Resolutions of almost complete intersections (Q1183280): Difference between revisions

In a local (respectively \(\mathbb{N}\)-graded ring) \(R\) let \(I\) denote a (homogeneous) ideal such that (i) \(I=(a_1,\ldots,a_n)R\) and \(\{a_1,\ldots,a_{n-1}\}\) is a regular sequence, (ii) \(I\) is not a complete intersection, and (iii) there is an element \(b\in R\backslash(a_1,\ldots,a_{n-1})R\) with \((a_1,\ldots,a_{n-1})R:a_n=(a_ 1,\ldots,a_{n-1},b)R\) and \((a_1,\ldots,a_{n-1})R : b=(a_1,\ldots,a_{n-1},a_ n)R\). [Note that if \(R\) is a Gorenstein ring and the Koszul homology \(H_1(I;R)\) is a free \(R/I\)-module (e.g. if \(R/I\) is a Gorenstein ring), then (i) implies condition (iii); see \textit{C. Peskine} and \textit{L. Szpiro}, Invent. Math. 26, 271--302 (1974; Zbl 0298.14022).] Let \[ F_*: \ldots\to F_i\xrightarrow{d_i} F_{i-1}\xrightarrow{d_{i-1}} \ldots\xrightarrow{d_2} F_1\xrightarrow{d_1} F_0\rightarrow 0 \] denote the minimal resolution of \(R/I\) over \(R\). Then it is shown that (i) \(\operatorname{rk} F_i=2^{n-1}\) for all \(i\geq n\), (ii) the maps \(\{d_i\}\) can be chosen so that \(d_{i+2}=d_i\) for all \(i>n\). This result follows from a generic base by base change and an explicit description of the resolution \(F_*\). As an application it turns out that if \(\mathrm{proj }\dim_ RR/I<\infty\) and \(R/I\) is a Gorenstein ring, then \(I\) is a complete intersection, a results shown by \textit{E. Kunz} [J. Algebra 28, 111--115 (1974; Zbl 0275.13025), respectively by \textit{D. A. Buchsbaum} and \textit{D. Eisenbud} in Algebra Commut., Geometria, Convegni 1971/72, Symp. Math. 11, 193--204 (1973; Zbl 0294.13009), using the following theorem of T. Gulliksen [see \textit{T. H. Gulliksen} and \textit{G. Levin}, ``Homology of local rings'', Queen's Pap. Pure Appl. Math. 20. Kingston, Ontario: Queen's University (1969; Zbl 0208.30304)]: An ideal \(I\) is generated by a regular sequence provided \(\mathrm{proj }\dim_ RR/I<\infty\) and \(H_1(I;R)\) is a free \(R/I\)-module. So the authors' approach yields a direct proof (i.e., without Tate resolution) of the result of E. Kunz above.