Homotopical algebra is not concrete (Q1793967)

A category \(\mathcal{D}\) is \textit{concrete} if it admits a faithful functor to the category of sets; if this is the case, the morphisms of \(\mathcal{D}\) can be interpreted as maps of sets. If \(\mathcal{D}\) is small, or accessible, or monadic over the category of sets, then \(\mathcal{D}\) is concrete. The homotopy category of topological spaces is known \textit{not} to be concrete. The argument given by [\textit{P. Freyd}, in: Sympos. Math., Roma 4, Teoria Numeri, Dic. 1968, e Algebra, Marzo 1969, 431--456 (1970; Zbl 0248.18008)] is outlined in \S3 of the paper under review. The three main ingredients are: a certain collection of abelian groups with distinguished elements satisfying a certain annihilation condition for homomorphisms between them, Moore spaces, and (singular) homology. The authors prove an appealing abstract version of this result (Theorem 4.8): If \(\mathcal{M}\) is a pointed model category that admits a ``weak classifying object of type \(n\)'' for some \(n \geq 2\), then the homotopy category of \(\mathcal{M}\) is not concrete. To explain the hypothesis, recall that the homotopy category of a pointed model category admits an adjoint pair of endofunctors, suspension \(\Sigma\) and looping \(\Omega\). The set of morphisms \(A \rightarrow \Omega^n X\) in the homotopy category is an abelian group (for \(n \geq 2\)), denoted \(\pi_n^A(X)\) and called the \(n\)th homotopy group of \(X\) with coefficients in \(A\). A weak classifying object of type \(n\) is an abstract version of Eilenberg-MacLane spaces; it is a functor \(k\) from the category of groups to the homotopy category of \(\mathcal{M}\) such that the composition \(\pi_n^A \circ k\) is \textit{full}, and is related to the identity functor by a natural transformation that is either a pointwise epimorphism or a pointwise monomorphism. The proof of the main theorem is somewhat dual to the proof given by Freyd, in the sense that instead of cofibration sequences and suspension one considers fibration sequences and looping. There is a variant of the result for ``quasi-stable'' pointed model categories (Theorem 4.11). The authors use their result to give short proofs that the homotopy category of chain complexes, the homotopy category of Bousfield-Friedlander spectra, and the local homotopy category of simplicial presheaves on a site are not concrete. At the end of the paper, the authors explain why the homotopy category of 1-types is not concrete either. This needs an extra argument due to the lack of model structures on the category of 1-types (which is not finite cocomplete). As a consequence, one obtains a conceptual proof of the fact that the homotopy categories of the category of groupoids and the category of categories (with categorical equivalences as weak equivalences) are not concrete.