dg: is for differential graded

Suppose that $F: C\rightarrow D$ is a dg-functor between small dg-categories such that:

  1. F: Objects of $C$ $\rightarrow$ Objects of $D$ is injective.

  2. $Hom_{C}(a,b)\rightarrow Hom_{D}(F(a),F(b))$ induces an isomorphism in homology for any $a, b \in C$

Let $\hat{C}$ be the category of dg-$C$-modules. Is it true that the induced dg-functor $$\hat{F}:\hat{C}\rightarrow \hat{D}$$ is faithfull in the sense that

for any $x,y\in \hat{C} $, $Hom_{\hat{C}}(a,b)\rightarrow Hom_{\hat{D}}(\hat{F}(a),\hat{F}(b))$ induces an isomorphism in homology.

Edit: I'm also interested in the particular case when $F:C\rightarrow D$ is an embedding of dg-categories e.g. when $C$ is a full dg-subcategory of $D$.

  • $\begingroup$ The example given in this question may be useful: mathoverflow.net/questions/244699/… (it is for sets instead of chain complexes, but can probably be adapted). $\endgroup$ Jun 21 '17 at 10:48
  • $\begingroup$ Which induced dg functor? There are two, and the most natural one goes in the other direction. $\endgroup$ Jun 25 '17 at 6:17

Let me start with an answer for your last claim, namely, that a fully faithful dg-functor $F \colon \mathcal C \to \mathcal D$ induces a fully faithful dg-functor $\hat{F} \colon \hat{\mathcal C} \to \hat{\mathcal D}$ between the dg-categories of right dg-modules. This is true, and the reason lies in general facts about Kan extensions. In any case, a sketch of proof would be as follows.

First, $\hat{F}$ is usually defined as the "induction" dg-functor $\operatorname{Ind}_F$, and in particular $$ \operatorname{Ind}_F(M)(D) = M \otimes_{\mathcal C} \mathcal D(D,F(-)) = \int^C M(C) \otimes\mathcal D(D,F(C)), $$ if you also like coends. To show that $\operatorname{Ind}_F$ is fully faithful, we first recall that it is by definition the left adjoint of the restriction dg-functor $\operatorname{Res}_F \colon \hat{\mathcal D} \to \hat{\mathcal C}$, which maps $N$ to $N \circ F$. There is a unit morphism $$ M \to \operatorname{Res}_F \operatorname{Ind}_F(M), $$ and if we check that this is an isomorphism, then $\operatorname{Ind}_F$ will be fully faithful. But now we can compute: \begin{align*} \operatorname{Res}_F \operatorname{Ind}_F(M)(C) &= M \otimes_{\mathcal C} \mathcal D(F(C),F(-)) \\ &\cong M \otimes_{\mathcal C} \mathcal C(C,-)\\ & \cong M(C), \end{align*} where I used fully faithfulness and for the last step the "co-Yoneda lemma" or "density theorem".

Now, your main claim that a quasi-fully faithful dg-functor induces something which is again quasi-fully faithful between the dg-category of dg-modules is a little more nuanced. Working with this "cohomological" assumptions, you'll want derived stuff all around. So, instead of working with dg-modules, it's best to work with h-projective/cofibrant/semi-free dg-modules, namely your favourite version of the "derived dg-category". Let us work with the dg-category of h-projective dg-modules, just to fix ideas. What is true is that if you have a quasi-equivalence $G \colon \mathcal A \to \mathcal B$ between dg-categories, then $\operatorname{Ind}_G$ will induce a quasi-equivalence $$ \operatorname{Ind}_G \colon \operatorname{h-proj}(\mathcal A) \to \operatorname{h-proj}(\mathcal B). $$ This is quite a well-known fact: if you want to check it yourself, you can work instead with semi-free dg-modules and prove that the unit and counit maps are quasi-isomorphism "by hand", or you can use the fact that you have compact generators: a reference is Drinfeld's famous article (Remark 4.3).

Now, that was the hard part. If you have just a quasi-fully faithful dg-functor $F \colon \mathcal C \to \mathcal D$ as in your post (you don't even need injectivity on objects, just that you have quasi-isomorphisms on hom-complexes), then clearly you can factor it as a quasi-equivalence and a (strictly) fully faithful dg-functor. So, in the end, I think you'll end up with a quasi-fully faithful dg-functor $\operatorname{h-proj}(\mathcal C) \to \operatorname{h-proj}(\mathcal D)$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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