Algebraic linkage and homological algebra (Q2184196)

A fundamental result in liaison theory [\textit{C. Peskine} and \textit{L. Szpiro}, Invent. Math. 26, 271--302 (1974; Zbl 0298.14022)] and [\textit{J. C. Migliore}, Introduction to liaison theory and deficiency modules. Boston, MA: Birkhäuser (1998; Zbl 0921.14033)] is the mapping cone procedure. It says that if two codimension \(r\) subschemes \(C,D \subset \mathbb P^n\) are linked by a complete intersection \(X\) of hypersurfaces of degrees \(d_1, \dots, d_r\) and \(\phi: K_\bullet \to E_\bullet\) is a map from the Koszul resolution of \({\mathcal I}_X\) to a locally free length \(r-1\) resolution for \({\mathcal I}_C\) induced by the inclusion \({\mathcal I}_X \subset {\mathcal I}_C\), then the mapping cone associated to \(\phi^\vee (n+1-\sum d_i)\) is a locally free length \(r\) resolution for \({\mathcal O}_D\). This result is essential for major liaison theorems in codimension \(r=2\), such as the Lazarsfeld-Rao property for even linkage classes [\textit{M. Martin-Deschamps} and \textit{D. Perrin}, Sur la classification des courbes gauches. Paris: Société Mathématique de France (1990; Zbl 0717.14017)] and [\textit{E. Ballico} et al., Am. J. Math. 113, No. 1, 117--128 (1991; Zbl 0754.14032)] and Rao's correspondence between even linkage classes and stable equivalence classes of vector bundles [\textit{P. Rao}, Math. Ann. 258, 169--173 (1981; Zbl 0493.14009)]. It is also crucial in extending these results to liaison of subschemes that are not locally Cohen-Macaulay [\textit{S. Nollet}, Trans. Am. Math. Soc. 348, No. 3, 1137--1162 (1996; Zbl 0859.14018)]. The authors extend the mapping cone procedure to a statement in homological algebra. Let \(\mathcal A, \mathcal B\) be abelian categories and \(T = (T^i)_{i \geq 0}\) a contravariant \(\partial\)-functor from \(\mathcal A\) to \(\mathcal B\). Suppose that \(n \geq 2\) and objects \(F,G \in \mathcal A\) have length \(n\) resolutions \(F_\bullet, G_\bullet\) in \(\mathcal A\) such that \(T^l (F_i)=0\) and \(T^l (G_i)=0\) for \(0 \leq i < n\) and \(0 < l \leq n-i\) or \(i=-1\) and \(0 \leq l < n\). If \(F \to G\) is a morphism and the induced morphism \(T^n(G) \to T^n (F)\) is injective with cokernel \(K\), then the mapping cone of the morphism \(T^0 (G_\bullet) \to T^0 (F_\bullet)\) of complexes in \(\mathcal B\) is a resolution for \(K\). The result is interpreted in local algebraic and geometric settings. The algebraic consequence is that if \(M\) and \(N\) are modules of projective dimension at most \(n\) over a local ring \(R\) with annihilators of depth \(n\) and \(\phi: M \to N\) is an \(R\)-module map inducing a chain map \(\phi_\bullet: M_\bullet \to N_\bullet\) and an injection \(\text{Ext}_R^n (N, R) \to \text{Ext}_R^n (M, R)\) with cokernel \(K\), then the mapping cone of \(\phi_\bullet^\vee\) is a projective resolution for \(K\). The geometric version says that if \(\phi: \mathcal F \to \mathcal G\) is a morphism of sheaves of codimension \(n\) support on a smooth variety \(X\), \(\mathcal F\), \(\mathcal G\) have homological dimension at most \(n\) and the morphism \({\mathop{\mathcal Ext}}^n(\mathcal G, \omega_X) \to {\mathop{\mathcal Ext}}^n (\mathcal F, \omega_X)\) is injective with cokernel \(\mathcal K\), then the mapping cone of the induced morphism of minimal projective resolutions for \(\mathcal F, \mathcal G\) gives a projective resolution for \(\mathcal K\). Reviewer's Remark: The geometric version is stated with sheaves of \(n\)-dimensional support, but it should be codimension \(n\). The same typo appears in the proof.