# The locally model bicategory of $\cal V$-profunctors

Not related to this old question of mine, but takes the question from a different perspective.

Let $\mathcal V$ be a monoidal model category (following the def of Hovey, for example). Then there is a bicategory $\text{Prof}(\mathcal V)$ of $\cal V$-valued profunctors, which has the following interesting property:

every hom-category has a model structure[1].

This is, I think, the paradigmatic example of a "locally model bicategory". There can be others: I'm interested in examples for this notion and related results.

I'm trying to get a list of properties to impose to the general notion (a "locally model 2-category", i.e. a 2-category "enriched" over model categories -this is not a true definition, as there is no sensible monoidal structure on model categories-).

Let's for example consider the following explicit question:

Let $\varphi\colon {\bf A}\looparrowright{\bf B}$ be a profunctor, and $\bf X$ be a category; then precomposition by $\varphi$ gives $$\text{Prof}({\bf B},{\bf X}) \overset{-\diamond \varphi}\to \text{Prof}({\bf A},{\bf X})$$ which has left and right adjoints $\text{Lan}_\varphi$, $\text{Ran}_\varphi$ defined by the co/ends $$\text{Ran}_\varphi\psi(b,x) \cong \int_a \hom(\varphi(a,b), \psi(a,x))$$ (Lan is similar). Does $-\diamond \varphi \dashv \text{Ran}_\varphi$ form a Quillen adjunction?

===

[1] in fact, many! Let's take $\mathcal V = \bf sSet$ and declare that I want to study the injective model structure on $\text{Prof}(\mathbf{sSet})(\mathbf A,\mathbf B)=[\mathbf A^\text{op}\times \mathbf B,\mathbf{sSet}]$.

• I don't think that precomposition by a profunctor has a left adjoint in general; it is the left adjoint. Precomposition by a functor has both a left and a right adjoint, but that string of three adjoints arises from the single adjunctions associated to the two representable profunctors. – Mike Shulman Oct 1 '16 at 20:16
• You're right indeed. It is nevertheless possible to define $\rm Ran$ and $\rm Rift$, right? – Fosco Oct 1 '16 at 21:32
• Yes. That is, the bicategory of profunctors is closed. – Mike Shulman Oct 2 '16 at 20:50
• Also, it's not clear to me that the hom-categories of $\mathrm{Prof}(\mathcal{V})$ have model structures absent any further assumptions on $\mathcal{V}$. If $\mathcal{V}$ is cofibrantly generated, then you can try to define a projective model structure at least, but even then I don't know how to show that it actually works unless the hom-objects of your $\mathcal{V}$-categories are cofibrant, or if $\mathcal{V}$ satisfies the "monoid axiom". – Mike Shulman Oct 2 '16 at 21:03
• I'm not claiming I don't need further assumptions on $\cal V$! – Fosco Oct 2 '16 at 22:03

Given two model categories $\mathcal{M},\mathcal{N}$, one does know what would have been a left Quillen functor out of what would have been the tensor product $\mathcal{M} \otimes \mathcal{N}$ into a third model category $\mathcal{K}$, and that is a left Quillen bifunctor $\mathcal{M} \times \mathcal{N} \to \mathcal{K}.$ This leads to a natural notion of a category (weakly) enriched in model categories, namely, a bicategory $\mathcal{C}$ such that each mapping category $\mathrm{Map}_{\mathcal{C}}(X,Y)$ carries a model structure and such that each composition operation $$\mathrm{Map}_{\mathcal{C}}(X,Y) \times \mathrm{Map}_{\mathcal{C}}(Y,Z) \to \mathrm{Map}_{\mathcal{C}}(X,Z)$$ is a left Quillen bifunctor. In the case of profunctors this can be achieved, for example, if one endows $\mathrm{Fun}(\mathbf{A}^{\mathrm{op}} \times \mathbf{B},\mathcal{V})$ with the model structure which is injective in the $\mathbf{A}^{\mathrm{op}}$ coordinate and projective in the $\mathbf{B}$ coordinate (call it the injective-projective model structure). Note that if one defines enrichment in model categories in this way then pre-composition and post-composition with a cofibrant morphism is a left Quillen functor. In particular, pre-composition or post-composition with a cofibrant profunctor $\varphi:\mathbf{A} \looparrowright \mathbf{B}$ (with respect to the injective-projective model structure) induces a left Quillen functor $\text{Prof}({\bf B},{\bf X}) \overset{-\diamond \varphi}\to \text{Prof}({\bf A},{\bf X})$ (both equipped with the injective-projective model structure).
• Why co-ending a cofibration in $\cal V$ should give a cofibration? – Fosco Oct 1 '16 at 16:41
• Yes, you should assume $\mathcal{V}$ to be combinatorial. The main point is that the coend operation $Fun({\bf B},\mathcal{V}) \times Fun({\bf B^{op}},\mathcal{V}) \to \mathcal{V}$ is a left Quillen bifunctor (see Remark A.2.9.27 in HTT), where the first factor is endowed with the projective model structure and the second factor with the injective model structure. -> – Yonatan Harpaz Oct 1 '16 at 19:30
• -> One can then deduce that the composition operation $Fun({\bf A^{op}} \times {\bf B},\mathcal{V}) \times Fun({\bf B^{op}} \times {\bf C} ,\mathcal{V}) \to Fun({\bf A^{op}} \times {\bf C}, \mathcal{V})$ is a left Quillen bifunctor as well (where both side carry the injective-projective model structure). – Yonatan Harpaz Oct 1 '16 at 19:33