On the non-existence of right almost split maps (Q2407537)

M. Auslander proved that a finitely presented module \(M\) over a ring \(R\) is the target of a right almost split map if and only if \(\mathrm{End}_R(M)\) is local. It is well known that the latter condition is necessary for the existence of a right almost split map. In the paper under review, it is shown that finitely presentedness is necessary, too, answering Auslander's question for a precise description of right almost split maps in terms of their target. For an infinite cardinal \(\kappa\), let \(\mathcal{S}_\kappa\) be the class of \(R\)-modules which are finitely \(\kappa\)-separable, that is, a directed union of \(<\kappa\)-presented direct summands. For an epimorphism \(f: B\to C\) where \(C\) is \(\kappa\)-presented, it is proved that there is a module \(L\in\mathcal{S}_\kappa\) such that \(f\) splits whenever each homomorphism \(L\to C\) factors through \(f\). As a consequence, it is shown that \(f\) cannot be right almost split unless \(C\) is \(<\kappa\)-presented. This leads to the main theorem that \(f\) is right almost split if and only if \(C\) is local and finitely presented. The theorem can be seen, more generally, in the context of Auslander's morphisms determined by an object. The author conjectures that the coimage of a right \(C\)-determined morphism with respect to a \(\kappa\)-presented module \(C\) splits whenever it is \(\kappa\)-pure epic. The author has verified this for \(\kappa=\aleph_1\). The general conjecture implies that the domain of a left almost split map is pure injective. Reviewer's remarks: Note that \textit{W. Crawley-Boevey} [Lond. Math. Soc. Lect. Note Ser. 168, 127--184 (1992; Zbl 0805.16028), Theorem~2.3] has shown that an indecomposable pure injective \(R\)-module \(M\) is the source of a left almost split map if and only if \(M\cong\mathrm{End}_E(X,E(S))\) with \(X\) finitely presented, \(E=\mathrm{End}_R(X)\), and a simple \(E\)-module \(S\).