5
$\begingroup$

I employ the vast majority of the post to develop the notion of quasi-functor between dg-categories: I think it is important to get the idea.

Let $k$ be a field, and let $\mathcal V =\mathbf C(k)$ be the category of cochain complexes of $k$-vector spaces and chain maps. $\mathcal V$ is a symmetric closed monoidal category, and also has a structure of model category (the weak equivalences are the quasi-isomorphisms, the fibrations are the surjective chain maps). Categories enriched over $\mathcal V$ are called dg-categories. $\mathcal V$ is also enriched over itself.

Let $\mathcal A, \mathcal B$ be categories enriched over $\mathcal V$. The category $\mathrm{Fun}_{\mathcal V} (\mathcal A, \mathcal B)$ of $\mathcal V$-functors and $\mathcal V$-natural transformations is also enriched over $\mathcal V$. The (enriched) Yoneda lemma holds, and we have the (fully faithful) Yoneda embedding: \begin{equation} \mathcal A \hookrightarrow \textrm{mod-}\mathcal A =: \mathrm{Fun}_{\mathcal V}(\mathcal A^{\mathrm{op}},\mathcal V). \end{equation}

The category $\textrm{mod-}\mathcal A$ of (right) $\mathcal A$-modules inherits a (levelwise) model structure from $\mathcal V$: for example, quasi-equivalences are given by levelwise quasi-equivalences.

Now, let me give the following definition: given $\mathcal A, \mathcal B$ two $\mathcal V$-categories, a $\mathcal V$-functor $F: \mathcal A \to \textrm{mod-}\mathcal B$ is called a quasi-functor if for any $A \in \mathcal A$, $F(A)$ is isomorphic to a representable right $\mathcal B$-module in the homotopy category $\mathrm{Ho}(\text{mod-}\mathcal B)$.

Now, observe that we can define a "tensor product" $\mathcal A \otimes \mathcal B$ of $\mathcal V$-categories. Moreover, $\mathcal V$-functors $\mathcal A \otimes \mathcal B^{\mathrm{op}} \to \mathcal V$ correspond exactly to $\mathcal V$-functors $\mathcal A \to \textrm{mod-}\mathcal B$. They are precisely the bimodules (or profunctors). Let me denote by $\mathcal A\text{-mod-}\mathcal B$ the $\mathcal V$-category of bimodules (covariant in $\mathcal A$, contravariant in $\mathcal B$). Since it is a category of modules, it is a model category, with the levelwise structure discussed above.

Now, I can define $\mathrm{rep}(\mathcal A,\mathcal B)$ as the full subcategory of $\mathrm{Ho}(\mathcal A\text{-mod-}\mathcal B)$ whose objects are the quasi-functors. Moreover, I can define $\mathrm{rep}_{\mathcal V}(\mathcal A,\mathcal B)$ as the full $\mathcal V$-subcategory of $\mathcal A\text{-mod-}\mathcal B$ whose objects are the quasi-functors which are also cofibrant as bimodules.

Where can we go from here? Well, the category $\mathcal V$-$\mathbf{Cat}$ of categories enriched over $\mathcal V$ has itself a model structure: it is a known result by G. Tabuada about dg-categories, more recently generalized. The homotopy category $\mathrm{Ho}(\mathcal V\text{-}\mathbf{Cat})$ is monoidal, and $\mathrm{rep}_{\mathcal V}(\mathcal A,\mathcal B)$ gives the internal hom. In the world of dg-categories, $\mathrm{rep}(\mathcal A,\mathcal B)$ is (equivalent to) $H^0(\mathrm{rep}_{\mathcal V}(\mathcal A,\mathcal B))$.

Finally, here is my question: generalize from this particular case $\mathcal V = \mathbf C(k)$ to a more general setting, possibly letting $\mathcal V$ be a monoidal model category (with some assumptions). Every definition given above should work without problems. Has someone developed a "general theory" of those quasi-functors? I've studied it in the case of dg-categories, but I guess it is just a particular case.

$\endgroup$
5
$\begingroup$

I think the right way to call this functors is "quasi-representable" functors (in the language of Toen) and "potentially distinguished" functors in the language of Dwyer-Hess. At least when $\mathbf{V}$ is one of these model categories 1) simplicial sets ($\mathbf{sSet}$), (compactly generated) topological space ($\mathbf{Top}$), symmetric spectra ($\mathbf{Sp}^{\Sigma}$) you can almost copy-paste Toen's argument to construct 1) the (derived) mapping space (using the moduli space of quasi representable functors) 2) the derived internal Hom of the model category $\mathbf{V}-Cat$.

$\endgroup$
  • $\begingroup$ My impression is that the Bergner model structure on simplicially enriched categories is not cartesian. So how do you construct the derived internal hom? $\endgroup$ – Zhen Lin Aug 9 '14 at 9:32
  • $\begingroup$ For all categories $\mathbf{V}$ (cited in my answer), the category $\mathbf{V}-Cat$ is symmetric closed monoidal category and a model category BUT the Monoidal closed structure and the model structure are NOT compatible in the sense of Hovey. What you can prove is that the homotopy category $Ho(\mathbf{V}-Cat)$ is a symmetric monodical closed category where the (derived) internal Hom is described in terms of $Rep_{\mathbf{V}}(A,B)$. It is a difficult theorem done by Toen. $\endgroup$ – Ilias Amrani Aug 9 '14 at 10:05
  • $\begingroup$ What I was saying in my answer is that in the case where $\mathbf{V}$ is the category of simplicial sets, topological spaces or Spectra you can repeat Toen's arguments i.e., the category $Ho(\mathbf{V}-Cat)$ is a symmetric monoidal closed category, where the internal Hom is given by $Rep_{\mathbf{V}}(-,-)$ and the mapping space is the moduli space of (right) quasi-representable functors. $\endgroup$ – Ilias Amrani Aug 9 '14 at 10:08
  • 1
    $\begingroup$ In the appendix of Higher Topos Theory there is a relatively easy construction of the internal hom (it follows from Lemma A.3.4.6 and Corollary A.3.4.14). On the other side, Toen does something much more difficult: he computes the mapping spaces. I am not aware of a generalization of this last result to $V \textrm{-} \mathbf{Cat}$. $\endgroup$ – Mauro Porta Aug 9 '14 at 10:56
2
$\begingroup$

In some sense, the "universal" version of this fact was proved by Blumberg-Gepner-Tabuada as Proposition 3.3 in this paper. That is, they proved the analogue for stable $\infty$-categories, which is the description of the $\infty$-category $Fun^{ex}(A, B)$, the internal hom in the $\infty$-category of idempotent complete stable $\infty$-categories and exact functors, via what they call "right-compact" $A$-$B$-bimodules, which are the analogue of right quasi-representable functors or quasi-functors. In fact, the proof is much nicer in this setting.

I claim that any "reasonable" version of Toën's theorem should follow from this version. By that I mean for any $\mathcal{V}$ such that $\mathcal{V}$-categories model some version of stable $\infty$-categories (so that my claim is basically tautological). For example, for spectral categories ($\mathcal{V}$-categories where $\mathcal{V}$ is the symmetric monoidal model category of symmetric spectra), this follows from the equivalence in Theorem 1.10 between idempotent complete stable $\infty$-categories, and spectral categories up to Morita equivalence. For dg-categories, it should follow similarly by using the equivalence between dg-categories and linear stable $\infty$-categories. (Note that the internal hom in the $\infty$-category of dg-categories is the same as the derived internal hom in the model category of dg-categories.)

$\endgroup$

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.