# Co/fibrant replacements via coend calculus

In the paper

Cordier, Jean-Marc, and Timothy Porter. "Homotopy coherent category theory." Transactions of the American Mathematical Society 349.1 (1997): 1-54.

the authors define a notion of coherent co/end $\oint T(a,a)$ as $$\int_{a',a''} T(a',a'')^{\delta A(a',a'')}$$ where $\delta A(a',a'')$ is a suitable "resolution" of the hom functor (linked to the bar construction; follow §1, and in particular the paragraphs above Definition 1.1).

The main claim of this paper is that category theory can be rewritten in a homotopy coherent fashion, building on this definition of homotopy coherent co/end. It is obvious why I like this point of view :-)

One of the first interesting claims is the following: fix a simplicial functor $F : A \to B$ between simplicial categories, and define $$\begin{gather} \overline F = \{\delta A(a,-),F\} = \oint_{x:A} Fx^{A(a,x)}\\ \underline F = \delta A(-,a)\otimes F = \oint^{x:A} Fx \otimes A(x,a) \end{gather}$$ This is the counterpart of the "ninja Yoneda lemma" in covariant and contravariant form.

Coend-fu now gives that "underlines and overlines absorb coherence" in that $$\text{CohNat}(F,G) \cong \text{Nat}(F, \overline G)\cong \text{Nat}(\underline F,G)$$ This gives natural transformations $$\eta_F : F \Rightarrow \overline F \qquad\qquad \eta^F : \underline F \Rightarrow F$$ as the images of the identity coherent natural transformation of $F$. The authors then prove that these maps are levelwise homotopy equivalences (Prop. 3.4).

Are $\overline F, \underline F$ replacements for $F$ in suitable model structures on categories of functors? Is $\overline F$ fibrant if $F \in [A, {\bf sSet}]_\text{inj}$? Is $\underline F$ cofibrant if $F \in [A, {\bf sSet}]_\text{proj}$?

• If A is the terminal simplicial category, the first constructions simply returns F (in this case, simply a simplicial set), whereas an injective fibrant replacement of F would be a Kan complex. Feb 18 '17 at 4:28
• Why you say so? If $A$ is terminal (by which I mean the category with a single object $*$ and $\Delta[0]$ as $A(*,*)$), then so is $\delta A$? Feb 18 '17 at 9:59
• In this case F=\bar F, so if F is a nonfibrant simplicial set, \bar F cannot be fibrant in the injective model structure. Feb 18 '17 at 19:02
• With respect to the other case (F̱) I recommend looking at Dugger's ‘Universal homotopy theories’, §2.6, where he discusses an explicit cofibrant replacement functor in the projective model structure. Feb 19 '17 at 23:38

Addressing this sort of question was one of the main goals of math/0610194. The best answer I was able to give is that if $B$ has a suitable model structure with respect to which $F$ is objectwise fibrant (resp. cofibrant), then $\overline{F}$ (resp. $\underline{F}$) belongs to a "right (resp. left) deformation retract" (an abstraction of the notion of fibrant (resp. cofibrant) replacement) that is suitable for constructing derived functors of functors such as homotopy limits (resp. colimits) and also the homotopy category of the functor category (which therefore involves homotopy coherent transformations).

A closely related perspective that you may also be interested in can be found in this paper by Gambino: the colimit of $\underline{F}$ is equivalently the colimit of $F$ weighted by a projective-cofibrant replacement of the terminal weight.

• Mike: Thanks, I was trying to find where I had seen this and was intending to check on your paper later. I had forgotten about Nicola Gambino's paper, but should have mentioned that as well. Feb 20 '17 at 10:08

This will not answer your question but is to mention that we only defined (and studied in depth) the functors $\underline{F}$ and $\overline{F}$ where the codomain $\mathbf{B}$ of the functor was a 'locally Kan' simplicially enriched category and thus fibrant in the usual model category structure on $\mathcal{S}$-cat. (This eliminates the awkwardness of the example that Dimitri mentions.) This means that the unit and counit maps $\eta_F$ etc, are levelwise homotopy equivalences not just weak equivalences. (N.B. Our philosophy was to show such things as directly as possible so there is no use of spectral sequences (at least explicitly in our contribution), for instance, by showing some map is a weak equivalence, followed by invocation of the Kan-ness of the objects involved so get it is a 'real' homotopy equivalence. This makes some of the arguments more constructive' but at the cost of being heavier on explicit, almost combinatorial or geometric, computation with induction up skeleta etc. as a tool.)

I suspect that some of the ideas explored in the article:

P. J. Ehlers and T. Porter, Ordinal subdivision and special pasting in quasicategories, Advances in Mathematics, 217 (2007), No 2. 489 - 518

may be useful, but do not, at the moment, see how to prove what you suggest to be the case. It is worth noting that Dan Dugger's construction, mentioned by Dimitri, is very similar to that given by `$F$ goes to $\overline{F}$', so adapting his proof may give the result.