Let $\mathcal{A}$ be an abelian category and consider the following categories:

  • $\mathbf{Ch} (\mathcal{A})$, the category of cochain complexes in $\mathcal{A}$.
  • The full subcategories $\mathbf{Ch}^\mathrm{b} (\mathcal{A})$ (cochains bounded), $\mathbf{Ch}^- (\mathcal{A})$ (cochains bounded above), and $\mathbf{Ch}^+ (\mathcal{A})$ (cochains bounded below).
  • The derived category $\mathbf{D} (\mathcal{A})$, i.e. the category of cochain complexes in $\mathcal{A}$ localised with respect to quasi-isomorphisms.
  • The full subcategories $\mathbf{D}^\mathrm{b} (\mathcal{A})$ (cohomology bounded), $\mathbf{D}^- (\mathcal{A})$ (cohomology bounded above), and $\mathbf{D}^+ (\mathcal{A})$ (cohomology bounded below).
  • $\mathbf{Q}^{-1} \mathbf{Ch}^\mathrm{b} (\mathcal{A})$, the category of bounded cochain complexes localised with respect to quasi-isomorphisms, and analogously $\mathbf{Q}^{-1} \mathbf{Ch}^- (\mathcal{A})$ and $\mathbf{Q}^{-1} \mathbf{Ch}^+ (\mathcal{A})$.

By the universal property of localisation, the inclusion $\mathbf{Ch}^\mathrm{b} (\mathcal{A}) \hookrightarrow \mathbf{Ch} (\mathcal{A})$ induces a functor $\mathbf{Q}^{-1} \mathbf{Ch}^\mathrm{b} (\mathcal{A}) \to \mathbf{D} (\mathcal{A})$; similarly we have functors $\mathbf{Q}^{-1} \mathbf{Ch}^- (\mathcal{A}) \to \mathbf{D} (\mathcal{A})$ and $\mathbf{Q}^{-1} \mathbf{Ch}^+ (\mathcal{A}) \to \mathbf{D} (\mathcal{A})$. It is easy to see that a cochain complex whose cohomology is bounded (resp. bounded above, bounded below) is quasi-isomorphic to a cochain complex that is itself bounded (resp. bounded above, bounded below). Thus, these functors factor as essentially surjective functors $\mathbf{Q}^{-1} \mathbf{Ch}^\mathrm{b} (\mathcal{A}) \to \mathbf{D}^\mathrm{b} (\mathcal{A})$, $\mathbf{Q}^{-1} \mathbf{Ch}^- (\mathcal{A}) \to \mathbf{D}^- (\mathcal{A})$ and $\mathbf{Q}^{-1} \mathbf{Ch}^+ (\mathcal{A}) \to \mathbf{D}^+ (\mathcal{A})$.

Question. When are these functors full and/or faithful?

My impression is that the correct definition of bounded derived category is the one denoted by $\mathbf{D}^\mathrm{b} (\mathcal{A})$ above, but it did not occur to me until just now that this might be different from $\mathbf{Q}^{-1} \mathbf{Ch}^\mathrm{b} (\mathcal{A})$. I imagine that if $\mathcal{A}$ has enough projective objects then $\mathbf{Q}^{-1} \mathbf{Ch}^- (\mathcal{A}) \to \mathbf{D}^- (\mathcal{A})$ is an equivalence of categories, and dually if $\mathcal{A}$ has enough injective objects then $\mathbf{Q}^{-1} \mathbf{Ch}^+ (\mathcal{A}) \to \mathbf{D}^+ (\mathcal{A})$ is an equivalence of categories – it should be one of those standard arguments about reducing zigzags using resolutions – but I have not checked, and even if this is correct, it does not answer the question about $\mathbf{D}^\mathrm{b} (\mathcal{A})$. It also leaves open the possibility that there is some impoverished $\mathcal{A}$ where these categories are genuinely different.


They are always fully faithful, and this doesn't require enough projectives or injectives.

A map in $\mathbf{D}(\mathcal{A})$ from $X$ to $Y$ is represented by a diagram $X\stackrel{s}{\leftarrow}Z\stackrel{f}{\to}Y$ where $s$ is a quasi-isomorphism (call this map $fs^{-1}$), and if $t:Z'\to Z$ is a quasi-isomorphism then $fs^{-1}=(ft)(st)^{-1}$.

But if $X$ (and hence $Z$) has cohomology bounded above, then there is always a quasi-isomorphism $Z'\to Z$ from a bounded above complex: for sufficiently large $n$, let $Z'$ be the "good" truncation $$\tau_{\leq n}Z:=\dots\to Z^{n-2}\to Z^{n-1}\to\ker(d^n)\to0\to0\to\dots$$ of $Z$, and let $t$ be the inclusion map.

So both ways of defining $\mathbf{D}^-(\mathcal{A})$ are "the same".

A dual argument works for $\mathbf{D}^+(\mathcal{A})$.

For $\mathbf{D}^b(\mathcal{A})$ use the first argument followed by the second.

  • $\begingroup$ Ah yes, it's coming back to me now! Prima facie this argument only shows the functor is full, but this argument plus the calculus of fractions shows faithfulness as well. Very good. $\endgroup$ – Zhen Lin Jul 30 '20 at 4:30

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.