Computing syzygies over \(V [X_1, \ldots, X_k]\), \(V\) a valuation domain (Q2254141)

A ring is called \textit{coherent} if every finitely generated ideal has a finitely generated syzygy module. Suppose \(V\) is a valuation domain, meaning that it is a domain whose elements are totally ordered by divisibility. Then \(V[x_1,\ldots,x_k]\) is coherent according to Theorem 7.3.3 in [\textit{S. Glaz}, Commutative Coherent Rings, Lecture Notes in Mathematics. 1371. Berlin etc.: Springer-Verlag (1990; Zbl 0745.13004)]. The proof of this theorem provided in the aforementioned reference is non-constructive and hinges on a difficult result found in [\textit{M. Raynaud} and \textit{L. Gruson}, Invent. Math. 13, 1--89 (1971; Zbl 0227.14010)]. The purpose of this paper is to provide the first fully general algorithm for computing a finite set of generators for the syzygy module of a finitely generated ideal \(I\subset V[x_1,\ldots,x_k]\), where \(V\) is a valuation domain. This provides the first constructive proof that \(V[x_1,\ldots,x_n]\) is coherent. Previously known algorithms depend on \(V\) being either Noetherian or of Krull dimension one. The key algorithm of the paper is one for computing the \(V\)-saturation of a finitely generated submodule of the free module \(V[x_1,\ldots,x_k]^n\), where \(n\geq 1\) is an integer. Let \(V\) be a valuation domain with unique maximal ideal \(\text{rad}(V)\). In Section 1, the authors describe an echelon form for vectors in \(V[x_1,\ldots,x_k]^n\) and provide a Gaussian algorithm (Algorithm 9) for obtaining the \(V\)-saturation of a finitely generated \(V\)-submodule of \(V[x_1,\ldots,x_k]^n\). At the beginning of Section 2, the authors give a compact criterion for a finitely generated \(V\)-submodule of \(V[x_1,\ldots,x_k]^n\) to be \(V\)-saturated (Theorem 12). Namely, if \(L\) is a finite list of vectors generating a \(V\)-submodule of \(V[x_1,\ldots,x_k]^n\), then this \(V\)-submodule is saturated iff \(\dim_K L=\dim_k \bar{L}\), where \(K\) is the quotient field of \(V\), \(k\) the residue field of \(V\), and \(\bar{L}\) the image of \(L\) modulo \(\text{rad(V)}[x_1,\ldots,x_k]\). In the remainder of Section 2, the authors lift the saturation algorithm of Section 1 to their main algorithm (Algorithm 18) for computing a finite set of generators for the \(V\)-saturation of a finitely generated \(V[x_1,\ldots,x_k]\)-submodule \(M\) of \(V[x_1,\ldots,x_k]^n\). This algorithm consists of repeated applications of Algorithm 9 from Section 1 to an appropriate filtration of \(M\). To show that this algorithm terminates, the authors use the `saturation defect series' of \(M\), which builds on Theorem 12. This is found by taking an exhaustive filtration \(\{L_i\}\) of \(M\) with each \(L_i\) a finitely generated \(V\)-submodule of \(M\). The difference between \(\dim_K L_i\) and \(\dim_k \bar{L}_i\) gives the \(i^{th}\) coefficient of the saturation defect series. If this series is identically zero, then \(M\) is \(V\)-saturated (Theorem 16). In the final section, the authors apply the algorithm of Section 2 to Prüfer domains, giving an algorithmic proof (Theorem 25) that \(R[x_1,\ldots,x_k]\) is coherent when \(R\) is a Prüfer domain.