Mathematics · Algebra and Number Theory
Cyclic Purity Versus Purity in Excellent Noetherian Rings
See on the mapAbstract
A characterization is given of those Noetherian rings R such that whenever R is ideally closed (= cyclically pure) in an extension algebra S, then R is pure in S. In fact, R has this property if and only if the completion (A, m) of each local ring of R at a maximal ideal has the following two equivalent properties:(i) For each integer N > 0 there is an m-primary irreducible ideal IN C mN.(ii) Either dim^ = 0 and A is Gorenstein or else depth A > 1 and there is no P e Ass(A) such that diia(A/P) = 1 and (A/P) (A/P) is embeddable in A.It is then shown that if R is a locally excellent Noetherian ring such that either R is reduced (or, more generally, such that R is genetically Gorenstein), or such that Ass() contains no primes of coheight < 1 in a maximal ideal, and R is ideally closed in S, then R is pure in S. Matlis duality and the theory of canonical modules are utilized.Module-theoretic analogues of condition (i) above are, of necessity, also analyzed.Numerous related questions are studied.In the non-Noetherian case, an example is given of a ring extension R - S such that R is pure in 5 but R'[T]] is not even cyclically pure in S[[T]].0. Introduction.All rings are commutative, with identity, and modules are unital.Recall that a submodule N of a module M (or the map N -* M) over a ring R is called pure (respectively, cyclically pure) if for every i?-module (respectively, every cyclic i?-module) E the map N E -> M E is injective.A case of particular interest is the one where N = R and E = S is an extension algebra of R: in this case the cyclic purity of R in 5 asserts that for every ideal I of R, IS D R = I (let E = R/I in the definition of cyclic purity).This condition is sometimes expressed by the phrase "i? is ideally closed in 5".Quite generally, if M/N is finitely presented, then N is pure in M if and only if N is a direct summand of M. We refer the reader to [3], [14], [15], [20, p. 64], and [24] for basic facts about purity.Our original objective was to give a really useful condition on a Noetherian