Transfinite Adams representability (Q5965067)

This long article is an important (both for theory and examples) contribution to the study of \textit{generalized Brown representability theorems} in (well generated) triangulated categories. Let us remind the classical \textit{Brown representability theorem} [\textit{E. Brown}, Ann. Math. (2) 75, 467--484 (1962; Zbl 0101.40603)]: every functor from \(\mathcal Sp^{\mathrm{op}}\) (where \(\mathcal Sp\) denotes the stable homotopy category) to \(\mathbf{Ab}\) which maps exact triangles to long exact sequences and preserves products is representable. A variation of this theorem (\textit{Adams representability theorem} [\textit{J.F. Adams}, Topology 10, 185--198 (1971; Zbl 0197.19604)]) asserts that, if \(\mathcal Sp^c\) denotes the full subcategory of \(\mathcal Sp\) of compact spectra, every functor \((\mathcal Sp^c)^{\mathrm{op}}\to\mathbf{Ab}\) which maps exact triangles to long exact sequences is the restriction to \(\mathcal Sp^c\) of a representable functor; moreover, every natural transformation between such cohomological functors is induced by a map of spectra (which does not need to be unique). The extension of Brown representability theorem to other triangulated categories \(\mathcal{T}\) than \(\mathcal Sp\) is true in fairly general situations - it holds in particular for \textit{well generated} triangulated categories (an assumption which is true in the usual examples). Nevertheless, Adams representability theorem fails for a lot of well generated triangulated categories (a sufficient condition for this to hold is the category \(\mathcal{T}^c\) to be essentially countable) -- an example is the derived category of the polynomial ring \(k[x,y]\), for \(k\) a field of big enough cardinality, as shown in [\textit{J. Christensen} et al., Topology 40, No. 6, 1339--1361 (2001; Zbl 0997.18007)]. In this paper, the authors study some transfinite variations of representability. Let \(\mathcal{T}\) be a well generated triangulated category and \(\alpha\) a regular cardinal. Let us denote by \(\mathcal{T}^\alpha\) the full subcategory of \(\alpha\)-compact objects in \(\mathcal{T}\) (for the definitions, the reader can look for example at [\textit{A. Neeman}, Triangulated categories. Annals of Mathematics Studies 148. Princeton, NJ: Princeton University Press (2001; Zbl 0974.18008)]). The article introduces the following properties: \(\mathrm{ARO}_\alpha\;\) Any cohomological (that is, taking exact triangles to long exact sequences) functor \((\mathcal{T}^\alpha)^{\mathrm{op}}\to\mathbf{Ab}\) that preserves products of less than \(\alpha\) objects is the restriction to \(\mathcal{T}^\alpha\) of a representable functor on \(\mathcal{T}\). \(\mathrm{ARM}_\alpha\;\) Any natural transformation \(\mathcal{T}(-,X)|_{\mathcal{T}^\alpha}\to\mathcal{T}(-,Y)|_{\mathcal{T}^\alpha}\) is induced by a morphism \(X\to Y\) of \(\mathcal{T}\). For \(\alpha=\aleph_0\), these properties are the same as Adams representability. To study properties \(\mathrm{ARO}_\alpha\) and \(\mathrm{ARM}_\alpha\), the authors take the following point of view. If \(\mathcal{C}\) is an essentially small category, let us denote by \(\mathbf{Mod}_\alpha(\mathcal{C})\) the category of functors \(\mathcal{C}^{\mathrm{op}}\to\mathbf{Ab}\) which preserve coproducts of less than \(\alpha\) objects. It is a nice abelian category; here the main case of interest is \(\mathcal{C}=\mathcal{T}^\alpha\). The restriction to this category of the representable functors on \(\mathcal{T}\) gives a functor \(S_\alpha : \mathcal{T}\to\mathbf{Mod}_\alpha(\mathcal{T}^\alpha)\); \(\mathrm{ARO}_\alpha\) means that the essential image of \(S_\alpha\) consists exactly of cohomological functors and \(\mathrm{ARM}_\alpha\) that \(S_\alpha\) is a full functor. One of the main theoretic results of the paper, which rely on a suitable obstruction theory, which is developed in its second part, is the following. Here, pd denotes the projective dimension of an object of an abelian category. Theorem. Corollary 2.4 and Proposition 1.2{\parindent=0.6cm\begin{itemize}\item[--] If \(F\) is an object of \(\mathbf{Mod}_\alpha(\mathcal{T}^\alpha)\) such that \(\mathrm{pd}(F)\leq 2\), then \(F\) lies in the essential image of \(S_\alpha\). \item[--] If \(X\) is an object of \(\mathcal{T}\) such that \(\mathrm{pd}(S_\alpha(X))\leq 1\), then any morphism \(S_\alpha(X)\to S_\alpha(Y)\) of \(\mathbf{Mod}_\alpha(\mathcal{T}^\alpha)\) is induced by a morphism \(X\to Y\) of \(\mathcal{T}\). \item[--] If the functor \(S_\alpha\) is full, then \(\mathrm{pd}(S_\alpha(X))\leq 1\) for each object \(X\) of \(\mathcal{T}\). \end{itemize}} The authors study the case of the derived category \(D(R)\) of an \textit{\(\alpha\)-coherent} ring \(R\). For the definition of this notion, and the one of the \textit{\(\alpha\)-pure global dimension} of \(R\), denoted by \(\text{pgd}_\alpha(R)\), we refer to the paper or to [\textit{C. Jensen} and \textit{H. Lenzing}, Model theoretic algebra: with particular emphasis on fields, rings, modules. Algebra, Logic and Applications, 2. New York etc.: Gordon and Breach Science Publishers (1989; Zbl 0728.03026)]. The main result is the following: Theorem 3.3. Assume that \(\alpha>\aleph_0\) and \(R\) is an \(\alpha\)-coherent ring. If \(D(R)\) satisfies \(\mathrm{ARM}_\alpha\), then \(\text{pgd}_\alpha(R)\leq 1\). If moreover \(R\) is hereditary, then {\parindent=0.6cm\begin{itemize}\item[--] the category \(D(R)\) satisfies \(\mathrm{ARM}_\alpha\) if and only if \(\text{pgd}_\alpha(R)\leq 1\); \item[--] the category \(D(R)\) satisfies \(\mathrm{ARO}_\alpha\) if and only if \(\text{pgd}_\alpha(R)\leq 2\). \end{itemize}} The paper gives a significant list of examples and counterexamples deduced from it. After that, it looks (Section 4) at some triangulated categories coming from topology (as \(\mathcal Sp\), but also its motivic analogue) or algebraic geometry (derived categories of sheaves, of projectives or injective modules), showing that, \textit{under the continuum hypothesis}, a lot of them satisfies \(\mathrm{ARO}_{\aleph_1}\). In Section 5, the authors discuss the still open conjecture of Neeman on \textit{Rosický functors} -- see [\textit{A. Neeman}, J. Topol. 2, No. 2, 262--276 (2009; Zbl 1179.18005)].