2

$\newcommand{\A}{\mathcal{A}}\newcommand{\Ch}{\mathbf{Ch}}$The following is an exercise from Weibel's An Introduction to Homological Algebra.

Show that if $\A$ has enough projectives, then so does the category $\Ch(\A)$ of chain complexes over $\A$.

The hint given is to use the previous exercise, where one characterises projective objects in $\Ch(\A)$. Namely, we saw that $P_{\bullet}$ is projective in $\Ch(\A)$ iff it is a split exact complex of projectives.


My attempt.
Let $M_{\bullet}$ be an arbitrary chain complex in $\Ch(\A)$. Then, since $\A$ has enough projectives, we can find epics $$P_{n} \to M_{n}$$ for all $n$. Moreover, using projectivity and the differentials of $M_{\bullet}$, we get maps $$d_{n} : P_{n} \to P_{n - 1}$$ such that $d^2= 0$ and they fit together to give a chain map $$P_{\bullet} \to M_{\bullet}.$$

The issue now is showing that the $P_{\bullet}$ is projective. (The above map being epic is clear since it is epic in each degree.)
In fact, I am quite sure that the construction above will actually not work in general. For example, if each $M_{n}$ was projective to begin with, then we could've picked each $P_{n}$ to be $M_{n}$ and the maps to be the identity maps. But an arbitrary complex of projectives need not be split exact.

This makes me think that one needs to do a different construction, but I don't see one.

hello
  • 396

1 Answers1

2

Proceed as follows:

  1. Components. Pick epic maps $Q_n \longrightarrow H_n$ and $P_n'' \longrightarrow B_n$ from projectives that then, by the proof of the Horseshoe lemma fit into a SEC $ 0\longrightarrow P_n'' \longrightarrow P_n' \longrightarrow Q_n \longrightarrow 0$ covering $0 \longrightarrow B_n \longrightarrow Z_n \longrightarrow H_n \longrightarrow 0$. This gives surjections $P_n' \to Z_n$.

  2. More components. There now exists another SEC $ 0\longrightarrow P_n' \longrightarrow P_n \longrightarrow P_{n-1}'' \longrightarrow 0$ covering the SEC$0 \longrightarrow Z_n \longrightarrow C_n \longrightarrow B_{n-1} \longrightarrow 0$.

  3. Differential. Note that $P_n = P_n'\oplus P_{n-1}''$ maps into $P_{n-1} = P_{n-1}'\oplus P_{n-2}''$ via the matrix $$d = \begin{bmatrix} 0 & 1 \\ 0 & 0 \end{bmatrix}$$ where $1$ is the inclusion $P''\subseteq P'$ in (1), and that $d^2=0$. This makes $P$ into a complex $(P,d)$ of projectives.

  4. Split. To see it is split, note that $P_n' = P_n''\oplus Q_n$ (another copy of $P'')$ and $P_{n+1} = (P_{n+1}''\oplus Q_{n+1})\oplus P_n''$ so you can project and include into this copy. In short: $$d(x,y,z) = (z,0,0), \quad s(x,y,z) = (0,0,x)$$ and then it is clear that $dsd=d$, so $(P,d)$ is split.

Pedro
  • 125,149
  • 19
  • 236
  • 403
  • I just realised the following: Your construction has $P_{\bullet}$ split. But for $P_{\bullet}$ to be projective in $\mathbf{Ch}(\mathcal{A})$, it needs to be split exact, right? So then we would require for $Q_{n}$ to be $0$? – hello Sep 15 '21 at 18:31
  • @hello See https://math.stackexchange.com/questions/638331/definition-of-exact-split-complex – Pedro Sep 16 '21 at 19:12
  • How exactly does that apply here? We don't have $sd + ds = 1$ here unless $Q_n = 0$, right? – hello Sep 16 '21 at 21:41
  • I am un-accepting the answer for the moment, hoping that it would "bump" it up the system somewhere and hopefully someone could help fix the small gap. – hello Sep 20 '21 at 14:07
  • @hello You can iterate the construction to make the resulting complex acyclic. Since Q covers H, it cannot be zero unless C is already acyclic. – Pedro Sep 20 '21 at 17:24