Mapping resolutions of length three. I (Q2104878)

This paper constructs some finite free resolutions of length three of the form \[ 0\to R^{b_3}\to R^{b_2}\to R^{b_1}\to R^{b_0}, \] with Betti numbers \((b_3,b_2,b_1,b_0)\) equal to \((1, n, n, 1)\) (for n even) and \((n-3,n,4,1)\). The construction is based on the construction of the generic ring \(\hat R_{\mathrm{gen}}\) given in [\textit{J. Weyman}, Ann. Inst. Fourier 68, No. 3, 1241--1296 (2018; Zbl 1498.13042)]. A generic ring \(\hat R_{\mathrm{gen}}\) and a generic complex \(\mathbb F_{\bullet}^{\mathrm{gen}}\) is constructed for each collection of Betti numbers \((b_3,b_2,b_1,b_0)\) in [loc. cit.]. Such ring and complex have a universal property, in the sense that any free resolution of that format in a given commutative ring can be obtained by the generic one by homomorphic image. This construction requires powerful tools from representation theory. A Kac-Moody Lie algebra is canonically associated to each format, and it turns out that the generic ring is Noetherian if and only if this Lie algebra is finite dimensional. This last fact identifies a special family of formats (called Dynkin formats) for which this condition holds. Ideals and modules whose minimal free resolutions are of Dynkin format are similarly called Dynkin. Dynkin ideals include Gorenstein ideals, almost complete intersections and few other formats. The first application of the complexes described above concerns the open sets \(U_{\mathrm{CM}}\) and \(U_{\mathrm{split}}\) of the spectrum of generic ring \(\hat R_{\mathrm{gen}}\). The subset \(U_{\mathrm{CM}}\) is the set of prime ideals \(P\) for which the localization of the dual complex \((\mathbb F_{\bullet}^{\mathrm{gen}})^*\) at \(P\) is acyclic, i.e. the points where the localization of the generic module \(H_0(\mathbb F_{\bullet}^{\mathrm{gen}})\) is perfect. There is also another open set \(U_{\mathrm{split}}\) which consists of points for which the complex \(\mathbb F^{\mathrm{top}}\) is split exact. Conjecturally these two open sets are equal. The conjecture is proven here for \(D_n\) type formats. (This result had already been established using linkage.) The point is that the argument in the present paper is independent of linkage and is likely to generalize to other Dynkin formats.