ (( M ⊗ M ) ⊗ M ) ⊗ M ( μ ⊗ 1 ) ⊗ 1 ( M ⊗ M ) ⊗ M μ ⊗ 1 M ⊗ M μ M
Triangle identity
M ⊗ M → μ M ↑ ε M ⊗ M ↑ μ M ⊗ M M ⊗ M ↘ ρ V − 1 ⊗ 1 ↗ μ ⊗ 1 ( M ⊗ I ) ⊗ M → ( 1 ⊗ e ) ⊗ 1 ( M ⊗ M ) ⊗ M ↑ a V ↑ a V M ⊗ ( I ⊗ M ) → 1 ⊗ ( e ⊗ 1 ) M ⊗ ( M ⊗ M ) ↖ 1 ⊗ λ V ↗ a M ⊗ M \begin{aligned}
&M \otimes M \xrightarrow{\ \mu\ } M\\
&\quad\uparrow^{\varepsilon_{M\otimes M}}\qquad\qquad\qquad\uparrow^{\mu}\\
&M \otimes M \qquad\qquad\qquad M \otimes M\\
&\quad\searrow^{\rho_{V}^{-1}\otimes1}\qquad\qquad\nearrow^{\mu\otimes1}\\
&(M\otimes I)\otimes M\xrightarrow{\ (1\otimes e)\otimes1\ }(M\otimes M)\otimes M\\
&\quad\uparrow^{a_V}\qquad\qquad\qquad\uparrow^{a_V}\\
&M\otimes(I\otimes M)\xrightarrow{\ 1\otimes(e\otimes1)\ }M\otimes(M\otimes M)\\
&\quad\nwarrow^{1\otimes\lambda_V}\qquad\qquad\nearrow^{a}\\
&M\otimes M
\end{aligned} M ⊗ M μ M ↑ ε M ⊗ M ↑ μ M ⊗ M M ⊗ M ↘ ρ V − 1 ⊗ 1 ↗ μ ⊗ 1 ( M ⊗ I ) ⊗ M ( 1 ⊗ e ) ⊗ 1 ( M ⊗ M ) ⊗ M ↑ a V ↑ a V M ⊗ ( I ⊗ M ) 1 ⊗ ( e ⊗ 1 ) M ⊗ ( M ⊗ M ) ↖ 1 ⊗ λ V ↗ a M ⊗ M
M ⊗ M → μ M ↑ ε M ⊗ M ↑ μ M ⊗ M → ε M ⊗ M M ⊗ M ↘ ρ V − 1 ⊗ 1 ↗ μ ⊗ 1 ( M ⊗ I ) ⊗ M → ( 1 ⊗ e ) ⊗ 1 ( M ⊗ M ) ⊗ M \begin{aligned}
&M\otimes M\xrightarrow{\ \mu\ }M\\
&\quad\uparrow^{\varepsilon_{M\otimes M}}\qquad\qquad\qquad\uparrow^{\mu}\\
&M\otimes M\xrightarrow{\ \varepsilon_{M\otimes M}\ }M\otimes M\\
&\quad\searrow^{\rho_V^{-1}\otimes1}\qquad\qquad\nearrow^{\mu\otimes1}\\
&(M\otimes I)\otimes M\xrightarrow{\ (1\otimes e)\otimes1\ }(M\otimes M)\otimes M
\end{aligned} M ⊗ M μ M ↑ ε M ⊗ M ↑ μ M ⊗ M ε M ⊗ M M ⊗ M ↘ ρ V − 1 ⊗ 1 ↗ μ ⊗ 1 ( M ⊗ I ) ⊗ M ( 1 ⊗ e ) ⊗ 1 ( M ⊗ M ) ⊗ M
Hexagon identities
We assume that we have a suitable definition of a braided monoidal bicategory in sense of being a one object tricategory equipped with some coherent braiding adjoint equivalence. In particular, V \mathcal{V} V is a braided monoidal bicategory if it is a monoidal bicategory equipped with a family of adjoint equivalences b x , y : X ⊗ Y ↔ Y ⊗ X : b x , y − 1 b_{x,y}: X \otimes Y \leftrightarrow Y \otimes X : b^{-1}_{x,y} b x , y : X ⊗ Y ↔ Y ⊗ X : b x , y − 1 that behaves as the braiding. The details are ultimately irrelevant for us but can be found carefully constructed in [16]. Other relevant sources include [21, 29].
Figure 3.
Figure 4.
Coherence diagrams for pseudomonoid objects
M ⊗ ( M ⊗ M ) → a M , M , M − 1 ( M ⊗ M ) ⊗ M → b M ⊗ M , M , M M ⊗ ( M ⊗ M ) → a M , M , M − 1 ( M ⊗ M ) ⊗ M ↓ 1 ⊗ μ ↓ μ ⊗ 1 ↓ 1 ⊗ μ ↓ μ ⊗ 1 M ⊗ M → a M , M − 1 M ⊗ M → b M , M M ⊗ M → a M , M − 1 M ⊗ M ↘ μ ↙ μ ↘ μ ↙ μ M \begin{aligned}
&M \otimes(M \otimes M) \xrightarrow{a^{-1}_{M,M,M}} (M \otimes M) \otimes M \xrightarrow{b_{M\otimes M,M,M}} M \otimes(M \otimes M) \xrightarrow{a^{-1}_{M,M,M}} (M \otimes M) \otimes M \\
&\qquad\downarrow{1\otimes\mu}\hspace{5.2cm}\downarrow{\mu\otimes1}\hspace{5.2cm}\downarrow{1\otimes\mu}\hspace{5.2cm}\downarrow{\mu\otimes1}\\
&M \otimes M \xrightarrow{a^{-1}_{M,M}} M \otimes M \xrightarrow{b_{M,M}} M \otimes M \xrightarrow{a^{-1}_{M,M}} M \otimes M\\
&\qquad\searrow_{\mu}\hspace{2.7cm}\swarrow_{\mu}\hspace{2.7cm}\searrow_{\mu}\hspace{2.7cm}\swarrow_{\mu}\\
&\hspace{7cm}M
\end{aligned} M ⊗ ( M ⊗ M ) a M , M , M − 1 ( M ⊗ M ) ⊗ M b M ⊗ M , M , M M ⊗ ( M ⊗ M ) a M , M , M − 1 ( M ⊗ M ) ⊗ M ↓ 1 ⊗ μ ↓ μ ⊗ 1 ↓ 1 ⊗ μ ↓ μ ⊗ 1 M ⊗ M a M , M − 1 M ⊗ M b M , M M ⊗ M a M , M − 1 M ⊗ M ↘ μ ↙ μ ↘ μ ↙ μ M
( M ⊗ M ) ⊗ M → b M , M ⊗ M M ⊗ ( M ⊗ M ) ↑ a M − 1 ↓ a M − 1 M ⊗ ( M ⊗ M ) → 1 ⊗ b M , M M ⊗ ( M ⊗ M ) → b M , M ⊗ 1 ( M ⊗ M ) ⊗ M ↓ 1 ⊗ μ ↘ 1 ⊗ μ ↙ μ ⊗ 1 ↓ μ ⊗ 1 M ⊗ M → a − 1 M ⊗ M → a − 1 ( M ⊗ M ) ⊗ M ↘ μ ↙ μ \begin{aligned}
&(M\otimes M)\otimes M \xrightarrow{b_{M,M\otimes M}} M\otimes(M\otimes M)\\
&\qquad\uparrow{a^{-1}_{M}}\hspace{9cm}\downarrow{a^{-1}_{M}}\\
&M\otimes(M\otimes M)\hspace{2cm}\xrightarrow{1\otimes b_{M,M}}\hspace{2cm}M\otimes(M\otimes M)\hspace{2cm}\xrightarrow{b_{M,M}\otimes1}\hspace{2cm}(M\otimes M)\otimes M\\
&\qquad\downarrow{1\otimes\mu}\hspace{2.5cm}\searrow_{1\otimes\mu}\hspace{4cm}\swarrow_{\mu\otimes1}\hspace{3.5cm}\downarrow{\mu\otimes1}\\
&M\otimes M \xrightarrow{a^{-1}}\ M\otimes M \xrightarrow{a^{-1}}\ (M\otimes M)\otimes M \\
&\qquad\searrow_{\mu}\hspace{5cm}\swarrow_{\mu}
\end{aligned} ( M ⊗ M ) ⊗ M b M , M ⊗ M M ⊗ ( M ⊗ M ) ↑ a M − 1 ↓ a M − 1 M ⊗ ( M ⊗ M ) 1 ⊗ b M , M M ⊗ ( M ⊗ M ) b M , M ⊗ 1 ( M ⊗ M ) ⊗ M ↓ 1 ⊗ μ ↘ 1 ⊗ μ ↙ μ ⊗ 1 ↓ μ ⊗ 1 M ⊗ M a − 1 M ⊗ M a − 1 ( M ⊗ M ) ⊗ M ↘ μ ↙ μ
Morphism hexagon axiom
( M ⊗ M ) ⊗ M → ( f ⊗ f ) ⊗ f ( M ′ ⊗ M ′ ) ⊗ M ′ ↓ a V ↓ μ ⊗ 1 ↙ χ ⊗ 1 ↓ μ ′ ⊗ 1 M ⊗ ( M ⊗ M ) → M ⊗ M → f ⊗ f M ′ ⊗ M ′ ↓ 1 ⊗ μ ⟸ a ⟹ ↓ μ ↙ χ ↓ μ ′ M ⊗ M → μ M → f M ′ \begin{array}{ccccc}
(M\otimes M)\otimes M & \xrightarrow{(f\otimes f)\otimes f} & (M'\otimes M')\otimes M' \\
\downarrow a_V & \downarrow{\mu\otimes1} & \swarrow{\chi\otimes1}\quad\downarrow{\mu'\otimes1} \\
M\otimes(M\otimes M) & \xrightarrow{\hspace{1cm}} & M\otimes M & \xrightarrow{f\otimes f} & M'\otimes M' \\
\downarrow{1\otimes\mu} & \Longleftarrow a \Longrightarrow& \downarrow\mu& \swarrow\chi\quad\downarrow\mu' \\
M\otimes M & \xrightarrow{\mu} & M & \xrightarrow{f} & M'
\end{array} ( M ⊗ M ) ⊗ M ↓ a V M ⊗ ( M ⊗ M ) ↓ 1 ⊗ μ M ⊗ M ( f ⊗ f ) ⊗ f ↓ μ ⊗ 1 ⟸ a ⟹ μ ( M ′ ⊗ M ′ ) ⊗ M ′ ↙ χ ⊗ 1 ↓ μ ′ ⊗ 1 M ⊗ M ↓ μ M f ⊗ f ↙ χ ↓ μ ′ f M ′ ⊗ M ′ M ′
( M ⊗ M ) ⊗ M → ( f ⊗ f ) ⊗ f ( M ′ ⊗ M ′ ) ⊗ M ′ ↓ a V ↙ ↓ a V M ⊗ ( M ⊗ M ) → f ⊗ ( f ⊗ f ) M ′ ⊗ ( M ′ ⊗ M ′ ) ↓ μ ′ ⊗ 1 ↓ 1 ⊗ μ ↙ 1 ⊗ χ ↓ 1 ⊗ μ ′ ↓ μ ′ M ⊗ M → f ⊗ f M ′ ⊗ M ′ → μ ′ M ′ ↓ μ ↙ χ ↘ μ ′ M → f M ′ \begin{array}{ccccc}
(M\otimes M)\otimes M & \xrightarrow{(f\otimes f)\otimes f} & (M'\otimes M')\otimes M' \\
\downarrow a_V & \swarrow\hspace{1cm} & \downarrow a_V \\
M\otimes(M\otimes M) & \xrightarrow{f\otimes(f\otimes f)} & M'\otimes(M'\otimes M') & \downarrow{\mu'\otimes1} \\
\downarrow{1\otimes\mu} & \swarrow{1\otimes\chi} & \downarrow{1\otimes\mu'} & \downarrow\mu' \\
M\otimes M & \xrightarrow{f\otimes f} & M'\otimes M' & \xrightarrow{\mu'} & M' \\
\downarrow\mu& \swarrow\chi& \searrow\mu' & \\
M & \xrightarrow{f} & M'
\end{array} ( M ⊗ M ) ⊗ M ↓ a V M ⊗ ( M ⊗ M ) ↓ 1 ⊗ μ M ⊗ M ↓ μ M ( f ⊗ f ) ⊗ f ↙ f ⊗ ( f ⊗ f ) ↙ 1 ⊗ χ f ⊗ f ↙ χ f ( M ′ ⊗ M ′ ) ⊗ M ′ ↓ a V M ′ ⊗ ( M ′ ⊗ M ′ ) ↓ 1 ⊗ μ ′ M ′ ⊗ M ′ ↘ μ ′ M ′ ↓ μ ′ ⊗ 1 ↓ μ ′ μ ′ M ′
Morphism left square
I ⊗ M → 1 ⊗ f I ⊗ M ′ λ V ↓ ↓ e ′ ⊗ 1 M → f M ′ \begin{CD}
I \otimes M @>{1\otimes f}>> I \otimes M' \\
@V{\lambda_V}VV @VV{e'\otimes1}V \\
M @>{f}>> M'
\end{CD} I ⊗ M λ V ↓ ⏐ M 1 ⊗ f f I ⊗ M ′ ↓ ⏐ e ′ ⊗ 1 M ′
I ⊗ M → e ′ ⊗ 1 M ′ ⊗ M e ⊗ 1 ↓ ↓ 1 ⊗ f M ⊗ M → f ⊗ 1 M ′ ⊗ M ′ \begin{CD}
I \otimes M @>{e'\otimes1}>> M' \otimes M \\
@V{e\otimes1}VV @VV{1\otimes f}V \\
M \otimes M @>{f\otimes1}>> M' \otimes M'
\end{CD} I ⊗ M e ⊗ 1 ↓ ⏐ M ⊗ M e ′ ⊗ 1 f ⊗ 1 M ′ ⊗ M ↓ ⏐ 1 ⊗ f M ′ ⊗ M ′
M ⊗ M → f ⊗ f M ′ ⊗ M ′ μ ↓ ↓ μ ′ M → f M ′ \begin{CD}
M \otimes M @>{f\otimes f}>> M' \otimes M' \\
@V{\mu}VV @VV{\mu'}V \\
M @>{f}>> M'
\end{CD} M ⊗ M μ ↓ ⏐ M f ⊗ f f M ′ ⊗ M ′ ↓ ⏐ μ ′ M ′
M ⊗ M → χ M ′ ⊗ M ′ λ ↓ ↓ λ ′ M → f M ′ \begin{CD}
M \otimes M @>{\chi}>> M' \otimes M' \\
@V{\lambda}VV @VV{\lambda'}V \\
M @>{f}>> M'
\end{CD} M ⊗ M λ ↓ ⏐ M χ f M ′ ⊗ M ′ ↓ ⏐ λ ′ M ′
III
I ⊗ M → 1 ⊗ f I ⊗ M ′ λ V ↓ ↓ λ V M → f M ′ \begin{CD}
I \otimes M @>{1\otimes f}>> I \otimes M' \\
@V{\lambda_V}VV @VV{\lambda_V}V \\
M @>{f}>> M'
\end{CD} I ⊗ M λ V ↓ ⏐ M 1 ⊗ f f I ⊗ M ′ ↓ ⏐ λ V M ′
I ⊗ M ′ → e ′ ⊗ 1 M ′ ⊗ M ′ λ V ↓ ↓ μ ′ M ′ = M ′ \begin{CD}
I \otimes M' @>{e'\otimes1}>> M' \otimes M' \\
@V{\lambda_V}VV @VV{\mu'}V \\
M' @= M'
\end{CD} I ⊗ M ′ λ V ↓ ⏐ M ′ e ′ ⊗ 1 M ′ ⊗ M ′ ↓ ⏐ μ ′ M ′
Morphism right square
Figure 1.
III
Figure 2.
Higher morphisms of weak involutions and inverses
Morphisms of weak 2-involutions
Definition C.1.1. Let ( K , H , α , ζ ) (K,H,\alpha,\zeta) ( K , H , α , ζ ) and ( K ′ , H ′ , α ′ , ζ ′ ) (K',H',\alpha',\zeta') ( K ′ , H ′ , α ′ , ζ ′ ) be two weak 2-involutions. A 1-morphism of these is given by
a 1-cell R : K → K ′ R:K\to K' R : K → K ′ ,
an adjoint equivalence u , u ′ u,u' u , u ′ given by the following 2-cell:
K → R K ′ H ↓ ↓ H ′ K → R K ′ \begin{CD}
K @>{R}>> K'\\
@V{H}VV @VV{H'}V\\
K @>{R}>> K'
\end{CD} K H ↓ ⏐ K R R K ′ ↓ ⏐ H ′ K ′
a 3-cell δ \delta δ as depicted by the following diagram:
K → R K ′ H ↓ ↓ H ′ K → R K ′ ⟹ δ K → R K ′ ↓ 1 K ′ ↓ 1 K ′ K ′ = K ′ \begin{CD}
K @>{R}>> K'\\
@V{H}VV @VV{H'}V\\
K @>{R}>> K'
\end{CD}
\quad\Longrightarrow^{\delta}\quad
\begin{CD}
K @>{R}>> K'\\
@VV{1_{K'}}V @VV{1_{K'}}V\\
K' @= K'
\end{CD} K H ↓ ⏐ K R R K ′ ↓ ⏐ H ′ K ′ ⟹ δ K ↓ ⏐ 1 K ′ K ′ R K ′ ↓ ⏐ 1 K ′ K ′
( H ′ H ′ ) R → α ′ 1 1 K ′ R → ι R a ↓ ↑ r H ′ ( H ′ R ) → 1 u H ′ ( R H ) R 1 K a ′ ↓ ↑ 1 α ( H ′ R ) H → u 1 ( R H ) H → a R ( H H ) \begin{CD}
(H'H')R @>{\alpha'1}>> 1_{K'}R @>{\iota}>> R\\
@V{a}VV @. @AA{r}A\\
H'(H'R) @>{1u}>> H'(RH) @. R1_K\\
@V{a'}VV @. @AA{1\alpha}A\\
(H'R)H @>{u1}>> (RH)H @>{a}>> R(HH)
\end{CD} ( H ′ H ′ ) R a ↓ ⏐ H ′ ( H ′ R ) a ′ ↓ ⏐ ( H ′ R ) H α ′ 1 1 u u 1 1 K ′ R H ′ ( R H ) ( R H ) H ι a R ⏐ ↑ r R 1 K ⏐ ↑ 1 α R ( H H )
where the vertical 3-cell is δ \delta δ ,
the 3-cell δ \delta δ satisfy the weak 2-involution 1-morphism axiom (section D.2).
Higher morphisms of weak involutions and inverses
Definition C.1.2. A 2-morphism ( γ , ω ) : ( R , u , δ u ) ⇒ ( S , v , δ v ) (\gamma,\omega):(R,u,\delta_u)\Rightarrow(S,v,\delta_v) ( γ , ω ) : ( R , u , δ u ) ⇒ ( S , v , δ v ) is given by
a 2-cell γ : R → S \gamma:R\to S γ : R → S ,
a 3-cell ω \omega ω as depicted by the following diagram:
K → R K ′ H ↓ ↓ H ′ K → S K ′ K → R K ′ H ↓ ↓ H ′ K → S K ′ \begin{CD}
K @>R>> K' \\
@V H VV @VV H' V \\
K @>S>> K'
\end{CD}
\quad
\begin{CD}
K @>R>> K' \\
@V H VV @VV H' V \\
K @>S>> K'
\end{CD} K H ↓ ⏐ K R S K ′ ↓ ⏐ H ′ K ′ K H ↓ ⏐ K R S K ′ ↓ ⏐ H ′ K ′ the 3-cell ω \omega ω satisfy the weak 2-involution 2-morphism axiom (section D.3).
Definition C.1.3. A 3-morphism between ρ : ( γ , ω γ ) ⇒ ( λ , ω λ ) \rho:(\gamma,\omega_\gamma)\Rightarrow(\lambda,\omega_\lambda) ρ : ( γ , ω γ ) ⇒ ( λ , ω λ ) is given by
a 3-cell υ : γ ⇒ λ \upsilon:\gamma\Rightarrow\lambda υ : γ ⇒ λ ,
the following identity holds:
H ′ R → 1 γ H ′ S u ↓ ↓ v R H → γ 1 S H ≡ H ′ R → 1 λ H ′ S u ↓ ↓ v R H → λ 1 S H \begin{CD}
H'R @>{1\gamma}>> H'S \\
@V u VV @VV v V \\
RH @>{\gamma1}>> SH
\end{CD}
\quad\equiv\quad
\begin{CD}
H'R @>{1\lambda}>> H'S \\
@V u VV @VV v V \\
RH @>{\lambda1}>> SH
\end{CD} H ′ R u ↓ ⏐ R H 1 γ γ 1 H ′ S ↓ ⏐ v S H ≡ H ′ R u ↓ ⏐ R H 1 λ λ 1 H ′ S ↓ ⏐ v S H
Morphisms of weak 2-inverses
Definition C.2.1. A 1-morphism ( K 1 , K 2 , F , G , ϵ , η , ϕ , ψ ) → ( K 1 ′ , K 2 ′ , F ′ , G ′ , ϵ ′ , η ′ , ϕ ′ , ψ ′ ) (K_1,K_2,F,G,\epsilon,\eta,\phi,\psi)\to(K'_1,K'_2,F',G',\epsilon',\eta',\phi',\psi') ( K 1 , K 2 , F , G , ϵ , η , ϕ , ψ ) → ( K 1 ′ , K 2 ′ , F ′ , G ′ , ϵ ′ , η ′ , ϕ ′ , ψ ′ ) consists of
two 1-cells R 1 : K 1 → K 1 ′ R_1:K_1\to K'_1 R 1 : K 1 → K 1 ′ and R 2 : K 2 → K 2 ′ R_2:K_2\to K'_2 R 2 : K 2 → K 2 ′ ,
two 2-cells
K 1 → R 1 K 1 ′ K 2 → R 2 K 2 ′ ↓ F ↙ u 12 ↓ F ′ ↓ G ↙ u 21 ↓ G ′ K 2 → R 2 K 2 ′ K 1 → R 1 K 1 ′ \begin{array}{ccccc}
K_1 & \xrightarrow{R_1} & K'_1 & & K_2 \xrightarrow{R_2} K'_2 \\
\downarrow F & \mathrel{\swarrow u_{12}} & \downarrow F' & & \downarrow G \mathrel{\swarrow u_{21}} \downarrow G' \\
K_2 & \xrightarrow{R_2} & K'_2 & & K_1 \xrightarrow{R_1} K'_1
\end{array} K 1 ↓ F K 2 R 1 ↙ u 12 R 2 K 1 ′ ↓ F ′ K 2 ′ K 2 R 2 K 2 ′ ↓ G ↙ u 21 ↓ G ′ K 1 R 1 K 1 ′
two 3-cells as depicted by the following:
K 1 K 1 ↙ F ↓ R 1 ↘ R 1 ↙ F ↓ 1 K 1 ↘ R 1 K 2 ← u 12 K 1 ′ ← η K 1 K 2 ← 1 K 1 K 1 ↓ G ↘ R 2 ↙ F ′ ↘ 1 K 1 ′ ↓ G ↓ K 1 → R 1 K 2 ′ ← G ′ K 2 ′ K 1 → R 1 K 1 ′ \begin{array}{ccccc}
& K_1 & & & K_1 \\
& \swarrow F & \downarrow R_1 & \searrow R_1 & \swarrow F \quad\downarrow1_{K_1} \quad\searrow R_1 \\
K_2 & \xleftarrow{u_{12}} K'_1 & & \xleftarrow{\eta} K_1 & K_2 \xleftarrow{1_{K_1}} K_1 \\
\downarrow G & \searrow R_2 & \swarrow F' & \searrow1_{K'_1} & \downarrow G \quad\downarrow\\
K_1 & \xrightarrow{R_1} K'_2 & \xleftarrow{G'} K'_2 & & K_1 \xrightarrow{R_1} K'_1
\end{array} K 2 ↓ G K 1 K 1 ↙ F u 12 K 1 ′ ↘ R 2 R 1 K 2 ′ ↓ R 1 ↙ F ′ G ′ K 2 ′ ↘ R 1 η K 1 ↘ 1 K 1 ′ K 1 ↙ F ↓ 1 K 1 ↘ R 1 K 2 1 K 1 K 1 ↓ G ↓ K 1 R 1 K 1 ′
( G ′ F ′ ) R 1 → α G ′ ( F ′ R 1 ) → 1 u 12 G ′ ( R 2 F ) → α − 1 ( G ′ R 2 ) F R 1 → r ∗ R 1 1 K 1 → 1 η R 1 ( G F ) → α − 1 ( R 1 G ) F ↑ l ∗ ↑ η 1 ′ ↑ u 21 1 \begin{aligned}
(G'F')R_1 &\xrightarrow{\alpha} G'(F'R_1)\xrightarrow{1u_{12}}G'(R_2F)\xrightarrow{\alpha^{-1}}(G'R_2)F\\
R_1 &\xrightarrow{r_\ast}R_1 1_{K_1}\xrightarrow{1\eta}R_1(GF)\xrightarrow{\alpha^{-1}}(R_1G)F\\
&\quad\uparrow l_\ast\qquad\uparrow\eta'_1 \qquad\uparrow u_{21}1
\end{aligned} ( G ′ F ′ ) R 1 R 1 α G ′ ( F ′ R 1 ) 1 u 12 G ′ ( R 2 F ) α − 1 ( G ′ R 2 ) F r ∗ R 1 1 K 1 1 η R 1 ( GF ) α − 1 ( R 1 G ) F ↑ l ∗ ↑ η 1 ′ ↑ u 21 1
K 2 → R 2 K 2 ′ R 2 ↓ ↓ G ′ K 2 → R 1 K 1 ′ R 2 ↓ ↑ F K 2 K 2 ′ ⇒ K 2 → R 2 K 2 ′ R 2 ↓ ↓ G ′ K 1 ′ ↓ ↑ ε ′ K 2 ′ \begin{CD}
K_2 @>{R_2}>> K'_2\\
@V{R_2}VV @VV{G'}V\\
K_2 @>{R_1}>> K'_1\\
@V{R_2}VV @AA{F}A\\
K_2 @. K'_2
\end{CD}
\qquad\Rightarrow\qquad
\begin{CD}
K_2 @>{R_2}>> K'_2\\
@V{R_2}VV @VV{G'}V\\
@. K'_1\\
@VVV @AA{\varepsilon'}A\\
@. K'_2
\end{CD} K 2 R 2 ↓ ⏐ K 2 R 2 ↓ ⏐ K 2 R 2 R 1 K 2 ′ ↓ ⏐ G ′ K 1 ′ ⏐ ↑ F K 2 ′ ⇒ K 2 R 2 ↓ ⏐ ↓ ⏐ R 2 K 2 ′ ↓ ⏐ G ′ K 1 ′ ⏐ ↑ ε ′ K 2 ′
F ′ ( G ′ R 2 ) → u 21 F ′ ( R 1 G ) → a ′ ( F ′ R 1 ) G → u 12 ( R 2 F ) G a ↑ ↓ ↓ ↓ a ( F ′ G ′ ) R 2 → ε ′ − 1 1 K 2 ′ R 2 → l R 2 r ↑ R 2 ( F G ) R 2 \begin{CD}
F'(G'R_2) @>{u_{21}}>> F'(R_1G) @>{a'}>> (F'R_1)G @>{u_{12}}>> (R_2F)G\\
@A{a}AA @VVV @VVV @VV{a}V\\
(F'G')R_2 @>{\varepsilon'^{-1}}>> {}_1K'_2R_2 @>{l}>> R_2\\
@. @. @A{r}AA\\
@. @. R_2(FG)\\
@. @. R_2{}_1K_2\\
@. @. R_2
\end{CD} F ′ ( G ′ R 2 ) a ⏐ ↑ ( F ′ G ′ ) R 2 u 21 ε ′ − 1 F ′ ( R 1 G ) ↓ ⏐ 1 K 2 ′ R 2 a ′ l ( F ′ R 1 ) G ↓ ⏐ R 2 r ⏐ ↑ R 2 ( F G ) R 2 u 12 ( R 2 F ) G ↓ ⏐ a
such that the two weak 2-inverse 1-morphism axioms hold (section D.5).
Definition C.2.2. A 2-morphism ( R 1 , R 2 , u 12 , u 21 , ⇐ , ⇒ ) ⇒ ( S 1 , S 2 , v 12 , v 21 , ⇐ , ⇒ ) (R_1, R_2, u_{12}, u_{21}, \Leftarrow,\Rightarrow) \Rightarrow(S_1, S_2, v_{12}, v_{21}, \Leftarrow,\Rightarrow) ( R 1 , R 2 , u 12 , u 21 , ⇐ , ⇒ ) ⇒ ( S 1 , S 2 , v 12 , v 21 , ⇐ , ⇒ ) consists of
two 2-cells γ 1 : R 1 ⇒ S 1 \gamma_1 : R_1 \Rightarrow S_1 γ 1 : R 1 ⇒ S 1 and γ 2 : R 2 ⇒ S 2 \gamma_2 : R_2 \Rightarrow S_2 γ 2 : R 2 ⇒ S 2 ,
two 3-cells
K 1 → R 1 K 1 ′ F ↓ ↓ F ′ K 2 → S 2 K 2 ′ K 1 → R 1 K 1 ′ F ↓ ↓ F ′ K 2 → R 2 K 2 ′ (1) \begin{CD}
K_1 @>{R_1}>> K'_1\\
@V{F}VV @VV{F'}V\\
K_2 @>{S_2}>> K'_2
\end{CD}
\quad
\begin{CD}
K_1 @>{R_1}>> K'_1\\
@V{F}VV @VV{F'}V\\
K_2 @>{R_2}>> K'_2 \tag*{(1)}
\end{CD} K 1 F ↓ ⏐ K 2 R 1 S 2 K 1 ′ ↓ ⏐ F ′ K 2 ′ K 1 F ↓ ⏐ K 2 R 1 R 2 K 1 ′ ↓ ⏐ F ′ K 2 ′ (1)
K 1 → R 1 K 1 ′ F ↓ ↓ F ′ K 2 → S 2 K 2 ′ ⇒ K 1 → R 1 K 1 ′ F ↓ ↓ F ′ K 2 → R 2 K 2 ′ (2) \begin{CD}
K_1 @>{R_1}>> K'_1\\
@V{F}VV @VV{F'}V\\
K_2 @>{S_2}>> K'_2
\end{CD}
\quad\Rightarrow\quad
\begin{CD}
K_1 @>{R_1}>> K'_1\\
@V{F}VV @VV{F'}V\\
K_2 @>{R_2}>> K'_2 \tag*{(2)}
\end{CD} K 1 F ↓ ⏐ K 2 R 1 S 2 K 1 ′ ↓ ⏐ F ′ K 2 ′ ⇒ K 1 F ↓ ⏐ K 2 R 1 R 2 K 1 ′ ↓ ⏐ F ′ K 2 ′ (2)
such that the two weak 2-inverse 2-morphism axioms hold (section D.6).
\begin{array}{ccccc}
K_{2} & \xrightarrow{S_{2}} & K'_{2} & & K_{2} \xrightarrow{R_{2}} K'_{2}\\
\downarrow G & \swarrow v_{21} \searrow\downarrow G' & & \downarrow G \swarrow u_{21} \searrow\downarrow G'\\
K_{1} & \xrightarrow{S_{1}} & K'_{1} & & K_{1} \xrightarrow{R_{1}} K'_{1}\\
\multicolumn{3}{c}{\overset{R_{2}}{\curvearrowright}\;\overset{\gamma_{2}}{\Downarrow}\;S_{2}} & & \multicolumn{1}{c}{\underset{S_{1}}{\curvearrowleft}\;\overset{\gamma_{1}}{\Downarrow}\;R_{1}}
\end{array}
Definition C.2.3. A 3-morphism ( γ 1 , γ 2 , ω 12 , ω 21 ) ⇒ ( λ 1 , λ 2 , ω 12 , ω 21 ) (\gamma_{1}, \gamma_{2}, \omega_{12}, \omega_{21}) \Rightarrow(\lambda_{1}, \lambda_{2}, \omega_{12}, \omega_{21}) ( γ 1 , γ 2 , ω 12 , ω 21 ) ⇒ ( λ 1 , λ 2 , ω 12 , ω 21 ) consists of
two 3-cells υ 1 : γ 1 ⇒ λ 1 \upsilon_{1} : \gamma_{1} \Rightarrow\lambda_{1} υ 1 : γ 1 ⇒ λ 1 and υ 2 : γ 2 ⇒ λ 2 \upsilon_{2} : \gamma_{2} \Rightarrow\lambda_{2} υ 2 : γ 2 ⇒ λ 2 .
such that the following two identities holds:
\begin{array}{ccccc}
G'R_{2} & \xrightarrow{\lambda_{2}} & G'S_{2} & & G'R_{2} \xrightarrow{\gamma_{2}} G'S_{2}\\
\downarrow u_{21} & \swarrow\omega_{21} \searrow\downarrow v_{21} & & \downarrow u_{21} \swarrow\omega_{21} \searrow\downarrow v_{21}\\
R_{1}G & \xrightarrow{\lambda_{11}} & S_{1}G & \equiv& R_{1}G \xrightarrow{\gamma_{1}} S_{1}G\\
\multicolumn{3}{c}{\overset{\gamma_{2}}{\curvearrowright}\;\overset{\upsilon_{1}}{\Downarrow}\;\lambda_{1}} & &
\end{array}
\begin{array}{ccccc}
F'R_{1} & \xrightarrow{\lambda_{1}} & F'S_{1} & & F'R_{1} \xrightarrow{\gamma_{1}} F'S_{1}\\
\downarrow u_{12} & \swarrow\omega_{12} \searrow\downarrow v_{12} & & \downarrow u_{12} \swarrow\omega_{12} \searrow\downarrow v_{12}\\
R_{2}F & \xrightarrow{\lambda_{21}} & S_{2}F & \equiv& R_{2}F \xrightarrow{\gamma_{21}} S_{2}F\\
\multicolumn{3}{c}{\overset{\gamma_{1}}{\curvearrowright}\;\overset{\upsilon_{2}}{\Downarrow}\;\lambda_{21}} & &
\end{array}
Coherence diagrams for weak objects
Weak 2-involution
( ( H H ) H ) H → a ( H H ) ( H H ) → a H ( H ( H H ) ) ( α 1 ) 1 ↓ a 1 ↓ ↓ 1 ( 1 α ) ( 1 H ) H ( H ( H H ) ) H → a H ( ( H H ) H ) H ( H 1 ) ( 1 α ) 1 ↓ 1 ( α 1 ) ↓ 1 ζ ↑ ( H 1 ) H → a H ( 1 H ) l 1 ↓ r 1 ↓ 1 l ↓ 1 r ↓ H H → ε H H H H ← H ( H 1 ) α ↓ α ↓ 1 1 \begin{CD}
((HH)H)H @>{a}>> (HH)(HH) @>{a}>> H(H(HH)) \\
@V{(\alpha1)1}VV @V{a1}VV @VV{1(1\alpha)}V \\
(1H)H @. (H(HH))H @>{a}>> H((HH)H) @. H(H1) \\
@. @V{(1\alpha)1}VV @. @V{1(\alpha1)}VV @A{1\zeta}AA \\
@. (H1)H @>{a}>> H(1H) @. @. \\
@V{l1}VV @V{r1}VV @V{1l}VV @V{1r}VV \\
HH @>{\varepsilon_{HH}}>> HH @<{}<< H(H1) \\
@V{\alpha}VV @. @V{\alpha}VV \\
1 @. 1
\end{CD} (( H H ) H ) H ( α 1 ) 1 ↓ ⏐ ( 1 H ) H l 1 ↓ ⏐ H H α ↓ ⏐ 1 a ε H H ( H H ) ( H H ) a 1 ↓ ⏐ ( H ( H H )) H ( 1 α ) 1 ↓ ⏐ ( H 1 ) H r 1 ↓ ⏐ H H 1 a a a H ( H ( H H )) ↓ ⏐ 1 ( 1 α ) H (( H H ) H ) H ( 1 H ) 1 l ↓ ⏐ H ( H 1 ) α ↓ ⏐ H ( H 1 ) 1 ( α 1 ) ↓ ⏐ 1 r ↓ ⏐ 1 ζ ⏐ ↑
( ( H H ) H ) H → a ( H H ) ( H H ) → a H ( H ( H H ) ) ( α 1 ) 1 ↓ α 1 ↓ ↓ 1 α ( 1 H ) H → a 1 ( H H ) ( H H ) 1 → a H ( H 1 ) l 1 ↓ l ↓ ↓ r ↓ 1 r H H 11 H H α ↓ l ↓ r ↓ α ↓ 1 1 \begin{CD}
((HH)H)H @>{a}>> (HH)(HH) @>{a}>> H(H(HH)) \\
@V{(\alpha1)1}VV @V{\alpha1}VV @VV{1\alpha}V \\
(1H)H @>{a}>> 1(HH) @. (HH)1 @>{a}>> H(H1) \\
@V{l1}VV @V{l}VV @VV{r}V @VV{1r}V \\
HH @. 11 @. HH \\
@V{\alpha}VV @V{l}VV @V{r}VV @V{\alpha}VV \\
1 @. 1
\end{CD} (( H H ) H ) H ( α 1 ) 1 ↓ ⏐ ( 1 H ) H l 1 ↓ ⏐ H H α ↓ ⏐ 1 a a ( H H ) ( H H ) α 1 ↓ ⏐ 1 ( H H ) l ↓ ⏐ 11 l ↓ ⏐ 1 a H ( H ( H H )) ↓ ⏐ 1 α ( H H ) 1 ↓ ⏐ r H H r ↓ ⏐ a H ( H 1 ) ↓ ⏐ 1 r α ↓ ⏐
Weak 2-involution 1-morphism
( H ′ H ′ ) R , H ′ ( ( H ′ H ′ ) R ) , H ′ ( H ′ ( H ′ H ′ ) ) , H ′ H ′ ( H ′ R ) , H ′ ( ( H ′ R ) H ) , ( R H ) H , ( R H ) 1 , ( H ′ H ′ ) ( R H ) , H ′ ( ( R H ) H ) , ( H ′ R ) 1 , R ( H 1 ) , R H , ( H ′ ( H ′ R ) ) H , ( R ( H H ) ) , ( H ′ R ) H ) H , ( R 1 ) , ( ( R H ) H ) H , R ( ( H H ) H ) , H ′ ( ( H ′ R ) H ) , ( H ′ R ) H , ( H ′ R ) ( H H ) , R ( H ( H H ) ) , H ′ ( H ′ R ) , H ′ ( ( H ′ ) R ) , ( H ′ R ) H , H ′ ( H ′ R ) , H ( ( R H ) ( H H ) ) , ( H ′ H ′ ) R , H ′ ( 1 ( H ′ H ′ ) ) , 1 l , 1 ( 1 u ) , 1 ( u 1 ) , a r , ( 11 ) u , 1 r a , 1 ( 1 α ) , ( 1 u ) 1 , ( 11 ) α , ( 11 ) α , u ( 11 ) , 1 ( 1 α ) 1 l , 1 a 1 ( α 1 ) , ( u 1 ) 1 , ( 1 α ) 1 \begin{gathered}
(H'H')R,\quad H'((H'H')R),\quad H'(H'(H'H')), \quad H'H'(H'R),\quad H'((H'R)H),\\
(RH)H,\quad(RH)1,\quad(H' H') (RH),\quad H'((RH)H),\quad(H'R)1,\quad R(H1),\quad RH,\\
(H'(H'R))H,\quad(R(HH)),\quad(H'R)H)H,\quad(R1),\quad((RH)H)H,\quad R((HH)H),\\
H'((H'R)H),\quad(H'R)H,\quad(H'R)(HH),\quad R(H(HH)),\\
H'(H'R),\quad H'((H')R),\quad(H'R)H,\quad H'(H'R),\quad H((RH)(HH)),\\
(H'H')R,\quad H'(1(H'H')),\\
1_l,\quad1(1_u),\quad1(u1),\quad a_r,\quad(11)_u,\quad1_r a,\quad1(1_\alpha),\\
(1_u)1,\quad(11)_\alpha,\quad(11)_\alpha,\quad u(11),\quad1(1_\alpha)\,1_l,\quad1_a\,1(\alpha1),\\
(u1)1,\quad(1\alpha)1
\end{gathered} ( H ′ H ′ ) R , H ′ (( H ′ H ′ ) R ) , H ′ ( H ′ ( H ′ H ′ )) , H ′ H ′ ( H ′ R ) , H ′ (( H ′ R ) H ) , ( R H ) H , ( R H ) 1 , ( H ′ H ′ ) ( R H ) , H ′ (( R H ) H ) , ( H ′ R ) 1 , R ( H 1 ) , R H , ( H ′ ( H ′ R )) H , ( R ( H H )) , ( H ′ R ) H ) H , ( R 1 ) , (( R H ) H ) H , R (( H H ) H ) , H ′ (( H ′ R ) H ) , ( H ′ R ) H , ( H ′ R ) ( H H ) , R ( H ( H H )) , H ′ ( H ′ R ) , H ′ (( H ′ ) R ) , ( H ′ R ) H , H ′ ( H ′ R ) , H (( R H ) ( H H )) , ( H ′ H ′ ) R , H ′ ( 1 ( H ′ H ′ )) , 1 l , 1 ( 1 u ) , 1 ( u 1 ) , a r , ( 11 ) u , 1 r a , 1 ( 1 α ) , ( 1 u ) 1 , ( 11 ) α , ( 11 ) α , u ( 11 ) , 1 ( 1 α ) 1 l , 1 a 1 ( α 1 ) , ( u 1 ) 1 , ( 1 α ) 1
a , 1 α , 1 r , r , a , ( 1 α ) , 1 c , 1 u , r 1 , r 1 , 1 ( α 1 ) , α , ( 1 α ) , α , ( 11 ) α , α , α , α , ( u 1 ) 1 , ( u 1 ) α \begin{gathered}
a,\quad1_\alpha,\quad1_r,\quad r,\quad a,\quad(1\alpha),\quad1_c,\quad1_u,\quad r_1,\quad r_1,\\
1(\alpha1),\quad\alpha,\quad(1\alpha),\quad\alpha,\quad(11)\alpha,\quad\alpha,\quad\alpha,\quad\alpha,\quad(u1)1,\quad(u1)\alpha
\end{gathered} a , 1 α , 1 r , r , a , ( 1 α ) , 1 c , 1 u , r 1 , r 1 , 1 ( α 1 ) , α , ( 1 α ) , α , ( 11 ) α , α , α , α , ( u 1 ) 1 , ( u 1 ) α
Figure 5 (PDF p. 88).
Weak 2-involution 2-morphism
( H ′ H ′ ) R → α H ′ ( H ′ R ) → 1 α H ′ H ′ ( R ) → 1 ω H ′ H ′ R ( 11 ) γ ↓ 1 ( 1 γ ) ↓ ( 1 γ ) 1 ↓ γ ( 11 ) ↓ H ′ ( R H ) → 1 α H ′ R ( H ) → α H ′ ( R H ) → α H ′ R ( H ) ( 1 γ ) 1 ↓ ( 1 γ ) 1 ↓ ( γ 1 ) 1 ↓ γ ( 11 ) ↓ ( H ′ R ) H → α H ′ ( R H ) → 1 α R ( H H ) → 1 α R 1 → r R \begin{CD}
(H'H')R @>{\alpha}>> H'(H'R) @>{1\alpha}>> H'H'(R) @>{1\omega}>> H'H'R \\
@V{(11)\gamma}VV @V{1(1\gamma)}VV @V{(1\gamma)1}VV @V{\gamma(11)}VV \\
H'(RH) @>{1\alpha}>> H'R(H) @>{\alpha}>> H'(RH) @>{\alpha}>> H'R(H) \\
@V{(1\gamma)1}VV @V{(1\gamma)1}VV @V{(\gamma1)1}VV @V{\gamma(11)}VV \\
(H'R)H @>{\alpha}>> H'(RH) @>{1\alpha}>> R(HH) @>{1\alpha}>> R1 @>{r}>> R
\end{CD} ( H ′ H ′ ) R ( 11 ) γ ↓ ⏐ H ′ ( R H ) ( 1 γ ) 1 ↓ ⏐ ( H ′ R ) H α 1 α α H ′ ( H ′ R ) 1 ( 1 γ ) ↓ ⏐ H ′ R ( H ) ( 1 γ ) 1 ↓ ⏐ H ′ ( R H ) 1 α α 1 α H ′ H ′ ( R ) ( 1 γ ) 1 ↓ ⏐ H ′ ( R H ) ( γ 1 ) 1 ↓ ⏐ R ( H H ) 1 ω α 1 α H ′ H ′ R γ ( 11 ) ↓ ⏐ H ′ R ( H ) γ ( 11 ) ↓ ⏐ R 1 r R
( H ′ S ) H → α H ′ ( S H ) → 1 α S ( H H ) → 1 α S 1 → s S \begin{CD}
(H'S)H @>{\alpha}>> H'(SH) @>{1\alpha}>> S(HH) @>{1\alpha}>> S1 @>{s}>> S
\end{CD} ( H ′ S ) H α H ′ ( S H ) 1 α S ( H H ) 1 α S 1 s S
Coherence diagrams for weak objects
Figure III.
Weak 2-inverse
Figure 6.
Figure 7.
Coherence diagrams for weak objects
G F ↙ η ↓ ℓ ↘ 1 → ℓ 11 → 1 η 1 ( G F ) → ℓ ⋅ 1 ( 1 G ) F → ( η 1 ) 1 r ↓ η 1 ↓ a ⋅ ↓ a ⋅ ↓ → 1 r G ( F 1 ) → 1 ( η ) G ( F ( G F ) ) → 1 a G ( ( F G ) F ) → a 1 ( G ( F G ) ) F → ( 1 ϵ ) 1 ( G F ) ( G F ) → a G ( F ( G F ) ) → 1 ϕ G ( ( F G ) F ) r ↓ a ↓ a ↓ a ↓ ( G F ) 1 → 1 η ( G F ) ( G F ) → a G ( F ( G F ) ) → 1 ( η ) G ( F 1 ) → 1 ( ϵ ) G F \begin{CD}
&& GF \\
& \swarrow^{\eta} \downarrow^{\ell} \searrow^{1} && \\
1 @>{\ell}>> 11 @>{1\eta}>> 1(GF) @>{\ell\cdot1}>> (1G)F @>{(\eta1)1}>> ((GF)G)F \\
@V{r}VV @V{\eta1}VV @. @V{a\cdot}VV @V{a\cdot}VV \\
GF @>{1r}>> G(F1) @>{1(\eta)}>> G(F(GF)) @>{1a}>> G((FG)F) @>{a1}>> (G(FG))F @>{(1\epsilon)1}>> (G1)F \\
&& (GF)(GF) @>{a}>> G(F(GF)) @>{1\phi}>> G((FG)F) \\
@. @V{r}VV @V{a}VV @V{a}VV @V{a}VV \\
&& (GF)1 @>{1\eta}>> (GF)(GF) @>{a}>> G(F(GF)) @>{1(\eta)}>> G(F1) \\
&&&& G((FG)F) @>{1(\epsilon)}>> G(1F) \\
&&&&&& GF
\end{CD} GF ↙ η ↓ ℓ ↘ 1 ℓ 1 r ( GF ) ( GF ) ( GF ) 1 1 ( ϵ ) GF 11 r ↓ ⏐ G ( F 1 ) a 1 η 1 η 1 ( η ) G ( F ( GF )) r ↓ ⏐ ( GF ) ( GF ) 1 ( GF ) η 1 ↓ ⏐ G ( F ( GF )) 1 ϕ a ℓ ⋅ 1 1 a G (( F G ) F ) a ↓ ⏐ G ( F ( GF )) ( 1 G ) F G (( F G ) F ) 1 ( η ) ( η 1 ) 1 a 1 a ↓ ⏐ G ( F 1 ) a ⋅ ↓ ⏐ ( G ( F G )) F ( 1 ϵ ) 1 a ↓ ⏐ a ⋅ ↓ ⏐
1 → η G F → 1 G F η ↓ ↓ 1 G F → 1 G F \begin{CD}
1 @>{\eta}>> GF @>{1}>> GF \\
@V{\eta}VV &&&& @VV{1}V \\
GF @>{1}>> GF
\end{CD} 1 η ↓ ⏐ ↓ ⏐ 1 GF η 1 GF GF 1 GF
III
Weak 2-inverse 1-morphism
Figure 8 (PDF p. 93).
Figure 9 (PDF p. 94).
Coherence diagrams for weak objects
Figure 10.
Coherence diagrams for weak objects
Figure 11.
Weak 2-inverse 2-morphism
R 1 → ι ′ 1 R 1 → η 1 ( G ′ F ) R 1 → α G ′ ( F R 1 ) → ι 12 G ′ ( R 2 F ) → α ′ ( G ′ R 2 ) F → υ 21 ( R 1 G ) F γ 1 ↓ ( γ 1 ) 1 ↓ 1 ( γ 1 ) ↓ 1 ( γ 2 ) ↓ ( 1 γ 2 ) 1 ↓ ( γ 11 ) 1 ↓ ( γ 1 ) 11 ↓ S 1 → ι ′ 1 S 1 → η 1 ( G ′ F ) S 1 → α G ′ ( F S 1 ) → ι 12 G ′ ( S 2 F ) → α ′ ( G ′ S 2 ) F → υ 21 ( S 1 G ) F \begin{CD}
R_1 @>{\iota'}>> 1R_1 @>{\eta_1}>> (G'F)R_1 @>{\alpha}>> G'(FR_1) @>{\iota_{12}}>> G'(R_2F) @>{\alpha'}>> (G'R_2)F @>{\upsilon_{21}}>> (R_1G)F \\
@V{\gamma_1}VV @V{(\gamma_1)1}VV @V{1(\gamma_1)}VV @V{1(\gamma_2)}VV @V{(1\gamma_2)1}VV @V{(\gamma_{11})1}VV @V{(\gamma_1)11}VV \\
S_1 @>{\iota'}>> 1S_1 @>{\eta_1}>> (G'F)S_1 @>{\alpha}>> G'(FS_1) @>{\iota_{12}}>> G'(S_2F) @>{\alpha'}>> (G'S_2)F @>{\upsilon_{21}}>> (S_1G)F
\end{CD} R 1 γ 1 ↓ ⏐ S 1 ι ′ ι ′ 1 R 1 ( γ 1 ) 1 ↓ ⏐ 1 S 1 η 1 η 1 ( G ′ F ) R 1 1 ( γ 1 ) ↓ ⏐ ( G ′ F ) S 1 α α G ′ ( F R 1 ) 1 ( γ 2 ) ↓ ⏐ G ′ ( F S 1 ) ι 12 ι 12 G ′ ( R 2 F ) ( 1 γ 2 ) 1 ↓ ⏐ G ′ ( S 2 F ) α ′ α ′ ( G ′ R 2 ) F ( γ 11 ) 1 ↓ ⏐ ( G ′ S 2 ) F υ 21 υ 21 ( R 1 G ) F ( γ 1 ) 11 ↓ ⏐ ( S 1 G ) F
R 1 → ι ′ 1 R 1 → η 1 ( G ′ F ) R 1 → α G ′ ( F R 1 ) → ι 12 G ′ ( R 2 F ) → α ′ ( G ′ R 2 ) F → υ 21 ( R 1 G ) F γ 1 ↓ ( γ 1 ) 1 ↓ 1 ( γ 1 ) ↓ 1 ( γ 2 ) ↓ ( 1 γ 2 ) 1 ↓ ( γ 11 ) 1 ↓ ( γ 1 ) 11 ↓ S 1 → ι ′ 1 S 1 → η 1 ( G ′ F ) S 1 → α G ′ ( F S 1 ) → ι 12 G ′ ( S 2 F ) → α ′ ( G ′ S 2 ) F → υ 21 ( S 1 G ) F \begin{CD}
R_1 @>{\iota'}>> 1R_1 @>{\eta_1}>> (G'F)R_1 @>{\alpha}>> G'(FR_1) @>{\iota_{12}}>> G'(R_2F) @>{\alpha'}>> (G'R_2)F @>{\upsilon_{21}}>> (R_1G)F \\
@V{\gamma_1}VV @V{(\gamma_1)1}VV @V{1(\gamma_1)}VV @V{1(\gamma_2)}VV @V{(1\gamma_2)1}VV @V{(\gamma_{11})1}VV @V{(\gamma_1)11}VV \\
S_1 @>{\iota'}>> 1S_1 @>{\eta_1}>> (G'F)S_1 @>{\alpha}>> G'(FS_1) @>{\iota_{12}}>> G'(S_2F) @>{\alpha'}>> (G'S_2)F @>{\upsilon_{21}}>> (S_1G)F
\end{CD} R 1 γ 1 ↓ ⏐ S 1 ι ′ ι ′ 1 R 1 ( γ 1 ) 1 ↓ ⏐ 1 S 1 η 1 η 1 ( G ′ F ) R 1 1 ( γ 1 ) ↓ ⏐ ( G ′ F ) S 1 α α G ′ ( F R 1 ) 1 ( γ 2 ) ↓ ⏐ G ′ ( F S 1 ) ι 12 ι 12 G ′ ( R 2 F ) ( 1 γ 2 ) 1 ↓ ⏐ G ′ ( S 2 F ) α ′ α ′ ( G ′ R 2 ) F ( γ 11 ) 1 ↓ ⏐ ( G ′ S 2 ) F υ 21 υ 21 ( R 1 G ) F ( γ 1 ) 11 ↓ ⏐ ( S 1 G ) F
S 11 → ι η S 1 ( G F ) S_{11} \xrightarrow{\iota_\eta} S_1(GF) S 11 ι η S 1 ( GF )
S 1 ( G F ) → α ′ ( S 1 G ) F S_1(GF) \xrightarrow{\alpha'} (S_1G)F S 1 ( GF ) α ′ ( S 1 G ) F
R 11 → ι η R 1 ( G F ) R_{11} \xrightarrow{\iota_\eta} R_1(GF) R 11 ι η R 1 ( GF )
R 1 ( G F ) → α ′ ( R 1 G ) F R_1(GF) \xrightarrow{\alpha'} (R_1G)F R 1 ( GF ) α ′ ( R 1 G ) F
S 1 → γ 1 1 R 1 → η 1 ( G F ) R 1 → α G ′ ( F R 1 ) → ι 12 G ′ ( R 2 F ) → α ′ ( G ′ R 2 ) F → u 21 ( R 1 G ) F γ 1 ↓ ι 1 ↓ ι 1 ↓ α ′ ↑ R 1 → γ 1 ′ S 1 → η 1 S 1 ( G F ) ( R 1 G ) F γ 11 ↓ γ 11 ↓ ( γ 11 ) 1 ↑ R 1 → η 1 R 1 ( G F ) → α ′ ( G ′ R 2 ) F \begin{CD}
S_1 @>{\gamma_1}>> 1R_1 @>{\eta_1}>> (GF)R_1 @>{\alpha}>> G'(FR_1) @>{\iota_{12}}>> G'(R_2F) @>{\alpha'}>> (G'R_2)F @>{u_{21}}>> (R_1G)F \\
@V{\gamma_1}VV @V{\iota_1}VV @V{\iota_1}VV @. @. @. @A{\alpha'}AA \\
R_1 @>{\gamma_1'}>> S_1 @>{\eta_1}>> S_1(GF) @. @. @. (R_1G)F \\
@. @V{\gamma_{11}}VV @V{\gamma_{11}}VV @. @. @. @A{(\gamma_{11})^1}AA \\
@. R_1 @>{\eta_1}>> R_1(GF) @>{\alpha'}>> (G'R_2)F @.
\end{CD} S 1 γ 1 ↓ ⏐ R 1 γ 1 γ 1 ′ 1 R 1 ι 1 ↓ ⏐ S 1 γ 11 ↓ ⏐ R 1 η 1 η 1 η 1 ( GF ) R 1 ι 1 ↓ ⏐ S 1 ( GF ) γ 11 ↓ ⏐ R 1 ( GF ) α α ′ G ′ ( F R 1 ) ( G ′ R 2 ) F ι 12 G ′ ( R 2 F ) α ′ ( G ′ R 2 ) F ( R 1 G ) F u 21 ( R 1 G ) F α ′ ⏐ ↑ ( γ 11 ) 1 ⏐ ↑
Coherence diagrams for weak objects
( F ′ G ′ ) S 2 → α F ′ ( G ′ S 2 ) → 1 S 2 F ′ ( S 1 G ) → α − 1 ( F ′ S 1 ) G → v 121 ( S 2 F ) G → α S 2 ( F G ) → 1 S 2 S 2 ( 1 1 ) γ 2 ↓ 1 ( γ 2 ) ↓ ( γ 1 1 ) ↓ ( γ 21 ) 1 ↓ γ 2 ( 1 1 ) ↓ γ 2 ↓ ( F ′ G ′ ) R 2 → α − 1 F ′ ( G ′ R 2 ) → 1 R 2 F ′ ( R 1 G ) → α − 1 ( F ′ R 1 ) G → u 121 ( R 2 F ) G → α − 1 R 2 ( F G ) → 1 R 2 R 2 \begin{CD}
(F'G')S_{2} @>{\alpha}>> F'(G'S_{2}) @>{1_{S_{2}}}>> F'(S_{1}G) @>{\alpha^{-1}}>> (F'S_{1})G @>{v_{121}}>> (S_{2}F)G @>{\alpha}>> S_{2}(FG) @>{1_{S_{2}}}>> S_{2} \\
@V{(1\,1)\gamma_{2}}VV @V{1(\gamma_{2})}VV @V{(\gamma_{1}1)}VV @V{(\gamma_{21})1}VV @V{\gamma_{2}(1\,1)}VV @V{\gamma_{2}}VV \\
(F'G')R_{2} @>{\alpha^{-1}}>> F'(G'R_{2}) @>{1_{R_{2}}}>> F'(R_{1}G) @>{\alpha^{-1}}>> (F'R_{1})G @>{u_{121}}>> (R_{2}F)G @>{\alpha^{-1}}>> R_{2}(FG) @>{1_{R_{2}}}>> R_{2}
\end{CD} ( F ′ G ′ ) S 2 ( 1 1 ) γ 2 ↓ ⏐ ( F ′ G ′ ) R 2 α α − 1 F ′ ( G ′ S 2 ) 1 ( γ 2 ) ↓ ⏐ F ′ ( G ′ R 2 ) 1 S 2 1 R 2 F ′ ( S 1 G ) ( γ 1 1 ) ↓ ⏐ F ′ ( R 1 G ) α − 1 α − 1 ( F ′ S 1 ) G ( γ 21 ) 1 ↓ ⏐ ( F ′ R 1 ) G v 121 u 121 ( S 2 F ) G γ 2 ( 1 1 ) ↓ ⏐ ( R 2 F ) G α α − 1 S 2 ( F G ) γ 2 ↓ ⏐ R 2 ( F G ) 1 S 2 1 R 2 S 2 R 2
Coherence diagrams for weak objects
( F ′ G ′ ) S 2 → α F ′ ( G ′ S 2 ) → θ 21 ( F S 1 ) G → α S 2 ( F G ) → ℓ S 2 ( 11 ) γ 2 ↓ ϵ 1 ↓ θ 21 ↓ ↓ γ 2 ↑ ( F G ′ ) R 2 → ϵ 1 1 R 2 → θ 2 1 S 2 → ℓ R 2 \begin{CD}
(F'G')S_2 @>{\alpha}>> F'(G'S_2) @>{\theta_{21}}>> (F S_1)G @>{\alpha}>> S_2(FG) @>{\ell}>> S_2 \\
@V{(11)\gamma_2}VV @V{\epsilon_1}VV @V{\theta_{21}}VV @V{}VV @A{\gamma_2}AA \\
(FG')R_2 @>{\epsilon_1}>> \mathbf{1}R_2 @>{\theta_2}>> \mathbf{1}S_2 @>{\ell}>> R_2
\end{CD} ( F ′ G ′ ) S 2 ( 11 ) γ 2 ↓ ⏐ ( F G ′ ) R 2 α ϵ 1 F ′ ( G ′ S 2 ) ϵ 1 ↓ ⏐ 1 R 2 θ 21 θ 2 ( F S 1 ) G θ 21 ↓ ⏐ 1 S 2 α ℓ S 2 ( F G ) ↓ ⏐ R 2 ℓ S 2 γ 2 ⏐ ↑
References