How to construct a `concrete' superdecomposable pure-injective module over a string algebra. (Q2464511)

It had been conjectured by the reviewer, among others, that wild finite-dimensional algebras might be characterised by their having pure-injective (= algebraically compact) modules which are superdecomposable, that is, which have no indecomposable direct summand. This was shown to be false by \textit{G. Puninski} [Proc. Am. Math. Soc. 132, No. 7, 1891-1898 (2004; Zbl 1133.16015)], who proved existence of such modules over non-domestic string algebras, at least over any countable such ring; more precisely, he proved that a certain dimension (``width'') which was defined by \textit{M. Ziegler} [in Ann. Pure Appl. Logic 26, 149-213 (1984; Zbl 0593.16019)], is undefined over these rings. From this, a theorem from Ziegler [op. cit.] gives, under the hypothesis that the ring is countable, existence of a superdecomposable pure-injective. The latter proof, however, is just an existence proof; in this paper Puninski shows how to produce an element whose hull (in the sense of Ziegler [op. cit.] and \textit{M. Prest} [Model theory and modules, Lond. Math. Soc. Lect. Note Ser. 130, Cambridge: Cambridge University Press (1988; Zbl 0634.03025)]) is superdecomposable. The key result, one which has already found applications over other types of ring, is that if \(R\) is any ring and if there is, in the lattice of pp formulas, over that ring, an interval which is freely generated by two chains, each of which contains a densely ordered subchain, then there is a superdecomposable pure-injective module over that ring. Puninski then takes a particular example of a non-domestic string algebra, finds explicitly such an interval and such chains, and uses his proof of the result to construct an element, with superdecomposable hull, in a product of finite-dimensional string modules (and here the cardinality of the underlying field plays no role). Although this gives a method which is very generally applicable, it does not give a general theorem, so the existence of superdecomposable pure-injectives over arbitrary uncountable string algebras remains (annoyingly) open. Puninski ends by asking more detailed questions about the structure of these superdecomposable modules, speculating that there might yet be a connection between the representation type of an algebra and the structure of these modules.