scientific article; zbMATH DE number 3904721

From MaRDI portal
Publication:3682585





[For the entire collection see Zbl 0509.00009.] Let W be a Dedekind domain with quotient field Q, F a finite rank torsionfree W-module and let QG denote the divisible hull of G. Finite rank torsionfree W-modules G and H are said to be quasi-equal if \(QG=QH\) and are quasi-isomorphic if G and H are isomorphic to quasi- equal submodules of some divisible W-module. The largest divisible submodule of G is denoted d(G), and G is said to be reduced if \(d(G)=0\). If p is a nonzero prime ideal of W then p-rank G is defined as length(G/pG). Let I be a W-algebra such that p-rank I\(=1\) for each prime ideal p of W. Then G is said to be I-split if \(I_{^*W}G=I\otimes_ WG/d(I\otimes_ WG)\) is a finitely generated I-module. This happens for instance if I is complete in the W-topology. The object of this paper is give a unified treatment of the I-split W-modules. The starting point is a theorem of Kurosh [\textit{L. Fuchs}, Infinite abelian groups. II (1973; Zbl 0257.20035)], which states that if W is a discrete rank 1 valuation ring with completion I and \(\bar G\) is the completion of G then G can be recovered from \(\bar G\) and QG as \(\vartheta^{-1}(\bar G)\) where \(\vartheta\) : QG\(\to Q\bar G\) is the canonical map. A splitting ring for W is a reduced commutative W-algebra I such that p-rank I\(=1\) for each prime ideal p of W and \(I\cong W_ 1\times...\times W_ t\), where the \(W_ i\) are Dedekind domains such that \(QW_ i\) is a field. Let \(\Gamma\) denote the category of I-split W-modules and -homomorphisms, and let \(Q\Gamma\) denote the category of I-split W-modules and quasi- homomorphisms. The first section contains several basic results on these categories. For example the pure injectives in these categories are characterized and some criteria are given for torsionfree W-modules to be I-split. - In the second section the Arnold duality functor is studied. This is an exact contravariant functor \(A: Q\Gamma\) \(\to Q\Gamma\). Although its definition is somewhat technical, it is independent of I in a certain natural sense, and has the property that there is a natural isomorphism in \(Q\Gamma\) from \(QG\otimes_ WH\) to \(Hom_ W(AG,QH)\) which takes \(d(G\otimes_ WH)\) onto \(QHom_ W(AG,H)\) for any torsionfree W-module H and finite rank I-split W-module G. - In the third section it is shown that if I has finite rank, then \(\Gamma\) is naturally isomorphic to the dual of a full subcategory of the category of finitely generated modules over the hereditary Noetherian ring \(\Lambda =End_ W(Q\oplus I)\). - In section four the author derives, under the assumption that I has finite rank, many of the properties of another contravariant functor. This functor F takes \(\Gamma\) to itself and is defined by \(F=Hom_ W(-,C^+Q)\) where \(C^+Q\) is a certain maximal pure submodule of I. It follows that if G is I-split then FG has no pure injective submodules, that the natural homomorphism \(G\to F^ 2G\) is a split epimorphism, and if G has no pure injective submodules then F takes an exact sequence \(0\to H\to X\to G\to 0\) into an exact sequence. - In section five the author uses the functors A and F to give a simplified and self-contained treatment of the special case of the Coxeter functors needed for abelian groups. In fact these functors \(C^+\) and \(C^-\) can be defined as AF and FA respectively. - In section six the author studies those W-modules which admit the structure of an integral domain. It turns out that every finite rank domain is I-split for some I, and conversely, every I-split module (for rank \(I\leq \infty)\) is a pure submodule and a homomorphic image of a finite direct sum of domains. The author also considers the question of how many I-split domains there are which are indecomposable in \(Q\Gamma\), and how to find them.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3682585)