A theorem of M. C. R. Butler for Dedekind domains (Q1897804)

The paper deals with an extension of a well-known theorem of Zassenhaus and Butler: Any torsion-free ring of finite rank which is locally free as an algebra over the ring of integers is the endomorphism ring of a locally free torsion-free abelian group of the same rank. A \(K\)-module \(M\) over a domain \(K\) is locally finitely generated (free) if for each prime \(p\) of \(K\) the localization of \(M\) at \(p\) is finitely generated (free). The authors establish an interesting and quite satisfying answer to the following question. Let \(K\) be a subring of an algebraic number field with infinitely many primes. If \(E\) is a torsion-free \(K\)-algebra of finite rank that is locally finitely generated as a \(K\)-module, is there a full \(K\)-locally finitely generated (right) \(E\)-submodule \(M\) of \(QE\) such that \(\text{End}(M)\cong E\) canonically? Say that \(E\) is realizable if the answer is yes. The only \(E\)'s which are not realizable are of very special form and determined in this paper. The proof is based on an analysis of Butler's proof, earlier results of Reid's and an interesting application of ``classical algebra''. Finally the result is applied to provide the existence of \(E\)-modules over (\(E\)-rings) \(K\). Recall that \(M\) is an \(E\)-module over \(K\) if all \(Z\)-homomorphisms from \(K\) to \(M\) are \(K\)-homomorphisms, and the ring \(K\) is an \(E\)-ring if it is an \(E\)-module over \(K\).