3
$\begingroup$

EDIT: The original title was: If a morphism of diagrams of complexes is composed of quasi-isomorphisms, is the induced arrow a quasi-isomorphism?

Let $J$ be a small category and $C$ be the category of complexes over an abelian category. If $F,G:J\to C$ are functors and $\tau:F\Rightarrow G$ is a natural transformation such that $\tau_i:Fi\to Gi$ is a quasi-isomorphism for every $i\in J$, is $\varinjlim \tau: \varinjlim F \to \varinjlim G$ a quasi-isomorphism?

What about the case of pushouts?

This question arises to me as I try to check that the category of complexes with the subcategory of quasi-isomorphisms is a "category of weak equivalences" in the sense of Waldhausen (well, not exactly, as I'm dealing with complexes of some special exact categories, but for the present case it doesn't matter).

EDIT: The general case has been answered by Zhen Lin below, but apparently the case when one of the arrows of a pushout is monic, the result should be true. After some hours now of trying, I'm not able to do it. I've tried several things using long exact sequences of homology and also Mayer-Vietoris for complexes, but I just can't conclude. Any help would be appreciated!

$\endgroup$
  • $\begingroup$ No in general (including pushouts). If that were the case then we wouldn't need to think about homotopy colimits. $\endgroup$ – Zhen Lin Mar 14 '14 at 16:46
  • $\begingroup$ @Zhen: What if one of the arrows is a monomorphism (in the case of the pushout)? $\endgroup$ – Bruno Stonek Mar 14 '14 at 16:47
  • $\begingroup$ Apparently yes, but I do not know the details. $\endgroup$ – Zhen Lin Mar 14 '14 at 16:51
  • $\begingroup$ @ZhenLin: Thanks. I'll roll up my sleeves and diagram-chase in that particular case then, I guess. I'm intrigued about your comment about homotopy colimits though, if you would like to expand on it (maybe as an answer?) it would be great :) $\endgroup$ – Bruno Stonek Mar 14 '14 at 16:53
  • $\begingroup$ It's not totally clear what you mean by "with one arrow monic." In ZhenLin's example, it seems to me that the map $$\mathbb{Z} \oplus \mathbb{Z} \to D$ is a monomorphism of complexes... Do you want one arrow monic in both pushout diagrams? $\endgroup$ – chthonian Mar 17 '14 at 4:55
1
$\begingroup$

In general the answer is no, even for pushouts. The whole point of homotopy colimits is to fix this deficiency of ordinary colimits.

For example, let $D$ be the two-term chain complex with $D_0 = \mathbb{Z} \oplus \mathbb{Z}$, $D_1 = \mathbb{Z}$ and differential given by $1 \mapsto (1, -1)$. Consider the following pushout diagram, $$\begin{array}{ccc} \mathbb{Z} \oplus \mathbb{Z} & \rightarrow & \mathbb{Z} \\ \downarrow & & \downarrow \\ D & \rightarrow & S \end{array}$$ where $\mathbb{Z} \oplus \mathbb{Z} \to D$ is given by $\mathrm{id}$ in degree 0 and $\mathbb{Z} \oplus \mathbb{Z} \to \mathbb{Z}$ is given by $(x, y) \mapsto x + y$. Then $S$ is a two-term chain complex with $S_0 = S_1 = \mathbb{Z}$ and zero differential. On the other hand, the morphism $D \to \mathbb{Z}$ given by $(x, y) \mapsto x + y$ in degree 0 is a quasi-isomorphism, and the corresponding pushout diagram is $$\begin{array}{ccc} \mathbb{Z} \oplus \mathbb{Z} & \rightarrow & \mathbb{Z} \\ \downarrow & & \downarrow \\ \mathbb{Z} & \rightarrow & \mathbb{Z} \end{array}$$ but $S$ is not quasi-isomorphic to $\mathbb{Z}$.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thank you. I will read your answer with detail later. I was just aware of "homotopy colimits" in the category of topological spaces, not in any more general sense. $\endgroup$ – Bruno Stonek Mar 14 '14 at 17:12
0
$\begingroup$

The diagram chase I had tried missed only one ingredient: the fact the cokernels of a pair of parallel arrows in a pushout are compatibly isomorphic. I didn't know this fact, actually. It's straightforward to prove using the characterization of a pushout by an appropriate right exact sequence.

Here's the proof, which I wrote up in $\LaTeX$:

enter image description here

enter image description here enter image description here

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Yeah, I missed a period at the end of a sentence. Sorry about that :P $\endgroup$ – Bruno Stonek Apr 5 '14 at 9:46

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.