Global effective versions of the Briançon-Skoda-Huneke theorem (Q2346753)

The authors give global effective versions of the Briançon-Skoda-Huneke theorem: Theorem. Let \(V\) be a germ of a reduced analytic set of pure dimension \(n\) at the origin in \(\mathbb{C}^N\). There is a number \(\mu_0\geq 0\) such that if \(a_1\), \dots, \(a_m\) are germs of holomorphic functions at \(0\), \(l\geq 1\), and \(|\phi| \leq C |a|^{\mu+\mu_0+l-1}\) in a neighborhood of \(0\) in \(V\), where \(C\) is a positive constant, \(\mu=\max\{m,n\}\) and \(|a|^2=|a_1|^2+\dots+|a_m|^2\), then \(\phi\) belongs to the ideal \((a_1, \dots, a_m)^l \subset \mathcal{O}_0\). If \(V\) is smooth, then one can take \(\mu_0=0\) and this is the classical theorem of \textit{J. Briancon} and \textit{H. Skoda} [C. R. Acad. Sci., Paris, Sér. A 278, 949--951 (1974; Zbl 0307.32007)]. The general case is due to \textit{C. Huneke} [Invent. Math. 107, No. 1, 203--223 (1992; Zbl 0756.13001)], who proved it by purely algebraic methods. An analytic proof was given by \textit{M. Andersson} et al. [Ann. Inst. Fourier 60, No. 2, 417--432 (2010; Zbl 1200.32007)]. Effective versions of the Briançon-Skoda-Huneke theorem can be used to deduce effective versions of the Nullstellensatz. Let \(F_1, \dots, F_m\) and \(\phi\) be polynomials in \(\mathbb{C}[z_1, \dots, z_N]\) and assume that \(\phi\) vanishes on the common zero set of the \(F_j\). Then Hilbert's Nullstellensatz asserts that there are polynomials \(G_1, \dots, G_m\) such that \[ F_1 G_1 + \dots + F_m G_m =\phi^\nu \] for some large enough integer \(\nu\). It is an interesting and intensively studied problem to bound the complexity of the solutions \((\nu, G_1, . . . , G_m)\), e.g., in the sense of finding bounds for \(\nu\) and the degrees of the \(G_j\) in terms of the degrees of the \(F_j\). We should mention some important contributions of \textit{J. Kollár} [J. Am. Math. Soc. 1, No. 4, 963--975 (1988; Zbl 0682.14001)], \textit{L. Ein} and \textit{R. Lazarsfeld} [Invent. Math. 137, No. 2, 427--448 (1999; Zbl 0944.14003)], and \textit{Z. Jelonek} [Invent. Math. 162, No. 1, 1--17 (2005; Zbl 1087.14003)]. An account of the state of the art can be found in the introduction of the paper under review, which contains new important contributions in the area. The first main theorem of Andersson and Wulcan is as follows: Theorem A. Assume that \(V\) is a reduced algebraic subvariety of \(\mathbb{C}^N\) of pure dimension \(n\) and let \(X\) be its closure in \(\mathbb{P}^N\). {\parindent=6mm \begin{itemize} \item[(i)] There exists a number \(\nu_0\) such that if \(F_1, \dots, F_m\) are polynomials of degree \(\leq d\) and \(\phi\) is a polynomial and \[ |\phi| / |F|^{\mu+\mu_0} \text{ is locally bounded on } V, \] then there are polynomials \(Q_1, \dots, Q_m\) such that \[ F_1 Q_1 + \dots + F_mQ_m=\phi \] holds on \(V\) and \[ \deg(F_jQ_j) \leq \max\big\{\deg\phi + (\mu+\mu_0) d^{c_\infty} \deg X, (d-1) \min\{m,n+1\}+\text{reg}\;X\big\}. \] \item [(ii)] If \(V\) is smooth, then there is a number \(\mu'\) such that if \(F_1, \dots, F_m\) are polynomials of degree \(\leq d\) and \(\phi\) is a polynomial and \[ |\phi|/|F|^\mu \text{ is locally bounded on } V, \] then there are polynomials \(Q_1, \dots, Q_m\) such that \[ F_1Q_1 + \dots F_mQ_m=\phi \] on \(V\) and \[ \deg(F_jQ_j) \leq \max\big\{\deg\phi + \mu d^{c_\infty} \deg X+\mu', (d-1) \min\{m,n+1\}+\text{reg}\;X\big\}. \] If \(X\) is smooth, then one can take \(\mu'=0\). \end{itemize}} Here, \(\mu=\max\{m,n\}\), \(d^{c_\infty}\) is, in the sense of Fulton-MacPherson, the maximal codimension of the so-called distinguished varieties of the sheaf \(\mathcal{J}_f\) generated by the sections in \(\mathcal{O}(d)|_X\) corresponding to \(F_1, \dots, F_m\), and \(\text{reg }X\) is the Castelnuovo-Mumford regularity of \(X\subset \mathbb{P}^N\). One has \(0\leq c_\infty \leq \mu\), and let \(c_\infty=-\infty\) if there are no distinguished varieties. If one applies Theorem A to Nullstellensatz data, i.e., \(F_j\) with no common zeros on \(V\) and \(\phi=1\), then one almost gets back the optimal effective Nullstellensatz of Jelonek [Zbl 1087.14003]. One can use Theorem A also to deduce versions of the effective Nullstellensatz due to \textit{F. S. Macaulay} [The algebraic theory of modular systems. Cambridge: University press (1916; JFM 46.0167.01)] and \textit{M. Hickel} [Ann. Inst. Fourier 51, No. 3, 707--744 (2001; Zbl 0991.13009)]. The second main theorem in the paper under review is a generalization to non-smooth varieties of the geometric effective Nullstellensatz of \textit{L. Ein} and \textit{R. Lazarsfeld} [Invent. Math. 137, No. 2, 427--448 (1999; Zbl 0944.14003)]. To explain that, consider a projective variety \(X\) and an ample line bundle \(L \rightarrow X\). Then there is a smallest number \(\nu_L\) such that \(H^q(X,L^{\otimes s})=0\) for \(q\geq 1\) and \(s\geq \nu_L\). When \(X\) is smooth, then \(\nu_L\) is less than or equal to the least number \(\sigma\) such that \(L^{\otimes \sigma}\otimes K_X^{-1}\) is strictly positive by the Kodaira vanishing theorem. In particular, if \(X=\mathbb{P}^n\), then \(\nu_{\mathcal{O}(1)}=-n\). Theorem B. Let \(X\) be a reduced projective variety of pure dimension \(n\). There is a number \(\mu_0\), only depending on \(X\), such that the following holds: Let \(f_1, \dots, f_m\) be global holomorphic sections of an ample Hermitian line bundle \(L\rightarrow X\), and let \(\phi\) be a section of \(L^{\otimes s}\), where \[ s\geq \nu_L + \min\{m,n+1\}. \] If \[ |\phi| \leq C |f|^{\mu+\mu_0}, \] then there are holomorphic sections \(q_1, \dots, q_m\) of \(L^{\otimes (s-1)}\) such that \[ f_1 q_1 + \dots + f_m q_m =\phi. \] If \(X\) is smooth, then one can choose \(\mu_0=0\), and we get back the effective Nullstellensatz of Ein-Lazarsfeld. Let us give some remarks on the proofs of Theorem A and Theorem B. They rely on known geometric estimates and some new results on multivariable residue calculus. In the situation of Theorem B, when \(X\) is smooth, the key idea is to construct from the Koszul complex of the \(f_j\) a residue current \(R^f\) with support on their common zero set such that its annihilator determines locally the ideal \((f_1, \dots, f_m)\): if \(\phi\) is a section of \(L^{\otimes s}\) such that \(R^f\phi=0\) and \(L^{\otimes s}\) is positive enough, so that a certain sequence of \(\overline\partial\)-equations can be solved on \(X\), one obtains a holomorphic solution \((q_1, \dots, q_m)\) to the division problem. The solvability of the involved \(\overline\partial\)-equations is achieved by taking \(s\) large enough. This approach was basically introduced in [\textit{M. Andersson}, Ann. Inst. Fourier 56, No. 1, 101--119 (2006; Zbl 1092.32002)], and further developed in [\textit{M. Anderesson} and \textit{E. Götmark}, Math. Ann. 349, No. 2, 345--365 (2011; Zbl 1216.32002)] and [\textit{E. Wulcan}, Math. Ann. 350, No. 3, 661--682 (2011; Zbl 1234.32001); J. Commut. Algebra 2, No. 4, 567--580 (2010; Zbl 1237.14061)]. Let us mention in this context also the following important result of \textit{M. Andersson} and \textit{E. Wulcan} [Ann. Sci. Éc. Norm. Supér. (4) 40, No. 6, 985--1007 (2007; Zbl 1143.32003)], where they generalized the construction of the Coleff-Herrera product from complete intersection ideals to arbitrary ideals of holomorphic functions on complex manifolds. From a hermitian resolution of such an ideal \(\mathcal{J}\), they constructed explicitly a vector-valued current \(R^\mathcal{J}\) such that its annihilator in the holomorphic functions equals \(\mathcal{J}\): \(\text{ann}_{\mathcal{O}} R^\mathcal{J}=\mathcal{J}\). The really new achievement in the paper under review is the extension of the framework from [Zbl 1092.32002; Zbl 1216.32002; Zbl 1234.32001; Zbl 1237.14061] to the singular setting. For this, one has to replace \(R^f\) by variations of a product \(R^f\wedge \omega_X\), where \(\omega_X\) is a structure form of \(X\). The important notion of structure forms was introduced in [\textit{M. Andersson} and \textit{H. Samuelsson}, Invent. Math. 190, No. 2, 261--297 (2012; Zbl 1271.32009)]. Structure forms are basically the generators of the Barlet-Henkin-Passare holomorphic \(n\)-forms on \(X\) (\(\overline\partial\)-closed meromorphic \((n,0)\)-currents with support on \(X\)). For the proofs of the main theorems, it is then essential to control the singularities of \(\omega_X\).