Krull dimension and integrality of symmetric algebras (Q1262902): Difference between revisions

In the first part the authors give a new and elegant proof of the dimension formula of Huneke-Rossi for the symmetric algebra S(E) of a finitely-generated module E over a noetherian ring R: Let \(\mu (E_{{\mathfrak p}})\) denote the minimal number of generators of the localization \(E_{{\mathfrak p}}\); then \(\dim(S(E))\) is the supremum of \(\dim(R/{\mathfrak p})+\mu (E_{{\mathfrak p}})\), \({\mathfrak p}\in Spec(R)\). If R is an equi-dimensional and catenarian domain and, furthermore, S(E) is unmixed, then one obtains an even simpler formula: \(\dim (S(E))=\dim (R)+ \text{rank}(E).\) The second part contains a discussion of the following question: When is S(E) an integral domain ? Except for the case proj\(\dim(E)=1\) not much is known, and the results of the authors rather add to the impression that the problem is difficult and a completely satisfactory answer can hardly be expected. A general necessary condition: If R is catenarian and S(E) is a domain, then, by the dimension formulas above, E must satisfy the condition \({\mathcal F}_ 1\): \(\mu (E_{{\mathfrak p}}\leq \text{rank}(E)+ \text{height}({\mathfrak p})-1\) for all \(\mathfrak p\in Spec(R)\), \({\mathfrak p}\neq 0\). Using a free resolution of \(S^ 2(E)\) one can then derive the following negative result: If R is a domain of dimension 4 and E is a reflexive module of projective dimension 2, then S(E) is not a domain. A positive result: If E is torsion-free, satisfies \({\mathcal F}_ 1\), has finite projective dimension, and its first syzygy module has rank 2, then S(E) is a domain. Using their approximation complexes the authors discuss several other cases too technical to be reported here. The articles is illustrated by many concrete examples.