Residual intersections and the annihilator of Koszul homologies (Q2629780)

If \(R\) is a Noetherian ring, \(I\) is an ideal of \(R\), and \(s\) is an integer which is at least as large as the height of \(I\), then the ideal \(J\) is an algebraic \(s\)-residual intersection of \(I\) if \(J\) has height at least \(s\) and \(J=(\mathfrak a:_RI)\) for some \(s\)-generated subideal \(\mathfrak a\) of \(I\). In a geometric setting, if \(X\) and \(Y\) are irreducible closed subschemes of a Noetherian scheme \(Z\) with \(\mathrm {codim}_Z (X)\leq \mathrm{codim}_Z(Y) =s\) and \(Y \not \subset X\), then \(Y\) is a residual intersection of \(X\) if the number of equations needed to define \(X \cup Y\) as a subscheme of \(Z\) is as as small as possible, namely, \(s\). The notion of residual intersection is a generalization of linkage (or liaison); it was introduced by \textit{M. Artin} and \textit{M. Nagata} [J. Math. Kyoto Univ. 12, 307--323 (1972; Zbl 0263.14019)] and fine-tuned by \textit{C. Huneke} [Trans. Am. Math. Soc. 277, 739--763 (1983; Zbl 0514.13011)] and \textit{C. Huneke} and \textit{B. Ulrich} [J. Reine Angew. Math. 390, 1--20 (1988; Zbl 0732.13004)]. The notes and references to chapter 9 in [\textit{W. Fulton}, Intersection theory. 2nd ed. Berlin: Springer (1998; Zbl 0885.14002)] give a history of the study of residual intersections from a geometric point of view. When proving theorems about residual intersections, especially in the algebraic setting, it is standard practice to impose a local growth condition, \((G_s)\), on the number of generators of localizations of \(I\). The aim of the present paper is to reprove and extend the old theorems about residual intersections without imposing the local growth condition. The new tool that makes this possible is a family of complexes \(\{_i\mathcal Z^+_\bullet\}_{i=0}^{\infty}\) which approximate the \(i^{\mathrm{th}}\) symmetric power of \(I/\mathfrak a\) for positive \(i\). The complex that is called \(_0\mathcal Z^+_\bullet\) in the present paper had previously been constructed and exploited by \textit{S. H. Hassanzadeh} [Trans. Am. Math. Soc. 364, No. 12, 6371--6394 (2012; Zbl 1277.13008)]. The approximation complexes of \textit{J. Herzog} et al. [Lect. Notes Pure Appl. Math. 84, 79--169 (1983; Zbl 0499.13002)] are an early ancestor of the approximation complexes which are constructed in the present paper. The present paper makes significant progress on five conjectures concerning the Cohen-Macaulayness of \(R/J\), the canonical module of \(R/J\), the minimal number of generators of \(\mathfrak a\), the unmixedness of \(J\), and the assertion that the Hilbert function of \(R/J\) depends only on \(I\) and the degrees of the generators of \(\mathfrak a\). These conjectures are due to Artin-Nagata [loc. cit.] (implicitly), Huneke-Ulrich [loc. cit.] (explicitly), \textit{B. Ulrich} [in: Free resolutions in commutative algebra and algebraic geometry: Sundance 90. Proceedings of a conference, held in May, 1990 in Sundance, UT, USA. . 133--138 (1992; Zbl 0790.13005)], and \textit{M. Chardin} et al. [Compos. Math. 125, No. 2, 193--219 (2001; Zbl 0983.13005)] and \textit{M. Chardin} et al. [Compos. Math. 151, No. 9, 1663--1687 (2015; Zbl 1329.13023)]. One further theme in this paper concerns the uniform annihilator of Koszul homologies. It is shown that in any residual intersection \(J=\mathfrak a:I\), the sliding depth condition passes from \(I\) to \(\mathfrak a\); and as a consequence, it is shown that \[ I\subseteq \bigcap_{1\leq j} \mathrm {Ann} (\mathrm {H}_j(\mathfrak a)), \] where \(\mathrm {H}_j(\mathfrak a)\) is the \(j^{\mathrm{th}}\) Koszul homology with respect to any minimal generating set of \(\mathfrak a\). This result leads to counterexamples to the statement \textit{if \(R\) is a Noetherian ring and \(\mathfrak a\) is an ideal of finite projective dimension \(n\), then \((\mathrm {Ann } \mathrm {H}_1(\mathfrak a))^{n+1}\subseteq \mathfrak a\)}, which appears in the literature.